You are on page 1of 282

STP 1468

Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects

R. A. Kishore Nadkarni, editor

ASTM Stock Number: STP 1468

ASTM International 100 Barr Harbor Drive PO Box C700 West Conshohocken, PA 19428-2959
INTERNATIONAL

Printed in the U.S.A.

Library of Congress Cataloging-in-Publication Data


Elemental analysis of fuels and lubricants: recent advances and future prospects/R.A. Kishore Nadkarni, editor. p. c m . ~ S T P ; 1468) Includes bibliographical references and index. ISBN 0-8031-3494-0 (alk. paper) 1. Fuel--Analysis. 2. Lubrication and lubricants--Analysis. I. Nadkami, R.A. II. Series: ASTM special technical publication; 1468. TP321.E46 2005 665.5'38---dc22 2005022779

Copyright 9 2005 ASTM International, West Conshohocken, PA. All rights reserved. This material may not be reproduced or copied, in whole or in part, in any printed, mechanical, electronic, film, or other distribution and storage media, without the written consent of the publisher.

Photocopy Rights Authorization to photocopy items for internal, personal, or educational classroom use, or the internal, personal, or educational classroom use of specific clients, is granted by ASTM International provided that the appropriate fee is paid to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923; Tel: 978-750-8400; online: http://www.copyright.com/.

Peer Review Policy


Each paper published in this volume was evaluated by two peer reviewers and at least one editor. The authors addressed all of the reviewers' comments to the satisfaction of both the technical editor(s) and the ASTM Committee on Publications. To make technical information available as quickly as possible, the peer-reviewed papers in this publication were prepared camera-ready as submitted by the authors. The quality of the papers in this publication reflects not only the obvious efforts of the authors and the technical editor(s), but also the work of the peer reviewers. In keeping with long-standing publication practices, ASTM maintains the anonymity of the peer reviewers. The ASTM Committee on Publications acknowledges with appreciation their dedication and contribution of time and effort on behalf of ASTM.

Printed in Baltimore,MD September2005

Foreword
This publication, Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects, contains selected papers presented at the symposium of the same name held in Tampa, Florida, on 6-8 December 2004. The symposium was sponsored by Committee D02 on Petroleum Products and Lubricants. The symposium chairman and editor was R. A. Kishore Nadkarni.

Contents
Overview Zen and the Art (or is it Science) of a Perfect Analysis--R. A. K. NADKARNI vii
i

A T O M I C EMISSION SPECTROSCOPY

Analysis of Gasoline and Diesel Fuel Samples by Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES), Using Pneumatic Nebulizer and Standard Spray Chamber---c. c. ONYESO Elemental Analysis of Lubricating Grease by Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES)--B. s. FOX The Use of Microwave Digestion and ICP to Determine Elements in Petroleum Samples--J. D. HWANG,M. HORTON,AND D. LEONG Advances in ICP-MS Technologies for Characterization and Ultra-Trace Speciation as a Tool for the Petroleum Industry--J. PASZEK,K. J. MASON,
A, S. MENNITO, AND F. C. MCELROY

17

24

33

42

Direct Trace and Ultra-Trace Metals Determination in Crude Oil and Fractions by Inductively Coupled Plasma Mass Spectrometry-S. DREYFUS. C. PECHEYRAN, C. MAONIER, A. PRINZHOFER. C. P. LIENEMANN. AND O. F. X. DONARD

51

Fuel Analysis by Filter Furnace Electrothermal Atomic Absorption Spectrometry--P. TII~rARELLL M. PRIOLA. S. RICCHIUTO, D. A. KATSKOV, AND
P. NGOBENI

59

Rotrode Filter Spectroscopy: A Recently Improved Method to Detect and Analyze Large Wear and Contaminant Particles in Fluids--M. LUKAS.
R. J. YURKO. AND D. P. ANDERSON

71

S U L F U R D E T E R M I N A T I O N AND X - R A Y FLUORESCENCE

Trace Levels of Sulfur in the Fuels of the Future: Analytical Perspective-R. A. K. NADKARNI

85

vi

CONTENTS

Analysis of Fuels, Lubricants, and Greases Using X-Ray Fluorescence Spectrometry--J. WOLSKA, B. VREBOS, AND P. BROUWER Determination of Sulfur Content in Crude Oil Using On-Line X-Ray Transmission Technology--s. FESS Low-Level Sulfur in Fuel Determination Using Monochromatic W D X R F ASTM D 7039-04--z. w. CHEN, F. WEI, I. RADLEY,AND B. BEUMER Latest Improvements on Using Polarized X-Ray Excitation EDXRF for the Analysis of Low Sulfur Content in Automotive Fuel--D. WISSMANN Rapid Determination of Sulfur in Liquid Hydrocarbons for At-Line Process Applications Using Combustion/Oxidation and UV-Fluorescence Detection--s. TARKANICAND J. CRNKO Pyro-Electrochemicai On-Line Ultra Low Sulfur Analyzer--J. R. RHODES DP-SCD and LTMGC for Determination of Low Sulfur Levels in Hydrocarbons--R. L. GRAS,J. C. LUONG. R. V. MUSTACICH,AND R. L. SHEARER

98

108

116

128

137 152

164

MERCURY DETERMINATION

Sampling and Analysis of Mercury in Crude Oil--s. M. WILHELM.


D. A. KIRCHGESSNER. L. LIANG, AND P. H. KARiHER

181

Determination of Total Mercury in Crude Oil by Combustion Cold Vapor Atomic Absorption Spectrometry (CVAAS)--B. s. Fox. K. J. MASON.AND
F. C. MCELROY

196

Mercury Measurements in Fossil Fuels, Particularly Petrochemicals-P. B. STOCKWELL, W. T. CORNS, AND D. W. BRYCE

207

OTHER HETEROATOMS

Recent Advances in Gas Chromatographic/Atomic Emission Hetero-Atom Selective Detection for Characterization of Petroleum Streams and eroducts--F, p. DISANZO AND J. W. DIEHL Improvements in the Determination of Fluorine in Fuel and Lubricants by Oxidative Combustion and Ion-Selective Electrode Detection--L J. NASH Phosphorus Additive Chemistry and its Effects on the Phosphorus Volatility of Engine Oils--T. w. SELBY, R. J. BOSCH. AND D. C. FEE Analysis of the Volatiles Generated During the Selby-Noack Test by 31p NMR Spectroscopy--R. J. BOSCH, D. C. FEE, AND T. W. SELBY Index

221

232

239

255 275

Overview
In spite of being a mature science, elemental analysis continues to play a vital role in product manufacturing and quality characterization in many sectors of all industries. Research divisions in both industry and academia continue devising new ways of lowering the elemental detection limits so that even the minutest amounts of elements in products could be determined in as accurate and precise a fashion as possible. The ASTM International D02 Committee on Petroleum Products and Lubricants through its Subcommittee 3 on Elemental Analysis has played a large and crucial role in the last several decades in standardizing numerous elemental analysis methods used in the oil industry. Currently there are about 75 standard test methods under the jurisdiction of SC 3, and additionally at least 6 more are under active development and moving towards standard designations. I have no doubt that this activity will continue in the future. These standards comprise virtually all known modem techniques for elemental analysis of petroleum products and lubricants. The first ASTM D02 symposium on this subject was held in New Orleans in December 1989 at which 20 papers were presented. Of these, 13 were published as a book, Modern Instrumental Methods of Elemental Analysis of Petroleum Products and Lubricants, ASTM STP 1109. The current and second "quindecennial" (i.e., every 15 years) was held in Tampa, Florida in December 2004. This was attended by over 120 people. Thirty papers were presented on diverse subjects from 64 authors from nine different countries: Brazil, France, Germany, Italy, the Netherlands, South Africa, Switzerland, U.K., and U.S. Of these, 12 papers were from the oil industry, 15 from the instrument manufacturers, l0 from national research organizations, and 4 from the universities. The objective of this symposium and this book is to acquaint the readers with the latest advances in the field of elemental analysis and to focus on what avenues of future research to explore in this area. The subjects included are various elemental analysis techniques such as atomic absorption spectrometry, inductively coupled plasma emission and mass spectrometry, isotope dilution mass spectrometry, X-ray fluorescence, ion chromatography, gas chromatography-atomic emission detection, other hyphenated techniques, hetero-atom microanalysis, sample preparation, reference materials, and other subjects related to matrices such as petroleum products, lubricating oils and additives, crude oils, used oils, catalysts, etc. Of the 30 papers presented at the symposium, 23 papers were published in the Journal of ASTM International (JAI), and are included in this ASTM publication. As far as possible, the papers have been arranged by analytical techniques used, although in some cases there is some overlap: ICP-AES, XRF, sulfur, mercury, other hetero-atoms. The first article is from the plenary lecture given at the symposium by the symposium chairman Kishore Nadkarni. It covers total quality management practices advocated for obtaining a "perfect" analysis. Proper staff training, sampling, calibration and quality control practices, adherence to test method details, participation in proficiency testing, accreditation from national bodies, benchmarking, etc., are some of the critically important approaches that need to be taken to achieve the ideal state of analytical Zen perfection.

Atomic Spectroscopy
Among the seven atomic spectroscopy papers in this book, five concern various aspects of ICP-AES, a technique widely used for the determination of metals in petroleum products vii

viii

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

and lubricants. Onyeso (Ethyl Corporation) presents an ICPAES method for the determination of additive elements and wear metals, principally manganese, in gasoline and diesel fuels, with simple dissolution in kerosene and using yttrium internal standard. Accessories such as direct injection nebulizer, ultrasonic nebulizer, chilled spray chamber, etc., were not necessary for this analysis. Fox (ExxonMobil Research and Engineering) presents an ICPAES method for the determination of additive elements and wear metals in lubricating greases. Since such samples cannot be directly nebulized in the ICP plasma, alternate sample dissolution techniques were employed: dry sulfated ashing, microwave assisted dry ashing, microwave assisted acid digestion with both open and closed vessels. This method is being developed into an ASTM standard test method and is expected to be published by YE05. Hwang and Leong (ChevronTexaco) also discuss the use of microwave acid digestion for sample preparation before ICPAES measurements. Elemental speciation using mass spectrometry in conjunction with ICPAES is a latest advance in atomic spectroscopy, which is becoming popular in analytical research labs. Mason et al. (ExxonMobil Research and Engineering) show how linking ICP-MS to various liquid chromatographic techniques has enabled determination of ppm levels of metals in hydrocarbons to ppb level measurements in refinery effluent streams. Hyphenated ICP-MS techniques were used to provide speciation information on nickel and vanadium in crude oils and assist in development of bioremediation options for selenium removal in wastewater treatment plants. Similar ICP-MS technique without sample demineralization was used by Lienemann, et al. (lnstitut Francais du Petrole) to determine the trace and ultra-trace amounts of metals in crude oils and fractions. Lukas et al. (Spectro Inc.) describe an improvement made in rotating disc electrode atomic emission technology by incorporating a filter device in the rotrode, which enables to detect particles greater than 10 i~m size. Tittarelli et al. (SSC, Milan) employed a transverse heated filter atomizer with atomic absorption spectrometry to determine a number of trace elements in automotive and jet fuels. Sub-ppm detection limits were obtained. The use of filter furnace reduces the risk of elemental loss during drying and pyrolysis steps, and decreases the interferences due to molecular absorption and light scattering.

X-Ray Spectroscopy
Similar to atomic emission spectroscopy, equally widely used technique for elemental analysis in the oil industry is X-ray fluorescence (XRF). There are four papers in this book using this technique, three of which deal with the determination of sulfur in gasoline and diesel. Wolska et al. (Panalytical BV) compared performance of three XRF technologies: high power and low power WDXRFs and a bench top EDXRE There are large differences in the sensitivities and hence varying lower limits of detection or qualification and sample throughput, for these technologies.

Sulfur Analysis
One of the most important analyses done today on petroleum products, particularly gasoline, reformulated gasoline, and diesel, is for low levels of sulfur. Government regulations on sulfur emissions from automobiles and other combustion sources have steadily increased; hence, the increasing interest in devising precise and accurate methods for trace and ultra-

OVERVIEW

ix

trace amounts of sulfur in fuels of the future as evident from seven papers on this subject published in this book. Nadkarni (Symposium Chairman) reviewed the alternate methods available for sulfur determination in fuels. Out of about 20 ASTM standard test methods available, only about five (D 2622 WDXRF, D 3120 microcoulometry, D 5453 UV-fluorescence, D 6920 pyroelectrochemical, D 7039 MWDXRF) are appropriate for ultratrace amounts of sulfur in gasoline or diesel. However, in their actual industrial use only D 2622 and D 5453 predominate. Chen et al. (XOS Inc.) describe a newly developed technology instrument based on monochromatic WDXRF for low sulfur analysis of fuels. The instrument has a significant advantage over existing WDXRF instruments in terms of increased sensitivity and improved signal to noise ratio. This technique has been recently given the ASTM designation D 7039. Another new instrument recently developed for sulfur by XRF determination is described by Wissmann (Spectro, Inc.). This method uses polarized EDXRF, considerably reducing background scatter, and achieving detection limit comparable to that of WDXRF. Recent developments in detector technology and in closed coupled static geometry have resulted in further improvement of sensitivity for this application. This method is also in the developmental stage for ASTM method designation. Shearer et al. (Ionic Instruments and Dow Chemicals) describe a novel technique developed tbr low levels of sulfur in hydrocarbon matrices using a low thermal mass temperature programmable and dual plasma chemiluminiscence detector. The method with appropriate modification can measure individual sulfur species similar to ASTM method D 5623.

On-line Sulfur Analysis


Increasingly refineries, plants, and pipeline operators are focusing on obtaining quick turnaround for sulfur analysis rather than wait ['or time-delayed laboratory analysis. A large number of such installations are being operated in the industry around the world. Three papers in this book discuss applications of such on-line technology for sulfur determination in fuels. In an on-line application of X-ray transmission technology, Fess (Spectro, Inc.) describes the basis of this technology and its application to classification and blending of crude oils that contain between 0.1 and 3.3 m % sulfur. Commercial instruments based on this technology are being used in the field. In a second on-line application paper, Tarkanic and Crnko (Antek/PAC) describe an online instrument based on ASTM Test Method D 5453, UV-Fluorescence Detection. The latter is a widely used method in the oil industry for low and ultra-low levels of sulfur. The online instrument appears to be very stable and fast (< 1 min per analysis) over extensive periods of field operations. In a third on-line application paper tbr sulfur analysis, Rhodes (Rhodes Consulting), ASTM Test Method D 6920 is applied for on-line application. This method uses pyro-combustion followed by electrochemical detection.

Mercury Determination
Although adverse effects of mercury emissions on environment and humans has been known lbr decades, in recent years there has been concern regarding the mercury content of crude oils, and its emission through petroleum refining process. There are three articles in this book discussing this issue. Wilhelm et al. (Mercury Technology Services/EPA et al.) provide a review of the presence of mercury in various parts of the world, its speciation, and alternate methods of determining

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

low ppm and sub-ppm levels. Fox et al. (ExxonMobil Research and Engineering) describe a method for the determination of ppb levels of mercury in crude oils and distillation cuts using combustion cold vapor atomic absorption spectrometry technique. Stockwell et al. (PS Analytical Ltd.) describe the technique of atomic fluorescence spectrometry for the determination of mercury both before and after mercury removal from petrochemicals. The technique has been used for on-line measurements in installations operating around the clock for at least 2 years.

Other Heteroatoms
DiSanzo and Diehl (ExxonMobil Research and Engineering) used GC-AED for the determination of elements such as carbon, nitrogen, sulfur, oxygen, and phosphorus in fuels and petroleum fractions. A simplified version of comprehensive GC x GC is coupled with atomic emission detector to reduce the hydrocarbon matrix interference using simple and rugged modulation along with rugged wide bore capillary columns. The technique together with other spectroscopic techniques such as GC-MS can provide information on many selected elements and compounds that may be present in fuels as additives or contaminants. In a pair of papers, Selby et al. (Savant, hzc. and Astaris LLC) describe using phosphorus as an indicator of volatility of engine oils. Phosphorus is volatilized during Noack volatility test (ASTM D 5800). The volatile material is trapped and analyzed for total phosphorus using ICP-AES, and for phosphorus species using 3~p NMR spectroscopy. An oxidative combustion followed by ion selective electrode detection method is proposed by Nash (Antek/PAC) for the determination of fluorine in fuels and lubricants. An ASTM method based on this technique is in development stage.

Unpublished Symposium Papers


Some papers were presented at the Symposium; however, they were not submitted for publication by the authors. Nevertheless, they represent interesting approaches to some specific elemental analysis issues in the petrochemical industry. It would be useful if the authors eventually publish these articles for the benefit of others in the industry. These presentations include the following: I. Kelly et al. (NIST) describe an isotope dilution thermal ionization mass spectrometry method for the determination of sulfur in fossil fuels. The method is being used in NIST for certification of a number of liquid fuels at low sulfur concentration levels. 2. Kelly et al. (NIST) also describe a "designer" calibration standard method for sulfur determination in fossil fuels for users to prepare NIST traceable working standards with known concentrations and uncertainties. 3. Manahan and Chassaniol ( Cosa Instruments and Dionex) describe an oxidative combustion followed by ion chromatographic conductometric method for the determination of a number of nonmetallic elements such as sulfur and halogens in liquid and gaseous hydrocarbons. A standard based on this technique is under development in ASTM for designation as a standard method. 4. Long et al. (NIST) describe another method for mercury determination in crude oils using isotope dilution-cold vapor-inductively coupled plasma-mass spectrometry technique. The method has very high sensitivity, very low blank and high accuracy. The technique is being used to determine mercury in a large number of crude oil samples from Department of Energy strategic petroleum reserve in the mercury concentration range of 0.02-10 ng/g.

OVERVIEW

xi

5. Finally, Mason et al. (ExxonMobil Research and Engineering) describe the approaches used for assay of fresh and spent reformer catalysts to determine the precious metals (platinum and rhenium) in them. Methods such as WDXRF, ICPAES, and classical wet chemistry methods are used for such analysis. Precise and accurate methods are critical for these analyses, since small errors in analysis can have a large impact in commercial transactions of these catalysts between the catalyst vendors and the oil companies. Hopefully, the papers included here will provide the readers with the current state-of-theart and future research trends in the field of elemental analysis in the oil industry. Most modern techniques used in the field are represented here.

Acknowledgment
I want to thank various ASTM staff members (particularly David Bradley, Dorothy Fitzpatrick, Crystal Kemp, Hannah Sparks, and Roberta Storer) for their prompt response and cooperation that made the symposium and subsequent efficient publication in JAI and of this volume possible. My thanks are also due to the reviewers who did a very good job of providing technical reviews of all original paper submissions. Their invaluable assistance in reviewing the papers made the final publication a much better quality product.

R. A. Kishore Nadkarni
Chairman, D02.SC 3 and Symposium Chairman

Journal of ASTM International, March 2005, Vol. 2, No. 3 Journal of ASTM International, March 2005, Vol. 2, No. 3 Paper ID JAI12964 Available online at www.astm.org

R. A. Kishore Nadkarni 1

Zen and the Art (or is it Science) of a Perfect Analysis


ABSTRACT: An analytical laboratory in any industry plays a crucial role in product quality management and ultimate customer satisfaction. Some factors need to be considered for an aspiring laboratory to become a perfect performer. These range from sampling, calibration, contaminationcontrol, and use of valid test methods to statistical quality assurance. Some approaches may be utilized to achieve a perfect analysis including: staff training, participation in proficiency testing, use of standard reference materials in the analytical sequence, internal and external audits, agency accreditation, continuous improvement program, benchmarking, etc. Laboratories managed in this way show demonstrated superiority in data precision and accuracy over the labs which do not practice such quality management. Well-managed industrial laboratories can have insignificant laboratory sigma compared with manufacturingvariability in the plant production. For a flawless perfect analysis, determinationto excel, mental discipline to stay the course, willingness to overcome inertia and resistance, and focus on producing a perfect analysis at all levels of laboratory staff are essential. KEYWORDS: analysis, quality management,perfect analysis As Robert Pirsig wrote in his landmark iconic autobiographical novel "Zen and the Art of Motorcycle Maintenance," the art of motorcycle maintenance is primarily a mental phenomenon [1]. One may have the tools, but unless there is mental preparation to achieve high goals, the tools alone will not help. A similar mindset is needed to achieve excellence in a laboratory to make it into a perfect laboratory that produces flawless performance. Tools may be available, but if there is no organizational passion and will to excel, the laboratory will not become a perfect laboratory. The culture of excellence must be pervasive throughout the laboratory organization from the laboratory manager to the laboratory technician. Higher management especially needs to show through visible actions that only the best will do. Perfection cannot be achieved through shortterm stop-gap measures. A long-term improvement plan must be in place and followed upon to be effective. What is a perfect laboratory? It is a laboratory which delivers the product (i.e., accurate and precise data) on time; if necessary, continuously improves on itself; makes the analysis "Right the First Time," thus eliminating repeat analysis and giving erroneous information to the customers; communicates with its customers and sometimes educates them when necessary. This laboratory cares about the success of its customers' business.

Customer Services
A laboratory is a microcosm of its parent organization. The product delivered from a laboratory is quality data. Hence, the primary objective of a laboratory should be to be the best in quality. A laboratory needs to deliver a consistent product on time which meets or exceeds customers' expectations, and which increases customers' confidence in the laboratory's
Manuscript received 7 September 2004; accepted for publication19 October 2004; published March 2005. i Millennium Analytics,Inc.,East Brunswick, NJ 08816.

Copyright9 2005 by A S T M International, Ban"Harbor Drive,PO Box C700, West Conshohockcn,PA 19428-2959. I00

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

reliability and dedication to quality. A customer needs to know that the laboratory cares. "Total Care" is the sum of impressions formed during contact with the customers. A perception by the customer that the laboratory is a caring organization can convert a customer from being forced to be a customer to becoming a customer by choice. Substantial or continuing violations of a customer's justified expectations will cause the customer to feel that the laboratory, organization simply does not care.
Pillars to Build a Perfect Laboratory

There are at least twelve components which help to produce flawless laboratory performance. In the approximate order in which an analysis is performed, these include but may not be conEmed to the following (See Fig. 1 on page 3): 1. Training 2. Representative Sampling and Contamination Control 3. Calibrations 4. Technical Details of Test Methods 5. Statistical Quality Assurance 6. Use of Certified Reference Materials 7. Documentation 8. Internal and External Audits 9. Proficiency Testing 10. External Agency Accreditation 11. Benchmarking 12. Ethics

Training
As Mark Twain reportedly said, "Training is everything. Cauliflower is nothing but a cabbage with a college education." The point is that without adequate training, staff cannot produce the best results. Training courses should include periodic refresher and new technology courses to improve the technical ability of staff members. These courses will benefit the staff members by helping them to think through the analysis rather than mechanically doing the test, to identify ways to improve the methodology, to obtain better precision and accuracy, and to improve the turnaround time. Some of the areas in which a laboratory staff member must be fully trained include safety, data security, laboratory instrumentation, test methods used, calibration protocols, statistical quality assurance, use of certified reference materials, long term analytical needs and goals, and ethical behavior. In today's culture, it is still up to the supervisors and management to filter the Zen attitude down to the working level people. The very fundamental first step toward obtaining a perfect analysis is through staff training. Dr. Derek Bok, former president of Harvard University once commented, "If you think education is expensive, try ignorance."

Sampling
Obviously the first critical step in any analytical sequence is the integrity and validity of a sample being analyzed. More often than not this sampling step does not involve actual laboratory

NADKARNI ON PERFECT ANALYSIS

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

staff; usually the chain of custody for a sample starts with the receipt of the sample in the laboratory. Once the laboratory acquires the sample, however, it is the laboratory's responsibility to have a system for unique identification of each sample, sample handling, storage and retention procedures, as well as safe disposal procedures. Identification of the population from which the sample is to be obtained, selection and withdrawal o f valid gross samples of this population, and reduction of each gross sample to a laboratory sample suitable for the analytical technique to be used are some of the key steps to be considered in obtaining a representative sample for analysis [2]. Equally important is documented chain of custody procedures to authenticate and maintain the sample integrity. Several ASTM standards deal with sampling aspects for the analysis of petroleum products and lubricants: * 9 9 o 9 D 4057: D 4177: D 4840: D 5842: D 5854: Manual Sampling of Petroleum and Petroleum Products Automatic Sampling of Petroleum and Petroleum Products Sampling Chain of Custody Procedures Sampling and Handling of Fuels for Volatility Measurements Mixing and Handling of Liquid Samples of Petroleum and Petroleum Products

Additionally, some ASTM standards give instructions for specific sampling requirements for specific analytical tests. Attention must be paid to these caveats to obtain reliable test results.
Contamination Control

Gross contamination of the sample in any analysis and in particular for trace analysis is a serious problem which, if unchecked, will completely negate the validity of the analytical results. The problem can become particularly insidious as one is working in the range of ppm and subppm levels of analytes. Contamination from particulates in the air, impurities in reagents, trace elements from the sample containers as well as glass- or plastic-ware used during analyses are all potential sources of contamination [3]. An accompanying "blank" sample used throughout the analysis sequence may or may not accurately measure the extent of contamination, since such contamination from air or glassware, etc. may not be uniformly present when in contact with the blank and a real sample. The point is that both a blank determination and a rigorous protocol for contamination control in the laboratory are essential for obtaining perfect results, particularly in the area of trace analysis. An excellent source book for discussion on contamination control is given in [4].
Calibration or Verification

Virtually all analytical test methods require some form of calibration or verification before actual samples are analyzed. Different test methods require different calibration intervals. Thus, a decision about appropriate calibration frequency must be made on a ease by case basis. There is a tendency among many laboratories to do the bare minimum calibrations similar to their approach toward quality control requirements. This is not the way to achieve superior performance. Moreover, if an instrument is out-of-calibration, under no circumstances can data from that instrument be reported to the customers. Appropriate calibration standards must be utilized during analysis. A wide variety of such standards are available from commercial sources, NIST, etc. Many laboratories have capabilities

NADKARNI ON PERFECT ANALYSIS

of preparing reliable in-house standards. Calibration standards identical to the samples being analyzed would be ideal, but failing that, at least some type of standards must be used to validate the analytical sequence. In physical measurements this is usually achievable, but it is often difficult or sometimes almost impossible in chemical measurements. Even the effects of small deviations from matrix match and analyte concentration level may need to be considered and evaluated on the basis of theoretical or experimental evidence. Sometimes the use of standard additions technique to calibrate the measurement system is a possibility. But because an artificially added analyte may not necessarily respond in the same manner as a naturally occurring analyte, this approach may not be always valid, particularly in speciation work9 An important aspect of calibration is the decision on calibration intervals, i.e., the maximum period between successive recalibrations. Two basic and opposing considerations are involved: the risk of being out of tolerance at any time of use and the cost in time and effort. The former should be the major concern because of the dilemma of what to do with the data obtained during the interval between the last known in and the first known out o f calibration. However, an overly conservative approach could be prohibitively expensive. A realistic schedule should reduce the risk of the former without undue cost and disruption to work schedules. The factors that need to be considered in a realistic schedule include:
9 0

9 9 9 9 9 9 9

Accuracy requirement for the measured data Level of risk involved Experience of the laboratory in use o f the equipment or methodology Experience of the measurement community Manufacturer's recommendations External requirements for acceptability o f data Cost of calibration and quality control

Quality control measurements can help a great deal in deciding what calibration frequency intervals should be used.

Test Method Details


A laboratory must have fully documented test methods that are used for analyses, and the staff members must be competent in the details in each test method that they will be using for analyses. Experience has shown that a major source o f analytical error is deviation from the prescribed standard test method, whether intentional or inadvertent. Most of the time the details given in a standard test method are there for a purpose. A laboratory wishing to deviate from the standard test method must document the deviation and show that the modified version produces statistically equivalent or (preferably) better results in terms o f precision and accuracy. In most laboratories, staff members periodically and sometimes frequently change; hence, it is important to have a system in place for periodic checking that the laboratory practice is indeed in conformance with the test method requirements9

Statistical Quality Control and Assurance


One cannot control what one cannot measure, and one cannot improve what one cannot control. Every measurement system is beset with variation and noise, and the only way to control

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

and reduce variation is by identifying its cause, establishing its extent, and interpreting what it is indicating. Variability arises because no two things are absolutely alike either in nature or in laboratory measurements. All one can try to do is to minimize it as much as possible. The primary step in measuring variance is the use of statistical quality control (SQC) or assurance (SQA) charts. Quality assurance must be viewed as an integral part of a complete analytical sequence and not as an added burden with additional costs. Calibration("--) Sample Analysis<---) SQA The only way to prove data integrity and reliability is through SQA. The frequency of quality control (QC) analysis will depend on the type of analysis, the instrument involved, stability of the measurement process, importance of decisions based on the test results, and magnitude of changes between batches. The same considerations described in the calibration section above need to be involved in making the decision on SQA frequency. But at least one QC standard must be analyzed with each set of samples to be analyzed. Preferably one QC standard each should be analyzed bracketing a series of sample analyses. Ideally, a few QC standards should be interspersed with samples to continuously monitor the data quality. Indeed there are some ASTM test methods where such protocol is mandatory, e.g., D 4951 and D 5185 ICP-AES methods for metals analysis. A general rule of thumb used in many laboratories is one QC sample per 5-10 actual samples. Although there are statistical protocols for reducing the QC frequency, this is overridden if a customer (and many do) requires QC accompanying their sample analyses. If a QC analysis indicates poor performance, immediate remedial action must be taken before continuing the analysis or reporting the data to the customers. Control Charts--A picture is worth a thousand words, and this is definitely true of the QC data as well. A chart based on statistical control limits must be plotted and interpreted on time. A control chart by itself can only indicate that there is an upset and not the cause of the upset. Laboratory staff must investigate, identify, and eliminate the cause(s) of the upset and bring the process back to a state of statistical control. If a control chart is merely plotted and filed away without taking action when required it is simply a waste of time and effort. There are a number of statistical run rules which are used to indicate out-of-control situations. Some laboratories use only one rule of a data point beyond 3 sigmas. This in itself may not be adequate, since it will ignore several short term and more importantly long term indications of system instability. Based on the QC data, monthly or quarterly laboratory capability should be calculated to compare the laboratory precision with that given in the standard test method where available. One way of estimating lab capability is the the equation below: Test Performance Index = Laboratory Standard Deviation / Test Method Repeatability Further discussion on the numerical criteria to be used for evaluating TPIs can be found in ASTM standard D 6792. Some laboratories prefer to compare their standard deviation against the test method reproducibility or to use an inverse calculation from that given above. In either case an individual best laboratory will consistently produce a superior precision compared to the industry average precision. Such periodic performance feedback should be a key feature of any continuous improvement program. Such a continuous improvement (CI) program can be initiated at any level and may include obtaining a better instrument or changing to a better alternate test method. Since the lab worker

NADKARNI ON PERFECT ANALYSIS

staff is usually busy in day-to-day analyses, the initiative for such CI program usually needs to come from supervisory or managerial staff. Such a program can continue indefinitely looking at various ways of improving diverse areas of laboratory operations. A project ends, but a process continues!

Use of Certified Reference Materials


Where available, reference materials need to be used for calibration or quality control. The National Institute of Standards and Technology (NIST) and many commercial sources supply such materials. Many high performing technology laboratories are capable of preparing their own reference materials using the same approach used by NIST in certifying their Standard Reference Materials (SRM). The calibration reference materials (RM) should be traceable to national standards, and the traceability must be preserved in the laboratory's documentation system. It is not a good idea to use the same material for both a calibration standard and quality control. The latter should be similar in matrix to the type of samples being analyzed, although this is not always possible due to the lack of availability of suitable certified materials. At least for QC it is easy enough to obtain a reasonably large quantity of plant product, analyze it by the tests of interest multiple times, and calculate the average value and the sigma limits to initiate the control chart.

Documentation
Laboratory analytical reports must contain all information necessary for a customer to understand or interpret the results. This may include the unique laboratory sample ID, equivalent customer sample identification, test methods used and any modifications done to them during analysis, actual analytical results, supporting QC results where available, signature of responsible authorized laboratory staff, etc. Records of calibration or QC records must be maintained for an adequate period of time. Errors in Data Reporting--Reporting blunders occur more frequently than one might think. We have often seen the laboratories taking part in the ASTM interlaboratory erosscheck programs reporting the data without paying adequate attention to the report formatting. A perfect analysis may have been done, but if the results of this perfect analysis are reported incorrectly, nothing is gained. We have repeatedly found that the numbers are often transposed, or wrong decimal places are used. When we check back with these laboratories, almost always it turns out that there was indeed an error in reporting. Some of the glaringly obvious examples that we have come across include: 9 The total metal content is <100 ppm, yet the ash (D 482) is reported as 3 m%. 9 The ash (D 482) and sulfated ash (D 874) results are widely different by an order of magnitude. 9 The kinematic viscosity (D 445) of a fuel sample is reported as 6300 cst @ 100~ that would make this a practically solid sample. 9 One laboratory reported 1000~ as the flash point (D 93) of a fuel sample! 9 The pour point (D 97) of a fuel oil sample was reported as 54~ that would make it a solid oil sample at room temperature.

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

9 9

The sulfur content (D 2622) of a reformulated gasoline sample is reported as 94 %! This should have been 94 ppm. The water content (D 6304) of a lubricating oil sample is reported as 9 %. This should have been 9 ppm.

All of these errors could have been prevented easily by reviewing the data before reporting. One would not send a defective product from a plant to a customer. Why then send the defective data out? It takes only a few moments to go over the final report form to check for such errors; a laboratory's credibility would be much better for having done so.

Internal and External Audits


Every laboratory striving for perfection needs to do periodic audits to check that its quality systems are still working properly. Audits o f test methods should be conducted (perhaps annually) to confirm adherence to the documented test methods. The performance of the entire test should be observed and checked against the official specified test method. Many errors in results are derived from taking liberties with the test method's detailed requirements. It needs to be recognized that the ISO 9000 audits alone by outside registrars do not adequately cover the technical aspects of a laboratory performance. It is imperative that internal audits be done by the laboratory staff since they are far more familiar with the required technical details of each test that the laboratory performs. Through such thorough reviews, a laboratory can eliminate defects, improve on its strengths, and become a perfect laboratory satisfying customer expectations. Findings and recommendations of internal or external audits must be promptly reviewed and acted upon, otherwise there is little point in spending efforts in doing audits.

Proficiency Testing
A laboratory needs to participate in relevant crosschecks organized within the company circuit, with its customers if requested, or with the industry organizations. A completely unbiased assessment of a laboratory's capability for precision and accuracy can only be judged by a blind crosscheck. If the laboratory results are equivalent to those of other laboratories in the industry, it assures the laboratory itself and its customers that the laboratory is producing reliable results. On the other hand, if the results are less than satisfactory, it also helps the laboratory to correct its operation, which may be deficient in some aspect. Participating in a crosscheck without following up on deficiencies if detected is a waste of resources. One such program widely used in the oil industry is described by Nadkami and Bover [5]. Participation in proficiency testing should not be considered as a substitute for in-house quality control and vice versa. These are two independent and necessary activities for a well-managed laboratory.

Laboratory Accreditation
Laboratory accreditation in the U.S. is handled by more than 150 accrediting private and quasi-government bodies. Thus, there is a bewildering array to choose from as to which is the best accreditation for a laboratory. This will depend to some extent on the industry and the customers with which a laboratory does business. The most well known program is ISO 9000

NADKARNI ON PERFECT ANALYSIS

registration. Often when a chemical plant or a refinery gets ISO 9000 registered, the testing laboratory associated with that site is also registered. It must be recognized however that having an ISO 9000 registration does not necessarily guarantee the best laboratory data unless supplemented by additional internal quality protocols [6,7]. Two other meaningful accreditation programs for the testing laboratories are by the American Association for Laboratory Accreditation (A2LA) [8] and ISO Guide 17025 [9]. There is a similar National Measurement Accreditation Service (NAMAS) in the United Kingdom and parts of Europe, which is administered by the U.K. National Physical Laboratory [10]. An advantage of these three accreditation schemes is that they involve actual test method performance, laboratory quality system, and other issues vital to a credible laboratory's performance. Given the widespread utilization of such accreditation schemes, many customers will forgo their own quality audits of contractor laboratories if the latter have such external certifications.

Benchmarking
Benchmarking is the search for those best practices that will lead to a company's superior performance, and it incorporates these practices in one's own organization. To become the best, one must learn from the best. This learning should not be limited only to the organizations in one's industries; companies outside a specific industry may also have superior standards which can be utilized beneficially. Ideas should be shamelessly stolen from the best, provided they are not company proprietary materials. Japanese people call benchmarking "Dontotsu," literally meaning "striving for the best of the best." Xerox Corporation was a pioneer in industrial benchmarking. An excellent text on this subject has been published by Xerox [1 t]. Some Baldrige National Quality Award winning companies incorporate benchmarking as an essential tool in their quality management process. Perfect laboratories also need to look at their counterparts to fred out how to improve themselves to the level o f the best laboratories. Many laboratories calculate the so-called Laboratory Productivity Index (LPI), defined as the total tests/total laboratory workers per month or per quarter, etc. However, experience shows that these indices may vary widely from laboratory to laboratory. It is not certain that LPI is a reliable measure for interlaboratory comparison, since: a. Different tests need different amounts of time to complete. b. The same analysis may be done automatically by analyzers in some laboratories and manually in others. c. The test requirements can be different depending on the types of products analyzed. d. The clientele and the test slate can be different for research and production laboratories. LPI may be useful to track laboratory productivity in one site but needs caution when comparing one laboratory against another.

Ethics
One would think that it is a simple fact that the results obtained in the laboratory should be reported honestly; however, several incidents reported in last few years make one wonder

10

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

whether ethics should be a mandatory training course in colleges and laboratories for analytical chemists! In April 1996, the U.S. Department of Justice t'med a major additives company $4.75 million for providing additive packages to the government falsely certifying that they met the military product specifications and passed specified testing. A Linden, NJ commercial laboratory was fined $1 million in 2001 for doctoring the laboratory reports that showed that they met EPA's Clean Air Act standards for cleaner burning fuels. The company was also put on probation for three years, and its former president was convicted of fraud. Some of the causes of the erroneous and false data can be ascribed to sloppiness in analysis, lack of training, use of improper techniques, inappropriate data manipulation, taking shortcuts in analysis, failure to follow prescribed steps in the test methods, improper calibration of instruments, etc. In a practice called "dry labbing," a worker merely copies the data from a previous sample and submits it as a new sample analysis with a new identification number. Commercial, business, and regulatory decisions based on such inaccurate data can often lead to wrong decisions for all affected parties. Nancy Wentworth, Director of quality staff at EPA's Office of Environmental Information, sums up advice to laboratories as: Get the right data; get the data right; and keep the data right [12].

Laboratory QualityImprovementthrough Quality Management


Using all or many steps outlined above, is it possible to achieve ultimate laboratory perfection? Is such a thing as a "perfect laboratory" or "perfect analysis" possible at all? The answer is in the affirmative if efforts are spent on the underlying foundation of good laboratory practices. These are: 9 9 9 9 The tests must be conducted exactly as written without any deviations. The calibrations, if part of a test, must be carried out properly; if necessary, each time the test is carried out. Quality control must be an integral part of the analytical sequence. The QC data must be plotted, the trends analyzed, and corrective actions taken when so indicated. At least annually, all tests performed in a laboratory must be audited by personnel other than those performing these tests or managing the laboratory to check that all details of the test method are properly followed. Any deviations found must be corrected promptly.

Performance of a group of laboratories which are well-controlled in terms of the above requirements has been shown to be superior to that of other industry laboratories based on the inhouse repeatability and their performance in ASTM interlaboratory crosscheck programs. Their repeatability in the laboratory is even superior to that advocated in the ASTM test methods themselves. Such precise data have implications beyond simple good analysis. A properly calculated product specification is usually set based on both the manufacturing and analytical testing variance. Total sigma = (Manufacturing sigma2 + Analytical sigma2) ~

NADKARNI ON PERFECT ANALYSIS

11

If a testing laboratory is being managed to its top performance levels, often the analytical sigma becomes virtually insignificant compared to the manufacturing variability resulting in the width of a product specification being dependent almost solely on the manufacturing variance. Thus, a superior analytical precision helps in both manufacturing a reliable product, eliminating quality complaints particularly in regards to the data, and keeping the customers happy and loyal.

Lessons from Baldrige Award Winners


Since the inception of the Malcolm Baldrige National Quality Award in 1988, a number of U.S. organizations has been recognized by the U.S. Department of Commerce as the best in total quality management in the country. There are a number of lessons to be learned from these winners which are equally valid in the laboratory environment to convert it into a perfect laboratory producing perfect analysis. The basic strategies among the Baldrige winners have been summarized as [13]: I. Leadership and management commitment 2. Total customer delight 3. Long term efforts 4. Teamwork with employees, suppliers, and customers 5. Employee involvement and satisfaction 6. Continuing training for all 7. Statistical measurement of progress 8. Continuous quality improvement 9. Benchmarking against the best in class 10. Total and open communications with employees, suppliers, and customers

Concluding Remarks
A paradigm shift in the mindset is necessary at all levels of laboratory staff to produce a superior product. Think of a ham and egg breakfast; the chicken is involved, but the pig is committed. You have to be the latter. Nor are the results an instant pudding. Again, think of a bamboo farmer; the sapling is watered for 4 years, and suddenly at the end the tree shoots up to 60 feet in 90 days! A constancy of purpose, clear vision, a prepared mind, commitment to long term pursuit of goals, and dedication to all technical aspects of pillars of good analysis are what will differentiate an ideal laboratory from an also-ran laboratory. The choice is yours to make. As W. Edwards Deming wrote, "Charles Darwin's law of survival of the fittest, and that the unfit do not survive holds in free enterprise as well as in natural selection. It is a creel and unrelenting law. Actually, the problem will solve itself. The only survivors will be the companies with constancy of purpose for quality, productivity, and service" [14]. Similar to Robert Pirsig's hypothesis of Zen and the art of motorcycle maintenance [1], a mental exercise is at the heart of managing a "perfect laboratory". The mind needs to be prepared to perform a flawless analysis and whatever is required to make it so; otherwise a perfect analysis is simply not possible. Determination to excel, mental discipline to stay the course, willingness to overcome inertia and resistance of others, and focus on producing perfect analysis at all levels of laboratory staff are essential. Improvement in laboratory quality performance is a process, not a project. A project ends, but a process continues.

12

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Inherently we all know the difference between a good laboratory and a not-so-good laboratory. The difference is like the difference between a star in a play and the understudy. The not-so-good laboratory (and the understudy) do their jobs and produce results, but a good laboratory (and the star) are needed to produce exceptionally perfect results. The technical quality management of an analytical laboratory can be a hard pill to swallow, but it need not be. It can help prevent excessive waste in time and money, improve productivity by eliminating duplication and waste, and most importantly lead to increased customer satisfaction and increased profits. Unless strong and sustained measures are taken, improvement in data quality cannot be achieved or maintained. No pain, no gain! Only a prepared mind can meet the challenge. The bottom line is that one needs to have a dream, a passion, and a vision to want their laboratory to be the best there is - the flawless perfect laboratory. Unless one nurtures and pursues that vision, as Yogi Berra said, "You can't be there if you don't know where you are going". Additional information on this subject may be found in Refs [15-18].
References

[1] [2] [3] [4] [5]

[6] [7] [8] [9] [10] [11] [12] [13] [14] [ 15]

Pirsig, R. M., "Zen and the Art of Motorcycle Maintenance," Williams Morrow and Co., Inc., New York, 1974. Kratochvil, B. and Taylor, J. K., "Sampling for Chemical Analysis," Analytical Chemistry, Vol. 53, No. 8, 1981, 924A-938A. Mitchell, J. W., "Ultrapurity in Trace Analysis," Analytical Chemistry, Vol. 45, No. 6, 1973,492A-500A. Zief, M. and Mitchell, J. W., "Contamination Control in Trace Element Analysis," John Wiley & Sons, New York, 1976. Nadkarni, R. A. and Bover W. J., "Bias Management and Continuous Improvements through Committee D02's Proficiency Testing," ASTM Standardization News, Vol. 32, No. 6, 2004, pp. 36--39. Nadkami, R. A., "ISO 9000 Quality Management Standards for Chemical and Process Industries," Analytical Chemistry, Vol. 65, 1993, 387A-395A. Nadkami, R. A., "What Every Chemical Laboratory Should Know About the ISO9000 Quality Standards," Analytical Chemistry, Vol. 66, 1994, 10(3-15G. Locke, J. W., "Quality System for Testing Laboratories," A2LA Reprint, 1992. de Leemput, Van, "Testing Their Calibre - A Standard to Ensure Laboratories are up to Standard," ISO Bulletin, 16-18, June 2000. Broderick, B. E., "Laboratory Accreditation: The Operation of an Established Scheme," Mikrochimica Acta, III, 17-21, 1991. Camp, R. C., "Benchmarking," ASQC Quality Press, Milwaukee, WI, 1989. Hogue, C., "Ferreting out Erroneous Data," Chemical and Engineering News, pp. 5 and 49-50, April 2002. Nadkami. R. A., "A Not-So-Secret Recipe for Successful Total Quality Management," Quality Progress, 91-96, November 1995. Deming, W. Edwards, "Out of Crisis," MIT Press, Cambridge, MA, 1982. "Quality Assurance for the Chemical and Process Industries: A Manual of Good Practices," American Society for Quality, Milwaukee, WI, 1987.

NADKARNI ON PERFECT ANALYSIS

13

[16] Nadkarni, R. A., "The Quest for Quality in the Laboratory," Analytical Chemistry, Vol. 63, No. 13, 1991, 675A-682A. [17] Nadkarni, R. A., "Quality Management: A Must for the Labs Too," Today's Chemist at Work, 34-38, December 1996. [18] ASTM D 6792-04: Standard Guide for Quality System in Petroleum Products and Petroleum Testing Laboratories, ASTM International, West Conshohocken, PA.

ATOMIC EMISSION SPECTROSCOPY

Journal of ASTM International, November/December 2005, Vol. 2, No. 10 Paper ID JAI12965 Available online at www.astm.org

Chris C. Onyeso, Ph.D.t

Analysis of Gasoline and Diesel Fuel Samples by Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES), Using Pneumatic Nebulizer and Standard Spray Chamber
ABSTRACT: A method has been developed for accurate quantitative determination of additive elements

and wear metals in gasoline and diesel fuel in the concentration range of 0 to 50 ppm using inductively coupled plasma atomic emission spectrometry (ICP-AES). This method requires an ICP capable of detecting metals at the 0.02 mg/kg level, and capable of linear calibration over the range of 0.05 to 20.0 mg/kg with a correlation coefficient of 0.9999 or better. No additional ICP-AES accessory or lengthy sample preparation is required. The PerkinElmer Optima 4300 has been shownto meet this requirement.
KEYWORDS: gasoline, diesel fuel, additive elements, ICP, wear metals, and analysis. Introduction

Trace metals in fuels, except in the case of additives, are usually undesirable and normally occur in very low concentrations, requiring sensitive techniques for their determination. Also, there is a need to determine the concentration of additive elements such as manganese (Mn) in gasoline. The industry method currently available for the determination of manganese (Mn) in gasoline is ASTM D 3831 [1], which requires bromine reduction of the gasoline followed by measurement with Atomic Absorption Spectrometer (AA). This method has inherent problems with high olefin content fuels and cannot accurately measure low concentrations of Mn in gasoline.

Use of lCP-AES
ASTM test methods D4951 [2] and D5185 [3] are commonly used for the determination o f additive elements and wear metals in fresh and used lubricating oils respectively. The problem with the use of ICP for gasoline analysis is that gasoline is highly volatile and extinguishes the ICP plasma [4,5]. A few examples of ICP analysis o f gasoline are found in the literature, but these methods involve the use of expensive ICP accessories such as direct injection nebulizer [6], ultrasonic nebulizer with micro-porous membrane desolvator [7], chilled spray chamber [8,9], a thermostated condenser between the spray chamber and the plasma torch [10], or lengthy sample preparation with possibility o f contamination, such as emulsification with surfaetants [ 11-14]. In this paper, an accurate method for determining the concentration of manganese and other metals in gasoline and diesel fuel by ICP without the use of chilled spray chamber, direct injection nebulizer, ultrasonic nebulizer with micro-porous membrane desolvator, thermostated condenser, or emulsification is discussed.

Manuscript received 15 November 2004; accepted for publication 13 May 2005; published November2005. Presented at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advancesand Future Prospects on 6-8 December 2004 in Tampa, FL. I Ethyl Corporation, Richmond,VA.
Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

17

18

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Analysis
Calibration standards are prepared from certified Conostan| standards, using kerosene containing Yttrium internal standard as the solvent. The kerosene containing internal standard is also used as the solvent for the gasoline and diesel fuel samples. The calibration standards and the gasoline samples are aspirated into the ICP instrument. Each element present in the sample emits light at discrete wavelengths. The intensity o f light at each element's wavelength is compared against a calibration curve generated from standards containing metals at k n o w n concentrations to provide quantitative determination. This method also outlines the preparation of the instrument, samples, and standards used in the analysis. The method was originally developed for the determination of manganese (Mn) in gasoline and diesel fuel, and later extended to include additive elements and wear metals.

Instrument
9 9 9 9 9 9 9 9 9 9 PerkinElmer Optima 4300. The appropriate wavelength (nm) for each element (analyte) was chosen. Internal Standard is Yttrium (Y). Quartz Torch (no slot) for organic solvents. Quartz Torch (single slot) for aqueous solution. Spray chamber - cyclonic spray chamber for organic solvent. Scott spray chamber for aqueous solution. Nebulizer - GemCone nebulizer for organic solvent. Crossflow nebulizer for aqueous solution. Injector Tube - Quartz injector tube (1.2 m m i.d.) for organic solvent - Sapphire injector tube (2.0 mm i.d.).

Spectrometer
Resolution was set at High with three replicates and integration time o f five seconds.

Plasma Parameter
9 9 Plasma View: radial. Peristatic Pump set at a flow rate o f 2.50 ml/min.

Calibration
9 9 9 The manganese calibration standards concentrations are in mg/kg (ppm). The instrument was calibrated using the following standards: blank (0 ppm), 5 (5.2333 ppm), 10 (10.7836 ppm), 15 (15.6748 ppm), and 20 (20.5006 ppm). A good calibration curve with a correlation coefficient o f 0.99996 was obtained (see Fig. 1 below).

Sample Analysis
After calibration, gasoline samples o f known Mn content (as added) were analyzed. Results are shown in Table 1.

ONYESO ON PNEUMATIC NEBULIZER

19

FIG. 1--Calibration curve. TABLE 1--Determination of Mn in gasoline samples by ICP. Gasoline Sample Given Cone. Measured Cone. Difference

(mg/kg)
1) Citgo 0 2) Motiva 0 3) Philips 10 4) Philips 2 5) Philips 6 6) Gasoline from unknown source 18 7) Gasoline from unknown source . . . . . . 8.3

(mg/kg)
0 0 10.3 1.8 6.4 19 9.2 0 0 0.3 -0.2 0.4 1.0 0.9

Sample Preparation
Samples were prepared by diluting 1.0 g of gasoline or diesel fuel with 99.0 g of kerosene containing Yttrium (Y) internal standard. Example: Gasoline sample wt. = 1.0601 g Kerosene wt. = 98.9727 g Gasoline is very volatile; the weighing of the One,gram and the addition of kerosene must be done very quickly to avoid loss of weighed gasoline by evaporation. Weighing the kerosene first and then adding in the gasoline has been found to reduce gasoline loss. Based on accurate results obtained on the gasoline samples with known Mn content (the gasoline samples were spiked with a known amount of Mn), many gasoline and diesel fuel samples were analyzed for Mn content and the ICP results (Alton) compared with those obtained by AA at a contract Lab (Tables 2 and 3). The target values are the spiked values.

20

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

...... TABLE 2--Determination of Mn in diesel fuel. Sample Type Contract Lab. D3831 Alton R & D Target Value (AA) mg/kg ICP mg/kg mg/kg Diesel 1: Run 1 2.7 2.3 Run2 1.9 3.1 Average 2.3 2.7 3.0 Diesel 2: Run 1 2.5 3.0 Run 2 3.2 3.1 Average 2.9 3.1 4.5 Diesel 3: Run 1 4.3 4.3 Run 2 4.4 4.3 Average 4.4 4.3 6.0 Diesel 4: Run 1 0.0 0.0 0.0 0.0 Average 0.0 0.0 0.0 TABLE 3--Determination of Mn in gasoline samples. Sample Contract Lab Alton R & D Target Value D3831 mg/kg ICP mg/kg mg/kg Gasoline 1: Run 1 0.0 0.0 Run 2 0.0 0.0 Average 0.0 0.0 0.0 Gasoline 2: Run 1 5.6 7.0 4.5 6.0 Average 5.1 6.5 8.0 Gasoline 3: Run 1 14.7 16.2 Run 2 12.4 14.7 Average 13.6 15.5 18 Gasoline 4: Run 1 20.7 22.9 Run 2 21.0 19.2 Average 20.8 21.1 25 Gasoline 5: Run 1 36.2 31.0 Run2 33.3 31.8 Average 34.8 31~.4. 40.0 In six out of the nine diesel and gasoline samples analyzed, the ICP results were closer to the target manganese values than the AA results. The differences seen between AA and ICP results are not unreasonable when one considers that the data sets were obtained using two different analytical methods in two different laboratories.

Precision
The manganese method has a repeatability of 0.24 mg/kg. At 12.0 mg/kg Mn level, the precision of the method is +/- 1.16 mg/kg. Diesel fuel was diluted 100-fold to keep the method consistent; detection limits and precision could be better for diesel at lower dilutions since diesel is not as volatile as gasoline.

ONYESO ON PNEUMATIC NEBULIZER

21

Expansion of the Method to Include Additive Elements and Wear Metals


Having successfully used this method to determine Mn in gasoline and diesel fuel, it was desirable to extend the method to include determination of additive and wear metals. The method was calibrated with a Conostan| standard containing the desired elements, and a good calibration curve was obtained as shown below (Table 4). TABLE 4--Multi-element calibration and limits of detection. Analvte Corr. Coef. Limit of Detection mg/kg Mn (293.305) 0.999997 0.9 A1 (308.215) 0.999817 2.1 Ca (315.887) 0.999962 1.8 Cu (324.752) 0.999938 1.3 Fe (238.204) 0.999937 1.2 Mg (285.213) 0.999987 0.7 Pb (217.000) 0.999972 3.1 V (290.880) 0.999975 0.6 Zn (206.200) 0.999617 3.2 With this calibration, gasoline, furnace oil, and diesel fuel samples were analyzed in duplicate. Table 5 shows elements detected in these samples. The minimum detection limits for the analytes (elements) were determined at 100-fold dilution and are shown in Table 4 above. TABLE 5--Elements detected in samples. Sample Mn (mg/kg) Run 1 Run 2 FBG-10513 (gasoline) 17 17 R05905 (gasoline) 22 21 EC194995-2 (gasoline) 12 11 EC194995-1 (gasoline) 21 20 Furnace Oil 0 0 Low Sulfur Diesel 0 0 R06241201 Diesel 20 21

Avg 17 21.5 11.5 20.5 0 0 20.5

Zero results indicate concentrations below the minimum detection limit. The gasoline, diesel fuel, and furnace oil samples listed above were analyzed for the calibrated elements listed in Table 4. A1, Ca, Cu, Fe, Mg, Pb, V, and Zn were below the method detection limit in Furnace Oil, Low Sulfur Diesel without additive, diesel fuel, and four gasoline samples. Only Mn was detected in those gasoline and diesel fuel samples with additives. The above results have shown that this method can be used to determine additive and wear metals in gasoline, diesel fuel, and furnace oil in concentrations that are within the detection limit of the instrument.

Aqueous Analysis
Another way that gasoline can be analyzed by conventional ICP without the use of additional accessories or lengthy sample preparation is to analyze it in aqueous solution. For gasoline and

22

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

diesel fuel samples containing solid impurities or particulates that are insoluble in kerosene, aqueous analysis is a very viable option. Determination of metals after total destruction o f organic matter is a very reliable procedure for petroleum and petrochemical samples. The gasoline samples were digested with nitric acid in a microwave digestion oven. The resulting acidic solution was diluted with de-ionized/distilled water down to about 5-10 % HNO3 and analyzed by aqueous ICP method (Table 6). Sample dilution in the aqueous analysis is about 1 to 50, unlike organic analysis, which is 1 to 100. TABLE 6--Aqueous ICP analysis of gasoline ...... Gasoline Samples (results in mg/kg)* Elements Gasoline 1 Gasoline 2 Gasoline 3 Gasoline 4 Gasoline 5 Fe 3 7 4 25 23 Mo 0 0 1 0 0 Zn 0 0 0 0 0 Pb 3 0 4 15 0 B 0 0 0 4 1 Si 11 22 29 31 9 *Sn, Cr, P, Ni, Ba, and Mn were below the detectionlimit in these six gasoline samples.

Gasoline 6 7 4 23 3 1 12

The differences between the results shown for Fe, Pb, Ca, and Si in Tables 5 and 6 are due to the fact that the gasoline samples are different, the dilution factors are different, and the matrices are different (kerosene versus water).
Conclusions

An ICP method for the accurate determination of additive elements and wear metals in gasoline and diesel fuel has been developed. The gasoline and diesel sampleswere dissolved in kerosene with yttrium internal standard and analyzed by ICP without any additional accessory. Samples were bracketed with matrix matched standards, and an internal standard was also used. Based on good calibration curves (correlation coefficients of 0.9999 and higher), lack of spectral interference, and accurate results obtained on gasoline samples with known amount of Mn, the results for the samples with an unknown amount of metals must be correct. For the organic ICP method, the standard deviations for four measurements of each of the twelve (12) gasoline samples ranged from 0.00 to 0.21. This study has demonstrated the ease of gasoline and diesel fuel analyses both in organic and aqueous matrices by ICP without the use of additional expensive accessories, such as direct injection nebulizer, ultrasonic nebulizer with micro-porous membrane desolvator, chilled spray chamber, a thermostated condenser between the spray chamber, and the plasma torch or lengthy sample preparation.

Acknowledgments
Thanks go to Ken Garelick and Penny Hurt for their assistance in the preparation and analysis of the samples and to Afton Chemical Company.

ONYESO ON PNEUMATIC NEBULIZER

23

References

[1] [2] [3]

[4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

ASTM Standard D 3831, "Standard Test Method for Manganese in Gasoline by Atomic Absorption Spectrometer," ASTM International, West Conshohocken, PA. ASTM Standard D 4951-02, "Standard Method for Determination of Additive Elements in Lubricating Oils by ICP-AES," ASTM International, West Conshohocken, PA. ASTM Standard D 5185-02, "Standard Test Method for Determination of Additive Elements, Wear Metals and Contaminants in Used Lubricating Oils and Determination of Selected Elements in Base Oils by ICP-AES," ASTM International, West Conshohocken, PA. Botto, R. I., Spectrochim. Acta, Part B, Vol. 42, 1987, p. 181. Kreuning, G. and Maessen, F. J. M. J., Spectrochim. Acta, Part B, Vol. 44, 1989, p. 367. Botto, R. I., Canadian Journal of Analytical Sciences and Spectroscopy, Vol. 47, No. 1, 2002, pp. 1-13. Botto, R. I., J AnaL At. Spectrom., Vol. 8, No. 1, 1993, pp. 51-57. Ryan, A., Chemistry in New Zealand, Vol. 61, No. 6, 1997, pp. 9-11. Hausler, D. W. and Taylor, L. T., AnaL Chem., Vol. 53, 1981, p. 1223. Maessen, F. J. M. J., Seeverens, P. J. H., and Kreuning, G., Spectrochim. Acta, Part B, Vol. 39, 1984, p. 1171. Saint Pierre, T. D., Dias, L. F., Pozebon, D., Aucelio, R. Q., Curtius, A. J., Welz, B., Spectrochim. Acta, Part B, 2002, 57B(12). Brenner, L B., Zander, A., Kim, S., Shkolnik, J. J., AnaL At. Spectrom., Vol. 11, No. 2, 1996, pp. 91-97. A1-Swaidan, H. M., Atomic Spectroscopy, Vol. 14, No. 6, 1993, pp. 170-173. Reimer, R. A. and Miyazaki, A., Journal of Analytical Sciences, Vol. 9, No. 1, 1993, pp. 157-159.

Journal of ASTM International, September2005, Vol. 2, No. 8 Paper ID JAI12966 Available online at www.astm.org B r i a n S. F o x 1

Elemental Analysis of Lubricating Grease by Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES)
ABSTRACT: Lubricating grease is a mixture with complex properties that poses far more difficult analytical challenges than do its individual components. At many petroleum testing laboratories, Inductively Coupled Plasma Atomic Emission Spectroscopy, or ICP-AES, is used for the analysis of additive elementsand wear metals in oil. However,it may also be used to measure the concentrationof both additive metals and contaminantsin lubricatinggrease. However,unlike some of the other matrices that are routinely analyzed, grease cannot be simply diluted for direct sample introductioninto the instrument without some form of acid digestion. This paper will discuss how ICP-AES is used as a grease analysistool. Several samplepreparationschemes will be covered, includingclassicalsulfatedash on a hot plate, microwaveassisted dry ash, and microwaveassisted acid digestion. Comparisondata from different digestiontechniqueswill illustratepotentialproblemsthat may be encounteredby the analystin each of these methods. The advantages of closed-vesselmicrowavedigestion for the analysisof elements that are often volatilizedtrader normal digestionconditionswill be discussed. KEYWORDS: lubricatinggrease, additives,ICP-AES

Introduction
In 1885, Grant McCargo formulated one of the earliest petroleum grease mixtures b y blending oil and soap together. Today, lubricating grease finds use in almost all bearings around the world. Lubricating grease is composed o f approximately 90 % additized oil and soap or other thickening agent [1]. In the absence of an oiling system, grease provides a film of additized oil to protect the metal surfaces that are in contact with one another. The thickener in the grease helps to keep it located in the bearing assembly, where it can do its job, Some of the other additives in lubricating grease mirror the functions that they perform in typical lubricating oil. In lubricating oil, the additive and wear metal concentrations are determined by several atomic spectroscopy methods, among them being ASTM D 4951 or ASTM D 5185. However, these tests have no counterparts for testing grease samples. The focus of most standardized grease testing has been upon performance and appearance, rather than determining the concentration of chemical components. In part, this is likely due to the relative volume of grease demand when compared to other lubricants. But the primary reason may be that grease is a rather difficult matrix with which to work from the perspective o f the analytical laboratory. Lubricating oil may be diluted in a solvent and introduced directly into an Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES) for rapid elemental measurements. Grease, on the other hand, does not dissolve in any common laboratory solvent that is compatible with ICP-AES.

Manuscriptreceived 18 November2004; accepted for publication7 March 2005; publishedSeptember2005. Presented at ASTM Symposiumon ElementalAnalysisof Fuels and Lubricants:Recent Advancesand Future Prospects on 6-8 December2004 in Tampa, FL. i AdvancedResearch Associate, ExxonMobilResearch and Engineering,Paulsboro, NJ 08066.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohockon, PA 19428-2959.

24

FOX ON ANALYSIS OF LUBRICATING GREASE

25

In addition to ICP-AES, an X-ray fluorescence spectroscopic technique is sometimes used to quantify elemental concentrations in grease samples. However, it suffers from two major drawbacks. The first is the inability to detect two important lubricating grease components, lithium and boron. Lithium compounds are used as a thickener in the vast majority o f lubricating grease types on the market. Boron is sometimes used with lithium as a complexing agent. The second drawback emerges in the case o f the analysis o f used grease obtained from a bearing. A grease sample that is displaced b y fresh grease and obtained from outside a bearing housing m a y contain dirt from the environment along with wear metals. The grease that is of primary interest is that which is sampled from the load zone o f a disassembled bearing [2]. Obtained from this source, there is often insufficient material available to fill the X-ray sample cup for a proper analysis. Despite the difficulty in working with grease samples, there are well over a dozen elements o f interest to the lubricating grease supplier in the form o f additives, contaminants, and wear metals. Table 1 shows some o f these elements in lubricating grease and the role that each may play. Determining their concentrations can be an important aspect o f grease manufacture. In addition, a reliable analysis technique can also assist in the process o f troubleshooting problems with new and used grease in the field. For example, some o f the thickeners used in the various types o f lubricating grease m a y be incompatible with one another [3]. Quantifying the signature elements o f these thickeners could determine the level o f accidental mixing o f two incompatible grease types. TABLE 1--Potential
Element

elementsfound in lubricatin~ ~rrease.


Concentration Range

Potential Sources

Aluminum, A1 Antimony, Sb Boron, B Barium, Ba Bismuth, Bi Cadmium, Cd Calcium, Ca Chromium, Cr Copper, Cu Iron, Fe Lead, Pb Lithium, Li Magnesium, Mg Manganese, Mn Molybdenum, Mo Nickel, Ni Phosphorous, P Sulfur, S Silicon, Si Tin, Sn Zinc, Zn

Clay or other thickener Extreme pressure additive Complexing additive for thickener Rust inhibitor Extreme pressure additive Wear metal, contaminant Thickener and extreme pressure additive Wear metal, contaminant Wear metal, contaminant Wear metal, clay thickener Extreme pressure additive Thickener Clay thickener minor element Clay thickener minor element Molybdenum disulfide / extreme pressure additive Wear metal Additives such as zinc dithiophosphate(ZDTP) Additives such as ZDTP Contaminant, thickener trace element clay Contaminant Additives such as ZDTP

<20 to 3000 #g/g <20 to 1000 #g/g <20 to 1000 #g/g <20 to 1000 #g/g <10 to 1000 #g/g Trace to 1000 #g/g <20 to 3000 #g/g Trace to 1000 #g/g Trace to 1000 #g/g <10 to 5000 #g/g <20 to 1000/zg/g <20 to 3000 #g/g Trace to 100 #g/g Trace to 100 #g/g <20 to 3000 #g/g Trace <20 to 1000 #g/g <20 to 1000 #g/g <10 to 1000 #g/g Trace to 100/zg/g <20 to 1000/~/~

This paper examines various techniques that are available to prepare lubricating grease samples for analysis b y ICP-AES. Each o f the techniques has specific advantages and m a y be applied to different grease sample types, depending upon composition and expected levels o f analyte concentration. Special attention will be given to the use o f high-pressure, microwave assisted acid digestion. This technique provides the optimal solution for complete digestion o f

26

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

the organic matrix, while minimizing loss of additive elements that may be volatilized during other preparation schemes.
Instrumentation

The measurements were performed using a Thermo Elemental ]RIS Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES). A 2 kW crystal-controlled radio frequency (RE) generator operating at 27.12 MHz powers the plasma source. An Echelle optical system with a 381-ram focal length diffracts the light from the plasma source before it is focused onto the Charge Injected Device (CID) camera detector [4]. A fully de-mountable ABC TJA IRIS Radial torch was equipped with a Twister cyclonic spray chamber for the non-hydrofluoric (HF) acid analyses. Both are available from Glass Expansion Inc, Pocasset, Massachusetts, USA. When analyzing samples prepared with a dilute HF acid solution, an alumina center tube and HF resistant spray chamber (Glass Expansion) were substituted. The nebulizer used was a Teflon Mira Mist that is available from Burgener Research Inc., Mississauga, Ontario, Canada. The microwave digestion oven used for sample preparation is an MDS 2000. It is available from CEM Corporation, North Carolina, USA. The microwave digestion vessels are high pressure, Teflon lined, 45-mL Model 4782 bombs that are available from Parr Instrument Company, Moline, Illinois, USA. For microwave assisted dry-ashing of lubricating grease samples, a CEM Microwave Ashing System MAS 7000 was used. The open vessel microwave digestion apparatus that was used for this work is a Prolabo Microdigest 401.
Reagents and Materials

All water used is 18 Mfl de-ionized water polished by a Simplicity ion exchange system available from Millipore Corporation, USA. Trace Metal Grade sulfuric, nitric, and perchloric acids that were used in the procedures are available from Fisher Scientific, Pittsburgh, Pennsylvania, USA. Commercially prepared, multi-element, aqueous ICP-AES standards and the blank acid matrix were obtained from VHG Labs, Inc., Manchester, New Hampshire, USA. The multi-element, oil standards used for validation are available from Conostan Division, Conoco Specialty Products, Ponca City, Oklahoma, USA. For sulfated ash digestions, either 250-ml Vycor beakers or dishes may be used (Fisher Scientific). The microwave digestion vessels used were high pressure, Teflon lined, 45-mL Model 4782 bombs that are available from Parr Instrument Company, Moline, Illinois, USA. The polypropylene volumetric flasks used to dilute digested samples to volume are available from Fisher Scientific.
Procedure

The following descriptions of the digestion methods are simplified versions. Depending upon sample difficulty, other steps may beadded to obtain a clear solution that is relatively free of the organic matrix and un-dissolved material.

Sulfated Ash Digestion


The sample size depends upon expected analyte concentration. Approximately 1-2 g of

FOX ON ANALYSIS OF LUBRICATING GREASE

27

material are weighed into a Vycor container of suitable size. The sample is then charred on a hot plate until it is reduced to about 0.5 g. A heat lamp is used to assist the process. Then 1-2 mL of sulfuric acid are added to the residue and heated until the fumes cease to evolve. The charred sample is placed in a muffle furnace at 530~ until the black color is gone. This typically takes about 2 h, but it was originally felt that a sample may be left overnight to obtain complete decomposition. Approximately 5 mL of hydrochloric acid are added, and the sample is gently heated to dissolve remaining solids. The solution is brought to volume in a 50-mL volumetric flask with water.

Microwave Assisted Dry Ashing


Approximately 1 g of sample is weighed into a crucible. Vycor or platinum crucibles have both been used for this work. The MAS 7000 furnace insert allows the use of metal vessels in the microwave unit. The crucible is placed into the microwave furnace. The temperature is ramped up to 525~ over a two-hour period. It then holds the temperature for one hour. After the sample cools, approximately 5 mL of a 1:1 nitric acid-water solution are added to the crucible and heated gently to dissolve the ash. It is critical that nitric acid only be added after the organic matrix has been ashed and cooled. The sample is quantitatively transferred to a 25-mL volumetric and brought to volume with water.

Open- Vessel Microwave Digestion


Approximately 1 g of sample is weighed into the digestion vessel, and 10 mL of nitric acid are added. The microwave unit is set to 30 W to cause a gentle reflux of the nitric acid. After about one hour or when the sample is nearly decomposed, the drop-wise addition of perchloric acid will typically complete the digestion. Perchloric acid is an extremely strong oxidizer that may react violently with organic material. Experienced personnel only should use it with utmost caution. The sample is heated until about 5 mL of solution remain in the vessel. The contents are washed into a 50-mL volumetric flask and brought up to volume with water.

Closed- Vessel Microwave Digestion


From a safety perspective, it is critical that no more than 0.1 g of sample be used for this technique. The sample is weighed into the Teflon digestion vessel. Approximately 4 mL of nitric acid are added. The vessel is capped and placed into the microwave oven. Four vessels are simultaneously processed. The microwave is set at 125 W for 15 min. The oven then ramps up to 190 W for another 15 rain. Care must be taken not to keep internal temperature and pressure within the capability of the vessels. Excessive heat and pressure will cause the digestion bombs to deform and potentially leak. After the cycle is finished, the vessels are placed into an ice bath for at least one hour to cool. The dissolved sample is washed into a 25-ml volumetric flask and brought to volume with water.

ICP-AES Instrument Conditions


The instrument operating conditions are shown in Table 2. Optimization of conditions is important to attain a robust plasma that is minimally affected by differences in acid concentration between standards and samples. Calibration was done using a series of external standards with

28

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

yttrium (Y) as the internal standard. The internal standard was continuously added to standards and samples via a "tee" fitting and mixing loop. TABLE 2--ICP-AES operating conditions. Forward Power 1150 W Coolant Gas, Argon 16 1 min "l Auxiliary Gas, Argon 0 1 min "1 Nebulizer Gas, Argon 0.65 1 min -l Sample uptake rate 1.1 ml min 1
Discussion

Lubricating grease does not possess the required qualities to make it a suitable Standard Reference Material, or SRM. The oil in the grease tends to separate from the thickener after a few months on the shelf. Most of the following work uses samples with expected concentrations, based upon known quantities of well-characterized additive. Commercial certified standards were also used to validate the digestion procedure and to determine detection limits. Table 3 shows the elements in lubricating grease, which was prepared with an organic thickener as opposed to a metallic soap. All of the results are expressed in weight percent. Four repeat analyses were made using each digestion scheme. The expanded uncertainty of each measurement is calculated by multiplying the value for of one experimentally derived standard deviation of the data times the critical value. The critical value of 3.18 is determined from a student's double-sided t-distribution [5] with 3 degrees of freedom and a confidence level of 95 %. The average of the sulfated ash digestion indicates that nearly 12 % of the phosphorous is lost from the sample. The loss is likely due to volatilization of phosphorous as opposed to sample spatter, based upon the fact that the zinc result is closer to the expected value. The dry ash technique shows less average phosphorous loss. However, it is still slightly more than that of Parr Bomb or closed vessel, high-pressure microwave digestion scheme. ........ TABLE 3--Additive concentration results in lubricating grease with organic thickener. Expected 2-h Sulfated Open-Vessel Parr Bomb Element Concentration Ash Dry Ash Microwave Microwave Sb, Mass% <0.001 <0.001 <0.001 n/a <0.002 B, Mass% <0.001 <0.001 <0.001 n/a <0.001 Li, Mass% <0.001 <0.001 <0.001 n/a <0.001 P, Mass% 0.19 0.168 + 0.013 0.181 + 0.009 n/a 0.186 + 0.005 Zn, Mass% 0.212 0.222 + 0.007 0.217 + 0.006 n/a 0.212 + 0.003 The microwave digestion procedure is limited to a maximum sample size of about 0.1 g. This is due to the pressure generated by the evolution of carbon dioxide and other gaseous digestion by-products from the acid solution. Despite the small sample size, microwave digestion provides additive concentration results that exhibit little or no increase in variability from those procedures for which sample size is not a constraint. However, this may not be the case for trace analyte concentrations where some loss of precision will inevitably occur. The grease in Table 4 represents a more typical lithium soap formulation. Phosphorous loss is almost negligible in this grease sample. However, boron suffers a nearly 25 % loss from the

FOX ON ANALYSIS OF LUBRICATING GREASE

29

expected value when the sample is prepared with the sulfated ash technique. A nearly complete loss of boron has also been observed when digesting a commercial multi-element oil standard by the sulfated ash technique. Open vessel microwave digestion results were consistent with expected values. However, this technique has limited utility due to safety concerns about using perchloric acid to complete the digestion. Another problem with the perchloric acid reagent was the presence of boron. Despite the use of trace metal grade acid, an unacceptable level of boron was detected. This forced the use of a blank correction that could lead to expanded uncertainty. Subsequent to the generation of these data, a new source of perchloric acid was found that appears to have suitably low levels of boron contamination. The four digestion schemes were applied to a second lubricating grease sample with lithium soap thickener. These results are shown in Table 5. The sulfated ash digestion resulted in loss of antimony, boron, and phosphorous, while lithium and zinc remained consistent with expected values. TABLE 4--Additive concentration results in lithium thickener lubricating grease. Expected 2-h Sulfated Open-Vessel Parr Bomb Element Concentration Ash Dry Ash Microwave Microwave Sb, Mass% <0.001 <01001 <0.001 <0.001 <0.002 B, Mass% 0.03 0.023 +0.009 0.027 +0.006 0 . 0 3 1 0.029+0.002 Li, Mass% 0.19 0.189+0.003 0.185+0.004 0.191 0.184+0.006 P, Mass% 0.19 0.194+0.008 0.192+0.011 0.189 0.193+0.006 Zn, Mass% 0.212 0.223 + 0.006 0.205 + 0.006 0.216 0.221 + 0.003 *Result is reportedto two significantfiguresdue to loss of precision fromacid blank correction. .... TABLE 5--Additive concentration results in lithium thickener lubricating; grease. Expected 2-h Sulfated Open-Vessel Parr Bomb Element Concentration Ash Dry Ash Microwave Microwave Sb, Mass% 0.026 0.02i + 0.008 0.022 + 0.007 0.025 0.025 + 0.002 B, Mass% 0.03 0.025 + 0.011 0.027 + 0.005 0.031 0.034 + 0.002 Li, Mass% 0.18 0.190 + 0.003 0.186 + 0.004 0.190 0.192 + 0.006 P, Mass% 0.195 0.166 + 0.016 0.192 + 0.011 0.192 0.191 + 0.009 Zn, Mass% 0.197 0.190 + 0.004 0.193 _+0.006 0.204 0.206 + 0.004 *Result is reportedto two significantfiguresdue to loss of precision fromacidblank correction. The dry ash did not appear to result in nearly as much analyte loss due to volatilization. However, this procedure has shown a higher propensity to spatter or ignite, resulting in ejection of analyte from the container. There is some evidence for this in the fact that the dry ash technique sometimes shows the lowest results for non-volatile grease components. It also can be more difficult to dissolve the ash from a dry combustion, especially in used grease samples. This may be the result of the formation of recalcitrant oxides that require severe heating of the acid to complete the dissolution of the ash. In the absence of a certified reference material that matches the sample, a generally accepted validation technique involves subjecting an analogous certified material to the digestion procedure and extrapolating the results to samples. This validation exercise was done on a commercial, multi-element oil standard. The results are shown in Table 6. As expected, an average of nearly 65 % of the boron was lost due to volatilization during the sulfated ash

30

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

procedure. However, there was little or no significant loss of the phosphorous over the four repetitions. The Parr bomb microwave digestion consistently gave good recovery for all analytes. TABLE 6---Additive elementsfrom commerciallyprepared metal in oil standard. Element Certified Concentration 2-h Sulfated Ash Parr Bomb Microwave B, Mass% 0.030 0.011 + 0.012 0.031 + 0.001 Ca, Mass% 0.100 0.098 + 0.004 0.101 + 0.003 Mg, Mass% 0.100 0.098 + 0.006 0.098 + 0.006 P, Mass% 0.100 0.102 _+0.008 0.099 + 0.004 Zn, Mass% 0.100 0.105 _+0.006 0.102 + 0.003 The satisfactory phosphorous recovery shown in Table 6 may lead a researcher to believe that phosphorous volatility is not an issue. However, Table 7 shows the effect that the sulfated ash scheme has on an additive material. A zinc dithiodialkylphosphate (ZDDP) additive was tested by the ASTM D 5185 method. The results obtained were consistent with the expected values. The sulfated ash procedure yielded an average loss of nearly 25 % of the phosphorous with much higher variability between measurements than the microwave digestion. TABLE 7--Comparison of ZDDP additive elements by ICP-AES. Expected 2-h Parr Bomb Element Concentration ASTM D 5185 Sulfated Ash Microwave B, Mass% <0.001 <0.001 <0.001 <0.002 Ca, Mass% <0.001 <0.001 <0.001 <0.002 Mg, Mass% <0.001 <0.001 <0.001 <0.002 P, Mass% 9.5 9.4 + 0.4 7.3 + 1.5 9.4 + 0.5 Zn, Mass% 10.6 10.4 + 0.5 10.4 + 0.5 10.3 + 0.6 In this series of analyses, the average loss of phosphorous was more severe when ashing the pure ZDDP additive than that observed while ashing the additized grease samples. It is possible that the presence of the metallic soap in the grease may contribute to minimizing the loss of volatile components in the short term. As the sulfated ash conditions become even more severe, a greater portion of volatile analyte is lost. When the samples are allowed to heat overnight at 530~ it was observed that nearly all of the antimony and boron are lost. More than half of the phosphorous was also lost. The advantages of retaining volatile analytes, while ensuring complete digestion of a grease sample when using the high-pressure Parr Bomb, start to become apparent. One of the major disadvantages of the closed vessel microwave digestion scheme is the sample size limit of approximately 0.1 g. Since 4 mL of acid are used to digest the sample, the analyst is either faced with an extremely large dilution factor or must deal with the effects of a strong acid concentration in the final solution. Because these digestion schemes may result in widely varying acid concentrations in the final solution, the ICP-AES conditions require careful optimization for this work. Researchers have seen that increasing acid concentration often causes a depression in signal intensity for some lines when using pneumatic sample introduction systems [6]. The effect may be especially prominent under non-robust plasma conditions. The ICP-AES conditions were optimized using

FOX ON ANALYSIS OF LUBRICATING GREASE

31

the Mg II 280.270-nm/Mg I 285.213-um as described by Mermet [7]. Varying the nitric acid concentration from less than 4 % v/v to over 18 % v/v caused less than a 2 % change in signal intensity for the lines of the yytrium internal standard. The cleanliness of the laboratory becomes an even more germane issue as the sample size decreases. Trace contamination becomes magnified when dilution factors are approximately 250 or more. The sample size constraint of the high-pressure microwave digestion scheme tends to raise the minimum detection limits of the analytes in a sample. Table 8 shows the result of digesting a certified Conostan standard. The elements in the table represent several of those that may be found at relatively low concentrations in lubricating grease, either due to contamination or wear. For the standard with a certified concentration of 100 #g/g, the recovery is good for all listed elements with the relative difference within 4 %. Uncertainty and recovery began to deteriorate for the standard with a certified concentration of 10 #g/g. Replacing the pneumatic nebulizer with an ultrasonic nebulizer (USN) would likely result in generally lower minimum detection limits for analytes in the samples. TABLE 8 Elemental analysis of certi~ed standards after closed-vessel microwave digestion. Element Certified, Found, RPD, % Certified, Found, RPD, % A1 Cr Cu Fe Mn Mo Na Ni Pb Si Sn Zn 100 100 100 100 100 100 100 100 100 100 100 100 103 + 8 102 + 3 103 + 3 104 + 5 100 + 4 100 + 2 97 + 10 102+6 101 + 8 98 + 7 99 + 9 103 + 6 3 2 3 4 0 0 3 2 1 2 1 3 10 10 10 10 10 10 10 10 10 10 10 10 13 + 2 l0 + 1 9+ 2 13 + 2 10 + 1 11 + 1 9+3 10+2 8+ 2 16 + 3 7+ 2 14 + 2 26 0 10 26 0 10 10 0 22 46 35 33

Certain clay-based greases may have silica as a trace element. In addition, silica is often found as dirt in used grease samples. Table 8 indicates good recovery for silicon despite using only nitric acid for the digestion of the multi-element oil standard. This level of recovery is not likely to be seen when analyzing samples that have significant amounts of inorganic silicon. We have found that the addition of about 100-200 #1 of 48 % hydrofluoric acid (I-IF) to the nitric acid in the Parr bomb prior to digestion yields clear solutions when analyzing either clay-based or used grease samples. Feng et al. [8] have found that addition of a small quantity of HF enhances aluminum and silicon recoveries and may be analyzed without passivation by boric acid. They also found that potential calcium precipitation in the presence of trace fluoride did not negatively impact their calcium recovery. Since both boron and calcium are important additives, these findings are of interest. To date, we have not done much work to validate mixed acid, closed vessel microwave digestions with lubricating grease, but the application to accurate silica quantification in both clay-based and used grease is promising.

32

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Conclusion

Modem lubricating grease is evolving with new additives and formulations to increase its life in bearings, gearboxes, and other applications. Extended service life means greater attention to additive decay and contamination. The development of elemental analysis techniques for lubricating grease are somewhat hampered by its difficulty as a sample matrix. The lack of standard reference materials and the tendency of grease to de-oil challenge the ability of the analyst to validate the measurement system. The fact that lithium thickener still dominates the grease market means that any atomic spectroscopy technique to be considered should include the capability of determining the concentration of this important analyte. The complex nature of the grease sample matrix makes it difficult to get into a form that is compatible with ICP-AES. However, the elemental composition may be effectively measured if a suitable digestion technique can be developed. Four digestion techniques were examined. Each of them has advantages that should be considered when analyzing a given sample. However, the closed-vessel microwave digestion seems to provide more advantages than classical hot plate digestion or dry ashing. It tends to yield clean solutions, ready for ICP-AES analysis, in less time than the other schemes, while minimizing the chance of contamination by the laboratory environment. Despite its sample size constraints, closed-vessel microwave digestion offers the benefit of good recovery and precision for analytes that may be volatilized under other severe decomposition techniques that are open to the atmosphere.
References

[1] Marino, J. H., "From Mystery to Profit: Grease Lubrication," Lubrication and Fluid Power, Vol. 2, Issue 2, May 2001, pp. 11-14. [2] Herguth, B., "Grease Analysis Monitoring Grease Serviceability and Bearing Condition," Practicing OilAnalysis, March-April 2002, pp. 18-25. [3] Dresel, W. H., "Performance of Mixtures of Lubricating Greases with Different Thickeners," Paper presented October 29, 2002 at National Lubricating Grease Institute (NLGI) 69th Annual Meeting, Preprint #0217. [4] Iris Plasma Spectrometer Standard Vertical Hardware Guide, Thermo Electron Corporation 1994. [5] Anderson, R. L., Practical Statistics for Analytical Chemists, Van Nostrand Reinhold Company Inc., 1987, p. 299. [6] Carre, M., Lebas, M., Marichy, M., Mermet, M., Poussel, E., and Mermet, J. M., "Influence of the Sample Introduction System on Acid Effects in Inductively Coupled Plasma Atomic Emission Spectrometry," Spectro Chimica Acta, 50B, 1994, pp. 271-283. [7] Mermet, J. M., "Use of Magnesium as a Test Element for Inductively Coupled Plasma Atomic Emission Spectrometry Diagnostics," Analytica Chimica Acta, 250, 1991, pp. 85-94. [8] Feng, X., Wu, S., Wharmby, A., and Wittmeier, A., "Microwave Digestion of Plant and Grain Standard Reference Materials in Nitric and Hydrofluoric Acids for Multi-Elemental Determination by Inductively Coupled Plasma Mass Spectrometry," s Anal. At. Spectrom., 14, 1999, pp. 939-946.

Journal of ASTM International, November/December2005, Vol. 2, No. 10 Paper ID JAI12967 Available online at www.astm.org

J. David Hwang, Ph.D., 1Meaghan Horton, 2 and David Leong3

The Use of Microwave Digestion and ICP to Determine Elements in Petroleum Samples
ABSTRACT: ICP-AES analyses are prone to errors caused by changes in the power level, nebulization rate, plasma temperature, and sample matrix. This is especially true for ICP-AES and ICP-MS organic

applications in the petroleum industry where extremely complicated sample matrices are common. As a result, accurate analyses of petroleum lubricants, lubricant additives, fuels, crude oils, residues, and related materials often require bracketing with matrix matched standards, the use of internal standards, and a flexible ICP-AES/ICP-MS system. Many new innovative analytical instruments were developed to better improve elemental analysis in the last few decades. However, sample preparation remains the formidable challenge because it is always labor-intensive, cumbersome, and does not permit fast operation. It is encouraging to see the application of microwave radiation field as the energy source for sample digestion to pursue more reliable, but simpler, safer, more rapid, and less expensive methods, particularly in the petroleum industry. KEYWORDS: microwave digestion, sample preparation, petroleum and petrochemical products, inductively coupled plasma-atomic emission spectrometry (ICP-AES)
Introduction

Although petroleum consists predominantly of hydrocarbons, many elements play a vital role in petroleum production and finished products. Metals such as nickel, iron, vanadium, sodium, and non-metals like sulfur and nitrogen in crude oils have an adverse effect on the refinery and processing operations, generally acting as catalyst poisons. Vanadium and sodium in fuel oils produce deleterious effects on refractory linings, metal surfaces o f modem furnace, boilers, and turbines. Data on metals and sulfur concentration in fuel oils are also important for evaluating emissions for oil-fired power plants. OrganometaUic compounds o f calcium, zinc, and other metals are useful additives to petroleum products, such as lubricant oils [1]. Wear metals like iron, lead, silver, copper, and chromium in used lubricant oils are often determined as indicators o f possible or incipient beating damage. Common analytical procedures for the analysis o f petroleum and petrochemical products b y ICP may be divided generally into two main groups: a) Determined b y direct introduction (e.g., aspiration) into analyzer after a simple dilution or dissolution o f the sample by suitable organic solvent or solvent mixture. b) The metal determination is performed after the sample "prepared" b y dry or wet ashing digestion has been converted into an aqueous solution.

Manuscript received 25 October 2004; accepted for publication 23 May 2005; published November 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. Senior Staff Scientist, Chevron Energy Technology Company, Richmond, CA, USA. 2 Research Assistant, Chevron Energy Technology Company, Richmond, CA, USA. 3Analytical Chemist, Chevron Energy Technology Company, Richmond, CA, USA.

Copyright 9 2005 by A S T M International,I00 Ban" Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. 33

34

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Direct introduction of organic sample solution may have transport interference problems [due to differences in physical properties (e.g., viscosity and surface tension) of the samples and calibration standards], background problems (caused by organic nature of the sample which may result in spectral interference), matrix effects (arising from the complexity of hydrocarbons and other contents), and standardization (calibration problems resulting from different type of organometallic compounds present in the sample). The analysis after digestion removes almost all the effects of direct organic introduction because digestion usually converts all organometallic compounds into an inorganic form [2]. The determination of metals performed after total destruction of organic matter is still [he most reliable procedure for determining metals in petroleum and petrochemical samples. Sample dissolution is one of the most common operations in analytical laboratories. This is particularly true for petroleum analytical laboratories in which quantitative trace element analysis of organic materials and minerals are needed and performed on a daily basis. Because most analytical techniques require that samples be introduced in liquid form, thousands of sample dissolutions are performed every day in analytical laboratories all over the world. Despite the importance and widespread applicability of sample dissolution, most traditional digestion procedures are time-consuming, labor-intensive, and the reagents used (such as perchloric acid) are dangerous and potentially harmful to laboratory analysts [3,4]. Many new innovative analytical instruments were developed to better improve elemental analysis in the last few decades. However, sample preparation remains the formidable challenge and the bottleneck of laboratory fast operation. Fortunately, microwave dissolution, a promising alternative to the furnace and hot plate, has appeared and makes it possible to speed the preparation of solid, sludge, organic matrix, and other difficult samples by combining the rapid heating ability of microwave energy with the use of sealed digestion vessels [5-7]. The advantages of microwave dissolution include faster digestion that results from the high temperature and pressure attained inside the sealed containers. The use of closed vessels also makes it possible to eliminate uncontrolled trace element losses of volatile species that are present in a sample or that are formed during sample dissolution. It is well known that significant amounts of elements such as arsenic, boron, chromium, mercury, antimony, selenium, and tin are lost at relative mild temperature with some open vessel acid dissolution procedures [8,9]. Another advantage of microwave dissolution is to have better control of potential contamination in blank as compared to open vessel procedures. This is due to less contamination from laboratory environment, unclean containers, and smaller quantity of reagents used. Spectrometric oil analysis was first applied in the early 1940s in an effort to detect wear in diesel locomotive engines by the railroad industry. In the early 1960s, the US Air Force adopted spectrometric oil analysis to monitor the wearing in turbojet aircraft engines [10]. It has been reported that direct dilution methods may not be accurate to determine metals in oils by either Flame Atomic Absorption (FAA) or Inductively Coupled Plasma (ICP) because of significant matrix effects [11-13]. It has been found that viscosity modifier polymers interfere with the formation of aerosol, thus affecting the sample delivery to the plasma torch which causes the ICP signal suppression as much as 50 %. By using both internal standard and appropriate dilution rate, adequate ICP results were obtained. However, the particle size of the wear metals in used oils may create different problems for spectrometric analysis. The detection and identification of metallic particles in the lubricant are very critical because their presence indicates engine oilwetted component wear. The inability to detect large wear particles has led to engine component failure without prior "warning" by spectrometric methods. Several severe incidents have been

HWANG ET AL. ON MICROWAVE DIGESTION

35

reported in which large metallic wear particles produced by severe wear mechanisms have not been quantitatively determined by spectrometric techniques employed [14]. Although "trend analysis" by rotating-disk electrode atomic emission spectrometry (RDE-AES) may serve the purpose to monitor the wear metals in engine, particle-size-independent techniques [15-17], and other dissolution approaches (such as microwave digestion) would give more accurate elemental determination in used oils. The use of high solid nebulizers such as V-groove nebulizer can minimize the clogging and blocking problems, but the short residence time (-2-3 milli-seconds) in ICP (DCP or flame) would not be able to fully decompose all particulates in used oils, which would result in lower analytical results. Conventional ashing and air-refluxing methods have been widely used in the Petroleum industry's analytical laboratories for many years. These traditional methods are always timeconsuming, labor-intensive, and tedious. Additionally, contamination is very likely to occur from the acids (especially sulfuric acid) and/or vessels used during the sample preparation. Moreover, sulfur determination may not be achieved because sulfuric acid is always used in these methods. In 1975, the microwave was finally applied to digest samples in the analytical laboratory [18-20]. The idea that something as revolutionary and simple as the microwave could also be applied to help with digestions was unheard of until then. Using a heating source as fast and reliable as a microwave would digest samples at a greater speed without having to be constantly attended to and would decrease the time it took to get their results back. A solution to the analytical scientists' problem was finally found in simplicity. When the idea of using microwaves started being tested, people used regular household microwaves; this soon proved to be troublesome due to a number of problems that would often be encountered. The microwaves used in homes were made to heat large quantities of food, which was a different case than analytical samples, which tend to be smaller. Analytical samples could also be explosive, give off toxic fumes, or lose valuable elements during the heating process. Utilizing a high pressure microwave system, many innovative and rapid analytical methods were developed in our atomic spectroscopy lab recently. The high operating temperature and pressure attainable in this microwave digestion system ensures total decomposition of complex petroleum crude oils and petrochemical products, with no residual organic content. Sample preparation time had been shortened from days (or, in some cases, even more than a week) to minutes. Our analytical methods can now also be applied to the determination of sulfur in crude oils and polymer samples because sulfuric acid is no longer required in the microwave digestion process. The microwave digestion system also eliminates contamination previously happening due to the use of more traditional digestion vessels and from airborne particulates in the lab environment. In this work, several approaches such as direct dilution, wet-ashing, and a newly developed microwave digestion technique were used to prepare and analyze petroleum samples for the purposes of method and result comparison. The samples used for this study were a NIST lubricant additive Standard Reference Material (SRM) 1848, Residue Fuel oil SRM 1634B, 1634C, and a polymer resin.

Experimental Section
Microwave Oven

A Milestone laboratory microwave oven, MLS 1200 MEGA model, with a 6 vessel carousel rotor was used for all microwave digestion work in this study. The MLS-1200 MEGA

36

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

microwave system consists of the microwave labstation, EM-45 exhaust module, and model 240 control terminal. Power to this system is a 1200 W microwave generating magnetron with 1000 W delivered inside the working chamber. Power emission is microprocessor controlled from 10 to 1000 W in 10 W increments. The microwave power frequency is 2450 MHz. To ensure explosion proof safety, the hardware of the unit is all stainless steel. An external coating was applied to ensure protection against acids and organic solvents. The inside of the working chamber is coated with a 5-layer PTFE coating and is provided with a direct connection to the EM-45 module to exhaust acid vapors from the working chamber itself. The working chamber was purged with nitrogen to create an oxygen free working environment to prevent any fire incident from handling petroleum hydrocarbons. The Microwave Digestion Rotor (MDR) technology has brought microwave digestion to "almost" routine analytical laboratory procedures. It provides fast, reliable, and very safe digestions. In our laboratory, >95 % samples can be completely digested in a one-step digestion procedure. A microwave vessel consists of a PEEK sleeve, adaptor, and spring, as well as a TFM vessel, ring, and cover. The use of TFM and PEEK components would minimize trace element contamination and provide very safe operations.

TraceCLEANAuto System
All microwave Teflon vessels were soaked in 1:1 HC1 overnight prior to cleaning by nitric acid using a TraceCLEAN Auto System from Milestone Inc. TraceCLEAN auto system is specifically designed for cleaning o f vessels and labware based on the principle of acid vapor cleaning. The bottom of the glass vapor container is filled with the nitric acid vapor for cleaning. The bottom plate of the vessel holder sits over the acid level and is automatically lifted up and down (controlled by a built-in control terminal). The bottom plate has several different holes to insert vapor tubes of different lengths or blind stoppers. The labware is placed onto the vapor tubes for cleaning. The acid is heated by an electric temperature controlled heating plate. The acid vapors enter through the vapor tubes into the vessels. The acid vapors then condense on the cold vessel walls and form droplets, thus washing down the vessel surface. The acid droplets gather on the PTFE bottom of the vessel holder and flow into the collecting channel of the glass vapor container. After the system has cooled, the dirty acids can be pumped out through an outlet connection.

Wet Ashing
A specially made, cylinder shape, quartz vessel (12 cm height and 3 cm ID) is used for wet ashing digestion. Common laboratory hot plate and programmable muffle furnace are also needed to conduct wet ashing procedures.

Inductively CoupledPlasma Atomic Emission Spectrometer (ICP-AES)


All the analytical measurements were made on a Thermo Electron Iris ICP. The Iris ICP uses Echelle optics and a unique Charge Injection Device (CID) solid-state detector to provide the flexibility of choosing wavelengths (as a scanning sequential ICP) and also the high throughput (as a simultaneous multi-channel ICP). The ICP system provides complete and continuous wavelength coverage over the analytical range from 165 to 1000 nm. When the optional XUV configuration is chosen, the range is extended to 130 nm, which allows C1 determination at

HWANG ET AL. ON MICROWAVE DIGESTION

37

134.724 nm. All operations of the ICP spectrometer are controlled by the TEVA (Thermo Electron Validated Analysis) software, a 32-bit applications program that runs under Microsoft| Windows | software.
TM

Reagents
All inorganic acids used for microwave digestion are o f "Ultrex | grade from LT. Baker | Chemical Co. Hydrogen peroxide used is Suprapur | from EM Science. For wet ashing, acids used are ACS grade from Fisher Scientific Co, and the flaming sulfuric acid used is from J.T. Baker | Chemical Co. De-ionized water of 18 Mf//cm specific resistivity is always used; it is prepared with either the Millipore | Milli-Q water and a Rios/Elix reverse osmosis purification system or a Nanopure Diamond water system from Barnstead | For direct dilution organics ICP, USP/FCC grade mineral oil and reagent ~rade O-xylene from Fisher Scientific are used. Organometallic standards from Conostan are used for calibration and internal standard.
Procedure

Microwave digestion conditions were established and optimized for our Milestone microwave MLS 1200 Mega system with a 6-vessel carousel MDR. Optimal conditions for microwave digestion depend on sample weight, composition, volume of digestion reagents, and microwave system used. A method was developed and used to digest >95 % of the samples in our laboratory. Approximately 0.1 g of the samples was accurately weighed into a pre-cleaned microwave Teflon digestion vessel. Eight ml of nitric acid, and 2 ml of hydrogen peroxide were added into the digestion vessel. The digestion vessels were placed in the 6 vessel MDR rotor and into the closed, high pressure microwave digestion system to completely convert the samples into an aqueous form. An accurate amount of internal standard (scandium) was added to the samples and brought up to a final volume with deionized water. The prepared samples were then analyzed using IrisT M ICP-AES or ElanT M DRC ICP-MS. For the wet ashing approach, approximately 1-2.5 g of samples were accurately weighed into a quartz digestion vessel. Two ml of fuming sulfuric were added into the container and placed on a hot plate (around 120~ and heated underneath an infrared lamp for 2-3 h. The container was then transferred into a programmable muffle furnace to be heated and ashed for 72 h until the oxidation of carbon was completed. The ash was dissolved in 8 ml of aqua regia with gentle boiling. The digested solution was finally filled up to a 25 ml mark with deionized water when the digestion was completed prior to ICP-AES analysis. For direct dilution organics ICP approaches, modified ASTM D 5185 and D 4951 methods were used in our laboratory. O-xylene is the solvent of choice, and internal standard is used for all organic ICP applications.
Results and Discussion

In Table 1, the direct dilution ICP approach gave lower results because the short residence time (2-3 milli-seconds) of sample aerosol in the ICP plasma was not able to fully decompose the large dry aerosol generated by the nebulization of the used oil. Contrary to direct dilution ICP, microwave digestion "dissolved" and "converted" all analytes into aqueous solution prior to ICP analysis. Thus, the results are much better and reliable. Different from samples #1 and #2, #

38

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

3 is a fresh oil, and microwave digestion results correlate with direct dilution (modified ASTM D 5185 method) quite well. TABLE 1--Comparison of microwave digestion and direct dilution ICP for used~fresh oils

(ppm).
Analytical Approach Microwave Digestion/ICP Direct ICP Microwave Digestion/ICP Direct ICP Microwave Digestion/ICP Direct ICP Sample # Used Oil 1 Used Oil 1 Used Oil 2 Used Oil 2 Fresh Oil 3 Fresh Oil 3 ca 8.88 % 1.67 % 2.53 % 1.21% 0.62 % 0.62 % Fe 5100 960 5214 670 28.9 25.2 Mg Mo S 6.23 % 1.65 % 2.55 % 1.61% 0.74 % 0.74 % Si 1229 85 6.55 % 27.3 % 18 14

760 ........i'106 170 1273 284 25.4 21.2 211 305 177 106 109

For lubricant additive (NIST SRM 1848), as shown in Table 2, boron results are not available because of the formation of volatile boron compounds (e.g., boric acid) when acid was added and heated. Loss has been associated with some boron species, but not with others. Loss of boron was not noted with a Conostan| boron standard. Sulfur results are not available because sulfuric acid is used in the ashing process. Other element results align more closely with the alternative preparation methods. While the precision for boron looks good for the microwave digestion preparation, compared to the two direct dilution methods, the results are slightly lower. This could be attributed to a safety feature of the microwave vessel that causes it to vent under excessively high pressure. Thus, volatile boron species may be lost. Calcium results by the direct dilution additive method compare well with wet ashing and microwave digestion results. Calcium results are 4-5 % lower when direct dilution was applied. All four methods compare well for magnesium and zinc. TABLE 2--Comparison of analytical methods using ICP-AES to determine elements in lubricant additive (%). Microwave Wet Ashing Direct Dilution Di~esti0n B 0.137 +/- 0.019 0.123 +/- 0.008 a N.A. b 0.142 +/- 0.004 Ca 0.359 +/- 0.011 0.359 +/- 0.010 0.358 +/- 0.004 0.348 +/- 0.009 Mg 0.821 +/- 0.038 0.835 +/- 0.014 0.847 +/- 0.034 0.858 +/- 0.016 P 0.788 +/- 0.028 0.793 +/- 0.017 0.750 +/- 0.022 0.750 +/- 0.036 S 2.327 +/- 0.0043 2.223 +/- 0.061 N.A. c 2.236 +/- 0.12 Zn 0.873 +/- 0.022 0.833+/- 0.026 0.877 +/- 0.020 0.847 +/- 0.019 aThe microwavevessel was "popped;"thus someboronloss mightoccur. bThe boronspecieswouldreact with acid to formvolatileboric acid and lost duringthe samplepreparation. cSulfuric acid was used in wet ashing; thus sulfur is not available. The closed vessel, high-pressure microwave digestion gave good results for calcium, magnesium, phosphorous, sulfur, and zinc. It has been proven that closed-vessel microwave digestion can be used to accurately prepare samples containing "volatile" boron and silicon Element Certified Value

HWANG ET AL. ON MICROWAVE DIGESTION

39

compounds. However, in this case, the lower Boron results may have been caused by the vessels "releasing" under high pressure, causing the loss of volatile boron (boric acid). Had the vessel not vented, the Boron result would have been higher and much closer to the certified value. Sulfur information was available because sulfuric acid was not required in the microwave digestion step. More importantly, the microwave digestion approach only took 30 min. The direct dilution approach works well for lubricant additives and fresh oils. However, this approach cfinnot handle used oils containing particulate or volatile compounds. Accurate analysis for sulfur by ICP can be problematic if the samples contain volatile sulfur such as in sulfide form. There is a "vapor enrichment" phase separation that occurs during the pneumatic nebulization process in ICP. Hence, the use of closed vessel, high-pressure microwave digestion can be used to accurately prepare samples containing "volatile" sulfides. ICP can be used to determine the accurate sulfur result. For petrochemical application, as shown in Table 3, microwave digestion ICP results of cobalt were pretty much in line with wet ashing approach. The microwave digestion approach took about 30 min, but it would take at least 4 days or even up to 2 weeks for the wet ashing approach. TABLE 3---Comparison of cobalt content in polymer resin. Wet Ashing Cobalt Content (%) Sample # 1 0.97 Sample #2 0.96 Sample #3 0.97 Microwave Digestion Sample # 1 1.03 Sample #2 1.05 ....... Sample #3 1.04 As another example, a certified fuel residue SRM 1634 C was used to "validate" the microwave digestion with other proven methods. Thus, microwave digestion is still able to produce better results than other traditional methods, such as wet ashing or direct dilutions. In Table 4, the microwave data is able to achieve better results than wet ashing for both 1634B and 1634C SRMs. However, the direct dilution data are just about the same as microwave digestion data. The use of microwave digestion has not only helped our laboratory improve the productivity (and turn-around time), but also provided a reliable tool to handle some difficult analytes and solve very tough problems. It appears that the main advantages of microwave digestion over conventional wet- or dry-ashing procedures are speed (wet ashing approach took up to two weeks, but microwave digestion can handle the samples within hours), simplicity of microwave digestion procedure, less sample and reagents needed, and much better contamination control. More than 95 % of the samples can be completely digested with our single microwave digestion method. Others need HC1 and/or HF added to completely digest "high" content metals or silica (in dirt of used grease samples). However, it is difficult to completely digest >0.5 g of petroleum hydrocarbon samples by microwave digestion. Thus, for certain samples, wet ashing may be the method of choice when larger samples are required to get consistent sampling and representative samples. The other drawback for microwave digestion is to have "poor" method detection limit because a small quantity of samples is used resulting in higher dilution rates (-100) than the wet

40

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

ashing approach (~20). To ensure accurate and reliable analytical results, microwave vessels need to be cleaned thoroughly prior to use. Common airborne contaminants, such as iron, sodium, silicon, and calcium are very difficult to completely remove unless a clean room is used. For boron and silicon analysis, deionization is not sufficient to remove these weak bases effectively. Deionization with reverse osmosis water purification system should be used instead to ensure that the purified water is free of borate and silica. TABLE 4---Comparison of different methods determining trace metals in fuel residue. Nickel (ppm) Vanadium (ppm) Sulfur (%) NIST Certified Value 28 +/- 2 55.4 +/- 1.1 2.80 +/- 0.05 Microwave Digestion 1634B 26.4 55.1 2.72 1634B Dulpicate 27.0 55.5 2.73 Direct Dilution 1634B 27.9 56.9 2.96 1634B Duplicate 27.7 56.7 2.95 1634B Triplicate 27.8 56.8 2.98 NIST Certified Value 17.54 +/- 0.21 28.19 +/- 0.40 Supplemental Value Microwave Digestion 1634C 17.1 29.5 2.11 1634C Duplicate 17.0 29.8 2.13 Direct Dilution 1634C 17.1 29.5 2.11 Wet Ashing 1634C (2.5 g) 2.95 21.61 NA 1634C Duplicate (2 g) 8.74 24.56 NA 1634C Triplicate (1 g) 15.05 27.49 NA

Summary
Microwave digestion has been proven to help speed up and improve the productivity and quality of the sample preparation in our laboratory. It has become the method of choice for many unknown and/or complicated matrix samples. "Obtaining a representative sample" can be a problem when smaller quantities (<0.2 g) are used for microwave digestion. Other applications, such as speciation, organic extraction, and sample preparation for wet chemistry are some areas to which we can apply this powerful tool in order to further improve our laboratory practice.

Acknowledgments
The authors would like to thank Karsia Yip, and other ICP Laboratory colleagues for their contribution and assistance.

References
[1] Mortier, R. M. and Orszulik, S. T., "Chemistry and Technology of Lubricant," Blackie Academic & Professional, 1997.

HWANG ET AL. ON MICROWAVE DIGESTION

41

[2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]

Hwang, J. D. and Wang, W. J., "Application of ICP-AES to Analysis of Solutions," Applied Spectroscopy Reviews, Vol. 30, No. 4, 1995, pp. 231-350. Bock, R., "A Handbook of Decomposition Methods in Analytical Chemistry," International Textbook Company, 1979. Matusiewicz, H. and Sturgeon, R. E., "Present Status of Microwave Sample Dissolution and Decomposition for Elemental Analysis," Prog. AnaL Spectrom., Vol. 12, 1989, pp. 2139. Kingston, H. M. and Jassie, L. B., AnaL Chem., Vol. 58, 1986, pp. 2534-2541. Kingston, H. M. and Jassie, L. B., "Introduction to Microwave Sample Preparation-Theory and Practice," American Chemical Society, 1988. Kingston, H. M. and Haswell, S. J., "Microwave-Enhanced Chemistry-Fundamentals, Sample Preparation, and Applications," American Chemical Society, 1997. Hoffman, J. L. and Lundell, G. E. F., Journal of Research of the National Bureau of Standards, Vol. 22, April 1939, pp. 465--470. Zief, M. and Mitchell, J. W., "Contamination Control in Trace Element Analysis," John Wiley & Sons, 1976. Eisentraut, K. J., Newman, R. W., Saba, C. S., Kauffman, R. E., and Rhine, W. E., AnaL Chem., Vol. 56, 1984, 1087A. Sychra, V., Lang, I., and Sebor, G., Prog. Analyt. Atom. Soectrosc., Vol. 4, 1981, pp. 341426. Bowman, J. A. and Willis, J. B., Anal Chem., Vol. 39, 1967, p. 1210. Bansal, J. G. and McElroy, F. C., "Accurate Elemental Analysis of Multigrade Lubricating Oils by ICP Method, Effect of Viscosity Modifiers," SAE Technical paper 932694, 1993. Kagler, S. H., Jantzen, E., and Fresenius Z.,AnaL Chem., Vol. 310, 1982, p. 401. Saba, C. S. and Eisentraut, K. J., AnaL Chem., Vol. 49, 1977, p. 454. Saba, C. S. and Eisentraut, K. J., AnaL Chem., Vol. 49, 51, 1979, p. 1927. Brown, J. R., Saba, C. S., and Rhine, W. E., Anal. Chem., Vol. 52, 1980, p. 2365. Abu-samra, A., Morris, J. S., and Koirtyohann, S. R., Anal. Chem., Vol. 47, 1975, p. 1475. Barrett, P., Davidowski, L. J., Penaro, K. W., and Coopeland, T. R., Anal Chem., Vol. 50, 1978, p. 1021. Nadkami, R. A., AnaL Chem., Vol. 56, 1984, p. 2233.

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JAI12968 Available online at www.astm.org

Joseph Paszek, 1, Kelly J. Mason, 1Anthony S. Mennito, 1 and Frank C. McElroy 1

Advances in ICP-MS Technologies for Characterization and Ultra-Trace Speciation as a Tool for the Petroleum Industry
ABSTRACT: Particular elements can, even at trace levels, impact product quality, deposit formation, and regulatory compliance. Characterization o f trace and ultIa-trace level constituents in aqueous and/or organic matrices has often required the need to employ more than one analytical technique. In recent years, improvements in ICP-MS technologies have made it a preferred technique for multi-element trace level characterization in a variety o f matrices. Coupling ICP-MS with various liquid chromatography techniques has enabled a unique capability to speciate organo-metallics at ppb levels. This paper will show how ICP-MS has been successfully used to measure ppb levels in refinery effluent streams. It will demonstrate how hyphenated ICP-MS techniques assisted in development of bioremediation options for selenium removal in wastewater treatment plants. KEYWORDS: selenium, selenocyanate, selenite, selenate, ICP-MS, and ion chromatography

Introduction

Today's analysts no longer focus on just trace elemental analysis, but rather they face the challenge of being able to speciate and quantify many high profile elements. ICP-MS coupled with suitable chromatographic separation techniques offers a unique combination of simultaneous multi-element capabilities with ppb and ppt detection limits and the ability to perform elemental speciation. In this paper, we will present how Inductively Coupled Plasma Mass Spectrometry (ICP-MS) has been successfully used with liquid chromatography to provide trace level speciation information. Selenium enters the refinery as contaminates of processed crude oils. As the refiner processes the crude oil, the selenium contaminate migrates from the crude oil into the refinery's striped sour water (SSW) streams. Once in the SSW stream, the selenium contaminate, which is an environmental concern, is now part of the wastewater stream. The problem for the refiner is how to efficiently remove the selenium prior to discharging the water to an appropriate catch, such as a bay or municipal water treatment plant. It has been shown that the predominate state of the selenium in these SSW streams is as the selenocyanate 2(.SeCN-) [1]. The selenocyanate in SSW is oxidized by refinery biotreaters to selenite (SeO3 "), which can then be removed via an iron based treatment process. One of the most effective of these iron-based processes is iron co-precipitation process. However, the selenium can also occur as s e l e n a t e (8eO42-), which is not removed during the iron coprecipitation process [2]. In order to design effective treatment protocols to remove the selenium, more than just the concentration of the metal must be determined. Without knowledge of the types of selenium species present in the water, design of an effective removal treatment protocol
Manuscript received 25 October 2004; accepted for publication 11 April 2005; published October 2005. Presented at ASTM Symposium on Elemental Analysis o f Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. i ExxonMobil Research and Engineering, 1545 Route 22 East, Annandale, NJ 08801. * Primary author to whom correspondence should be addressed.

Copyright9 2005 by ASTMInternational,100 Ban"HarborDrive,PO Box C700, West Conshohockoa,PA 19428-2959. 42

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

43

is impossible. The need to know that the species of selenium present, as well as the fact that the metal would be present in the untreated SSW at levels ranging from a few parts-per billion to low parts-per-million, make the analysis very difficult. While ICP-MS has a demonstrated capability for being able to provide accurate and precise selenium measurements at these levels, there still remained the problem of the identification of the selenium species present in the samples. Research efforts defined here focused on developing a quick and reliable separation technique to separate the selenium species present in the various SSW and wastewater samples prior to quantification by ICP-MS. Building upon a study that was performed for the Western States' Petroleum Association (WSPA) [1] to optimize iron-based selenium removal processes, a scheme was developed to measure total selenium, and the individual selenium species present in the SSW and aqueous streams. Total selenium was measured by ICP-MS after a peroxide/acid digestion pretreatment of the samples. Speciation of the selenium was accomplished using an anionic ion chromatographic separation followed by quantification of the selenite, selenate, and selenocyanate using ICP-MS. The goal of the analytical program was to develop a measurement method with a detection limit of 10 ixg SeL "1. By the end of the program, we were able to not only demonstrate that this goal was achievable, but that detection limits of 2 ~tg selenite'L"1 and 0.2 ~tg selenate'Ll were possible.

Instrumentation
The instrumentation used for this work to quantify and speciate selenium in WWS is listed in Table 1. TABLE 1--Instrumentation usedfor Selenium Speciation Study. Instrumentation/Software Manufacturer/Model Perkin Elmer/Sciex Elan 6000 ICP-MS ICP-MS Chromatographic Integration Program - (P/E software) Integration Software Dionex Model 2000i Tertiary Ion Chromatography System IC Analytical Column Dionex AS-4-SC (4.0 m m x 250 mm.) Dionex AG-4-SC (4.0 mm x 50 mm.) Guard Column Gilson Peristaltic Pump Spray Chamber Scott Meinhard Nebulizer

Reagents and Materials


All eluents were purged with helium. In addition, helium was used to pressurize and blanket eluent bottles as well as operate the IC pneumatics. All of the standards materials and eluent constituents were obtained from Aldrich Chemical. The stock 1000-mg Se'L"1 selenite and selenate standards were made from their corresponding sodium salts, while the selenocyanate stock was made from the potassium salt. The germanium internal standard was made by diluting a 1000 m g L "l aqueous commercial standard to 10 mg/L in 300 mM NI-hOH. The reason for using ammonium hydroxide was to avoid the precipitation of germanium oxide when the internal standard was mixed with the IC effluent. The eluents were made by dissolving the appropriate amount of pure material in

44

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

distilled deionized water that was prepared using a Millipore Milli-Q deionization system. Complete dissolution of the p-cyanophenol in the eleuant required the use of an ultrasonic bath. The sodium hydroxide solutions were made fresh and blanketed under helium in the eluent reservoirs. The 50 % (m/m) sodium hydroxide concentrate was discarded every 30 to 45 days to minimize the influence of any dissolved carbonate that may have accumulated in the sodium hydroxide during that time period. The 30 % (w/w) hydrogen peroxide used in the digestion was commercially available reagent grade material. The concentrated nitric acid used was Ultrex grade acid available from Baker Chemicals. The 1 % (v/v) used in the digestion procedures was prepared by diluting 10 ml of the Ultrex grade to 1 L using deionized distilled water (DDW).

Experimental
An analytical scheme was developed [3] to characterize both the concentration and nature of the selenium species present in the various streams measured (Fig. 1). Because an in-depth discussion of all of the techniques would be beyond the scope of this paper, the reviewed work will focus on the efforts to measure total selenium as well as selenite, selenate, and selenocyanate. A Sciex Elan Model 6000 Inductively Coupled Plasma-Mass Spectrometer system equipped with a water-cooled spray chamber was used for all of the selenium measurements. The operating conditions used for the ICP-MS are presented in Table 2, while conditions for the IC are presented in Table 3.

Scheme for Selenium Speciation

FIG. 1--Scheme for selenium speciation.

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

45

TABLE 2--General ICP-MS instrument conditions. Parameter Nominal Value RF Power 1.2 KW Nebulizer Gas flow 0.6 L/min-1 Plasma Gas Flow 15 L/min-1 Auxiliary Gas Flow 1.0 L/min"1 Sample Flow Rate 0.5 ml/min"1 Nebulizer Meinhard TABLE 3 - - I C conditions used for selenium speciation. %El %E2 %E3 5 5 5 50 50 0 0 5 5 0 95 0 95 0 95 0 50 0 50 40 60 40 60 0 95 0 95 E1 = Ammonium Carbonate (8 ~ ) E2 = 50 mM p-cyanophenol in 78 mM NI-I4OH E3 = Distilled Deionized Water ..... Comments

Time (min.) 0 0.1 0.3 5.0 7.0 7.1 18.0 18.1 20.0

Inject Load

End Run

Initially, the interface between the Ion Chromatography (IC) and ICP-MS used y-tubing fitting where the effluent from the IC could feed one arm of the y-tubing fitting. A peristaltic pump fed germanium (Ge) internal standard into the other arm of the y-tubing fitting. The output of the y-tubing fitting was connected to a Meinhard nebulizer of the ICP-MS. However, with this arrangement, observed fluctuations in the intensity of the internal standard were attributed to incomplete mixing of the eleuant and internal standard. In the final setup, which was used for all of the speciation measurements in this program, the effluent from the IC system was interfaced to the ICP-MS via a short length of 0.010 mm ID PEEK tubing connected to one of the side arms of a plastic "T" fitting. The other arm of the T was connected to the output of the peristaltic pump supplying the Ge internal standard. The position of the arm supplying the internal standard was at a 90 ~ angle to the flow of the IC effluent to the ICP-MS. This arrangement was used to ensure adequate mixing of the internal standard with the stream carrying the selenium species.
Results and Discussion

The samples of striped sour water (SSW) used for this study were refinery sour water streams that have had the bulk of the ammonia and hydrogen sulfide removed or striped. No preservatives were added to the samples prior to analysis. All of the samples were kept refrigerated with a minimum amount ofheadspace in the bottles to minimize the reactivity of the selenium species present in the various samples.

46

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Because the Se in the SSW samples could be present in many ways, such as particulate, all total Se measurements were carried out on unfiltered samples of the water. These samples were digested prior to any total Se measurements by using a procedure that employed a combination of hydrogen peroxide and nitric acid to convert all of the selenium to soluble selenium. In the total selenium procedure, an aliquot of the unfiltered sample is treated with 3-4 ml of hydrogen peroxide and heated to 95~ to digest the bulk of the organic materials present in the sample. The volume of the solution is reduced to half its original volume, and an additional 3-4 ml of hydrogen peroxide is added to the solution. When the peroxide-treated solution has been further reduced to approximately one-fiRh its original volume, one more addition of peroxide is performed. The solution is then heated until all effervescence subsides and the solution become colorless. At this point, 0.1 ml or more of the 1 % (v/v) nitric acid is added to dissolve any and all precipitated material. Once the solution is clear and colorless, it is transferred to a 50-ml volumetric flask and made up to volume using DDW. The final step used to treat the digested SSW samples prior to analysis by ICP-MS was to remove the bulk of the sodium present in these samples. Removing the sodium prior to measurement would minimize the effects high sodium levels could have on the plasma characteristics, thereby impacting the overall precision of the total selenium measurements. Sodium was removed by passing the digest solutions through a Bio-Rad AG50W-X8 cation exchange column prepared according to manufacturer's directions. The first 1-2 column bed volumes are discarded and the rest retained and measured by ICP-MS. All connections between columns and instruments were made using 0.007 mm ID PEEK tubing. A schematic of the instrument arrangement, as well as representative chromatograms of the selenium speciation, are presented in Figs. 2 and 3.

FIG. 2--Schematic of the LC/IC-ICP-MSarrangement.

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

47

50000

100 ppb Standard


SeO= ="
~000~

Se04 ;!-

~UUUt

_=
20000

O0

40C

230

3C0

~4~0 lime

~CO

.300

?CO

300

9C0

FIG. 3--Selenium Speciation Standard. The 82Se mass was used to measure the selenium for both the total and speciated measurements. The reasons for choosing this line, which is not one of the more abundant lines (-8.7 %), were based on minimizing the mass overlap interference from Ar2, which would have quenched the more abundant 78Se and 8~ signals. However, even the 82Se mass line suffered from mass overlaps from Kr, HBr, and SO3 that, if left uncorrected, could affect the overall accuracy of the selenium measurements. In order to correct the 82Se signal, the responses for these masses were monitored in all of the measured samples and mathematically corrected for in the 82Se signal. Monitoring the 78Se mass line in conjunction with the 82Se line could be used as a means of testing the level of interference from sulfur species present in the samples being measured. Although we did not pursue it as part of this program, we were able to demonstrate that the lower limit of quantification could be extended by the use of a short concentrator cartridge mounted in place of the injection loop on the IC sample introduction section. This concentrator cartridge was typically a short 5-cm column containing the same substrate as the analytical column. Using this approach, we were able to load 1 ml of the SSW water sample without any selenium breakthrough. This approach effectively gave us a 40-fold improvement in the lower selenium detection limit. The concentrations of the mixed standard (selenate, selenite, and selenocyanate) were confirmed by measuring the total selenium concentration of the solution. The mixed standards were also carried through the digestion procedure as a quality control check of the recovery. Recoveries were generally in the range of 95-98 %. Examples of the spike information, as well as a comparison of total Se data versus the sum of the calculated species can be found in Table 4. Additionally, selected samples of the SSW stream were analyzed on different days to confirm the repeatability of the measurements. Figure 4 shows repeat measurements for the three selenium species. The duplicate measurements for three selenium species for various samples differed by no more than 4 % of the mean.

48

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE Sample ID
1

4--Comparison of speciated and digested total Se measurements.


Se (I.tgL as "l) SeO3 Total (Digested) 102 10 <10 112 103 70 5 <10 75 70 71 5 <10 76.0 80 142 19 <10 161 154 500 500 500 1500 1430 Note: % Recovery = (Total Speciated/Total Di~ested)*100 SeO4 SeCN Total % Recovery 109 107 95 105 105

2 3 4 5

FIG. 4

Selenium species repeatability.

Once the ability of the IC-ICP-MS system to repeatedly separate the three selenium species was confirmed, efforts focused on investigating process variables which would influence the concentration o f various selenium species, specifically selenocyanate (SeCN) in the SSW. One such process variable studied was pH. The study of the pH effects on the conversion of the SeCN to other selenium species, such as selenate and selenite, was undertaken to define the pH boundaries for any pre-treatment or treatment conditions for the SSW. Knowledge o f this behavior would influence the design of any systems, e.g., iron co-precipitation, built to remedate the Se in the SSW streams. A series o f samples o f a single SSW stream was buffered at several different pH levels to determine the effect pH and time had on the SeCN. The study showed that the pH o f the streams studied had to be strongly acidic, pH less than or equal to 2, to cause any significant change in the concentration of the SeCN (Fig. 5). A plot o f the results for the buffered solutions shows that, even after several days, there is not an appreciable loss of the SeCN signal in the samples whose pH was greater than 2 (Fig. 6). However, at a pH of 2 there was an appreciable loss o f SeCN over the course of several days with the level dropping below the quantification limit after six days. Knowing not only the level o f the selenium present but also its form and how it reacted to various chemical treatments was critical to the success o f the remediation efforts. The

PASZEK ET AL. ON ULTRA-TRACE SPECIATION

49

ruggedness of the separation technique coupled with the sensitivity and precision of the ICP-MS guided the development of effective, efficient means for treating refinery streams.
z-60.7-

~oao-

SeCNin HC1Matrix pH = 2

InitialSeCN= 1180ppb

~o"

I Day1, SeCN=590ppb 1

lo

.... ~ .... , .... , .... , .... i .... , .... ~ .... ""'lb

.... ,,,,,I

.... ""',6

.... ' .... ~' .... ,,,,,&,,,,I,,,,~

.... ,,,'

FIG. 5--SeCN degradation at p H = 2.

FIG. 6---Low p H SeCN degradation.


Conclusion

ICP-MS offers sensitivity, accuracy, and precision with multi-element capabilities, which is not available with other analytical techniques. The unique qualities of ICP-MS, when coupled with proper chromatographic techniques, provides the analyst with concentration and nature of the elemental species present in their samples. Other elemental techniques, which have been coupled with chromatographic systems, do not exhibit the same robustness as has been demonstrated with the ICP-MS [4,5]. Exploiting this ability, we were able to use ICP-MS coupled with an anionic chromatography separation, to determine ideal conditions for remediation of selenium contamination in various aqueous refinery streams.

50

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

References

[1] Brown and Caldwell, Western States Petroleum Association Removal Of Selenium By IronBased Treatment Processes Final Report, California Public Health Foundation, May 1995. [2] Cotton, F. A. and Wilkinson, G., Advanced Inorganic Chemistry, A Comprehensive Text, Second Edition, John Wiley & Sons, Inc., 1966. [3] Roehl, R., "Recent Advances in the Speciation of Selenium in Petroleum Refinery and Municipal Waste Waters Using ICP-MS and ICP-MS Coupled with Liquid Chromatography," Transcript of Oral Presentation, 1996 Winter Conference on Plasma Spectrochemistry, Ft. Lauderdale, FL, January 8-13, 1996. [4] Zhang, X. H. and Koropchak, J. A.,AnaL Chem., 71, 3046, 1999. [5] Roychowdhury, S. B. and Koropchak, J. A., AnaL Chem., 62, 484, 1990.

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JAI12969 Available online at www.astm.org

S. Dreyfus, 1,2 C. Pdcheyran, 1 C Magnier, 2 A. Prinzhofer, 2 C. P. Lienemann, 2 and O. F. X. Donard 1

Direct Trace and Ultra-Trace Metals Determination in Crude Oil and Fractions by Inductively Coupled Plasma Mass Spectrometry
ABSTRACT: A method was developed to analyze direct trace and ultra-trace metal elements in crude oil and its fractions (maltenes-asphaltenes) by ICP-MS after sample dilution in xylene. Efficient introduction of organic compounds requires addition of O2 for complete combustion of the sample; carbon deposit on cones (interface) and extraction lenses was minimized by optimization of argon to oxygen ratio in the plasma. A PFA-100 (100 ~tl.mn MicroFlow Nebulizer and a Scott "double pass" chilled spray chamber 1) were associated for an optimal introduction of petroleum products. The Q-ICP/MS was equipped with platinum cones to limit drifts caused by cone corrosion during an organic analysis. A standard addition method was realized for the calibration procedure in order to control the matrix effects. Method validation was completed by analyzing three certified reference materials from the National Institute for Standards and Technology (NIST). NIST 1085b, NIST 1084a, and NIST 1634c accuracies were approximately 10 %. Detection limit of ultra trace elements in xylene were at pg.g-t level. A clean PFA filter system was developed to separate the oil fractions (maltenes and asphaltenes) by precipitation of asphaltenes (heavy fraction) in n-heptanc. For optimal detection conditions, the samples were diluted in xylene according to their pre-estimated element concentration range. Total mass balance shows a recovery close to 100 % for Ni, V, Cu, Mo, Ag, Sn, Ba, and Pb. These results show that elements analyzed are highly concentrated in the asphaltenic fraction.

KEYWORDS: inductively coupled plasma mass spectrometry, organic solvents, microflow nebulizer, trace and ultra trace metals in crude oils, maltenes, and asphaltenes
Introduction

In recent years, intensive studies have been dedicated to trace and ultra trace metals determination in petroleum products. The need o f the petroleum industry in this field is highly related to exploration and exploitation activities: geochemical characterization o f source rocks and basins, and corrective actions during oil production and refining (e.g., prevention o f catalyst poisoning, corrosion, and pollution control). The existence o f metals in fossil fuels was first established b y Alfred Treibs in the 1930s [1]. In the 1960s, E. W. Baker has started his studies on the characterization o f petroporphyrins [2]. Since then, m a n y works have been published on the geochemistry o f organometallic complexes [e.g., Ni(II), VO(II), Cu(II), Fe(iII), Ga(III), Mn(III) porphyrins] in geological materials (e.g., bitumens and marine sediments) [3,4]. Trace metals are incorporated into crude oils as organometallic compounds (e.g., geoporphyrins), but also as inorganic components originating from formation and production o f water (e.g., contamination of B a and As) and mineral matter (e.g., trace elements in clay minerals) [5]. The inorganic or Manuscript received 8 October 2004; accepted for publication 22 March 2005; published October 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. :LCABIE/CNRS UMR 5034, 2, Avenue P. Angot, 64053 Pan Cedex 9, France. E-mail: sebastien.dreyfus@mtiv-pau.fr. 2Institut Frangais du Pdtrole, 1 et 4, Av. de bois Prdau, 92852 Rueil Malmaison Cedex, France.

Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

51

52

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

metal organic species of crude oils (generally less than 1%) is important for two reasons. Firstly, a few organometallic species have damaging effects during processing and upgrading of crude oils. Secondly, trace metals in crude oils may be used as geochemical biomarkers in oil-oil or oilsource rock correlation, in the identification of source rock depositional environments, and in the quantification of oils' thermal maturity and biodegradation levels. The evolution of elemental signatures, such as Ni and V, in crude oils has already been used to determine the level of reservoir biodegradation. However, new parameters are needed to improve the general knowledge about migration, maturation, mixing, and biodegradation of the oil from an inorganic or bioinorganic angle. Sensitive methods are then required to analyze other elements occurring in the petroleum in very low concentration (e.g., Mo, Pb, Cu, Cd, Sn, Ba, etc.). The instrumental limitations of the most common techniques Used so far (AAS, ICP-OES) have been the main reason for the poor knowledge about inorganic constituents ofoil [6]. Recently, inductively coupled plasma mass spectrometry (ICP-MS) has been used to analyze trace elements in oil and other petroleum products [7-9]. Performances of ICP-MS allow the detection of very low concentrations in an organic matrix. A sample introduction method in the plasma is fundamental for the use of organic aerosol. The presence of organic vapors in the plasma causes carbon deposit on cones and extraction lenses. Such deposits may affect the transport efficiency of ions and plug cones, and they may even blow out the plasma. In addition, polyatomic interferences caused by carbon-argon-oxides species do not allow the quantification of certain elements (e.g., Mg, Cr," Fe) with a Q-ICP-MS. Moreover, a significant quantity of organic vapors introduced in the ICP can also increase matrix effects [10,11]. In order to minimize these problems, several solutions have been found, with more or less success. Mineralization of petroleum samples can be performed in order to remove the organic matrix [12]. However, this method has some disadvantages: the sample preparation is time consnming, volatile elements can be easily lost, and the blank contamination may be important. Ultrasonic nebulization (USN) used in addition to a desolvation membrane permits an efficient desolvation of the organic aerosol [6,13]. Although this nebulizer allows one to achieve excellent detection limits and to reduce carbon based interferences, certain metal species (e.g. volatiles species) can be lost during the desolvation process [13]. Furthermore, crude oil and its fractions can easily foul the USN and its desolvation membrane. A MicroFlow Nebulizer associated to a chilled spray chamber appears to be the best introduction tool for an optimal analysis of petroleum products by ICP-MS. This combination allows working with very low volumes of petroleum products and minimizes the quantity of organic solvent entering into the plasma. The major advantages of this technique are very low detection limits, no loss of metallic species, and no contaminations [ 14]. This methodology has never been applied to the analysis of trace elements in crude oil and its fractions by ICP-MS. In this study, we have developed a technique to analyze directly trace and ultra trace metal elements in crude oils and its fractions (maltenes-asphaltenes) by ICP-MS after sample dilution in xylene. This method was then used to produce a new set of data, using the above-mentioned methodology, on the distribution of metal trace elements between asphaltenes and maltenes in crude oils from a Brazilian petroleum basin.

DREYFUS ET AL ON DIRECT TRACE AND ULTRA-TRACE

53

Materials and Methods

Instrumentation
Samples were analyzed on a X7 Series CCT ICP-MS (Thermo Electron Corporation, Winsford, UK) equipped with a PFA-100 MicroFlow Nebulizer (Elemental Scientific Inc., Omaha, NE) fitted to a Scott-type double pass chilled spray chamber and with Platinum Sampler and Micro Skimmer cones (Thermo Electron Corporation, Winsford, UK). Oxygen was supplied continuously into the nebulizer gas flow for complete combustion of the sample and reduction of carbon deposits on the interface cones. A 1.14-mm injector orifice de-mountable torch was fitted to the instrument. It is essential when running organic solvents to increase the forward power to help break apart the organic bonds. The power had to be set to 1550 W. The Q-ICP/MS was equipped with platinum cones to limit drifts caused by cone corrosion during an organic analysis. The ICP-MS operating conditions are listed in Table 1. TABLE 1--ICP-MS operating conditions for xylene introduction. ICP-MS system ICP/MS X7 CCT THERMO| Electron RF power (W) 1550 Cones (orifice diameter, mm) Sampler (Platinum, 1.0); Skimmer (Platinum, 0.7) Spray chamber Scott type double pass at -7~ Nebulizer PFA-100 MicroFlow Injector diameter (nun) 1.14 Ar plasma (l/rain) 15.5 Ar outer (1/mn) 1.1 Ar nebulizer (l/mn) 0.5 02 nebulizer (1/mn) 0.066 51~, 58-,.~-. 60~T- 62~T- 6 3 ~ 65,'~ 9 8 ~ r 100~, 107-Isotopes measured (m/z) V, 1N1, I N I ~ 1~11, k.,U, U U , lVlO~ _rVI.O, Ag, l~ llSSn' ]2~ ' 137Ba 138Ba 2~ ' 207pb' 208pb ' ' Data acquisition parameters Resolution (amu) 0.8 Sweeps 100 Dwell time (ms) 20 Replicates 5

Reagents
Xylene "Baker Analyzed" ACS and n-Heptane "Baker Analyzed" were purchased from J. T. Baker (MaUinckrodt Baker Inc., Phillipsburg, NJ). Membrane filters in Regenerated Cellulose (0.45 ~tm, O: 47 ram, type 184) were obtained from Sartorius AG (Goettingen, Germany). Conostan| S-21 Blended Standards (100 ppm organic multi element standard in oil) was used in this work (ConocoPhillips, Houston, TX) for the samples' calibration. Method validation was completed with three Standard Reference Materials (SRM) from NIST (National Institute for Standards and Technology, Gaithersburg, MD): 1085b, 1084a lubricating oils, and 1634c (residual'fuel oil).

54

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Oil Fractionation
In order to limit filtration blank contamination, a clean PFA filter unit was developed to separate the oil fractions by precipitation of asphaltenes (heavy fraction) in n-Heptane. Membrane filters of size 0.45 ~tm in regenerated cellulose were used with the filter unit. One gram of crude oil was dissolved in 50 ml of n-H~ptane (in a 250 ml PFA balloon). The solution was then heated for 20 rain at 60~ under magnetic stirring and finally filtered under vacuum. Asphaltenes were accumulated on the filter and stocked in 50 ml polypropylene tubes at room temperature in the dark for further analysis. Maltenes, which are soluble in n-H~tane, were collected in a polyethylene Erlenmeyer and evaporated under pure nitrogen flow in 50 ml polypropylene tubes.
Results and Discussion

Sample Introduction and Measurement


The sample introduction technique used to introduce organic aerosol into the ICP-MS is critical for success. Experience shows that the minimization of solvent quantity brought in the plasma and the oxygen to argon ratio are the critical parameters for analysis of mixed organic products (e.g., diluted crude oil) by ICP-MS in order to reduce polyatomic interferences and matrix effects. The spray chamber had to be chilled in order to reduce the amount of solvent load entering the mass spectrometer. For the same reason, a 1.14 mm injector orifice de-mountable torch had to be fitted to the instrument. With this torch the aerosol enters faster in the plasma. Thus, the nebulization gas flow can be decreased without any loss of sensitivity in order to reduce the quantity of aerosol introduced in the plasma. As a result, the signal detected by the mass spectrometer is more stable, the signal to background noise ratio is better, and thus the limit of detections are lower (Fig. 1). Detections limits (LOD-3o) for all trace elements in xylene obtained with this torch are below the 0.3 ng.gq level (LOD are at pg.g-i level for ultra trace elements such as Pb, Ba, Sn, etc.). LOD were calculated from the sensitivity (slope) of the analyte addition curve and the standard deviation of a xylene blank. Rinse efficiency of the spray chamber is also an important factor. A good rinsing efficiency limits chamber fouling by heavy organic compounds and thus minimizes signal drifts and memory effects on analytes. Figure 2 shows the signal intensity of three isotopes analyzed in crude oil diluted 100 times in xylene as a function of stabilization, acquisition, and rinsing times. A stabilization time of 2-3 min is necessary before measuring intensities. For H3Cd isotope, the signal is still not stabilized after 3 min, and a drift of the intensity remains during all of the acquiring time. Matrix effects can be important during the analysis of petroleum products. They generally disturb intensities emitted by a detected isotope [10]. Organic matrix may modify element ionization in the plasma and consequently cause a variation of sensitivity. For these reasons, a standard addition method was performed for the calibration procedure to control the matrix effects. Concentrations of metals in analyzed organic samples must be compatible with ICP-MS potential in order to obtain reliable results. Experience shows that concentrations detected by the mass spectrometer ideally should be in the range of 1-100 ng/g. Thus, for optimal detection conditions, samples were diluted in xylene according to their pre-estimated' element concentration range.

DREYFUS ET AL. ON DIRECT TRACE AND ULTRA-TRACE

55

FIG. 1--Detection limits (3a) of trace elements in xylene obtained with the 1.14 mm injector orifice torch compared to the detection limits obtained with the standard torch (1.5 mm orifice injector).

FIG. 2--Intensity of98Mo, t13Cd, and 2~ rinsing times.


Validation of the Methodology

as a function of stabilization, acquisition, and

Accuracy and repeatability of the analytical procedure was assessed by analyzing certified elements in three SRM, diluted 2000 (NIST 1085b), 800 (N-IST 1084a), and 200 (NIST 1634c) times in xylene, and by using the standard addition method. Table 2 summarizes concentrations

56

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

o f trace elements and percentage error (E, %) obtained on NIST 1085b, 1084a, and 1634c and compared to the certified values. Accuracies are better than 20 % for all elements and thus permit to validate this analytical technique on the analysis of trace metals in petroleum products by ICP-MS. TABLE 2--Concentrations and percentage error (E, %) compared to the certified value obtained on NIST 1085b, 1084a, and 1634c. Element V (51 V) Ni (62 Ni) Cu (63 Cu) Zn (66 Zn) Mo (98 Mo) Ag (107 Ag) Cd (112 Cd) Sn (118 Sn) Ba (138 Ba) Pb (208 Pb) V (51 V) Ni (62 Ni) Cu (63 Cu) Mo (98 Mo) Ag (107 Ag) Sn (118 Sn) Pb (208 Pb) Certified Values' NIST 1085b (~tg.~') z 297.8 4.61 295.9 7.4 295.6 + 8.5 296.8 + 6.8 300.6 q- 3.2 304.6 8.9 302.9 + 5.1 299.4 ~- 4.8 300.1 + 2.4 297.7 6.8 NIST 1084a (~g.g-1)2 95.9 + 9.4 99.7 4- 1.6 100 + 1.9 100.3 4- 1.4 100.4 3.8 97.2 2.6 101 1.3 NIST 1634c (ixg.g'l)2 This Work (~tg..g -1)'241 4 303 a: 6 313 :~ 7 303 4 255 :~ 7 340 4 347 ~: 5 244 ~: 5 307 :~ 8 289 :L 7 81 + 8 108 3 117 9 4 92 5 103 9 6 105 -~ 4 95 5 E (%) -19 2 6 2 -15 11 14 -18 2 -3 -16 8 17 -8 2 8 -6

V (51V) 28.2 26.7 0.8 -5 Ni (62 Ni) 17.5 0.2 16.2 0.6 -8 1All concentrationsin ~tg.g errors on concentrationswere calculated from 95 % confidenceintervals~ "1, ~l'hose SRM are initially supplied for ICP-AES analysis. The certified values of NIST 1984a and NIST 1985b are too high for an ICP-MS analysis (respectively 100 ~tg.g and 300 ~tg.g'l). Both SRM need to be highly diluted "l before their introductionin the ICP-MS. This dilution may induce errors in the measurements.

Analysis of Crude Oil and lts Fractions, Total Mass Balance (TMB) and Repeatability of Oil Fractionation
Once the methodology for trace element determination in petroleum products by ICP-MS was validated, it was applied to several crude oils (supplied by Petrobr~ts) and their fractions originating from a Brazilian petroleum basin. In order to prove the reliability o f analytical results obtained after oil fractionation, total mass balance (TMB) was calculated for each element measured. Table 3 shows an example o f crude oil and its fractions analyzed by ICP-MS. TMB results were in the range o f 90-110 % for most o f the elements analyzed. In order to confirm the repeatability o f oil fractionation and analysis, each sample (oil + fractions) was analyzed three

DREYFUS ET AL. ON DIRECT TRACE AND ULTRA-TRACE

57

times with a different sample preparation (including oil fractionation, samples' dilution, and addition calibration procedure) for each time. This study indicates that trace metals, such as Ni, V, Cu, Mo, Ag, Sn, Ba, and Pb, are highly concentrated in the asphaltenic fraction. TABLE 3--Mean concentrations (in I~g/g), standard deviations (n = 3), and Total Mass Balance (TMB) of trace elements obtained after fractionation o/'15 crude oil sample. Element V Ni Cu Mo Ag Sn Ba Pb Crude oil (lg) 11.5 17.1 1.07 0.088 0.008 0.019 4- 0.003 0.029 0.004 0.45 0.97 0.075 Asphaltenes 6 % Maltenes 94 % 944-2.1 5.4 146 8.7 9.3+0.3 0.51 0.98 0.11 0.044 0.004 0.31 0.042 < LOD 0.48 4- 0.057 < LOD 4.7 0.23 15.6 0.4 0,08 0.006 TMB 100 % 93 99 98 114 98 109 111 105

Acknowledgments
Petrobr~s is acknowledged for providing crude oils. Thermo Electron is acknowledged for the loan of instrumentation.
References

[1] [2] [3] [4]

[5]

[6]

[7]

[8]

Treibs, A., "Porphyrins in Coal," Ana. Chem., Vol. 520, 1935, pp. 144-151. Baker, E. W., "Porphyrins," Organic Geochemistry, G. Eglinton and M. T. J. Murphy, Eds., 1969, pp. 464-497. Baker, E. W. and Louda, J. W., "Porphyrins Geochemistry of Atlantic Jurassic-Cretaceous Black Shales," Organic Geochemistry, Vol. 10, 1985, pp. 905-914. Eckardt, C. B., Wolf, M., and Maxwell, J. R., "Iron Porphyrins in the Permian Kupferschiefer of the Lower Rhine Basin, N.W. Germany," Organic Geochemistry, Vol. 14, 1989, pp. 659-666. Filby, R. H., "Origin and Nature of Trace Elements Species in Crude Oil, Bitumens and Kerogen: Implications for Correlation and Other Geochemical Studies," Geofluids: Origin, Migration and Evolution of Fluids in Sedimentary Basins, Geol. Soc. Sp. Pub., J. Parnell, Ed., Vol. 78, 1994, pp. 203-221. Duyck, C., Miekeley, N., Porto da Silveira, C. L., and Szatmari, P., "Trace Element Determination in Crude Oil and Its Fractions by Inductively Coupled Plasma Mass Spectrometry Using Ultrasonic Nebulization of Toluene Solutions," Spectrochimica Acta Part B: Atomic Spectroscopy, Vol. 57, 2002, pp. 1979-1990 A1 Swaidan, H. M., "The Determination of Lead, Nickel and Vanadium in Saudi Arabian Crude Oil by Sequential Injection Analysis/Inductively-Coupled Plasma Mass Spectrometry," Talenta, Vol. 43, 1996, pp. 1313-19. Hausler, D., "Trace Element Analysis of Organic Solutions Using Inductively Coupled Plasma-Mass Spectrometry," Spectrochimica Acta Part B, Vol. 42, 1987, pp. 63-73.

58

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

[9] Kahen, K., Strubinger, A., Chirinos, J. R., and Montaser, A., "Direct Injection High Efficiency Nebulizer-Inductively Coupled Plasma Mass Spectrometry for Analysis of Petroleum Samples," Spectrochimica Acta Part B, Vol. 58, 2002, pp. 397-413. [10] Botto, R. I., "Matrix Interferences in the Analysis of Organic Solutions by Inductively Coupled Plasma-Atomic Emission Spectrometry," Spectrochimica Acta, Vol. 42B, 1987, pp. 181-189. [11] Botto, R. I., "Method for Correcting for Acid and Salt Matrix Interferences in ICP-AES," Spectrochimica Acta, Vol. 40B, 1985, pp. 397-412. [12] Wondimu, T., Goessler, W., and Irgolic, K. J., "Microwave Digestion of "Residual Fuel Oil" (NIST SRM 1634b) for the Determination of Trace Elements by Inductively Coupled Plasma-Mass Spectrometry," Frenesius, Vol. 367, 2000, pp. 35-42. [13] Botto, R. I. and Zhu, J. J., "Use of an Ultrasonic Nebulizer with Membrane Desolvation for Analysis of Volatile Solvents by Inductively Coupled Plasma Atomic Emission Spectometry," Journal of Analytical Atomic Spectrometry, Vol. 9, 1994, pp. 905-912. [14] McElroy, F., Mennito, A., Debrah, E., and Thomas, R., "Uses and Application of Inductively Coupled Plasma Mass Spectrometry in the Petrochemiehal Industry," Spectroscopy, Vol. 13, No. 2, 1998, pp. 43-52.

Journal of ASTM International, July/August 2005, Vol. 2, No. 7

Paper ID JA112976 Available online at www.astm.org Paolo TittarellL 1Marco Priola, 1 Silvia Ricchiuto, ~Dmitri A. Katskov, 2 and Prince Ngobeni 2

Fuel Analysis by Filter Furnace Electrothermal Atomic Absorption Spectrometry


ABSTRACT: The determination of Mn, Fe, Co, and Zn was performed in gasoline, jet, and diesel fuel

samples by electrothermal atomic absorption spectrometry using the Transverse Heated Filter Atomizer (THFA). Thermal conditions were experimentally defined for the investigated elements. The elements were analyzed without the addition of chemical modifiers, using organometallic standards for the calibration. Gasoline samples were analyzed directly, while jet and diesel fuel samples were diluted 1+3 with n-beptane. The following LODs were obtained using 40 ~tL injections: 0.08 ~tg/LMn, 0.35 lag/LFe, 0.54 ~tg/LCo, and 0.06 I.tg/LZn. The limits of determination, calculated for diesel and jet fuel samples, were 1.2 ~g/kg Mn, 4.9 ~tg/kgFe, 7.6/ag/kg Co, and 0.8 I~g/kgZn. The corresponding limits of quantitation for gasoline samples were about four times higher.
KEYWORDS: electrothermal atomic absorption, transverse heated filter atomizer, gasoline, diesel fuel,

trace elements
Introduction

Electrothermal atomic absorption represents a suitable instrumental technique for the analysis o f elements in petroleum products [1-5]. The technique shows very low detection limits, similar to or even better than those found for other spectroscopic techniques, such as Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES) [6] and ICP-Mass Spectrometry (ICP-MS) [7-12]. Some problems were evidenced in the use o f ETAAS when elements like Ni [3] and Pb [13] are analyzed, due to the different behavior o f organo-metallic species during the thermal treatment o f standards and samples. Recently, the use o f transverse heated filter furnace for the analysis o f trace elements in automotive fuels was proposed [14]. This instrumental technique showed interesting results for volatile elements, such as Cd and Pb. The filter furnace approach was the subject o f several papers and o f continuous developments in order to improve the design [15], to extend the range to low volatile elements [16], to employ the furnace in transverse heated atomizers [17], and to determine trace elements in aqueous and organic solutions [ 18,19]. The most interesting aspects o f the filter furnace concept are: the capability to inject large volumes due to the large volume o f the zone allocated to the sampling, the limited non-specific absorption b y molecular species due to the filtering action o f the furnace, and no need for chemical modifiers due to the furnace design.

Manuscript received 6 October 2004; accepted for publication 29 December 2004; published July 2005. Presented at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. i Stazione Sperimentaleper i Combustibili, San Donato Milanese MI, 20097, Italy. 2 Tshwane University of Technology, Pretoria, 0001, South Africa.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

59

60

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Furthermore, in the case of petroleum products, the high surface area of the substrate in the furnace can be wet easily and homogeneously by the organic medium. In particular, the injection of large volumes, up to 100 IxL, can be loaded onto the fiber of the filter furnace and can be carried out without any particular problems, while the loading of organic solution samples greater than 15-20 ~tL onto conventional THGA platforms can lead to analytical problems, due to the spread of the liquid on the graphite surface [20]. The aim of this investigation was to extend the range of analyzed elements, to examine petroleum products other than gasoline and diesel fuel (namdy jet fuel and fuel oil), to verify the stability of some liquid fuels as far as the determination of trace dements is concerned, and to validate the THFA procedure using standard reference materials and the comparison with an independent technique (microwave digestion followed by ICP-MS measurements). The behavior of manganese, iron, cobalt, and zinc was investigated. Copper was also examined to compare the actual figures of merit with those obtained in a previous investigation on automotive fuels using THFA [ 14].

Experimental
Instrumentation
A Perkin Elmer SIMAA 6000 atomic absorption spectrometer, equipped with Perkin Elmer hollow cathode lamps (Co, Cu, Fe, Mn, and Zn), was employed. The instrument was set up according to the manufacturer's instructions [21] and run in the single element mode. Filter furnaces were assembled according to the procedure outlined in Ref. [17]. The design of the THFA is reported in Fig. 1. In particular, the platform was removed from the standard Transverse Heated Graphite Atomizer (THGA) tube, and a filter made of porous graphite in the form of a spool was inserted in the tube. The filter was made from a 6 mm rod of AGKSP (Union Carbide) graphite with 28 % porosity. The total length of the filter was 17 ram, and the collector length was 6 mm. The collector was filled with carbon fiber (Levertex G-868), by wrapping the fiber around the filter spool. The average fiber loading was about 15 mg. Before use, the fiber was treated with a stream of chlorine to remove trace contaminants.

FIG. 1--Transverse heated filter furnace: a) tube, b) filter, c) collector. The sample, injected through the dosing hole, distributes along the surface of the fiber that is much wider than that of a conventional platform. In particular, petroleum products homogeneously distribute onto the fiber surface, due to the wetting capability of the organic medium. During the heating of the external tube, the sample is forced to migrate through the porous graphite of the filter. The filtering action of the porous graphite strongly reduces the formation of molecular species and non-specific absorption during the atomization step. While in the case of THGA sampling vaporization and atomization occur in the same tube zone, in the

TITTARELLI ET AL. ON FUEL ANALYSIS

61

case of THFA sampling, vaporization and atomization occur in three distinct locations of the tube, i.e., collector, filter, and analytical zone. A Perkin-Elmer SCIEX ELAN 9000 ICP-MS was employed for element determination after microwave digestion of fuel samples, performed using an Anton Paar Multiwave 3000 microwave digestion system.

Reagents
The multielement standard in oil Conostan $21 (Conoco, Ponca City, OK) was employed for calibration and recovery studies. N-heptane was used as the solvent (Merck, Darmstadt, Germany). The standard 10 mg/kg content of each element was diluted with n-heptane to obtain the required concentration of element. The microwave digestion was performed using nitric acid Tracepure and hydrochloric acid Suprapure (Merck, Darmstadt, Germany) and hydrogen peroxide (J. T. Baker, Deventer, Holland). NIST 1634c Trace elements in fuel oil were employed to verify the accuracy of THFAETAAS and ICP-MS measurements.

Procedure ETAAS Measurements--The following wavelengths were employed: Mn 279.5 nm, Fe 248.3 nm, Co 242.5 nm, Zn 213.9 nm, and Cu 324.8 nm. Gasoline samples were examined directly, without dilution with solvent. Diesel fuel and jet fuel samples were diluted with n-heptane 1+3 by mass. The thermal program employed is reported in Table 1. The vaporization temperature (step 2) was set for all types of petroleum products, while pyrolysis and atomization temperatures were chosen according to the thermal properties of element under investigation. Forty ~tL of sample were usually injected, although THFA furnace shows excellent results in vaporizing volumes up to 100 ~tL [17,18]. However, the sample volume was limited to 40 IxL, to achieve a suitable compromise between improvement in detection limit and steadiness in analytical results.
TABLE 1--Instrumental conditions for THFA-ETAAS. Step Temperature, 0C ..... Ramp, s Holdl s
1 110 5 1

2 400 20 20 3 * 5 5 4 * 0 * 5 2450 1 5 *Temperatureand hold time set accordingto the element. Gas stop during atomization,40 #L injection. The determination of all elements was performed without the addition of chemical modifiers. Three replicate readings were performed for each determination. Determinations were also carried out in consecutive days on jet and diesel fuel samples to evaluate the stability of samples spiked with known amounts of trace elements.

ICP-MS measurements--Microwave digestion followed by ICP-MS was performed on SRM 1634c samples. About 0.2 g were digested in microwave oven using 6 mL nitric acid, 1 mL

62

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

hydrochloric acid, and 2 mL hydrogen peroxide, following the procedure outlined in Ref. [22]. The power program is reported in Table 2. The acidic solution was diluted to 25 mL with deionized water. Further dilution was required according to the elements under investigation. The instrumental conditions of ICP-MS measurements are reported in Table 3. Rhodium was used as the internal standard. TABLE 2--Powerprogramformicrowave digestion. Step Power, W Time, min 1 0-1100 30 2 1100 10 3* 0 20 * Systemcoolingwith air flow. TABLE 3--Instrumenta ! conditions for ICP-MS measurements. RF Power 1200 W Argon flow rates plasma 15 L/min nebulizer 0.5 L/min Scan mode peak hopping Sweeps/reading 20 Readings/replicate 1 Replicates 5 Dwell time 50 ms
Results and Discussion

The main difference between gasoline and diesel fuel, as far as ETAAS are concerned, is the sample volatility. Gasoline sample can be injected directly in the electrothermal atomizer and vaporized without formation of any residue, while diesel fuel undergoes partial pyrolysis during the thermal treatment before the atomization occurs. The pyrolysis can lead to the formation of a carbonaceous residue. The carbonaceous buildup affects the atomization of carbide forming elements such as Ni and V [3]. In this work, diesel fuel samples were diluted with a suitable solvent to reduce the formation of carbon residues.

Manganese
Pyrolysis and atomization curves of manganese are shown in Fig. 2. The pyrolysis curve was obtained using 2300~ as the atomization temperature. The selected pyrolysis temperature for Mn in THFA is about 1600~ Using this pyrolysis temperature, the optimum atomization temperature corresponds to 2100~ These pyrolysis and atomization temperatures do not correspond to the THGA suggested temperatures [21], as the furnace design is different and the rate of the vaporization process is also different. Some loss on manganese is observed above 2100~ due to the volatility of the atomic species at high temperature. A typical atomization peak of manganese is shown in Fig. 3. In this case, due to the very low Mn content in diesel fuel, manganese is added to the diesel, so that 0.2 ng Mn are injected together with the sample aliquot. The peak shape shows some tailing till the end of the atomization step. The background is very limited and almost continuous during the step.

TITTARELLI ET AL. ON FUEL ANALYSIS

63

0.~.$ -

0.L5

0.10

0.85

0.00 1300 17OO 2100 ('C) 2500

T~l~m~

FIG. 2--Pyrolysis (~) and atomization (/X) curves of manganese.

0.08

,% / \

--

atomicabsorption

<i

~0.04

0 0 2 4 6 8 10
Time (s) FIG. 3--Manganese atomization peak. Diesel#el spiked with 0.2 ng Mn. Pyrolysis 1600~ atomization 2100~ The figures of merit of Mn are shown in Table 4. The limit of detection (LOD) is calculated as three times the standard deviation of the average of six blank runs. The limit of quantitation (LOQ) is calculated as three times LOD. The data reported in Table 4 are referred to diesel samples that are diluted 1+3 with n-heptane. The corresponding LOQ for gasoline samples are four times lower, as gasoline is injected without preliminary dilution. TABLE 4---Limits of detection and quantitation for diesel fuel. LOD, #g/L a LOQ,/zgJkg o Mn 0.08 1.2 Fe 0.35 4.9 Co 0.54 7.6 Zn 0.06 0.8 a 3a of n-heptanr (n=6); b 3-LOD.

64

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Iron
Pyrolysis and atomization behavior of iron are reported in Fig. 4. The optimum conditions employed for further measurements are 1600~ and 2300~ for pyrolysis and atomization steps, respectively. Figure 5 shows a tailing of the atomization peak more pronounced than that observed for manganese. Some memory effect can be expected if a cleaning step is not performed at the end of the vaporization cycle. Background absorption is very limited.
o~.

3 ~
0.20"

~ o.t~-

.e! ~ O.lO. ~ o.115.


0.~ 1300

1700 2100 T e ~ ' a l m ~ ('C)

2500

FIG. 4--Pyrolysis ( I ) and atomization (lk) curves of iron.

FIG. 5--Iron atomization peak. Diesel fuel spiked with 0.6 ng Fe. Pyrolysis 1600~ atomization 2300~ As in the case of manganese, 0.6 ng iron are added to the diesel fuel, due to the very low content in most samples. As already reported for some elements, the content of trace elements in modem automotive fuels is very low, often below the quantitation limit of THFA-ETAAS [3]. The limits of detection and of quantitation for iron in diesel fuel are reported in Table 4.

TITTARELLI ET AL. ON FUEL ANALYSIS

65

Cobalt
The thermal properties of cobalt in THFA are shown in Fig. 6. The pyrolysis temperature can be set to 1600~ However, the continuous increase of the atomization curve indicates that cobalt atomization is not quantitative even at the maximum temperature allowable to the THFA tube (2500~ The atomization profiles (Fig. 7) evidence a remarkable tailing of peaks. A behavior similar to that of cobalt was observed for nickel [3]. Both elements in fact vaporize at high temperature. The continuous increase of the atomization curve and the contemporaneous tailing can be attributed, as for nickel, to the formation of element carbides in the graphite tube. This situation was evidenced in the analysis of vacuum gas oils in a conventional end-heated atomizer [3].
0.25.

0.'~0 -

0as.
~ 0.I0,

0.05 -

0.00 1000

6 1500

'2000

:aSl)O

? m p ~ a ~ (*r

FIG. 6--Pyrolysis (ll) and atomization (Z~) curves of cobalt.

FIG. 7--Cobalt atomization peak. Diesel fuel spiked with 1.6 ng Co. Pyrolysis 1600~ atomization 2400~ In the case of cobalt, 1,6 ng cobalt are added to the diesel sample, as the element shows a lower sensitivity in comparison to manganese and iron. Limit of detection and quantitation of cobalt are affected by the low sensitivity of THFA for low volatile elements (Table 4).

66

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

~nc

Zinc is extremely sensitive to temperature changes of pyrolysis and atomization steps as indicated in Fig. 8. The highest pyrolysis temperature without element loss can be set to about 700~ The atomization curve shows a maximum at about 1800-1900~ The sensitivity to thermal conditions is reflected in the atomization profiles that exhibit uneven behavior (Fig. 9). As for all elements investigated, background is very limited. Due to the high volatility of zinc, LOD and LOQ (Table 4) are very low and comparable with those obtained with conventional THGA tubes [21 ].
0.2

0,15

,.~ 0,1

~ O.O5

~0

1000

1500

2000

2500

Temloemtu-re (*C)
FIG. 8--Pyrolysis (R) and atomization (/x) curves of zinc.

FIG. 9--Zinc atomization peak. Diesel fuel spiked with 0.1 ng Zn. Pyrolysis 700~ atomization 1900~

Performance of THFA
The limits of detection obtained with THFA-ETAAS are listed in Table 5 with those obtained recently using various instrumental techniques.

TITTARELLI ET AL. ON FUEL ANALYSIS

67

TABLE 5--Detection limits of analytical techniques. ETV-ICP-MS USN-ICP-MS DIHEN-ICP-MS ETAAS Gasoline [9] SRM 1634c [ 1 1 1 xylene [ 12] gasoline [4] Mn 0.02 0.6 Cu 0.22 0.1 0.4 Pb 0.2 1.46 1.8 Fe 0.5 3.0 Ni 0.38 0.4 0.8 0.8 USN: UltrasonicNebulizer; DIHEN:DirectInjectionHigh EfficiencyNebulizer.

THFA heptane 0.08 0.27 [14] 0.37 [14] 0.35 0.6 [14]

In general, THFA-ETAAS compares well with conventional ETAAS and various ICP-MS approaches. Detection limits of copper and nickel were obtained in a previous work focused on automotive fuels investigation.

THFA Tubes Lifetime


The lifetime of a THFA tube was evaluated by consecutive 40 #L injections of a 40 #g/L copper solution in n-heptane, using instrumental conditions already defined in Ref. [14]. Figure 10 shows a slight increase in peak area absorbance, possibly due to some memory effect. The lifetime can be considered to last around 900-1000 firings. The visual inspection of THFA components shows limited corrosion of the ends of the filter. However, the used fiber can be removed and replaced by a new one.
0.35

0 . 3 0

~0.25 lad

0.20
200
400 600 Fi r ~ no . 800 i0~0 1200

FIG. 1O---Lifetime of THFA tubes.

Stability of Fuel Solutions


Jet and diesel fuel samples were spiked with known concentrations of manganese, copper, and iron and stored in dark bottles (20 #g/L spike). The element content was checked over a period of 25 days using calibration solutions prepared daily. The results obtained for manganese are reported in Fig. 11. While manganese content in diesel fuel samples remains constant during the timeframe investigated, the element content in jet fuels drops dramatically after about two

68

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

weeks. The same behavior holds for copper and iron. Hence, the time elapsed between jet fuel sampling and analysis should be kept to a minimum to avoid element loss and erroneous results. The same behavior of jet fuel can be postulated for gasoline.
25

0 0 5 10 15 ~0 25

....

30

Tiaue(da~) FIG. 11--Stability of manganese in jet fuel ( A ) and dieselfuels (~ , Vq).

Comparison with ICP-MS Measurements


Fuel oil SRM 1634c was examined to compare the accuracy of the results obtained by THFA-ETAAS with those obtained by ICP-MS after microwave digestion (Table 6). The average of three independent determinations is shown in the Table. Manganese and copper contents obtained by THFA are higher than those obtained by ICPMS. Vanadium and nickel were not examined by TI-IFA due to the high atomization temperature required and the memory effects associated with the incomplete atomization of these elements. TABLE 6--SRM 1634c. Comparison between THFA-ETAAS and ICP-MS. Certified, mg/kg THFA ICP-MS Mn 0.35 + 0.03 0.27 0.02 Cu 0.23 0.02 0.13 0.04 Pb 0.10 0.02 0.12 0.03 Co 0.1510 0.0051 0.16 0.01 0.12 0.02 V 28.19 0.40 28.0 1.0 Ni 17.54 0.21 16.3 0.4

Conclusions
The THFA approach is suitable for liquid fuel analysis. In particular, the injection of large sample volumes represents an advantage for ultra-trace analysis. The handling of large sample volumes still has to be exploited in its full capability to enhance the analytical performance of the technique. However, problems are encountered when stable elements, such as V and Ni, are analyzed. The validation of instrumental procedures for trace element analysis at /zg/kg levels is hindered by the lack of automotive/aviation fuel SRMs. Furthermore, light fuels (gasoline/jet

TITTARELLI ET AL. ON FUEL ANALYSIS

69

fuel) are prone to element loss. Nevertheless, the stability of diesel fuels is quite promising for possible SRMs development.
References

[1] ASTM Standard D 6732-02, "Standard Test Method for Determination of Copper in Jet Fuels by Graphite Furnace Atomic Absorption Spectrometry," ASTM International, West Conshohocken, PA. [2] Silva, I. A., Campos, R. C., and Curtius, A. J., "Determination of Lead and Copper in Kerosene by Electrothermal Atomic Absorption Spectrometry: Stabilization of Metals in Organic Media by a Three-Component Solution," Journal of Analytical Atomic Spectrometry, Vol. 8, 1993, pp. 749-754. [3] Anselmi, A. and Tittarelli, P., "Determination of Nickel in Petroleum Heavy Ends by Electrothermal Atomic Absorption Spectrometry," La Rivista dei Combustibili, Vol. 48, 1994, pp. 435-441 (in Italian). [4] Campos, R. C., Santos, H. R., and Grinberg, P., "Determination of Copper, Iron, Lead, Nickel in Gasoline by Electrothermal Atomic Absorption Spectrometry Using ThreeComponent Solutions," Spectrochimica Acta Part B, Vol. 57, 2002, pp. 15-28. [5] Aucelio, R. Q. and Curtius, A. J., "Evaluation of Electrothermal Atomic Absorption Spectrometry for Trace Determination of Sb, As and Se in Gasoline and Kerosene Using Microemulsion Sample Introduction and Two Approaches for Chemical Modification," Journal of Analytical Atomic Spectrometry, Vol. 17, 2002, pp. 242-247. [6] Botto, R. I., "Applications of Ultrasonic Nebulization in the Analysis of Petroleum and Petrochemicals by Inductively Coupled Plasma Atomic Emission Spectrometry," Journal of AnalyticalAtomic Spectrometry, Vol. 8, 1993, pp. 51-57. [7] van Heuzen, A. A., "ICP-MS in the Petroleum Industry," Application of Inductively Plasma Mass Spectrometry, A. R. Date and A. L. Gray, Eds., Blackie, London, 1989, Chapter 7. [8] A1-Swaidan, H. M., "Trace Determination of Vanadium and Nickel in Saudi Arabian Petroleum and Petroleum Products by Microemulsion ICP-MS," Atomic Spectroscopy, Vol. 14, 1993, pp. 170-173. [9] Kumar, S. J. and Gangadharan, S., "Determination of Trace Elements in Naphtha by Inductively Coupled Plasma Mass Spectrometry Using Water-in-Oil Emulsions," Journal of Analytical Atomic Spectrometry, Vol. 14, 1999, pp. 967-971. [10] Saint Pierre, T. D., Dias, L. F., Pozebon, D., Aucelio, R. Q., Curtius, A. J., and Welz, B., "Determination of Cu, Mn, Ni and Sn in Gasoline by Electrothermal Vaporization Inductively Coupled Plasma Mass Spectrometry, and Emulsion Sample Introduction," Spectrochimica Acta B, Vol. 57, 2002, pp. 1991-2001. [11] Duyck, C., Miekeley, N., Porto da Silveira, C. L., and Szatmari, P., "Trace Elements Determination in Crude Oil and Its Fraction by Inductively Coupled Plasma Mass Spectrometry Using Ultrasonic Nebulization of Toluene Solution," Spectrochimica Acta PartB, Vol. 57, 2002, pp. 1979-1990. [12] Kahen, K., Strubinger, A., Chirinos, J. R., and Montaser, A., "Direct Injection High Efficiency Nebulizer-Inductively Coupled Plasma Mass Spectrometry for Analysis of Petroleum Samples," Spectrochimica Acta Part B, Vol. 58, No. 200, pp. 397-413. [13] Kowalewska, Z., Bulska, E., and Hulanicki, A., "Organic Palladium and PalladiumMagnesium Chemical Modifiers in Direct Determination of Lead in Fractions from

70

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

[14]

[15] [16] [17] [18] [19]

[20] [21] [22]

Distillation of Crude Oil by Electrothermal Atomic Absorption Analysis," Spectrochimica Acta Part B, Vot. 54, 1999, pp. 835-843. Anselmi, A., Tittarelli, P., and Katskov, D. A., "Determination of Trace Elements in Automomotive Fuels by Filter Furnace Atomic Absortion Spectrometry," Spectrochimica Acta Part B, Vol. 57, 2002, pp. 403-411. Katskov, D. A., Marais, P. J. J. G., and Tittarelli, P., "Design, Operation and Analytical Characteristics of the Filter Furnace, a New Atomizer for Electrothermal Atomic Absorption Spectrometry," Spectrochimica Acta Part B, Vol. 51, 1996, pp. 1169-1189. Katskov, D. A., Marais, P. J. J. G., Katkovnik, V. J., and Tittarelli, P., "Adaptation of the Filter Furnace Atomizer for Atomic Absorption Determination of Less Volatile Metals," Spectrochimica Acta Part B, Vol. 52, 1997, pp. 1377-1394. Katskov, D. A., Marais, P. J. J. G., and Ngobeni, P., "Transverse Heated Filter Atomizer for Electrothermal Atomic Absorption Spectrometry," Spectrochimica Acta Part B, Vol. 53, 1998, pp. 671-682. Marais, P. J. J. G., Panichev, N. A., and Katskov, D. A., "Performance of the Transverse Heated Filter Atomizer for the Atomic Absorption Determination of Mercury," Journal of Analytical Atomic Spectrometry, Vol. 15, 2000, pp. 1595-1598. Mbileni, C., Ngobeni, P., Katskov, D. A., and Panichev, N., "Determination of Lead and Cadmium in Organic Solutions by Electrothermal Atomic Absorption Spectrometry with a Transverse Heated Filter Atomizer," Journal of Analytical Atomic Spectrometry, Vol. 17, 2002, pp. 236-241. Tserovsky, E. and Arpadjan, S., "Behaviour of Various Organic Solvents and Analytes in Electrothermal Atomic Absorption Spectrometry," Journal of Analytical Atomic Spectrometry, Vol. 6, 1991, pp. 487-491. The THGA Graphite Furnace, Techniques and Recommended Conditions, PerkinElmer, 13berlingen, 1999. Bettinelli, M., Spezia, S., Baroni, U., and Bizzarri, G., "Determination of Trace Elements in Fuel Oils by Inductively Coupled Plasma Mass Spectrometry after Acid Mineralization of the Sample in a Microwave Oven," Journal of Analytical Atomic Spectrometry, Vol. 10, 1995, pp. 555-560.

Journal of ASTM International, March 2005, Vol. 2, No. 3 Paper ID JAI12970 Available online at www.astm.org

Malte Lukas, 1 Robert J. Yurko, 2 and Daniel P. Anderson 3

Rotrode Filter Spectroscopy: A Recently Improved Method to Detect and Analyze Large Wear and Contaminant Particles in Fluids
ABSTRACT: Routine spectrochemical analysis of used oil samples with rotating disc electrode (RDE) or inductively coupled plasma (ICP) emission spectrometers has become standard practice in machine condition monitoring programs to determine elemental content. Unfortunately, both of these techniques become less responsive to particles as they increase in size. To overcome this limitation, several complementary methods and analytical techniques to detect large particles in used oil have been applied, including acid digestion, ferrography, and X-ray analysis. Unfortunately, these techniques are not practical or economical to implement in a high sample throughput machine condition-monitoring program. One technique that has demonstrated success as a complementary method for detecting large particles is rotrode filter spectroscopy (RFS). RFS extends the capability of rotating disc electrode (RDE) spectrometers to detect particles greater than 10 lam in
size,

KEYWORDS: Rotrode Filter Spectroscopy (RFS), rotating disc electrode spectrometer, RDE, used oil

analysis
Introduction

Spectrometric oil analysis has been applied for more than 40 years as a routine and costeffective condition monitoring technique [1]. It is used to determine the elemental concentration in parts per million o f wear metals, contaminants, and additives in a used oil sample. Commercial oil analysis laboratories report on as many as 20 different elements. With the knowledge o f the wear metal and contaminant limits for the machine or engine being monitored, a determination may be made as to whether or not that equipment is operating properly. A fact that has recently become more widely understood is that spectrometric oil analysis detection efficiency decreases as the wear particle size increases [2]. This partial limitation is not a problem with reciprocating engines that generate mostly small particles, but it is sometimes a problem in fatigue failures o f rolling element bearings, such as those in military gas turbines, which generate large particles at the outset o f failure without generating many small particles. In addition, some wear modes such as spalling, severe sliding wear, and cutting wear generate large particles that may go undetected b y spectroscopy [3]. This is true to a greater degree for certain types o f spectrometers than others. It is also more important for certain types o f machines and engines. Atomic absorption spectroscopy (AAS) and inductively coupled plasma (ICP) spectrometers suffer the most from particle detection inadequacies. Rotating disc electrode (RDE) spectrometers are responsive to somewhat larger particles, but the upper limit is still approximately 10 ~tm [4]. Ferrography and X-ray analyzers address this limitation and are Manuscript received 6 October 2004; accepted for publication 28 November 2004; published March 2005. Chief Technology Advisor, Spectro Inc., Littleton, MA 01460. 2 V.P. R&D and Engineering, Spectro Inc., Littleton, MA 01460. s V.P. Sales and Marketing, Spectro Inc., Littleton, MA 01460.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

71

72

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

capable of identifying large wear particles as a complementary technique. However, these instruments are not always practical or cost-effective in a used oil analysis laboratory environment that competes on the price of samples and may process hundreds or even more than 1000 samples per day. The RFS method makes use of the fact that carbon disc electrodes used in RDE spectrometers are porous and can be used as a filter to trap particles in oil samples. Early versions of the RFS fixture used standard disc electrodes with some success and acceptable sensitivity [5]. However, standard disc electrode geometry caused the process to be slow, and lack of manufacturing control of disc porosity resulted in poor repeatability. Due to its ease of use and inherent advantages, the RFS concept was recently revisited with the intent of further improving performance. The updated RFS technique uses custom ring-shaped graphite electrodes and a semi-automatic process to improve sensitivity and repeatability and to reduce sample preparation time. This paper describes the theory of operation of the RFS technique, the evolution of the concept based on experience, and how it is applied in commercial used oil analysis laboratories.

Theory of RFS
A microscopic inspection of the carbon disc electrodes used in rotating disc electrode (R.DE) spectrometers reveals that they are porous. Rotrode filter spectroscopy (RFS) makes use of this fact, and a funnel-based fixture (Fig. 1) is used to clamp the discs so that used oil samples can be drawn through the outer circumference of the disc when a vacuum is applied. The filtering process through the disc captures the particles in the oil. The oil is then washed away with solvent, and the disc is allowed to dry. The particles are left on the outer circumference of the disc electrode so that they are vaporized and detected when sparked on the RDE spectrometer.

FIG. 1--RFS disc electrode filtering oil through its circumference.

LUKAS ET AL. ON ROTRODE FILTER

73

The RFS technique is used as a comparative method due to the unavailability of oil standards with known gravimetric concentration of particles for each element measured by the spectrometer. In practice, a used oil sample is first analyzed using the standard RDE technique, which provides an analysis of dissolved and small wear particles. A second analysis of the same sample using the RFS technique detects large particles. The two analyses provide an indication of the wear particle size distribution in the sample. A sudden presence of large wear particles will not be seen by conventional analysis alone. Their presence, however, will be readily evident through the RFS analysis.

The RFS Fixture


The original RFS fixture to prepare electrodes for the analysis of large particles was first developed in the early 1990s (Fig. 2). The fixture used standard disc electrodes, which were clamped in the funnels to filter and capture large particles. It also had five stations so that five samples could be processed simultaneously. A vacuum pump was used to pull the sample and filter it through the electrode. A manual rinse with heptane was used to wash away any remaining lubricant. Sample preparation time of an electrode for RFS analysis varied with the viscosity and contamination of the oil sample. Preparation time could be as short as 4-5 min for relatively clean, used lubricating oil samples, such as from turbines, electric motor bearings, and hydraulic systems. Engine oil samples with high soot levels require the longest filtration times, sometimes half an hour or more.

FIG. 2--Early RFSfixture.

74

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

It has always been a problem that standard disc electrodes are not manufactured with consistent porosities, which affects the capture efficiency of the particles. This variability hurts repeatability and thus makes data trends, on which used oil analysis maintenance recommendations are based, more difficult to establish. Furthermore, long sample preparation times for highly sooted samples from diesel engines were not practical for some laboratories. These factors led to a new system that addressed these concerns and offered additional improvements. The updated and improved system is known as automated rotrode filter spectroscopy (A-RFS) (Fig. 3). It greatly improved repeatability through the use of controlled electrodes, automation, and a vacuum/pressure pump to reduce sample preparation times.

FIG. 3--A-RFS system. The A-RFS system is a semi-automated sample preparation instrument with a five-station fixture of funnels. It works on the same principle as the original RFS system. However, it uses a special electrode manufactured to more consistent porosity specifications, it has an automated cleaning cycle, and its filtration times are reduced by the application of a vacuum/pressure pump to both pull and push the used oi ! sample through the electrode. Conventional off-the-shelf electrodes were used as a filter in the original RFS method. Experimental data over time confirmed that the porosity of these electrodes was not controlled in the manufacturing process, and thus it had a detrimental effect on repeatability when new batches of electrodes were purchased. The new A-RFS sample preparation process begins with a specially designed RFS electrode. The RFS electrode is custom manufactured to consistent porosity specifications. Thinner wails improve repeatability and reduce filtration times. The dimensions of the standard disc electrode used for conventional RDE analysis and the new RFS electrode are shown in (Fig. 4).

LUKAS ET AL. ON ROTRODE FILTER

75

FIG. 4---Conventional electrode and RFS electrode. In routine operation, the A-RFS disc electrodes are mounted on an electrode clamp assembly for installation in one of the five preparation stations. The electrode clamp assembly is made of two separate pieces so that a new electrode can be mounted on it directly from its container, thus limiting the possibility of contamination through handling (Fig. 5a.). The clamps are numbered so they can be easily identified and associated with their matching oil sample. The electrode clamp assembly with a fresh carbon disc is then installed in one of the RFS sample funnels. An oil sample is poured into the funnel, and the funnel is clamped shut (Fig. 5b). With the start of the automated process, a vacuum/pressure pump creates a pressure drop of approximately 5 arm across the disc electrode, causing the sample to flow through the disc electrode. The filtration process captures and concentrates large wear particles on the surface of the electrode. A sensor determines when the sample has been filtered through the electrode and automatically starts the electrode cleaning cycle followed by a drying process. Heptane is used as a solvent to wash the remaining oil from the RFS electrode and to re-suspend any particles remaining in the funnel so that they can be filtered and captured. The RFS preparation process is complete when all of the oil has passed through the disc electrode, residual oil has been washed away, and the electrode is dried. The operator then removes the electrode clamp assembly and installs the shaft complete with the electrode in the spectrometer for analysis (Fig. 5c).

FIG. 5--Preparation process of the A-RFS electrode.

76

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

RFS Data Interpretation


Unfortunately, calibration standards prepared with actual wear metal particles at various sizes and concentrations are not available. Also, even if they were available, shelf-life, particle settling, and the actual distribution of particle sizes could be problematic. Thus, in order to provide the basis for benchmarking RFS calibration and to assure that data are consistent, spectrometers are calibrated with certified commercial or military metallo-organic calibration standards. This insures the spectrometer has the same response over time and permits trending of wear metals and contaminant particles. It is important to note that RFS data is therefore not in true parts per million (ppm), but in scalar thaits that the spectrometer can reproduce over time. RFS is a powerful analytical tool because it provides additional information on actual wear and contaminant particles in an oil sample. In routine operation, laboratories perform two analyses on a used oil sample. The first is conventional oil analysis that provides quantitative and qualitative analysis of dissolved and small particles in the sample. The second is RFS analysis that provides an indication of large particles and their elemental composition. Taken together, trends of the two analyses can be used as a trend to provide a much clearer picture as to the mechanical health of oil wetted systems. A typical data trend from an aircraft engine is shown in Tables 1 and 2. The first set of analyses (Table 1) is with conventional oil analysis for dissolved and small particles in the oil samples over a period of about 2 months. Conventional oil analysis shows no trends or significant concentrations of wear metals and contaminants. The second set of data (Table 2) is for the same samples but with RFS analysis for large wear particles. Note that the concentrations of several elements such as iron, magnesium, sodium, and silicon are significantly higher than conventional analysis. In addition, other wear metals such as aluminum, chromium, nickel, lead, and zinc are now detected. The RFS analysis also identifies a major trend change at 299 h denoted by a big jump in the iron, magnesium, and silicon concentrations. This signifies that there is a big increase in large particles. Conventional analytical data (Table 1) showed no trend or major concentrations for any of the elements. The aircraft was taken out of service at 312 h, the engine was inspected, and a beating was replaced. Continued service could have resulted in complete failure. TABLE 1--Aircraft engine oil analysis trend, conventional spectroscopy analysis. Date 12/10/03 12/12/03 12/15/03 12/16/03 12/18/03 01/05/04 01/08/04 01/15/04 H 262 268 271 278 293 299 307 312 Fe 0 1 0 1 1 1 0 0 Ag 0 0 0 0 0 0 0 0 A1 0 0 0 0 0 0 0 0 Cr Cu 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Mg 0 0 0 0 0 0 0 0 Na 0 0 0 0 0 0 0 0 Ni Pb 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Si 0 0 0 0 0 0 0 0 Sn Ti 4 1 4 1 5 1 4 1 4 0 3 1 3 0 3 0 B 0 0 0 0 0 0 0 0 Mo 0 0 0 0 0 0 0 0 Zn 0 0 0 0 0 0 0 0 V 0 0 0 0 0 0 0 0

LUKAS ET AL. ON ROTRODE FILTER

77

TABLE 2--Aircraft engine oil analysis trend, RFS analysis. Date 12/10/03 12/12/03 12/15/03 12/16/03 12/18/03 01/05/04 01/08/04 01/15/04 H 262 268 271 278 293 299 307 312 Fe 24 15 10 18 14 52 58 106 Ag 0 1 0 0 0 0 0 1 A1 2 6 3 6 0 3 7 4 Cr 1 1 1 1 1 3 3 8 Cu 1 1 0 2 0 1 3 2 Mg 18 9 10 7 4 30 47 19 Na 14 9 8 3 1 18 30 14 Ni Pb Si 2 2 43 1 2 19 0 1 22 1 2 18 1 2 12 4 0 80 3 1 113 9 3 45 Sn Ti 0 2 0 1 0 1 0 1 0 0 0 3 0 4 0 2 B 1 1 0 0 2 1 2 2 Mo Zn 1 7 1 2 0 8 1 6 1 1 0 8 1 11 1 12 V 0 0 0 0 0 0 0 1

It is common knowledge that an analysis is only as good as the oil sample. This is even more important with RFS analysis since RFS analyzes large wear and contaminant particles. It is thus important to always take the oil sample from the same place and in the same manner to improve analytical performance. Variations in RFS analysis also will be observed due to particle settling in the sump if samples are taken at varying times after a system is shut down. RFS analyses, as shown by this example, are capable of identifying and quantifying large particles that would normally be missed by conventional spectrometric oil analysis.

A-RFS Field Experience


RFS has shown itself to be a valuable additional tool to military oil analysis programs. However, it started in industrial applications and continues to provide rapid and valuable information for in-house and commercial laboratories. Experimental data from military application and a commercial case history will be reviewed in this section.

Experimental Data with Metal Powders


In order to verify that the RFS technique can detect particles greater than 10 gm, the U.S. Department of Defense Joint Oil Analysis Program Technical Support Center (JOAP-TSC) prepared samples using M50 alloy particles. M50 alloy is often present in aero-derivative turbines. Some failure modes of J52 engines appear to be sudden catastrophic failures where the root cause began with the lack of lubrication of the 489 beating area. Initially, analysis of the samples by RDE spectroscopy did not indicate the failure mode [6]. The composition and specifications for M50 alloy are as shown below (Table 3).

TABLE 3--Composition of M50 alloy." Fe Cr 89.8 3.75-4.5 in % concentration. Ni 0.3 max Si 0.2-0.6 Mo 3.9-4.75 V 0.8-1.25

78

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

The M50 alloy was sieved into known particle ranges and mixed at several concentrations in MIL-L-23699 oil in groupings of 25-38 ~tm and 45-63 Ixm. The samples were analyzed using conventional spectroscopy and also RFS [7]. Conventional atomic emission spectroscopy (Table 4) did not detect any of the M50 alloy mixed into the new MIL-L-23699 oil. This was to be expected and confirms the inability to detect large particles in an oil sample. The same samples also were analyzed with the new A-RFS technique (Table 5). The data verify the capabilities of RFS in detecting large particles and the presence of the M50 alloy, a frequently used alloy for roller beatings in aviation gas turbines. The response for particles in the 25-38 ~tm range is good. The data analyst can readily detect the presence of large particles, and the alloy can also be identified by its metallurgy. As the particle size increases to the 45-63 ~tm range, the excitation potential of the spectrometer is insufficient, and only a portion of the particles is vaporized. TABLE 4---Conventional spectrometric analysis of M50 alloy particles in MIL-L-23699 oil. M50 Particle Size / Vanadium Concentration 25-38 Ixm / 3 ppm 25-38 ~tm / 12 ppm 25-38 ~tm / 20 ppm 45-63 Ixm / 3 ppm 45-63 ~tm / 7 ppm 45-63 ~tm / 15 ppm Fe 0.0 0.9 1.5 0.0 0.1 0.0 Cr 0.0 0.1 0.1 0.1 0.1 0.0 Ni 0.0 0.0 0.0 0.0 0.0 0.0 Si 0.0 0.0 0.0 0.2 0.0 0.1 Mo 0.0 0.0 0.0 0.0 0.0 0.0 V 0.0 0.0 0.0 0.0 0.0 0.0

TABLE 5--RFS analysis of M50 alloy particles in MIL-L-23699 oil. M50 Particle Size / Vanadium Concentration 25-38 ~tm / 3 ppm 25-38 lam / 12 ppm 25-38 ~tm / 20 ppm 45-63 ktm / 3 ppm 45-63 ~tm / 7 ppm 45-63 ~tm / 15 ppm Fe 463 1454 3083 26.9 21.1 25.8 Cr 17.5 59.6 157 0.9 0.6 0.9 Ni 25.1 86.6 219 1.2 0.8 1.0 Si 15.0 17.5 42.9 1.4 2.0 2.1 Mo 21.2 81.6 227 0.9 0.7 0.4 V 3.2 12.1 37.1 0.0 0.1 0.3

Microscopic inspection of the RFS electrode as well as an automatic particle counter and shape classifier have been used to verify that the effective RFS capture range of particles from a used oil sample is from 2 ~tm to more than 100 ~tm. However, the upper particle size limit that can be completely volatilized by the spectrometer's arc is closer to 50 ~tm. This does not preclude the RFS analysis from detecting larger particles, because even partial volatilization of a particle above 50 ~tm will add signal and concentration to the analysis. The purpose of RFS analysis is to detect and trend the presence of large particles in a used oil sample that are missed by routine spectrometric analysis, and not to quantify the exact size or concentration of these

LUKAS ET AL. ON ROTRODE FILTER

79

particles. The mere presence of large particles is most significant, and the RFS analysis directs the data analyst to additional analytical tests. It also may be observed that the RFS technique is very sensitive. Just a few ppm of particles in the oil give a big response by the spectrometer. This is due in part to the RFS filtration process concentrating particles on the disc surface from a relatively large oil volume for subsequent vaporization and detection.

Commercial Case History


The following is a typical case history [8] from a Northeastern power company. The item being sampled is a condensate vacuum pump. In this case, the commercial laboratory provided oil analysis services for lubricant physical properties and metals including RFS for larger particles. Although the laboratory provides data for 20 wear metals, contaminants, and additives, a trend was observed only for iron and silicon (Fig. 6).

FIG. 6---Spectrometric analysis trends for normal (fine) and RFS (coarse) analyses. Normal spectrometric analysis, "fine," did not show significant wear. In most industrial systems such as this pump, concentrations of wear metals will fluctuate in accordance with oil added and can vary by as much as 10 ppm. The wear as indicated by normal spectrometric analysis therefore did not exceed expected limits, and did not result in any maintenance recommendations. On the other hand, the RFS analyses as indicated by "Coarse" in (later in Fig. 8) showed a clear trend resulting in maintenance recommendations. The silicon readings by themselves were not cause for alarm, but they were when combined with RFS data for iron. The RFS trend for iron jumped enough for' the laboratory to issue a severe alarm after the January 9, 2003 sample. The maintenance response was to "feed and bleed," i.e., drain some lubricant and top it off. The next two RFS analyses showed a corresponding reduction in the iron analyses. However, the May 22, 2003 sample indicated that

80

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

there was still a problem, and a severe alarm was again generated. This time the maintenance personnel opened up the pump and found that a bearing had a loose fit and was fretting on the bearing housing, bearing, and shaft (Fig. 7). The RFS data also prompted the laboratory to prepare a ferrogram of this sample. Ferrographic analysis verified that large particles were present and that cutting wear (Fig. 8) was taking place.

FIG. 7--Bearing housing, outer, and inner race fretting due to a loose fit.

FIG. 8--Ferrogram showing cutting wear.

LUKAS ET AL. ON ROTRODE FILTER

81

In this example, normal spectrometric analysis did not detect the large wear particles emanating from the bearing-fretting problem. The RFS analysis, however, did provide early waming to repair, clean up, and solve the problem before it could lead to a major failure with possible secondary damage. The next three samples show a return to a normal trend after the lubricant was cleaned and replaced.
Conclusion

RFS is an important enhancement to standard RDE analysis. It provides new information on the condition of the machine being monitored by expanding the particle detection capability of the RDE technique. Sometimes it is also used to screen samples to determine if more time consuming analysis, such as ferrography, is required. Rotrode Filter Spectroscopy (RFS) is a recently improved technique that has been shown in actual field applications to provide important additional information about large wear particles, information that may be missed with conventional techniques. It is applied with existing instruments to provide two analyses on the same used oil sample. The first, using the standard RDE atomic emission technique, detects and quantifies dissolved and small wear metals and contaminants. The second analysis, by RFS, qualifies and semi-quantifies larger particles. This paper has shown how the early version of the RFS system has been improved and updated to provide better sensitivity, repeatability, faster sample throughput, and improved particle capture capability. RFS analysis is not used by itself, but as a second test that expands the particle detection capability of spectroscopy to provide a comprehensive tool to direct the laboratory to identify and solve a maintenance problem. Frequently, the RFS analysis is also used as a screening tool to determine if additional tests, which are too time-consuming or expensive to perform on all samples, should be performed. Supplementary techniques such as analytical ferrography or particle shape characterization and particle counting are then applied to verify a severe wear condition.
References

[1] Sawyer, J. W., "Sawyer's Turbomachinery Maintenance Handbook," Turbomachinery International Publications, CT, 1980, pp. 5-8. [2] Lukas, M. and Giering, L. P., "The Effects o f Metal Particle Size in the Analysis for Wear Metals Using the Rotating Disc Atomic Emission Technique," International Symposium on OilAnalysis, Erding, Germany, July 1978. [3] Anderson, D. P., "Wear Particle Atlas (Revised)," Prepared for Advanced Technology Office, Naval Air Engineering Center, Lakehurst, NJ, June 28, 1982. [4] Rhine, W. E., Saba, C. S., and Kaufinan, R. E., "Metal Particle Detection Capabilities of Rotating-Disc Emission Spectrometers," Lubrication Engineering, Vol. 42, No. 12, 1986, p. 755. [5] Anderson, D. P. and Lukas, M. "Rotrode Filter Spectroscopy: A Method for MultiElemental Analysis of Particles in Used Oil Samples," Proceedings of the International Condition Monitoring Conference, Pensacola, FL, November 16-20, 1992. [6] Humphrey, G. R. "Filter Debris Analysis by Energy Dispersive X-Ray Fluorescence Applied to J52 Engines," Joint Oil Analysis Program International Condition Monitoring Conference, Pensacola, FL, April 18-22, 2004.

82

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

[7] Yurko, R. J. "Applying Rotrode Filter Spectroscopy (RFS) and LaserNet Fines (LNF) Technology to Detect Large Wear Particles of M50 Alloy from J52 Engine Beating Failures," Joint Oil Analysis Program International Condition Monitoring Conference, Pensacola, FL, April 18-22, 2004. [8] Case History Compliments of National Tribology Services, 5 Lakeland Park Drive, Peabody, MA 01960, U.S.A.

SULFUR DETERMINATION AND X-RAY FLUORESCENCE

Journal of ASTM International, March 2005, Vol. 2, No. 3 Paper ID JAI12978 Available online at www.astm.org

R. A. Kishore Nadkarni 1

Trace Levels of Sulfur in the Fuels of the Future: Analytical Perspective


ABSTRACT: Sulfur levels in both gasoline and diesel fuels are being restricted because of environmental pollution concerns throughout the world. The U.S. Environmental Protection Agency has mandated the levels of sulfur which may be allowed in most road and off-road vehicles. Although there are many standard test methods available for determining sulfur in petroleum products, presently only about four can qualify for their usefitlness at the expected low levels in the future fuels. A large interlaboratory study undertaken by ASTM to estimate the capabilities of these methods is described. Also, data from the ASTM interlaboratory crosscheck programs is reviewed to identify the trends that the industry laboratories are showing. Finally, a brief overview of some new methods being developed in the industry for such analyses is presented. KEYWORDS: sulfur analysis, gasoline analysis, diesel analysis, intedaboratory crosscheck programs, ultralow sulfur in diesels Origin and Occurrence of Sulfur in Petroleum Products Sulfur and its compounds are and will be present in virtually all petroleum products and lubricants from crude oils to ultra low sulfur fuels o f the future. The sulfur levels in these products vary from the highest amounts in the crude oils and some lubricants to trace levels in the currently proposed fuels. The inorganic elements present in crude oil originate from its marine animal and vegetative origins deposited with sediment in coastal waters in pre-historic times. Over the millennia, sulfur, oxygen, nitrogen, and other volatile compounds evolved out o f this primordial mixture due to bacterial action, leaving behind a mixture o f hydrocarbons containing some o f these elements and metals and other compounds. Apart from carbon, hydrogen, and oxygen, sulfur is one o f the main heteroatom present in the crude oils. Sulfur up to 6 m % has been reported in some crude oils. Sulfur content is one o f the more important criteria in evaluating the crude oil value and its processing. As the sulfur level increases, so do the complexity and the expense o f the refining operations. Low boiling point sulfur containing compounds such as hydrogen sulfide will be evolved during the distillation process, while other sulfur compounds will remain in the heavier distillation residue. Sulfur content is always determined as a part o f the "crude assay" o f the crude oils. Enormous amounts o f such data exist in each oil companies' crude assay database for the crude oils o f their interest. Some examples o f the sulfur content o f the major oil fields o f the world are given in Table 1.

Manuscript received 14 September 2004; accepted for publication 2 November 2004; published March 2005. i Millennium Analytics, Inc., East Brunswick, NJ 08816. Copyright9 2005byASTMInternational,100Ban"HarborDrive,POBoxC700,WestConshohocken,PA 19428-2959. 85

86

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

TABLE 1--Typical sulfur content of crude oils from some giant oil fields [1]. Country Abu Dhabi Algeria Canada Colombia Egypt Indonesia [ran [raq Kuwait Libya Nigeria Saudi Arabia Sulfur, m% 0.62-0.77 0.02-0.31 0.2-3.67 0.25-1.11 0.84-2.06 0.07-0.18 0.76-3.68 1.36--2.1 1.82-2.58 0.13-1.04 0.10-0.26 1.25-3.91 Country/State United States Alaska (N. Slope) California Louisiana Oklahoma Texas Wyoming Sulfur, m% 1.07 0.2-1.2 0.1-0.8 0.2-1.18 0.1-2.4 0.1-3.44

GasoBne and Diesel


Sulfur oxides formed during combustion of gasoline in the internal combustion engine are undesirable because o f damage both to the engine and to the environment. Also, o f major concern are the poisoning effects of sulfur on Nox reducing catalysts and adsorbers employed in mobile sources (vehicles) [8]. Sulfur acids increase rusting and corrosion of engine parts, piston ring, and cylinder walls. In the atmosphere, sulfur oxides will convert to sulfuric acid by reacting with moisture, with potential harm to vegetation and animal-human life. In diesel fuels, sulfur compounds can be present as mercaptans, sulfides, disulfides, and heterocyclic compounds, such as thiophenes. Their combustion by-products can cause wear in diesel engines and increase the amount o f deposits in the combustion chamber and on the pistons. Some typical sulfur concentrations reported in diesel fuels are given in Table 2. TABLE 2--Typical sulfur content of diesel fuels [2]. Diesel Fuel Diesel Fuel Oil Grade 1 D Diesel Fuel Oil Grade 2 D Diesel Fuel Oil Grade 4 D Kerosene Premium Diesel Railroad Diesel Marine Distillate Diesel Sulfur, m% 0.50 0.50 2.0 0.12 0.30 0.50 1.2

Fuel Oils
Depending upon the source of the crude oil, the refining process, and the fuel grade, varying a m o u n t s o f sulfur may be present in different types o f fuel oils. Combustion of the sulfur containing fuel oils produces sulfur oxides, which pollute the atmosphere and cause corrosion problems in boiler equipment. They may form sodium and vanadium complexes, and such deposits on external surfaces o f superheater tubes, economizers, and air heaters cause equipment corrosion and loss o f thermal efficiency.

NADKARNI ON TRACE LEVELS OF SULFUR

87

Lubricants In case of the lubricant additives and lubricating oils including automatic transmission fluids, gear oils, turbine oils, etc., additional amounts of sulfur in the form of metal sulfonates are added to enhance the performance characteristics of the oils in the internal combustion engines. Two principal sources of sulfur in engine oil are zinc dialkyl dithio phosphates and base stock. Some typical examples of sulfur levels in such lubricants are given in Table 3. See [4] for a discussion of the metal compounds used for this purpose and the effect they have on engine performance. Additional information on the presence and significance of sulfur in petroleum products can be found in [2-4]. TABLE 3--Sulfur levels in lubricants. LubricantProduct Automatic TransmissionFluids Gear Oils Lubricating Oils AutomotiveLubricantAdditives Sulfur, m% 0.1 - 0.3 1.5 - 2.0 0.3 - 0.7 0.5 - 3.5

Government Regulations
Because of the deleterious effects of sulfur emissions from motor vehicles, a number of government regulatory agencies, principally in North America and Europe, has been vigorously controlling and gradually reducing allowable sulfur contents of the fuels used for automotive, marine, aviation, off-road, utilities, and home heating. The allowed sulfur levels are steadily being tightened from a current level of about 330 mg/kg sulfur in gasoline and about 500 mg/kg in diesel to 30 and 15 mg/kg, respectively, in gasoline and diesel by 2006, and near zero by 2010. European diesel specification EN 590 mandates sulfur content of diesels from 350 mg/kg maximum in 2000 to 10 mg/kg maximum by January 1, 2005 in European Union countries. One method of sulfur removal from refinery streams is by hydrodesulfin'ization (HDS) in the refineries. This step also directly impacts the characteristics of low sulfur diesel fuels, such as density, aromatics content, cetane number, and cloud point. The magnitude of these changes will depend upon the type and setup of refinery HDS units. However, in the end some refractory compounds in fuel, e.g., 4,6-dimethyl dibenzothiophene, are very resistant to desulfurization, owing to the inaccessibility of the organically bound sulfur atom. Lower pressure HDS units which can work satisfactorily at 350 mg/kg sulfur levels, may have difficulty achieving reduction to 50 or 10 mg/kg sulfur level. In May 1999 the U.S. Environmental Protection Agency (EPA) released its proposed "Tier 2" emission standards for vehicles and gasoline sulfur standards for refineries [5-7]. They mandated the ASTM D 2622 X-ray fluorescence (XRF) method for sulfur determination, and in later announcements D 6428 became the mandatory method for sulfur determination in diesel fuels. Over the years through discussions and litigations, these rules have been modified to the point where alternate test methods can be used instead of only the mandated methods. The alternate methods allowed would be: (a) those developed by the voluntary consensus based standards writing bodies, such as ASTM, or (b) industry methods for which data will have to be provided to EPA to establish their correctness under Performance Based Measurement System (PBMS). Discussions between the industry and EPA are ongoing to further define the statistical basis for judging the available methods' acceptance and other related issues.

88

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Analytical Perspectives
Several test methods are available for determining sulfur in diverse petroleum product and lubricant matrixes from sub-mg/kg to m% concentration levels. One of the earliest methods issued by the D02 committee was for the sulfur determination by an oxygen bomb method (D 129) first issued in 1922. This was the 15th standard ever issued by the D02 Committee. The techniques currently used for the sulfur determination include classical wet chemistry (D 129, D 1266, D 1552, and D 2784), X-ray fluorescence (D 2622, D 4294, D 4927, D 6334, D 6443, D 6445, and D 7039), atomic spectroscopy (D 4951 and D 5185), and various thermal combustion microelemental methods using various detection techniques such as microcoulometry (D 3120 and D 3246), UV-fluorescence (D 5453 and D 6667), electrochemistry (D 4045 and D 6920), GC-FPD (D 7041), etc. A list of such available test methods is given in Table 4 along with other pertinent details as to the applicability of the method. Although nearly 20 test methods for sulfur are listed in Table 4, appropriate selection of a method for analysis would depend on the matrix, sulfur concentration level, and the desired precision of the analysis. Thus, out of the 20 methods above, only 4 typically qualify to be considered as potentially useful methods for the determination of low levels of sulfur in the ultra low sulfur fuels of the future: D 2622, D 3120, D 5453, andD 6920. Several of these ASTM test methods have also been separately issued by other international standards writing bodies such as IP, ISO, DIN, JIS, and AFNOR. These equivalent test methods are listed in Table 5 [9]. In an ever-expanding global market, it is useful in commerce to recognize the analytical methodology used in other industrialized nations.

ASTM Research Report D02-1456


In a very thorough research study done at SouthWest Research Institute (SWRI), San Antonio, TX, three sulfur test methods (D 2622, D 4294, and D 5453) were studied for fitness of use for the analysis of fuels at sulfur levels below 500 mg/kg [10]. All three were found to be equivalent for measurements in the 150-500 mg/kg range. But D 2622 and D 5453 were found to be of equivalent fitness for use down to 20 mg/kg, and D 5453 even to as low level as 1 mg/kg sulfur. It was also found that the latter method gives the best precision at such low sulfur levels. Later work in ASTM crosschecks confirms these conclusions.

European Sulfur Cross Check


In a very large crosscheck conducted in Europe, 69 laboratories from 9 countries analyzed fuels with sulfur ranges between 5 and 500 mg/kg in 8 gasoline and 7 diesel samples [11]. ISO equivalent methods ofASTM D 2622, D 4294, D 5453, D 3120, and D 1266 were employed. At these levels all five test methods were found to produce equivalent test results, but the precision of different methods varied considerably. The best reproducibility was found for methods D 2622 and D 5453. Based on this study, it was proposed that D 1266 and D 4294 methods be dropped from consideration for analyzing fuels at the 30 mg/kg sulfur levels in the 2006 and onward fuels. Only the D 5453 method was considered suitable for the determination of sulfur at <10 mg/kg. These conclusions are consistent with the work done at SWRI mentioned above.

NADKARNI ON TRACE LEVELS OF SULFUR

89

T A B L E 4--Methods for sulfur Technique ASTM Method Bomb combustion and BaSO4 D 129 precipitation D 1266 D1552 D 2622 D2784 D 3120

determination in petroleum products and lubricants.


Matrix Lube oils, adpacks, greases, other petroleum products Gasoline, kerosene, naphtha Petroleum cokes, other products All liquid petroleum products and oils LPG fuels Light liquid hydrocarbons Comments No longer widely used in the industry No longer widely used in the industry Nitrogen and chlorine may interfere Expensive iustrument, yet one of the most widely used methods Halogens may interfere Chlorine and nitrogen may interfere; precision is not very good Chlorine and nitrogen may interfere Not widely used in the industry Widely used in industry because of low cost instrumentation; poor precision at low sulfur levels Widely used in additives industry Widely used in additives industry for metals analysis Widely used in additives and oil industry for metal analysis Widely used in the industry for very low levels of sulfur; excellent precision at ultra low sulfur levels Standards must match oxygenated samples

Lamp combustion and gravimetry or titrimetry High temperature combustion and iodate titration or IR detection Wavelength Dispersive XRF Oxy-hydrogen lamp combustion and gravimetryorturbidometry Oxidative pyrolysis& microcoulometry Oxidative pyrolysis & microcoulometry Hydrogenolysis and rateometric colorimetry Energy Dispersive XRF

D 3246 D 4045 D 4294

LPG fuels Naphtha, kerosene, alcohol, jet fuel, etc. All liquid petroleum products and oils Additives and lube oils Additives and lube oils Additives, lube oils, and used oils Liquid hydrocarbons at low sulfur levels Gasoline and oxygenate blends Lube oils and additives Gasoline Gasolines, diesels, and oxygenates Gasoline and diesels Gasoline and diesels

D 4927 D 4951 D 5185 D5453

Wavelength Dispersive XRF ICP-AES ICP-AES Combustion U-v'-fluorescence detection Wavelength Dispersive XRF

D 6334 D 6443 D 6445 D 6920

Wavelength Dispersive XRF Energy Dispersive XRF Oxidative combustion and electrochemical detection D 7039 Monochromatic wavelength dispersive XRF D 7041 Combustion gas chromatography , with flame [0hotometricdetection

Precision poorer than other popular methods D 2622 or D 5453 New method; little field experience On-line sulfur method widely used in the field

T A B L E 5--International standards for the determination of sulfur. Analysis ASTM IP DIN AFNOR JIS ISO Bomb Method D 129 61 51-577 T60-I09 . . . . . . Lamp Method D 1266 107 ... M07-031 . . . . . . HT Method D 1552 ... M07-025 WDXRF D 2622 ... 51-400T6 K 2541 14596 M07-022 K 2276 16591 OxidativeMicroeoulometry D 3120 ... OxidativeMicroeoulometry D 3246 373 ... M07-052 ... EDXRF D 4294 336 M07-053 ... 8754 WDXRF D 4927 407 51-391T2 . . . . . . UV-Fluorescenee D 5453 ...... M07-059 . . . . . . (IP: Institute of Petroleum, U.K.; DIN: Deutsche Institut fur Normung (Germany); AFNOR: Association Francaise de Normalisation (France);/IS: Japan Industrial Standards; ISO: International Organization for Standardization)

90

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Reliability of EDXRF Method D 4294

This method is widely used in the industry for low levels o f sulfur determination because o f its comparatively low cost instrumentation and ease o f operation in spite o f the fact that the scope o f the method limits it to >150 mg/kg sulfur levels. Indeed, the method produces quite satisfactory results at these and higher levels o f sulfur. See Table 6 for comparison between four widely used methods which all produced equivalent results at m % levels o f sulfur in diverse products. T A B L E 6--Methods for sulfur determination at higher levels. ASTM D 2622 D 4294 D 4951 D 5453 Sample WDXRF EDXRF ICPAES UV-FL. CO 0403 1.008 0.019(14) 1.042 0.038(76) ... GO 0404 ... 1.856 0.069(6) 1.831 ~ 0.148(7) ... GO 0312 _. 1.832 0.060(8) 1.758 0.038(4) ... GG 0403 0.2089 0.0055(12) 0.2057 0.0098(22) ... RFG 0404 0.0416 0.0016(58) 0.0438 0.0022(35) ... 0.0411 MG 0404 0.0271 0.011(30) 0.0276 0.0040(51) ... 0.0258 RFG 0312 0.0238 0.0013(63) 0.0258 0.0029(38) ... 0.0230 All values in m% and expressed as average value standard deviation (number of results). A S T M interlaboratory crosscheck programs, however, have repeatedly demonstrated that at low levels o f sulfur this method gives biased high results compared with other reliable test methods, and it has very poor reproducibility sometimes approaching 100 % or more o f the m e a n sulfur value. The results o f such inaccuracy or imprecision could result in a false positive result for sulfur, indicating that the product is exceeding the regulatory sulfur limits and that the customer-supplier will have on-going quality disputes as to the true levels of sulfur in the shipped products i f the two parties are using two different test methods. Some recent data obtained b y three most widely used methods from A S T M crosschecks on low level sulfur in motor gasoline, diesel fuel, and reformulated gasoline are summarized in Table 7, which clearly indicate that the E D X R F method is not useful for determining low levels o f sulfur in the fuels o f the future. Because o f such poor quality o f data, the U.S. E P A did not include D 4294 as one o f the preferred methods for determining sulfur in fuels o f the future. With active help from instrument vendors, A S T M is looking into the possibility o f improving the accuracy and the precision o f n e w technology E D X R F methods. Several such initiatives are underway. T A B L E 7--Comparison of three widely used sulfur ASTM Sample D 2622 D 4294 WDXRF EDXILF Mogas 0208 13.4 4.8 (36) 31 21 (39) Diesel 0202 32.6 6 (80) 44 43 (137) RFG 0204 23.6 4.1 (64) 36 16 (27) RFG 0205 55.4 + 5.4 (67) 77 26 (30) RFG 0206 25.6 4.9 (72) 32 16 (25) RFG 0211 9.5 3.4 (63) 24 12 (27) RFG 0212 32.6 4.7 (66) 44 14 (31) RFG 0303 19.1 4.2 (67) 27 17 (34) RFG 0304 19.2 3.9 (66) 42.1 26.6 (34) RFG 0307 16.8 31

test methods.
D 5453 UV-F1. 11.7 1.9 (22) 32.0 4.7 (45) 22.9 3.2 (34) 54.3 6.8 (33) 22.4 2.6 (34) 8.2 + 1.8 (37) 31.6 3.9 (32) 16.3 3 (40) 16.9 2.3 (39) 15.0

All values are expressed as robust mean robust standard deviation in mg/kg (# labs).

NADKARNI ON TRACE LEVELS OF SULFUR

91

A Benchmark ASTM Interlaboratory Study


In 2000, the U.S. EPA published requirements for tier 2 fuels and also mandated the test methods that can be used for sulfur determination in these fuels. ASTM D 6428 was chosen as the mandatory method for diesel fuels with ASTM D 2622, D 3120, and D 5453 methods allowed as alternates if their accuracy and precision were shown to be equivalent to those of D 6428 method. The petroleum industry widely uses the latter three test methods for sulfur but had virtually no experience with the mandated D 6428 method, which is under the jurisdiction of ASTM D 16 Committee on Aromatic Hydrocarbons and Related Chemicals. The method was written for the sulfur determination of liquid aromatic hydrocarbons, their derivatives, and related chemicals. The precision quoted in this method was for aromatic hydrocarbons and not for petroleum products. Moreover, it was inadequate (10 analyses by one operator on two samples) and did not conform to the statistical protocols used in the D02 Committee test methods for calculating the precision and bias estimates. Consequently in the fall of 2001, D02.SC 3 undertook the responsibility to conduct a major Interlaboratory Study (ILS) to evaluate the capabilities of these four test methods - D 2622, D 3120, D 5453, and D 6428 - for the determination of low levels of sulfur in gasoline and diesel motor fuels of the future. In its scope and eventual extent this was one of the most exhaustive ILS ever undertaken by ASTM D02 Committee. There were a total of 16 samples each of gasolines and diesels. These also included biodiesels, oxygenated fuels, reformulated gasolines, and NIST gasoline and diesel SRMs. They contained varying levels of sulfur from 1-100 mg/kg. A total of 70 laboratories including those from the U.S. and abroad volunteered for this ILS, eventually, however, only 56 laboratories returned their results: 38 oil industry laboratories, 14 instrument vendor laboratories, and 4 commercial analytical service laboratories. Multiple vendor instruments were used in all methods. Several laboratories also used multiple test methods for the analysis. Some laboratories determined the required density determination in these samples. Thus, overall there were about 6000 data points used in the statistical evaluation of the data. The full report of this ILS is available as RR-D02-1547. The statistical evaluation of this data indicates the following important points: Although a similar European study [10] used somewhat different protocols and statistics, there was no dramatic difference in the precisions obtained in the European and this ASTM ILS. 9 The EPA mandated D 6428 test method had the poorest precision compared to three other methods used: D 2622, D 3120, and D 5453. 9 At the time of this study there were only 2 vendors available for D 6428 instruments. The repeatability but not the reproducibility of these 2 instruments was found to be significantly different. 9 The test method D 6428 would not be precise enough for regulatory compliance at the proposed sulfur limits. For example, the reproducibility at 30 mg/kg sulfur in gasoline would be 23, and that for diesel would be 16 at 15 mg/kg sulfur in diesel 9 The preferred methods for regulatory compliance would be D 2622 and D 5453 in terms of their accuracy and precision. Table 8 summarizes the precisions obtained from this ILS. As a follow-up to this landmark study, all three existing methods were revised to include the precisions obtained, and the D 6428 method of the D 16 committee was revised with the new precision data and issued separately as 9

92

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

D 6920 under the jurisdiction o f D02 committee. Eventually, E P A abandoned the requirement o f using mandatory D 6428 method in light o f the p o o r precision shown in this ILS. T A B L E 8--Precis!ons for sulfur ASTM Test Technique Method D 2622 D 3120 D 5453 D 6920

determination in ultra-low sulfur fuels of thefuture.


Range 30 mg/kg Gasoline r R 3.8 7.4 3.85 17.7 1.95 11.7 4.7 22.8 15 mg/kg Diesel r R 3.4 6.7 1.33 8.07 1.20 6.26 3.1 15.6

WDXRF Microcoulometry UV Fluorescence Eleetroehemistr/.

3 mg/kg-5.3 m% 3-100 mg/kg 1-800 mg/kg 1-100 m~Jk~

ASTM Interlaboratory Crosscheck Programs


Since 1993, A S T M D02 C o m m i t t e e ' s Coordinating Subcommittee 92 has been conducting crosschecks on a scheduled basis. This is probably the largest industry proficiency testing program in the world. A t present over 2000 laboratories w o r l d w i d e (of these 45 % are non-U.S. based laboratories) participate in 18 programs spanning diverse petroleum products and lubricants [12,13]. S o m e o f these programs include the determination o f sulfur from m g / k g to percent levels depending upon the product being analyzed. The most c o m m o n l y used methods for such analysis include D 2622, D 4294, and D 5453. Other test methods such as D 1266, D 3120, D4045, and D 6344 are used to a m u c h lesser extent. Overall, there appears to be good agreement between the m e a n results obtained by alternative methods for a particular product. H o w e v e r , there is considerable difference between the precisions obtained with the individual methods. S o m e examples o f data from m o r e recent cross checks are shown in Table 9. In most examples, only the data at sulfur level b e l o w 50 m g / k g are shown to illustrate the capability o f the alternate techniques. TABLE ASTM Sample Mogas 0208 Mogas 0312 Diesel 0202 Diesel 0402 Avgas 0307 Avgas 0403

9---Analysis for low levels of sulfur (nfuel products.


D 2622 .... 13.4 ~: 4.8 (36) 6.3 q- 3.1 (24) 32.6 + 6 (80) 9.3 q-4.3 (67) 63.6 5.4 (62) 9.6 q- 2.8 (60) D 4294 31 :~ 21 (39) 27 + 31 (39) 44 ~-43 (137) 26 a: 23 (91) 75 :~ 24 (117) 26 -~ 20 (94) ....... D 5453 11.7 -~ 1.9 (22) 3.74 q- 0.85 (37) 32.0 ~: 4.7 (45) 8.22 -~ 1.86 (61) 59.7 9 6.0 (43) 9.0 1.5 (56)

RFG 0204 23.6 4.1 (64) 36 16 (27) 22.9 d: 3.2 (34) RFG 0205 55.4 5.4 (67) 77 26 (30) 54.3 q- 6.8 (33) RFG 0206 25.6 4- 4.9 (72) 32 16 (25) 22.4 2.6 (34) RFG 0211 9.5 :~ 3.4 (63) 24 ~: 12 (27) 8.2 1.8 (37) RFG 0212 32.6 4.7 (66) 44 + 14 (31) 31.6 :~ 3.9 (32) RFG 0303 19.1 4.2 (67) 27 17 (34) 16.3 3 (40) RFG 0304 19.2 :t: 3.9 (66) 42.1 + 26.6 (34) 16.9 2.3 (39) RFG 0307 16.8 3.5 (58) 31 + 18 (28) 15.0 2.7 (40) RFG 0308 22.9 4.8 (60) 31 ~- 22 (31) 20.7 + 2.9 (42) RFG 0310 38.4 4.0 (65) 66 23 (36) 36.8 4.4 (46) RFG 0403 10.8 q- 4.2 (60) 28 :~ 19 (24). 9.31 :~ 1.42 (54) All values are in robust mean robust standard deviation in mg/kg sulfur (number of labs).

NADKARNI ON TRACE LEVELS OF SULFUR

93

This data excerpt clearly shows that the agreement between D 2622 and D 5453 is quite good, and both produce comparable precision. The EDXRF method D 4294, however, produces biased high results with reproducibility often approaching or exceeding 100 %, making it an unsuitable method for commerce or regulatory compliance at such low levels of sulfur. ASTM ULSl) ILCP Program Since the beginning of January 2004, the ASTM D02 Committee added a new Interlaboratory Cross Check Program for proficiency testing. The intent of this new program was to evaluate the industry laboratories' capability for determining ultra low levels of sulfur in diesel fuel. This is a monthly program, and only four test methods (D 2622, D 3120, D 5453, and D 6920) are allowed to be used. If other test methods have been approved by the U.S. EPA under its Performance Based Measurement System (PBMS) protocol, they may also be used. Generally the level of sulfur in these samples is between 1-15 mg/kg. It is also recommended that a certified reference material sample be analyzed along with the crosscheck sample and that data be returned to ASTM. A suggested standard for this work was NIST SP,,M 2723a with a certified sulfur value of 11 + 1.1 mg/kg. This program has proved quite successful in the short time that it has been offered since EPA is interested in using this database for evaluating individual industry laboratories' capability for such analyses. A summary of results from first six months' of crosschecks is given in Table 10. Based on these data it appears that: 9 9 9 Most laboratories use D 2622 and D 5453 test methods. Very few laboratories use the D 6920 method mandated by EPA at one time. Generally, the results by D 2622, D 3120, and D 5453 methods are comparable. All laboratories that used NIST SRM obtained correct results for it. However, only 20 % of the laboratories used this SRM. An equal number of laboratories did not use any CRM, and the rest used other commercial or in-house prepared CRMs. The reproducibility obtained in the crosschecks is similar to that given in the standard methods. In some samples (0402 and 0403) the sulfur levels were below the pooled limits of quantitation given in the test methods; yet, the labs were able to produce reasonably good results even though the uncertainty around the mean value was quite high. However, since the mean values were well below the regulatory compliance limits, the high uncertainty should not be a limiting factor for determining sulfur at these low levels. The sample 0406 had the lowest sulfur levels (sub-mg/kg) in this series of crosschecks. As a result, the uncertainty around the mean value was very large, sometimes approaching or exceeding 100 % of the mean value. Also, the agreement between the mean values by different methods was not very good. Obviously the capability limits of the currently available test methods become evident for sub-mg/kg sulfur determination. It would also be interesting to learn which test methods the refineries and blenders of these ultra low sulfur diesels are using to certify such low sulfur levels when it is clear from these crosschecks that the currently known test methods are not capable of such precision or detection. Discussions concerning all of these methods suggest that when the operator and equipment are prepared for the 10 mg/kg and below range, better precision can be obtained. This would be generally true for most instrumentation and test methods that

9 9

94

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

ideally the calibration range and the analyte concentration in the sample should closely match each other to produce the best results. TABLE l(Y--Results of ultra-low sulfur in diesel ASTM cross check programs. ASTM Sample # D 2622 D 3120 D 5453 D 6920 January 0401 15.2 2.3 (80) i3.35 3.46 ('i4) 15.1 + 1.5 (99) 13.8; 15.2 (2) February0402 2.35 + 1.76 (86) 1.53 + 0.52 (14) 1.58 + 0.47(104) 0.3; 0.9; 1.1 (3) March 0403 1.86 + 1.52 (87) 1.53 + 0.40 (10) 1.41 + 0.44(118) 0.96 + 0.54 (8) April 0404 15.41 + 2.13(96) 14.84 + 1.07 (14) 14.9+ 1.45(100) 16.20 2.96 (8) May 0405 10.48 1.79(95) 9.89 + 0.67 (16) 9.83 0.98(113) 0.3 - 9.8 (6) June 0 4 0 6 0.48+0.57(70) 0.90+0.63(14) 0.66 <0.1-1.2(6) July 0407 4.01 1.47 (89) 3.57 0.50 (12) 3.49 0.59(124) 2.91 1.18 (8) All values are expressed as robust mean + robust standard deviationin mg/kg (numberof results). For intercompany commerce as well as regulatory compliance, reproducibility of the data is a key parameter to deciding the validity of certificates of analysis sent from the suppliers to the customers. Hence, the reproducibility obtained for the four methods used in these crosschecks are compared with those given in the standards in Table 11. In the 21 data sets in this table, the test method reproducibility is better in 3 cases, about the same in 9 cases, and worse in 9 cases compared with the reproducibility obtained in these crosschecks. Thus, there is no clear indication that the field precisions have improved or worsened since the standards were issued. TABLE 11--Reproducibilities of the crosschecks versus standard test methods. ASTM Sulfur, D2622 D 3120 D 5453 ULSD # mgJkg C.C. STM C.C. STM C.C, STM 0401 ~ 15 6.38 3.64 9.98 7 . 2 2 4 . 0 9 6.29 0402 -2 4.89 4.75 1.49 1.13 1.31 1.35 0403 - 1.5 4.22 4 . 5 3 1.10 1.13 1.22 1.29 0404 N 15 5.92 8.27 2.57 8 . 0 9 3 . 9 4 6.23 0405 ~ 10 4.96 7 . 2 4 1.88 5.43 2 . 7 2 4.37 0406 ~0.5 1.59 3 . 7 9 1.77 0.80 1.11 1.01 0407 ~4 6.01 5 . 4 0 1.77 2.18 1.64 0.88 (C.C.: Crosscheck; STM: StandardTest Method)

Reference Materials for Sulfur Determination


One o f the key aspects in proving the applicability of an established or a newly developed method or analytical instnamentation is how well it performs in its accuracy and precision of measurements. Although the precision can be locally determined in a laboratory b y multiple analyses of a stable homogenous product over a period of time, well-characterized and preferably certified materials need to be available to determine the accuracy o f an analytical system. Although there are a number of commercial vendors who provide so-called "Certified Reference Materials" (CRM), their certified values are generally based merely on gravimetric blending of an analyte in an appropriate solvent, or they have been analyzed by one standard method. The National Institute of Standards and Technology (NIST) has pioneered in producing extremely well-characterized Standard Reference Materials (SRM) with infallible analytical values assigned to them. Generally, the NIST certification involves determination of an analyte by two independent primary analytical methods and collaboration by other secondary methods [14].

NADKARNI ON TRACE LEVELS OF SULFUR

95

M a n y sulfur S R M s have been produced b y N I S T o v e r the years that h a v e proven to be highly useful in determining the bias and accuracy o f existing and new test methods. Very recently, N I S T has been preparing some ultra-low sulfur fuel standards to help the industry in coping with the analyses o f such "clean" fuels. A partial list o f such available N I S T S R M s for sulfur in petroleum products is given in Table 12. There is also w o r k underway toward certifying other low level sulfur SRMs. TABLE Product Crude Oils Diesel Fuels

12--Sulfur SRMsfrom NIST.


NIST SRM # 2721 2722 1624d 2723a 2724b 2770 2294 2295 2296 2297 2299 2298 1616b 1617a 1819a 1848 2718 2719 1619b 1620c 1621e 1622e 1623c 1634c 2717a 1083 1084a 10854 S Concentration 1.5832 0.0044 m% 0.21037 0.00084 m% 0.3882 0.0020 m% 11.0 1.1 mg/kg 426.5 + 5.7 mg/kg 41.57 0.39 mg/kg 40.9 1.0 mg/kg 308 2 mg/kg 40.0 0.4 mg/kg 303.7 1.5 mg/kg 13.6 1.5 mg/kg 4.7 1.3 mg/kg 8.41 0.12 mg/kg 0.17307 0.00034 m% 0.04235 - 0.6135 m% 2.327 0.0043 m% 4.7032 0.0079 m% 0.8877 0.0010 m% 0.6960 0.0077 m% 4.561 0.015 m% 0.9480 0.0057 m% 2.1468 0.0041 m% 0.3806 0.0024 m% (2 m%) 2.9957 0.0032 m% (980 mg/kg) (1700 mg/kg) (4500 m~/k~)

Gasolines - RFG

Gasolines - High Octane Kerosines Lubricating Oil (X 5) Lubricating Oil Additive Petroleum Coke Residual Fuel Oils

Used Oils

Similarly, Laboratory o f G o v e r n m e n t Chemist (LGC) in U.K. has produced several fuel standards o f low sulfur levels: Product Gas Oil Gas Oil Diesel Diesel Diesel LGC CRM # 3000 3001 3023 3021 3022 Sulfur Concentration 94.4 2.4 m g / k g 448 11 52.4 1.1 10.67 0.27 ~ 32 In Preparation

96

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

On-Line Sulfur Determination


For many operations measuring on-line is a preferred choice rather than waiting for the offline analysis in a laboratory. At present, the U. S. EPA has not usually allowed on-line methods to be used for fuel products certification. Many on-line methods are extensions of wellestablished test methods such as D 2622, D 4045, D 4294, D 5453, etc., while others were developed for on-line analysis only. Some of the known on-line methods based on vendor supplied information are given below. This is not intended as a complete list of such available technology. 9 A technology based on combustion-UV Fluorescence method similar to ASTM D 5453 has been widely used in the field. An ASTM standard based on this application is being developed. A method based on combustion followed by gas chromatographic separation of sulfur species and flame photometric detection of S02 has been widely used in the field. This method was issued as ASTM standard D 7041 earlier this year. A hydrogenolysis and rateometric colorimetry technique based on ASTM D 4045 is being developed. An EDXRF method similar to ASTM D 4294 is available, but it may not have enough sensitivity to be used for levels in the ultra low sulfur diesel fuels.

9 9

There are also other vendors who are working on other technologies to market new on-line sulfur analyzers for an expected rise in demand for instrumentation in the refineries and blend terminals.

Emerging New Technology Test Methods for Sulfur


Given the increasing importance of precisely measuring lower and lower quantities of sulfur in fuels of the future, it is not surprising that a number of organizations are working to develop better methods for such analyses. Some of these methods, which have been presented for ASTM review, include: Monochromatic WDXRF: Traditional wavelength dispersive XRF instrument uses polychromatic excitation. This new technology uses monochromatic focused excitation. Preliminary work has shown a reproducibility of about 2 mg/kg at a level of 10 mg/kg of sulfur in gasolines and diesel fuels. Earlier this year this method was issued as ASTM D 7039. 9 Polarization EDXRF: Polarized excitation geometry used with energy dispersive XRF employing low power tubes and X-ray end window tube with palladium or rhodium anode and beryllium window is claimed to lower the sulfur detection limit to 1 mg/kg. The precision of the method has not yet been established. 9 EDXRF with proportional Counter: A low background proportional counter is key in a new energy dispersive XRF based method to measure lower levels of sulfur. This low background proportional counter suppresses the noise generated when incident X-rays are absorbed near the wall with resulting incomplete charge collection. An electrode shield close to the wall detects incomplete charge collection, and associated electronic detection 9

NADKARNI ON TRACE LEVELS OF SULFUR

97

circuitry rejects those events, resulting in lower spectral background and consequent lower detection limit compared to conventional proportional counters. There is a number of papers being presented in this DO2 Symposium which also deals with various aspects of new technology for low level sulfur determination in liquid fuels. Additional information on this subject may be found in Refs [15-18].
References

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

[15]
[16]

[171
[18]

Magee, E. M., Hall, H. J., and Varga, G. M., "Potential Pollutants in Fossil Fuels," EPA R2-73-249, National Technical Information Service Publication PB 225-039, 1973. Dyroff, G. V., "Manual on Significance of Tests for Petroleum Products," ASTM MNL 1, 5th ed., 1989. Boldt, K. and Hall, B. R., "Significance of Tests for Petroleum Products," ASTM STP 7c, 1976. Nadkami, R. A., "Modem Instrumental Methods of Elemental Analysis of Petroleum Products and Lubricants," ASTM STP 1109, ASTM International, West Conshohocken, PA, 1991. U.S. Environmental Protection Agency - Federal register 62 (133); 40 CFR Part 80 Part II, page 37337, July 11, 1997. U.S. Environmental Protection Agency- Federal Register 64 (1992); 40 CFR Parts 80, 85, and 86, page 26055, May 13, 1999. U.S. Environmental Protection Agency - Federal Register 65(28); 40 CFR Part 80, Page 6752 - 74, February 10, 2000; 66(12); 40 CFR Part 80, page 5002-5141, January 18, 2001. U.S. Environmental Protection Agency - Tier 2/Sulfur Regulatory Impact Analysis (EPA 420-R-99-023, December 22, 1999, Chapter IV: Technology Feasibility, page 311). Nadkami, R. A., "Guide to ASTM Test Methods for the Analysis of Petroleum Products and Lubricants," ASTM MNL 44, ASTM International, West Conshohocken, PA, 2000. Kohl, K, "Fitness for Use of Sulfur Methods," ASTM RR-D02-1456, 1999. Tittarelli, P., "Round Robin Exercise for Sulfur Test Methods for EN 228 and EN 590 Fuel Specifications, CEN/TC 19/WG 27, April 2000. Nadkami, R. A., "Does Your Lab Measure Up? ASTM Program Helps Build Proficiency," Lubes'N'Greases, April 2000, pp. 28-34. Nadkarni, R. A. and Bover, W. J., "Bias Management and Continuous Quality Improvements through Committee D02's Proficiency Testing," ASTM Standardization News, Vol. 32, No. 6, June 2004, pp. 36-39. May, W., Parris, R., Beck, C., Fassett, J., Greenberg, R., Guenther, F., et al., "Definition of Terms and Modes Used at NIST for Value-Assignment of Reference Materials for Chemical Measurements," NIST Special Publication 260-136, 2000. Nadkarni, R. A., "Determination of Trace Amounts of Sulfur in Petroleum Products," World Refining, Vol. 10, No. 5, June 2000. Nadkami, R. A., "Detecting Sulfur in Diesel Fuel," Today's Refinery, August 2000. Nadkarni, R. A., "Determination of Sulfur in Petroleum Products and Lubricants: A Critical Review of Test Performance," Amer. Lab., Vol. 32, No. 22, November 2000, pp. 16-25. Nadkami, R. A., "The Challenge of Sulfur Analysis in the Fuels of the Future," ASTM Standardization News, Vol. 32, No. 6, June 2004, pp. 32-35.

Journal of ASTM International, January 2006, Vol. 3, No. 1 Paper ID JA112972 Available online at www.astm.org
J. Wo]ska, 1 B. Vrebos, 1 a n d P. B r o u w e r 1

Analysis of Fuels, Lubricants, and Greases Using X-ray Fluorescence Spectrometry


ABSTRACT: X-ray spectrometry (XRF) is a versatile instrumental method for elemental analysis in a

wide variety of materials. The performance of three different XRF systems will be compared: a high power wavelength dispersive x-ray spectrometer (WDXRF), a low-power WDXRF, and a bench-top energy dispersive instrument (EDXRF). The data obtained show quite clearly the large difference in sensitivities between instruments. This translates into varying lower limits of detection (LLD) or quantification and sample throughput. For quantitative analysis well above the LLD, the accuracy of the results obtained on the different instruments is similar, or - for practical purposes of quality control in many applications - the differences in accuracy between the different instruments can be considered insignificant. Most of the applications and methods developed by a variety of laboratories and regulatory organizations are applicable to specimens of a specified nature and matrix. Some of the more versatile procedures include a method to account for differences in matrix between standard reference materials and unknowns. One of the effects that is usually neglected in this case is the wedge effect, which is caused by the fact that the volume that is excited by the beam and the volume from which characteristic radiation can be observed are not the same and vary (for a given spectrometer system) depending upon the density and the quantity of the sample. This effect will be illustrated and discussed.
KEYWORDS: Elemental analysis, x-ray spectrometry, analysis of fuels, lubricants and greases, XRF,

WDXRF, EDXRF
Introduction

As regulations regarding the environmental impact o f petroleum products like fuels and lubricants become more and more strict, the need for accurate and fast chemical analysis increases. Moreover, accurate and fast chemical analysis also strongly supports the production and ultimately the quality o f the end product. The demand to reduce the concentrations o f possibly harmful elements like sulfur imposes new challenges for detection technology. X-ray spectrometry (XRF) is a versatile, fast, and accurate instrumental method for elemental analysis in a wide variety o f materials. The performance o f three different XRF systems will be compared: a high power wavelength dispersive x-ray spectrometer (WDXRF), a low-power W D X R F , and a bench-top energy dispersive instrument (EDXRF). Because liquids comprise the majority o f samples in the petroleum industry, specific issues regarding the analysis o f liquids with X R F will be discussed. Such issues are related to the low mass attenuation coefficient and therefore the relatively large penetration depth for these specimens, for example. Manuscript received 8 October 2004; accepted for publication 13 June 2005; published January 2006. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. i Panalytical BV, Lelyweg 1, 7602 EA, Almelo, the Netherlands.

Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

98

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

99

Principles of Operation
A quick overview of the principles of operation of both WD-XRF and ED-XRF instruments will be given in this section. The main difference between these two techniques is in the detection system: wavelength-dispersive systems (Fig. 1, top) use a single crystal (monochromator) to select a particular wavelength of interest from the radiation emanating from the specimen. As a result of this, the task of the detector and its associated counting electronics is limited to counting pulses; in principle, all pulses detected correspond to the same wavelength. Some fairly straightforward and simple pulse height analyzers can take into account the presence of pulses corresponding to higher order reflections on the crystal and the escape peak. Energy-dispersive systems, on the other hand, lack these crystals (Fig. 1, bottom). As a result, all wavelengths in the beam emanating from the specimen are entering the detector. Hence, the detector and the counting electronics also have the task of 'sorting' the incoming photons according to their energy. Recently, several new implementations have been developed, combining the above principles with polarization and/or monochromatizing of the incident beam using curved crystals. In Fig. 1, top right, a WDXRF instrument with a monochromator in the incident beam is represented, while in Fig. 1, bottom right, the diagram of an EDXRF instrument using a polarizer is shown. These instruments will not be discussed further in this paper.

FIG. 1--Schematic diagrams of different XRF techniques.

100

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Lower Limit of Detection

The lower limit of detection has been a key issue with the industry. Regulatory bodies, such as the EPA in the USA and the EC have pushed the allowable levels of several compounds (such as sulfur) down significantly over the last two decades. Instruments and methods have to follow these directives, enabling the industry to provide proof that their products are meeting those requirements. Actually, both of the new developments mentioned earlier are driven by the fact that the instrumental lower limit of detection can be improved upon by lowering the intensity of the background radiation. The lower limit of detection (LLD) is dependent on the intensity of the background radiation as given by:

where LLD is the instrumental detection limit [ppm], S is the sensitivity [1/s/ppm], B is the intensity of the background [l/s], and T is the measuring time [s]. The parameter n determines the level of confidence and is usually taken equal to 3. Equation 1 clearly shows the effect of the counting time, the sensitivity, and the intensity of the background radiation on the value of the lower limit of detection. It is customary to work with mg/kg, although mg/L can also be used. The performance of the different X-ray fluorescence systems is largely determined by the power of the x-ray tube (read: sensitivity) and the specifications of the detector system (read: intensity of the background radiation). In general, high-end wavelength dispersive systems offer the best performance because they use high power excitation sources and they exhibit low background distortions due to the presence of the monochromator. In many instances of quality control, however, the mechanically much simpler energy-dispersive bench top instntments offer viable alternatives. TABLE 1--Typical LLDs in lubricating oils for different type o f XRF spectrometers. All . numbers . are in ppm and are. calculated for 100. s. . . .
System Highpower WDXRF Lowpower WDXRF Highpower EDXRF BenchMg 0.9 A1 0.4 Si 0.3 P 0.1 S 0.3 C1 0.8 K 0.2 Ca 0.2 Ti 0.4 Cr 0.3 Mn 0.3 Fe 0.2 Ni 0.1 Cu 0.1 Zn 0.1 Cd 0.4 Sn ""Ba' 0.6 0.7 Pb 0.3

15

13

12

0.5

0.3

0.9

1.3

1.9

1.1

0.6

0.5

3.5

4.1

3.4

12

10

10

12

10

1.1

1.6

0.8

0.7

05

0.2

0.2

0.2

0.8

2.3

0.4

100

34

20

15

10

10

30

16

23

10

top
EDXRF

The data in Table 1 show quite clearly the large difference in the lower limits of detection (LLD) between instruments. This translates also in significant differences of sample throughput. For quantitative analysis well above the LLD, the accuracy of the results obtained on the different instruments is similar, or - for practical purposes of quality control in many applications - the differences in accuracy between the different instruments can be considered insignificant. The high-power WDXRF and EDXRF instruments are the more expensive ones (and have comparable purchase prices), and the low-power WDXRF instrument is about a third o f these, while the bench-top EDXRF instruments cost about 10-20 % of the high-end ones.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

101

FIG. 2--Calibration graphs f o r sulfur obtained on two different spectrometers. The upper curve (squares) is achieved on a high-power WDXRF instrument, while the lower curve is achieved on a benchtop E D X R F spectrometer.
Matrix Effects

X-ray fluorescence analyzes the material in a condensed form (solid or liquid), compared to techniques that work in the gaseous or plasma state. As a result o f this, the intensity of the signal of a particular analyte is usually not directly proportional to its concentration, but is also subject (to a lesser but not negligible degree) to the concentration o f the other elements present ('the matrix'). The matrix elements influence the intensity in two distinct ways: absorption (generally causing a reduction o f intensity) and enhancement (an increase in intensity). The measured intensity is obviously the net result of these matrix effects. As a result, the calibration curves can be convex or concave (see Fig. 3). The physics of XRF, however, are well understood, and correction algorithms have been developed to deal with these effects in a mathematical way. Also, compensation methods (in which the effect is made to cancel out) have been developed. A popular one is the use o f an added internal standard. In this method, a fixed concentration o f a particular element is added to the specimen. The element is chosen so that it is affected by the matrix effect (absorption and enhancement) in a way similar to the analyte's radiation. The intensity ratio is then independent of the matrix effects. An example of such method is ISO 14596, which uses Zr as an internal standard for the determination of S in oil, or ASTM 5059, which uses Bi as an internal standard for determination of Pb in gasoline.

102

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 3--Influence o f absorption and enhancement effect on a calibration curve.

Sample Thickness
The matrix in oils consists primarily of low atomic number elements, which have a low absorption for the x-ray energies o f interest. This causes the primary beam to penetrate much deeper in the specimen than usually anticipated. Also, the characteristic radiation of the analyte elements experience low absolute absorption. Between specimens, however, the relative differences can still be significant, leading to matrix effects. The low absolute absorption causes the photons to travel a long way - and just like the photons of the primary beam, geometry becomes an issue as indicated in Fig. 4. The critical depth, d, is given by: 4.61 . d = sin V~ (2) PP where d = critical depth [cm];/d = mass attenuation coefficient [cm2/g];/9 = density [g/cm3]; and ~2 = take off angle.

FIG. 4--Critical depth.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

103

This thickness corresponds to the thickness required to absorb 99 % of radiation in a material with a given density and at a specified energy. The data in Table 2 give the critical depths (or infinite thickness as they are also referred to) for lead, iron, silica, lithium tetraborate, and water for photons corresponding to Mg Ka, CrKa, and Sn Kct. Mg K a radiation (with an energy of 1.25 keV) is readily absorbed within a very small distance - less than 20 microns in all materials listed. Some photons corresponding to Cr Kcx (energy of 5.41 keV) will travel a distance of roughly one millimeter in water. This means that it takes a layer of water 1 mm thick to absorb 99.9 % of the fluorescent photons. The small remaining fraction will eventually be absorbed deeper in the material. For Sn Kc~, the situation becomes quite different, as we are now talking about path lengths of centimeters. Please note that the distances travelled in oil are not much different from those given in water; even though there is some minor influence by the C/H ratio and the content of oxygen in the oil considered. TABLE 2--Critical depth for different materials for distances in microns, unless indicated otherwise. Material Mg Kcr Cr Kcz Lead 0.7 4.5 Iron 1 35 SiO2 8 110 Li2B407 13 900 H20 16 1000

Mg Ka, Cr Kv~ and Sn Ka. All


Sn K a 55 290 0.9 cm 4.6 cm 5.3 cm

In XRF, the specimens are commonly assumed to be infinitely thick for all photon energies considered (including the incident beam). This is certainly justified in industrial applications when dealing with alloys. So, many quantitative methods are implicitly derived under this assumption. However, as the matrix of the specimens consists of more and more low atomic number elements, this assumption is no longer justified, and applying the methods developed for .analysis of alloys will lead to significant bias. This is caused primarily by the fact that the incident beam is not fully absorbed within the specimen. If the fraction absorbed is constant between specimens and standards, a workable solution is achieved in many cases. The assumption that the complete incident beam is absorbed and causes fluorescence in the specimen is then merely replaced by the fact that a constant fraction of the beam is absorbed; as shown above in Fig. 4. Several methods have been derived with the explicit assumption that the specimens are not infinitely thick, but these are beyond the scope of this paper. An extensive review can be found in the work by van Grieken and Markowicz [1].

The Wedge Effect


Another effect compounding the situation is the wedge effect. The algorithms developed for specimens of less-than-infinite thickness are usually derived with the assumption that the thickness is small compared to the surface area. This implies that the surface of any layer within the specimen is constant. This assumption is no longer justified when the thickness (or rather the path length) is comparable to the diameter of the specimen; see Fig. 5. Consider the sample as being built up by many very thin layers of the same composition, parallel to the measured surface of the specimen. The elements in each of these layers are excited

104

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

by the photons of the primary beam and emit their characteristic radiation. The attenuation of the primary beam and of the secondary beam by other layers on the way from the x-ray source and toward the detector can be taken into account. As long as the penetration depth of the radiation is small compared to the surface area, the area of each layer (and thus the amount of material contributing to the measurement in each layer) can be considered constant.

FIG. 5--Schematic illustration of the wedge effect. As the penetration depth becomes larger and larger, eventually the volumes that are not excited or where fluorescent radiation generated is not detected become significant. This effect is called the shadow effect or the wedge effect (see Fig. 5) and finds its origin in geometrical aspects of the instrumentation. The fraction of each layer that contributes to the measured signal decreases as the layer is positioned further from the specimen surface. This decrease is additional to the decrease caused by absorption by intermediate layers. The intensities ofSn Kct and Cu Kc~ of butanol samples containing 500 ppm Sn and 500 ppm Cu each were measured, but with different volume (thickness). All samples had the same diameter of 37 mm. The penetration depth from Sn Kct in butanol is about 5 cm, versus only about one millimeter for Cu Kot (see Table 2, and noting that butanol consists of C, H, and O; the value for H20 is used for reference purposes only; the theoretical calculations are performed with the correct matrix type). The data in Fig. 6 and Table 3 show that the measured intensity reaches its maximum value for a thickness of 20 mm. A sample with a thickness of 8 mm already produces 85 % of the maximum reachable intensity. Also, from Fig. 7 it can be concluded that one way to avoid the wedge effect is to analyze small amounts of sample: if the thickness of the specimen is small (e.g., below 1 mm, although this depends on the energy of the analyte's radiation), the effect becomes negligible. When this approach is taken, however, it must be clear that always the same mass of specimens must be used for analysis, or a finite thickness correction must be used. For Cu-Kct radiation the effect is much less pronounced than for Sn-Kc~ radiation, because the penetration depth for Cu-Kot radiation is much lower than that of Sn-Ko; as illustrated in Fig. 7. Methods based on an internal standard (such as ISO 14596, which uses Zr as an internal standard for the determination of S in oil) will implicitly correct for tiffs wedge effect and for the matrix effect because the effect is the same for both the internal standard and the analyte lines.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

105

60

SO

.4,- ~

12O W 9 W C

40

f
I

30

o.

2o

10

15

20

25

T h i c k n e s s [ram]

F I G . 6--Measured intensity (solid line) and calculated intensities of Sn Ko~ Dashed line: calculated without wedge correction; dotted line: calculated with wedge correction of samples with different thicknesses. Note that the wedge effect is significant for Sn Ka in butanol (refer to the penetration depth of Sn K a in [-120 in Table 2). TABLE Thickness [mini 4.1 8.2 12.2 16.3 20.4
41.00

3--Penetration depth.

,,

. . . . Calculated without wedge effect 20 34.3 44.3 51.4 56.4

Intensities [kcps] Calculated with wedge effect 20 30.6 35.4 37 37

Measured 20 31.8 36 37.3 37.4

40.00.

J,

J_

<<-"
Q. CJ 39.00.

........

. . . . . . . . . .

,,~ 38,00 9

36.00 -

35.00 5 10 15 20 25

Thickness

[ram]

7--Measured intensity (solid line) and calculated intensities of Cu Ket (dotted line with wedge correction, dashed line without wedge correction) of samples with different thickness.
FIG.

106

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

In case of greases where introducing the internal standard into the sample itself is often not possible, as it is in liquids, the alternative may be analyzing the thin grease layer distributed in the sample cup. The thickness of each sample has to be the same, and instead of internal standard, a plate or foil from Nb in this case can be used. This practical approach is illustrated in Fig. 8. But for methods without internal standard, care must be taken to matrix-match the standards and the unknowns, and to work with constant weight in the sample cups. The alternative procedure is to account for the wedge effect and to use a full fundamental parameter approach to the analysis.

FIG. 8--Preparation of a grease sample.


Conclusion

XRF is a well-established and versatile instrumental technique for elemental analysis in a broad range of mostly industrial applications. Where high-power WDXRF yields the best results in terms of LLD, the more economical bench-top, on the other hand, often gives quite acceptable results, especially in production control. The wedge effect is significant for high energetic radiation in low-Z matrices. The effect can be modelled adequately by taking it into account in the Fundamental Parameter equations. When analyzing heavy elements in low-Z matrices, two scenarios can be used: either measure only a small amount of sample to reduce the wedge effect, or take the wedge effect into account if larger amounts are measured. In both cases, finite thickness correction is required. Measuring only small amounts will not affect the sensitivity drastically. A number of XRF methods currently exist in ASTM literature and are widely used. These include D 2622, D 6334, D 4294, D 5059, and D 7039 for fuels; and D 4927 and D 6443 for lubricants.

WOLSKA ET AL. ON FUELS, LUBRICANTS, AND GREASES

107

Reference

[1] "Quantification in XRF Analysis of Intermediate-Thickness Samples," Handbook of X-Ray Spectrometry, 2nd edition, R. van Grieken and A. Markowicz, Eds., Marcel Dekker, New York, Basel, 2002.

Journal of ASTM International, September 2005, Vol. 2, No. 8

Paper ID JAI12974 Available online at www.astm.org


Scott F e s s 1

Determination of Sulfur Content in Crude Oil Using On-Line X-ray Transmission Technology
ABSTRACT: This paper describes the function and use of X-ray Transmission (XRT) analytical instrumentationas applied to the classificationand blending of crude oil that contains0.1-3.3 % sulfur by weight. In the XRT method, sulfur absorbs X-rays transmitted between an X-ray source and detector. The techniquecompensates for density changes in the oil and minimizesthe effects of variations in the C/H ratio of the oil. When used on-line the instrument obtains the sulfur measurementin the highpressure flow conditions required to pump crude oil, typically 800 psig in pipelines. Automatic compensationfor variationsin oil densityand temperature are achievedusing integrateddensitometerand temperature sensor readings taken by the instrument as the sulfur is being measured. In this paper calibration,precision,and long-termstabilityresults are shown. Experimentaldata were obtainedusing a commerciallyavailableon-line XRT analyzer. KEYWORDS: sulfur, fuels, crude oil, refinery,pipeline

Introduction
As sulfur is tightly regulated in petroleum fuels, such as gasoline and diesel, the sulfur content of the oils must be closely monitored throughout the entire refining process, beginning with crude oil. Determining the sulfur content in crude oil is of importance for the classification and blending of the crude. Although government regulations control the allowable sulfur content in fuel oils, the sulfur content in crude oil is not directly regulated. Thus, the quality and price of crude oil depends on the sulfur content. Sweet crude contains a maximum of 0.50 % (wt/wt) total sulfur, whereas sour crude may have a maximum o f 2.99 % total sulfur. Some crude oils may contain as much as 6 % total sulfur. On-line analysis is highly desirable for the classification and blending of crude oil, requiring a fast and robust method of analysis. This paper examines the use of the XRT (X-ray Transmission) technique as applied to the analysis of sulfur in crude oil. XRT may also be referred to as XRA (X-ray Absorption).

Analytical Techniques and Instrumentation


X-ray Fluorescence (XRF) is a common technique for sulfur determination in hydrocarbon oils. At-line and laboratory XRF analysis is covered in the ASTM standard test methods for sulfur in petroleum products: ASTM Standard Method D 4294 (for Energy Dispersive XRF, EDXRF) [1] and ASTM Standard Method D 2622 (for Wavelength Dispersive XRF, WDXRF) [2]. Polarized EDXRF [3] is also used at-line for the determination of very low sulfur content (<10 ppm sulfur) in diesel and gasoline fuels. Manuscript received 8 October 2004; accepted for publication 27 January 2005; published September 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December2004 in Tampa, FL. i SPECTROAnalyticalInstruments,Inc., 1515N. Highway 281, MarbleFalls, TX 78654. Copyright9 2005byASTMInternational, Ban"HarborDrive,POBoxC700,WestConshohocken, 19428-2959. 100 PA 108

FESS ON ON-LINE X-RAY TRANSMISSION

109

EDXRF is well suited for the on-line measurement of sulfur content for less dense oils that are pumped at low pressures, typically at a maximum of 30 psig. Refineries, pipelines, and blending operations, however, require on-line analysis of thick, dense, viscous crude oil pumped at high pressure. These environments require a robust analysis technique with instrumentation capable of accurately measuring crude oil pumped at a pressure of 800 psig. XRF is not desirable for such an analysis. XRF measures the fluorescent X-ray emission of sulfur, which occurs at an energy of 2.308 keV (K-alpha emission line). Absorption effects are therefore much greater than in the XRT method. Matrix effects that occur in XRF can be minimized using XRT. Also, because the sulfur fluorescent X-rays are low energy, the path length that they travel before being completely absorbed is a tiny fraction of the path length of X-rays with energy 21 keV used in the XRT technique. Therefore, coatings that may build up on the flow cell window will greatly affect the XRF measurement. Furthermore, the thin windows necessary to efficiently transmit the soft fliaorescent X-rays cannot sustain pressure differentials as high as 800 psig. For these reasons, XRT is a more desirable technique for the on-line analysis of sulfur in crude oil. ~,a XRT, X-rays are emitted from an X-ray source and pass through the volume ofoil to a detector on the opposite side of the sample flow cell. X-rays are emitted from an X-ray source at a high enough energy to pass through the volume of oil to a detector on the opposite side of the sample flow cell. The presence of sulfur in the oil absorbs these source X-rays, and so calibration is obtained by relating concentration to X-ray count rate. This calibration is an inverse relationship: as the sulfur concentration increases, the X-ray intensity reaching the detector decreases. In this way the sulfur is indirectly measured by measuring the amount of Xray radiation transmitted through the oil. The transmission (or absorption) of X-rays is given by the following equation [4]: I/Io = exp-dt[ ~tm(1-Cs)+lasCs] where: I = final X-ray intensity, after absorber [counts] Io = initial X-ray intensity, before absorber [counts] d = density of absorbing material [g/cm3] t = thickness of absorber [cm] #m = mass absorption coefficient for the matrix [cm2/g] /~ = mass absorption coefficient for sulfur [cruZ/g] Cs = weight fraction of sulfur [% wt/wt] (1)

Table 1 shows the mass absorption coefficients for the hydrogen, carbon, and oxygen components of organic oil, as well as ~t for sulfur. Because the sulfur ~t is on the order of ten times greater than those for hydrogen, oxygen, and carbon, the transmission of X-rays are primarily proportional to the sulfur content. TABLE 1--Selected mass absorption coefficients at X-ray energy 21 ke V [5]. Element Mass Absorption Coefficient, ~t [g/cm2] H 0.37 C 0.41 O 0.79 S 5.82

1 10

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Possible interferences include the presence of chlorine in the oil and coating of the flow cell beryllium windows. Chlorine may interfere with the sulfur measurement if the chlorine is present in a concentration ratio to the sulfur of greater than 1:10 and the chlorine concentration varies independently of the sulfur. Flow cell window coating has negligible effect on the sulfur measurement in XRT. This is because a coating build-up of a few hundred microns is a negligible part of the entire X-ray path length. The SPECTRO Model 682T-HP on-line XR.T analyzer was used to obtain the data presented in this paper [6]. All measurements and data were obtained with the sample in a static condition, completely filling the sampling pipes (flow cell and densitometer paths). X-rays and X-ray transmission are not affected by sample flow.

Description of the Instrument


Figure 1 shows a drawing of the analyzer. The essential components are: sample flow cell; X-ray source; X-ray detector; and densitometer. Figure 2 shows the sample flow and data output of the analyzer.

Detector

X-raySource

CabinetDoor in the
Open position

SampleFlow Cell FIG.

l--Component diagram.

FESS ON ON-LINE X-RAY TRANSMISSION


R T R TO P O E S EU N R CS

11 1

I" 600# ANSI RF FLANGE--~

MANUALSAMPLEINPUT/VENT

A/02 = Density
CaUbratlon S e l e c t D/J2 = Remote S t o r t / S t o p I)/[ll = Common At~rm
D/It =

I]/Q2 = Resu~UDd~ted(PuLse)

SPECTR~] XRA/XRTANALYZER INPUT/OUTPUT

POVER 110/220VAC 3/1,5 A, 60/50Hz PURGE SUPPLY 9 ) 5 kg/r PSI) ) 0.22~ /hP r $CFH) Depends on Area Ct~s$1FIc~tlon

~
1" 60D# ANSI RF FLANGE

PROVIDED--,-BY SPECTRD M N A DRAIN AUL

POIE R VD D BY C S O E UT MR

P E S R RELIEF VALVE RSUE

I I E~
S V { ~
DRAIN

INLET FRDMP O E S R CS 10 - 60"C (50 - 140"F), 56 kg/c~e( O PSI) M X B O A

FIG. 2--Sample flow and data output.

Sample Flow Cell


The oil sample enters the bottom o f the flow cell tube and exits at the top. The inner diameter of the flow cell tube is 4.90 cm. The X-ray measurement is taken as the sample flows through the tube. Beryllium windows allow the X-rays to enter and exit the analysis volume, the X-rays traversing a total path length of 8.64 cm.

X-ray Source
The X-ray source is a 30 kV side-window X-ray tube with a rhodium anode target. X-ray tube voltage and emission current are set in order to yield source X-rays with energy above 25 keV and approximately 30 000 cps at the detector. For these tests the following X-ray tube parameters were used: tube voltage of 27 kV and emission current of 75 uA. To obtain a more monochromatic X-ray source in the range 20-30 keV, a palladium filter is mounted over the Xray tube. This gives source X-rays primarily at 21 keV, the K-line emission of palladium.

X-ray Detector
The detector is a gas-filled proportional counter with an argon fill gas. The bias voltage is set such that the detector operates in the normal proportional counter mode, detecting all X-ray energies between 3 keV and the tube voltage setting.

112

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Densitometer
Oil from the process stream simultaneously flows through the X-ray sampling flow cell and a densitometer. The densitometer measures the oil density and temperature and inputs these values into the data processing software for corrections to the calibration and final analysis results.

Data Output
Total X-ray counts from the detector are simply accumulated and output using a programmable logic controller (PLC). The total counts are expressed as count rate (total counts divided by analysis count time) and sent to the software for use in calibration and analysis algorithms.
Performance

This section describes typical results obtained for calibration, precision, sensitivity, and longterm stability. A comparison o f repeatability is also made with ASTM D 4294. All measurements were obtained using an analysis count time of 100 s, and results are quoted as % wt/wt.

Calibration
Calibration was performed b y analyzing a suite o f certified crude oil standards. [7] The assay values for these samples are given by the certification shown in Res 7. A broad range calibration was built for the range o f sulfur from 0.1-3.3 %, with the density increasing with sulfur content. To refine the calibration over the specific range used in blending operations, a second calibration was built over the range o f 0.1-1.5 % sulfur, with independently varying densities. ,,, TABLE 2--Broad range correlation results. Element: S SEE = 0.1624 Density [g/cm3j Assay Value [%J MeasuredValue [%J Sample 0.763563 0.10 0.089 1 0.836063 0.58 0.599 2 0.883313 1.01 1.098 3 0.862688 1.97 1.758 4 0.958500 2.48 2.689 5 1.009310 3.28 3.187 6 TABLE 3--Blending range correlation results. Element: S SEE = 0.051 Sample Density [g/era3] Assay Value [%] MeasuredValue [%] 11 0.902630 0.10 0.092 7 0.798000 0.30 0.370 12 0.900380 0.30 0.273 9 0.848630 0.50 0.461 10 0.882190 1.00 1.000 13 0.898880 1.00 1.033 8 0.801000 1.50 1.440 14 0.897188 1.50 1.538

FESS ON ON-LINE X-RAY T R A N S M I S S I O N

113

TABLE 4---Precision-based on broad range calibration. Element: S Sample Assay Value* Mean Value 1a Std. % [%] [%] Dev. Relative 1 0.10 0.104 0.005 4.6 6 3.28 3.225 0.003 0.1 * As givenby the certification. See Ref. [7].
Broad Range Correlation Plot
3.500 3.000 == 2.500 2.000 1.500

1.000
0.500 t
0.000
, ,

0.00

1.00

2.00 3.00 Assay Value

4.00

FIG. 3--Broad range correlation plot.


Blending Range Correlation Plot
1.800 ,

=e 1.200 ] 0.900 1

1
0.00 0.30 0.60 0.90 1.20 1.50 1.80 Assay V a l u e

FIG. 4--Blending range correlation plot.

Precision and Sensitivity


The precision results for selected samples shown in Table 5 were obtained by making 100 repeat analyses of each sample using a count time of 100 s per analysis. A detection limit was obtained using the empirical method. The Minimum Detection Limit (MDL) here is defined as the 3tr observed standard deviation of 100 analyses of a blank sample containing no sulfur. Using heavy mineral oil as the blank sample, the MDL for sulfur was observed to be 0.02 %.

114

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE 5--Precision-based on blending range calibration. Element: S Sample Assay Value* [%] Mean Value [%] lo Std. Dev. 11 0.10 0.093 0.006 7 0.30 0.344 0.005 10 1.00 0.991 0.006 13 1.00 1.022 0.005 14 1.50 1.536 0.004 * As givenby the certification. See Ref.7.

% Relative 7.0 1.4 0.6 0.5 0.3

Long-term Stability
Long-term stability data were gathered over 3090 continuous measurements of a static sample over a 92-h time period. The results are shown in Table 6.

Comparison to ASTM D 4294


Section 15.1.1 (Repeatability) of ASTM D 4294 expresses the r-value for the measurement of sulfur in oils using EDXRF to be: r = 0.02894(X+ 0.1691) (2) where Xis the sulfur concentration in mass %. Table 7 shows a comparison of the precision results of the XRT method (derived from Table 5) and the r-value (repeatability) given in D 4294. The table shows the 95 % confidence value (2.77 o) for the precision results using XRT as compared to the maximum 95 % confidence rvalue listed in D 4294. These results indicate that the XRT on-line method is equivalent to the repeatability performance quoted in ASTM D 4294 (for EDXRF) for sulfur concentrations above 0.10 %. TABLE 6---3090 repeat analyses at 100 s per analysis. Element: S Sample Assay Value* Mean Value la Std. % [%] [%] Dev. Relative 14 1.50 1.538 0.005 0.3 * As givenby the certification. See Ref. 7.

Sample 11 10 7 13 14

TABLE 7--XRT results compared to ASTM D 4294. XRT XRT 2.770 r-value from D MeanValue(%Sulfur) Std. Dev. 4294 0.093 0.017 0.008 0.991 0.017 0.034 0.344 0.014 0.015 1.022 0.014 0.034 1.536 0.011 0.049

FESS ON ON-LINE X-RAY TRANSMISSION

1 15

Conclusions

9 9 9 9

The X-ray Transmission technique is a viable method for the determination of sulfur content in crude oil. Precision for sulfur above 0.1% is equivalent to the r-value (repeatability performance) quoted in ASTM D 4294. The XRT method is better suited than XRF for the process control of sulfur in crude oil. Future work is required to determine the response time of the instrument to step changes in sulfur concentration and to test flowing conditions of the sample.

References

[1] ASTM D 4294, Standard Test Method for Sulfur in Petroleum Products by EnergyDispersive X-ray Fluorescence Spectroscopy, ASTM Annual Book of Standards, ASTM International, West Conshohocken, PA, 19428. [2] ASTM D2622, Standard Test Method for Sulfur in Petroleum Products by X-ray Spectrometry, ASTM Annual Book of Standards, ASTM International, West Conshohocken, PA. [3] SPECTRO Analytical Instruments GmbH & Co. KG, Boschstr. 10, 47533 Kleve, Germany. [4] Liebhafsky, H. A., Pfeiffer, H. G., Winslow, E. H., and Zemany, P. D., X-ray Absorption and Emission in Analytical Chemistry, 2nd ed., John Wiley & Sons, Inc., 1966, pp. 130-133. [5] Robinson, J. W., CRCHandbook of Spectroscopy, Vol. 1, CRC Press, Inc., 1974, p. 208. [6] SPECTRO Analytical Instruments, Inc., Marble Falls, TX. [7] Analytical Services, Inc., The Woodlands, TX. Standards certified and assayed using ASTM D 2622.

Journal of ASTM International, September 2005, Vol. 2, No. 8 Paper ID JAI12971 Available online at www.astm.org

Z. IV. Chen, 1 Fuzhong Wei,11an Radley, i and Berry Beumer 1

Low-Level Sulfur in Fuel Determination Using Monochromatic WD XRFmASTM D 7039-04


ABSTRACT: A monochromatic wavelength dispersive X-ray fluorescence (MWDXRF) technique has been successfully developed for measuring low-level sulfur in fuel. In this technique, two doubly curved crystal (DCC) optics were used to provide monochromatic excitation and fixed channel wavelength dispersive analysis. Using highly efficient DCC optics, compact bench top MWDXRF analyzers for sulfur analysis have been successfully produced. A new ASTM standard test method for low-level sulfur determination in diesel and gasoline, D7039-04, has been developed based on this technique. Data for ultra-low-sulfur fuel were collected and analyzed using this new method. In this paper, the results of repeatability, reproducibility, and bias are presented and discussed. The repeatability for 10-ppm sulfur fuel was shown to be around 1 ppm. The pooled limit of quantification (PLOQ) for ultra-low-sulfur diesel was found to be less than 1.5 ppm in this study. The reproducibility of 15-ppm sulfur diesel fuel was determined to be better than 3 ppm (95 % confident level). The limit of detection for a single analyzer was found to be 0.36 pprrL The effect of matrix composition was investigated. The correction due to matrix for low-level sulfur diesel and gasoline is generally not significant.
KEYWORDS: sulfur, diesel, X-rays, D7039, fluorescence, optics Introduction

To deliver low-sulfur fuel to meet the new EPA regulation, it is important to monitor the sulfur level during the fuel-refining process and transport in a pipeline terminal. X-ray fluorescence (XRF) analysis has been demonstrated to be a reliable and accurate method for online measurement o f sulfur in petroleum processes. However, conventional XRF using energy dispersive spectrometry (EDS), without the use o f X-ray optics, is limited in detection o f lowlevel sulfur in oil due to a relatively poor signal-to-background ratio (S/B). Another serious issue for any XRF/EDS system is that the strong scattering o f the primary beam from the sample can adversely inundate the energy dispersive detector resulting in large detector dead times during measurement. This limits the intensity of the primary beam due to the counting limitations o f the energy detector. Alternatively, XRF with a wavelength dispersive spectrometer (WDS) gives improved sensitivity and precision over EDS systems. However, high performance WDS systems require high power X-ray tubes with water-cooling and extensive maintenances. These systems are not suitable for distribution terminal quality control and online applications. XRF analysis using monochromatic excitation can provide a much better S/B than the polychromatic excitation for EDS and WDS XRF techniques described above by eliminating the scattering o f bremsstrahlung (continuum X-rays) from the X-ray tube. Laboratory monochromatic XRF methods have been hindered b y the lack o f efficient X-ray focusing Manuscript received 9 November 2004; accepted for publication 27 January 2005; published September 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. I XOS, 15 Tech Valley Drive, East Greenbush, NY 12061.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

116

CHEN ET AL. ON MONOCHROMATIC WDXRF

117

monochromators until recently. With innovative point-to-point focusing doubly curved crystal (DCC) optic devices [1,2], intense monochromatic focused beams can be achieved using low power compact X-ray sources. The basic configuration of a compact monochromatic WDS XRF sulfur analyzer with single channel analysis based on DCCs is shown in Fig 1. It consists of an X-ray tube, a point-focusing DCC for X-ray excitation, a sample mount, a focusing DCC for collection of characteristic X-rays from the sample, and an X-ray counter. The X-ray tube has power of 30-75 W and its spot size is in the range of several hundreds of microns. In this system, the first-point focusing DCC captures a narrow bandwidth of X-rays from the source, typically a characteristic line of the X-ray tube, and focuses an intense monochromatic beam to a small spot onto a sample. The target material of the X-ray tube is chosen to have proper characteristic line for the excitation of sulfur atoms. During sample excitation, sulfur K~x secondary characteristic fluorescence X-rays are emitted from the excitation volume of the sample. The second DCC, the collection crystal, collects the sulfur Kct X-rays and focuses them at the detector. The X-ray detector can be a proportional counter or a solid-state detector. The intensity of the detected X-rays has linear relationship with the sulfur elemental concentration in the specimen. With proper calibration, the sulfur concentration in a sample can be directly determined.
Sam )le

--

,,

<5 tor

Source Count DCC Optic2 DCC Optic1

FIG. l--Principle of Monochromatic WavelengthDispersiveXP~F sulfur analysis. for The most important advantage of the monochromatic wavelength dispersive X-ray fluorescence (MWDXRF) approach is that the S/B is improved by using the monochromatic excitation of the X-ray source characteristic line. The primary beam is highly monochromatic, and the bremsstrahlung photons from the X-ray tube with energies corresponding to the sulfur fluorescence peak can only reach the detector by an unlikely two-step successive scattering process. Therefore the S/B ratio is drastically improved compared to polychromatic excitation. The other advantages of this unique MWDXRF sulfur analysis system are that is has no moving parts, is very compact, and requires very little maintenance. An ASTM standard, D 7039-04, has been recently designated for this MWDXRF technique. In this paper, data for sulfur analysis in diesel and gasoline using D7039-04 are presented and discussed.

1 18

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Materials and Methods

Samples used in this study included commercially available sulfur in diesel standards, gravimetric samples prepared in-house, NIST SRMs, and refinery diesel and gasoline products. The commercial sulfur diesel standards were gravimetricaUy based and were obtained from AccuStandard in New Haven, CT, and Analytical Service Inc. in Woodlands, TX. The in-house samples were made by mixing n-butyl sulfide (CsH18S) with decalin (C10H18), tetralin (Cl0Hl2), or decane (CloH22). The NIST SRMs included 1616a, 2723a and 2299. Bench top MWDX_RF analyzers were used in this study. Measurements were performed in two different laboratories. Five different bench top instruments were used in the XOS R&D laboratory and one instrument was used for measurements acquired in a refinery lab. For laboratory measurements, samples were measured directly with minimal sample preparation. Prior to analysis, the samples were prepared by placing --4 mL of liquid sample into a Chemplex XRF disposable plastic liquid sample cell fitted with a Chemplex cardboardmounted mylar window (3.6-pza thickness). The analytical volume of the MWDXRF method is small due to the focusing nature of the incoming X-ray beam. The full size of the excitation beam is about 1 m m x 1 mm on the Mylar window and the sulfur X-rays 1/e escape depth for fuel is ~0.1 ram. Therefore, only a small amount of the specimen near the center of the film is analyzed during measurement. The sample cell was vented immediately after sample preparation if gasoline or another volatile sample was used. The Mylar film surface was kept dust free, flat, and wrinkle free near the center area of the sample cell to avoid errors during measurement.
Results and Discussion

Calibration and Linearity


The bench top analyzer was first calibrated using commercial gravimetrically based diesel standards. Two sets of commercial diesel standards and a set of in-house prepared standards were used to generate three different calibration curves for comparison. The first set of standards, made from No. 2 diesel (Vendor A), included sulfur concentration levels of 0 ppm (blank), 5 ppm, 10 ppm, 100 ppm, 500 ppm, 1000 ppm, and 3000 ppm. The second set, made from a synthetic diesel fuel matrix (Vendor B), included sulfur levels of close to 0 ppm (blank), 20 ppm, 50 ppm, 100 ppm, 200 ppm and 500 ppm. The in-house standards were made by mixing different ratios of decalin and n-butyldisulfide (CsHjsS) gravimetrieally with resulting sulfur concentration levels of close to 0 ppm (blank), 10 ppm, 48.6 ppm, 87.6 ppm, 320 ppm, and 946 ppm. The measured sulfur fluorescent X-ray intensity, in counts, for each standard is listed in Table I. Based on the repeatability statement in the ASTM D 7039-04 method, the precision of the MWDXRF method is a function of the concentration and it can be best fit by a square root function. Therefore, a weighted linear least square fit is needed to obtain a proper calibration curve over a wide range of concentration levels. Since X-ray counting is governed by Poisson statistics, and the uncertainty due to Poisson statistics is comparable to the precision statement of the ASTM D 7039-04 method, a satisfactory fitting result can be achieved by using the Poisson fluctuation as a weighted error bar. The linear regression function can be written as: Y = A+B*X, where Y is the X-ray counts obtained in 300 s, X is the concentration, A is the calibration intercept, and B is the slope of the calibration curve. Weighted linear regression

CHEN ET AL. ON MONOCHROMATIC WDXRF

119

results are compared with non-weighted fitting results (Table 2). The regression output includes the intercept, the slope, and the linear correlation probability R. TABLE 1 Sulfur X-ray intensity with corresponding concentration for three sets of calibration standards. No. 2 Diesel Set (VendorA) Concen~afio~ Measured pprn Counts 0.2 301 5 943 10 1 611 100 13 965 500 68 507 1000 136 784 3 000 407 712 SyntheticDiesel Set (Vendor B) Concengafio~ Measured ppm Cotm~ 0.5 291 20 3 042 50 7 186 100 14 020 200 27 772 500 68 464 . . . . . . . In-houseSet, Decalin-based C o n c e n ~ a t i o ~ Measured ppm Counts 0 244 10.0 1 582 48.6 6 905 87.6 1 2334 320 4 4114 946 1 29686 . . . . .

TABLE 2--Comparison of linear regressio n using non-weighted and weighted data. CalibrationSet Intercept, A, Counts Slope, B, Counts/ppm Linear Correlation Probability,R2 Weighted NonWeighted NonWeighted Nonweighted weighted ____weighted No. 2 Diesel Standard 271 13 418 :~132 136.2 0.2 a: 135.8 1.0000 1.0000 Synthetic Diesel Standard 234 19 342 :~57 137.1 0.5 136.4 0.3 0.9994 1.0000 Deealin-basedStandard 242 + 8 273 29 136.9 0.2 136.8 1.0000 1.0000 It is clearly shown that non-weighted linear fit gives high uncertainty of the intercept. This can adversely affect the accuracy o f the low ppm measurement. Particularly for the two commercial calibration standard sets, the biases for the low ppm measurements can be as much as 1 ppm if non-weighted fitting is used. The linearity of the calibration curve is exceptional for all three sets of standard. There is no significant difference among the three calibration curves. The in-house standard has the smallest fitting error. The final calibration for this study was set at A = 242 and B = 136.9. The calibration line with the data points is shown in Fig. 2. The relationship between the X-ray counts and the sulfur concentration is given: W (in ppm) = ((counts/T)-B/300)/(A/300), where T is the measurement time in seconds.

160k-

8k

~'120k"~ 8Ok.
@ ~ 4Ok- S

4k
Ok

i=242+136.9~ w

R=I,O000

Ok 0 2o0 46o soo eoo lobo


Sulfur Concentration W (ppm)
FIG. 2 - - T h e calibration curve for a bench top sulfur analyzer." the open circles are the

calibration data points and the solid line is the linear fit with weighted error bars.

120

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Repeatability and Precision


Repeatability measurements were performed at two sites: the XOS R&D laboratory and a refinery laboratory.

XOS Results
All the measurements were done with the bench top analyzer with the calibration curve shown in Fig. 2. Three bottles of diesel samples with sulfur levels of 5 ppm, 10 ppm, and 300 ppm were measured periodically over 20 days. For each measurement, a specimen was taken from a sample and placed into the bench top analyzer and a reading was obtained. Two successive measurements were performed on each day for each sample. The two successive measurements were done within 15 min. of each other. Forty measurements were obtained in total for each concentration level and the results are shown in Fig. 3. The difference between each pair of measurements, total 60 pairs, is shown in Fig. 4. Single site repeatability can be determined from the data shown in Figs. 3 and 4. The results are listed in (Table 3). The definition of repeat measurement standard deviation is based on ASTM D 6300. There is no significant difference between the paired repeat standard deviation and the standard deviation of the 40 total measurements. This indicates that there is no observable measurement drift over 20 days. The standard deviation of the 40 measurements for the 5-ppm and 10-ppm samples are very close to the Poisson standard deviation based on the Xray photons\counts. Conversely, at high concentrations, the Poisson fluctuation is small, and other instrumental factors become important. Table 3 shows that at 300 ppm the difference between the measured standard deviation and the Poisson standard deviation is significant. EPA recently issues the precision criteria for selecting an ASTM testing method for measuring sulfur in diesel fuel. With at least 20 tests over 20 days, the maximum standard deviations for the selected method are 0.7 ppm and 7 ppm for a 5-15-ppm sulfur diesel fuel and 200-500-ppm sulfur diesel fuel, respectively. The precision obtained above for both low-level and high-level sulfur is well within the EPA precision criteria.

Refinery Site Results


Data were also collected from a different bench top sulfur analyzer at a refinery laboratory over a period of one month. Again, all the measurements were carded out with one calibration curve. For the repeatability evaluation, two diesel products and a gasoline product were used. The gasoline and one of the diesel samples contain ~ 50 ppm sulfur. The other diesel sample has 10 ppm sulfur. For each sample, two successive readings were obtained within 10 rain. from two different specimens by the sample operator. A total of 26, 36, and 32 pairs of results were obtained over a month for the 50-ppm diesel, 10-ppm diesel and the gasoline samples respectively. The difference between each pair was calculated and the results are shown as histograms for the three samples (see Fig. 5). The repeat measurement standard deviation and the site repeatability were found and the results are listed in Table 4.

CHEN ET AL. ON MONOCHROMATIC WDXRF

121

310
e9 9 ~ 9

306

~149
302 0

# 9

t 9 :=
298

* 9

Ca
0

11.5

10.5 9.5

.*t

,t*,.= **e*tt*t

***~

"l
5

4,

9120

9/25

9/30
Date

10/5

10/10

10115

FIG. 3--Twenty pairs of repeats over 20 days for diesel samples with sulfur level at 5 ppm,

l Oppm, and 300 ppm.

FIG. 4---The Histograms for the difference (A) between each pair at 5 ppm, l Oppm, and 300

ppm sulfur level.


TABLE 3--Repeatability of a MWDXP,F bench top analyzer for single site. Diesel C, ppm cr of repeats, ppm Repeatability, ppm a of whole set, ppm Poisson c, ppm 5 0.24 0.71 0.22 0.23 10 0.25 0.74 0.34 0.30 300 1.5 5.3 2.9 1.5

122

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 5--Histograms for the difference between the two values of each pair. TABLE 4--The repeat measurement ,standard deviation and the repeatability at a refinery lab. Sulfur level, ppm RepeattL ppm Repeatability,r, ppm 50, diesel 0.88 2.7 50, gasoline 0.87 2.6 10, diesel 0.46 1.3 The precision o f the analyzer was also evaluated. Two gasoline samples with ~50 ppm and ~10 ppm sulfur and one diesel sample with ~10 ppm sulfur were used for the study. For each measurement, a specimen was obtained from the sample and placed into the analyzer. Each measurement was taken at least 7-h apart. A total of 26-28 measurements were carried out for each sample over the period o f one month. The results for the three samples are shown as histogram plots in Fig. 6. The site precision was determined and the results are listed in Table 5. Again there is no significant difference between the repeatability and the site precision, indicating no observable drift for the instrument. The measurement results were also in good agreement with the XOS data.

FIG. 6--Histogram plot for extensive repeat measurements on 10- and 50-ppm sulfur gasoline, and 1O-ppm sulfur diesel.

CHEN ET AL. ON MONOCHROMATIC WDXRF


TABLE Sulfur level, ppm 50, gasoline 10, gasoline 10, diesel

123

5--Siteprecision for low-sulfur gasoline and diesel.


Mean, ppm 45.8 7.0 8.4 ...... o, pprn 0.96 0.24 0.40 Precision, ppm 2.5 0.6 1.0

Repeatability and Reproducibility for Ultra-low-sulfur Diesel Cross Check Samples


An in-house round robin study was conducted using the ultra-low-sulfur diesel crosscheck samples from the A S T M D2 study group. Five bench top instruments were used with a different operator for each instntment. Calibration standards from different commercial sources were used for the calibration o f these instruments. There were a total o f 8 diesel samples from the crosscheck program. NIST SRM2723a was added to the 8-sample group so a total o f 9 samples were used in the study. Two successive repeat measurements were made for each sample on each analyzer. The measurement results are listed in Table 6. TABLE &-Measured sulfur concentrations (ppm) of the ultra-low-sulfur diesel cross check samples and SRM 2 723a with five analyzers. Analyzer A B C D E ULSD04 03 1.26 1.33 1.70 1.55 1.24 1.40 2.00 1.95 1.40 1.70 ULSD04 04 14.50 13.92 14.33 13.66 13.77 13.33 13.85 14.58 14.60 15.50 ULSD04 05 9,94 9,74 10.94 11.13 10.07 10.55 10.80 10.85 9.60 10.30 Sample ULSD04 ULSD04 06 07 0.39 3.26 0,47 4.11 0.63 4.11 0.68 4.14 0.11 4.09 0.23 4.55 0.80 3.38 0.63 3.68 1.00 3.90 1.10 4.10 ULSD04 08 14,30 14.42 15.16 15.54 16.30 16.18 16.03 16.88 14.00 13.40 ULSD04 09 0.31 0.17 0.69 0.48 0.43 0.23 0.65 0.65 0.60 0.60 ULSD04 010 15.51 15.77 15.42 15.32 15.28 15.52 15.25 13.93 15.80 16.30 SRM 2723a 10.00 9,80 10.50 10.40 10.70 10.40 10.80 9.80 10.10 10.90

Following the guidelines o f A S T M D 6300, repeatability (r) and reproducibility (R) were determined from the data shown in Table 6. The results are given in Table 7 along with the mean o f each sample and the standard deviations o f r and R. The repeatability results show significant improvement over that o f the A S T M D 7039-04. The reproducibility results are similar to that o f inter-laboratory study (ILS) in 2002 [3]. Further interlaboratory round robin studies will be needed to confirm the reproducibility. TABLE 7--Standard deviations, repeatability, and reproducibility for ultra-low-sulfur diesel fuel, in ppm. Sample ID Mean at OR r R U L S D 0 ULSD0 403 404 1.55 14.20 0.12 0.48 0.29 0.64 0.44 1.75 1.05 2.33 ULSD0 405 10.39 0.28 0.56 1.03 2.05 ULSD0 ULSD0 406 407 0.60 3.93 0.08 0.33 0.33 0.40 0.29 1.19 1.21 1.45 ULSD0 408 15.22 0.35 1.21 1.29 4.40 U LSD 0 U LS D 0 SRM 409 4010 2723a 0.48 15.41 10.34 0.10 0.46 0.42 0.20 0.63 0.40 0.38 1.68 1.53 0.72 2.28 1.45

124

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Accuracy and Bias


The accuracy and bias of the analyzer was evaluated using three NIST SRMs and a commercial gravimetric standard. One specimen was taken from each sample and 10 readings were recorded. The mean of the 10 data points was obtained and was compared with the reference value. The comparison is shown in Table 8. Biases for these reference samples were found to be non-significant except for the gasoline sample. TABLE 8 - - Accuracy and bias: comparison of the measured concentrations and the referenced values. Sample Reference Value, ppm Measured Value, ppm Bias,ppm Biassignificance SRM 1616a,kerosene 146.2 149 4-2.50 2.8 n SRM 2723a, diesel 11.0 4- 1.1 10.8 4-0.6 -0.1 n SRM 2299, gasoline 13.6 4-1.5 11.6 4-0.6 -2.0 y Commercialstandard, diesel 300 4- 3.0 303.2 4- 3.0 3.2 n

Matrix Effects
The absorption o f both the excitation beam and the characteristic sulfur X-rays o f the sample is matrix dependent. In general, the matrix effect is minor and there is no need for correction. However, if the sample measured has a high oxygen content or very different hydrogen-tocarbon H/C ratio compared to the calibrant, a correction will be necessary. Due to the monochromatic excitation, the matrix correction is simple if the major composition o f the matrix is known. The primary fluorescence intensity for an element in the matrix can be simply found from the first principle for monochromatic excitation [4]:

Ii(Ei) = ~(Ei)

Qi(Ei)I(E~

Wi

(1)

Co (/4Eo) + c/4Ei))
where rl(E) is detector efficiency, Q(Ei) is the fundamental parameter for element I and I(E) is the excitation beam intensity, Wi is the concentration o f the element by weight, Ix(E) is the average mass absorption coefficient at energy E for the matrix, and Go and G are geometrical factors related to the beam and sample configuration. The term ~M = (~t(Eo)+Gl.t(~)) is the only parameter related to the matrix. The current sulfur analyzer has a Cr Kct excitation beam and the fluorescence element is sulfur. The G factor for the beam configuration in the analyzer design is .--0.87. For diesel and gasoline, the matrix consists o f C, H and O. Assuming their weight percentage as Wc, WH, and Wo respectively, the matrix effect term can be written based on the mass absorption coefficients for C, H, and O at Eo = 5.4keV and E4 = 2.3 keV [5] and is the following: /.t M = 1 9 8 W c + 0.35W,~ + 4 4 0 W o

182Rc.,,0 =

Wo)

o.350-

wo) +

440W o

(2)

(Rc. u + 1)

(Rc, n + 1)

where Rc~ is the C/H mass ratio in the matrix. The matrix correction factor, CM, then can be defined as CM = ~M (specimen)/~m (calibrant), and the correct sulfur X-ray counts of the specimen is Ic=Cu*Ir, where Ir is the recorded sulfur net counts from the analyzer.

CHEN ET AL. ON MONOCHROMATIC WDXRF

125

To validate Eq 2, measurements were carded out for eight gravimetric diesel standards with known C/H ratios to determine the C factor experimentally. The experimental C factor can then be compared with the C factor calculated from Eq 2. Relatively high sulfur concentration samples were used since the relative precision o f the method is better for high concentration. The study results are listed in (Table 9). Measurements on gasoline samples with added ethanol were also performed to validate the effect o f oxygen. The results are also given in Table 9. Please keep in mind, decalin is the matrix for the calibration material and the correction factor for decalin samples is 1. TABLE 9--Effects of C, H, and 0 C/H ratio, Wo Sulfur Concentration RC,H W, ppm 6.18 0 300 5.51 Decane 5.42 Mixture of Decane and Decalin Decalin Mixture of Decalin and Tetralin Tetralin Gasoline Gasoline + 3.31% ethanol Gasoline + 6.71% ethanol Gasoline + 8.30 % ethanol 6.03 6.67 8.09 9.93 6.56 6.68 6.81 6.85 0 0 0 0 0 0 0.0115 0.0233 0.0288 951.5 950.55 946.1 948.5 951.5 205 195.4 188.4 185.2 130502 130372 129763 130091 130502 28312 27384 26294 25966 135731 133464 129694 128036 125231 28032 26768 25331 24706 0.961 0.977 1.000 1.016 1.042 1.005 1.018 1.033 1.046 0 146

composition.
Expected Counts, I~ 41312 20229 Recorded Counts, Ir 41720 20708 CM= IJ Ir 0.991 0.977

Sample Diesel, Kerosene, SRM1616a

The experimental values o f CM are plot vs. RC,Hat different oxygen levels and compared with the theoretical value from Eq 2 (See Fig. 7). The experimental results follow the theoretical curves quite well. Therefore the theoretical curve can be used as a general guideline for practical use. The C/H mass ratio is in the range o f 5.3-6.7 for diesel products and 5.7-7.3 for gasoline. I f the synthetic diesel calibration set (commercial standards with C/H mass ratio o f 6.18) is used, the biases due to the matrix effect will be no more than 2 % in the extreme case. The biases will be insignificant for low-sulfitr diesel and gasoline, and there is no need for matrix correction. For gasoline with oxygenates, the matrix correction will also be insignificant, provided the oxygen content is less than 2 %, which corresponds to no more than ~ 1 1 % o f MTBE or ~6 % o f ethanol.

Limits of Detection
The minimum limits o f detection o f an analyzer can be determined b y measuring the standard deviation o f the background. Decalin was used for background measurement. Twenty repeat measurements were carried out on this sample and the measured X-ray intensities are given in

126

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Fig. 8. Each measurement was done with a 300s counting time. The average X-ray counts in 300s for the background is B = 241 counts with standard deviation o f era = 16.9 counts. From the calibration curve in Fig. 2, the background standard deviation can be written in terms o f p p m : era = 16.9/A = 0.12 ppm. Therefore the minimum detection limit is given as: L O D = 3or B = 0 . 3 6 p p m

1.10.

Z.9%

Tg

1.1[~ 9

1,00/ / ~I~ 1 9
, oe*e.os, oo eo Mixture of decane, deealin and tetrstin

0.95. o 13 0.90

J J

J.

9 Gasoline+ ethanol,1.2%Oxygen 9 Gasoline, ethanol no O Diesel Kerosene C/H Mum


Ratio

FIG. 7--Matrix effects: measured correction factor and the theoretical correction curves f o r C/H mass ratio and total oxygen wt. %
300-

Mean = 241, u.=16.9

9 9

9 am n 9

W 9

~ U

240.

200too
i i i i

10

111

20

observalion

FIG. 8 - - M e a s u r e d background X-ray counts.

CHEN ET AL. ON MONOCHROMATIC WDXRF

127

Conclusion

Compact bench top analyzers have been successfully produced for low-sulfur determination in fuel. A new ASTM method, D 7039-04, was the basis for the study in this paper. The bench top analyzer provides a wide dynamic range for sulfur measurements. The linearity of the technique is excellent and it holds from 0-3000 ppm. During the calibration, linear regression with a weighted error bar is necessary for obtaining good accuracy for low ppm sulfur measurements. The repeatability for 10-ppm sulfur in fuel was shown to be around 1 ppm. The PLOQ for ultra-low-sulfur diesel was found to be less than 1.5 ppm in this study. The precision of an analyzer over one month without further calibration remains the same as that of repeat measurements taken over very short intervals, such as minutes. The reproducibility of 15-ppm sulfur diesel fuel was determined to be better than 3 ppm (95 % confident level). The limit of detection for a single analyzer was found to be 0.36 ppm. The correction factor due to matrix composition was measured and compared to the theoretical curve. The correction factor predicted by the theoretical curved is accurate within the measurement error. The correction due to the matrix for low-sulfur diesel and gasoline is generally not significant.

Acknowledgments
We would like to acknowledge the assistance of David Aloisi and Dr. Tomasin Miller with measurements.
References

[1] Chen, Z. and Gibson, W. M., "Doubly Curved Crystal (DCC) X-ray Optics and Applications," Powder Diffraction, Vol. 17, No. 2, 2002, p. 99. [2] Chen, Z. W. and Wittry D. B., "Microprobe X-ray Fluorescence with the Use of Pointfocusing Difffactors," Applied Physics Letters, Vol. 71, No. 13, 1997, p. 1884. [3] Lubeck, A. J., ASTM Research Report RR:D02-1552 [4] Jenkins, R., Gould, R.W. and Gedcke, D., Quantitative X-ray Spectrometry, Marcel Dekk~r, New York, 1995. [5] Goldstein, J. I., Newbury, D. E., Echlin, P., Joy, D. C., Roming, A. D., Jr. Lyman, C. E., Fiori, C., and Lifshin, E., Scanning Electron Microscopy and X-ray Microanalysis, Plenum Press, New York, 1992.

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JAI12975 Available online at www.astm.org
Dirk Wissmann I

Latest Improvements on Using Polarized X-Ray Excitation EDXRF for the Analysis of Low Sulfur Content in Automotive Fuel
ABSTRACT: Polarized X-Ray EDXRF has been used for the analysis of low levels of a broad range of elements in light matrices for a long time. By using the polarization of the excitation radiation, the background scattered by the sample is drasticallyreduced. For the analysis of low sulfur in automotive fuels, this meant a detectionlimit comparableto the WDXRFtechnique. Recent developments in detector technology and in closed coupled static geometry result in further improvementof sensitivityfor this application. This paper shows the developmentof this technique for the analysis of low sulfur content in petroleum products. It shows the results of a roundrobin test, performedby CEN TC 19 WG 27 in 2001. The technicalbasis of these data is compared to the recent technicaldevelopments,and future options are shown. The influence by sample preparation and sample presentation, which might be special for this analyticaltechnique,is investigatedand discussed. KEYWORDS: EDXRF,polarization,sulfur, automotivefuels Introduction XRF basically operates always in the way that the excitation radiation is transferred into the sample, where fluorescence radiation is produced. An energy dispersive detection system allows the separation of this fluorescence radiation by energy and the counting of the intensity collected by energy. If an X-ray tube is used as a source for the excitation, the radiation can be modified by using collimators, filters, secondary targets, crystals, and focusing optics. The applied detection systems can be characterized by the energy resolution, the maximum throughput, the peak-to-background ratio, and other parameters. Excitation Geometry When the radiation of the X-ray tube is directly transferred to the sample (maybe optimized by collimators and filters), this is called direct excitation. Figure 1 shows such a setup. If a direct excitation geometry is used for the analysis of automotive fuels, a Ti X-ray tube can be used. Typically those tubes use a power o f maximum 30 W. Figure 2 shows an example spectrum for a sample with a sulfur content of 100 mg/kg. The peak-to-background ratio achieved in the example from Fig. 2 is 2.5:1. Manuscriptreceived 8 October 2004; accepted for publication 11 April 2005; publishedOctober2005. Presentedat ASTM Symposiumon ElementalAnalysisof Fuels and Lubricants:Recent Advancesand FutureProspects on 6-8 December2004 in Tampa,FL. Spectro AnalyticalInstruments,160 AuthorityDr., Fitchburg,MA 01420. Copyright9 2005byASTMInternational, BarrHarborDrive,POBoxC700,WestConshohocken, 19428~ I00 PA 128 Copyright*2005 by ASTM International

WISSMANN ON POLARIZED X-RAY EXCITATION EDXRF

129

FIG. 1--Direct excitation geometry.

FIG. 2--Spectrum of a sample with a sulfur content of l O0 mg/kg using direct excitation; the Ti escape line appears at the displayed energy in combination with a Si based detector. Polarization can be implemented in this geometry by using a crystal, which reflects and polarizes the excitation radiation. The reflection by a crystal creates a monochromatic excitation radiation with an energy that can be close to the absorption edge of the dement of interest. With this reflection the radiation is also polarized, and by using a Cartesian geometry the background, which is created by scattered radiation, can be reduced [3]. Figure 3 shows the excitation setup. If the X-ray tube power is increased to 400 W and a Pd target tube is used in the geometry shown in Fig. 3 together with a HOPG (highly oriented pyrolithic graphite) crystal, Pd L-lines can be used for the excitation. This will lead to an improvement in the sensitivity, which can be seen in Fig. 4, where a spectrum, measured with a sample with sulfur content of 100 mg/kg, was analyzed. The peak-to-background ratio is improved to 3.7:1; the intensity of the S fluorescence line is increased by a factor o f - 3 in comparison to the example of direct excitation from above.

130

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 3--Polarization geometry.

FIG. 4---Spectrum, taken in polarization geometry (blue), in comparison to direct excitation (red). One of the limitations of such a setup certainly is the use of a higher power in the X-ray tube in comparison to a direct excitation setup. This is caused by losses due to longer paths in the excitation as well as by losses in the reflection using the crystal. Using a fine focus X-Ray tube helps compensate these effects. The excitation radiation of such a tube can be transmitted optimally to the sample by X-ray lenses or X-ray mirrors. As a mirror for X-rays, again a crystal can be used, which in this case is concave-shaped. The reflected radiation is polarized [2,4,5]. Figure 5 shows this in a linear plane schematic. By reflecting and focusing the X-ray beam, a higher flux of excitation radiation is transmitted to the sample. This increases the sensitivity, and lower power X-ray (less than 50 W) tubes can be utilized. A spectrum with a sulfur in oil sample at a sulfur concentration of 100 mg/kg is displayed in Fig. 6. The peak-to-background ratio is improved to 5.5:1, and the intensity of the sulfur fluorescence line is increased by a factor of -2 in comparison to the example of polarization excitation and by a factor o f - 6 in comparison to the example of direct excitation from above. Additional improvements in this geometry in the future can be seen in using differently shaped mirrors and shorter coupling of tube, sample, and detector.

WISSMANN ON POLARIZED X-RAY EXCITATION EDXRF

131

FIG. 5--Polarization geometry using curved crystal with fine focus X-ray tube.

FIG. 6--Spectra comparison focusing polarization geometry (black), polarization geometry (blue), direct excitation (red).
Detection

The detection of the fluorescence X-rays should be done by using a detection system with a spectral resolution enabling the easy separation between the excitation radiation (in this case energy of Pd-L-lines) and the fluorescence radiation of sulfur. Detection systems such as Si(Li) detectors and silicon drift detectors were found to be suitable for this application. They typically work in a range of 150-175 eV (at Mn K-alpha radiation). With this spectral resolution it is even possible to analyze low contents of chlorine as well using the same excitation. Figure 7 shows spectra of a blank sample (displayed in red), a second sample containing sulfur 100 mg/kg (displayed in blue), and a third sample containing 100 mg/kg chlorine (displayed in black). The fluorescence lines can be clearly separated, and even a concentration of 100 mg/kg has no significant effect on the background of the sulfur fluorescence line. The major differences in the choice of a detection system for this application between Si(Li) and Silicon drift detector is the Peak-to-Background (P/B) ratio, which typically is better using a Si(Li) detection system. Future developments in the SDD technology show that P/B ratios are improving.

132

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 7--Spectra of a blank sample (red), a second sample containing 100 mg/kg sulfur (blue), and a third sample containing 100 mg/kg chlorine (black).

Sample Preparation and Influence of Film


Liquid samples in all cases are analyzed by preparing a sample cell for liquid samples. This sample cell is closed at the analytical end by a film, which should be as transparent as possible for the excitation and fluorescence radiation and should be robust, not destroyed by the sample itself. Various types of film are available with various thicknesses. Also, impurities, which might be inside the film, must be considered, if they are at higher level or in varying concentrations. Table 1 shows the transmission of sulfur fluorescence radiation relative to the most transparent film for sulfur. TABLE 1--Transparency for Sulfur K radiation and impurities for different types offilm. X-Ray Film Thickness Transparency Impurities Mylar 1.5 ~tm 100 % Si, (P) Mylar 2.5 ~tm 94 % P, Ca Mylar 3.5 ~tm 87 % P, Ca Mylar 6.0 ~tm 75 % P, Ca Polypropylene 4.0 ~tm 96 % ... Polypropylene 6.0 ~tm 93 % ... Polycarbonate 4.0 ~tm 86 % ... Polycarbonate 5.0 ~tm 81% ... Hostaphan 3.5 ~tm 83 % Si, (P) In all cases of XRF analysis, a part of the excitation radiation will be scattered by the sample, including the X-ray film used. This effect certainly is smaller when using polarized excitation radiation but still can be seen because the degree of polarization typically is only 90 %. Figure 8 shows the Cartesian geometry with polarized excitation radiation.

WISSMANN ON POLARIZED X-RAY EXCITATION EDXRF

133

The influence by the scattering from the X-ray film depends on its material and thickness. The scattering of the excitation radiation will influence the intensity of the Pd-L-lines in the measured spectrum if a Pd- X-ray tube is used for the excitation. This again will have an impact on the background in the spectral region of the S fluorescence radiation, and by this it influences the achievable detection limit. Table 2 shows the background intensity in that spectral region for various X-ray films relative to the highest intensity measured using a blank sample.

FIG. 8--Principle setup of a polarization excitation system in Cartesian geometry.

TABLE 2--Background intensity at Sulfur K radiation for different types offilm, determined from the measurement of a blank sample. X-Ray Film Hostaphan Mylar Mylar Mylar Polycarbonate Polycarbonate ..... Polypropylene ..... Ttlickness 3.5 xm 3.5 lm 2.5 lm 1.5 xm 5.0 lm 4.0 lm 4.0 lm Backgroun~l , 100 % 91% 83 % 64 % 64 % 39% 40 %

In addition, the stability of the X-ray film is very important for a safe operation of the analyzer. For the analysis of sulfur in automotive fuel, two films are used, one for closing the sample cup itself and the second as a safety window in the holder for the sample cup. The use of the safety window with an additional film has become a standard technique, as this gives additional safety in the case of a leaking sample cup. Figure 9 shows the sample cell with the Xray film together with the sample cup and its safety window. The most common and most appropriate technique is to use Polycarbonate film for the sample cup and for the safety window.

134

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 9--Image of sample cup with X-ray film and sample cup with safety window.

Analysis Results
In 2001 CEN TC19 WG27 [1] performed a round robin test for the analysis of low sulfur levels in fuel, where 102 laboratories participated in the test. Out of these, five laboratories have used EDXRF instruments (from one manufacturer) using polarized X-rays for excitation. Tables 3 and 4 show a comparison of the achieved analysis results for diesel and petrol samples in comparison to those results achieved with WDXRF instruments with and without internal standard. In the evaluation of the round robin test the reproducibility r and the repeatability R were calculated for various analytical techniques. WDXRF was differentiated between low power (LP) and high power (HP) instruments with both methods with and without using internal standard. Tables 5 and 6 show the comparison for diesel and petrol samples. TABLE 3--Analytical results of diesel samples using polarization EDXRF versus WDXR (high power (> 3 kW) systems only). Ident. QCG OG PG RG SG TG VG WG ZG WD-XRF No IS 31.48 4.47 52.57 21.39 10.33 40.96 36.48 8.11 8.44 WD-XRF IS 32.78 4.79 54.13 22.15 10.74 42.21 37.51 8.46 8.80 EDXRF Polarized 32.44 5.41 53.75 22.49 11.15 42.48 37.73 9.16 8.73

WlSSMANN ON POLARIZED X-RAY EXCITATION EDXRF

135

TABLE 4---Analytical results of petrol samples using polarization EDXRF versus }VDXRF (high power (>3 kW) systems only). Ident. QCP AP BP CP DP EP FP GP WD-XRF No IS 20.65 15.34 38.49 1.23 22.30 6.44 56.32 44.16 WD-XRF IS 21.22 16.27 40.17 1.50 23.87 6.94 58.47 45.82 EDXRF Polarized 21.11 15.37 40.00 0,98 23.11 6.81 59.18 45.81

TABLE 5--Reproducibility r and the repeatability R for diesel samples at 10 and 30 mg/kg sulfur using different analytical techniques. r 10 1.85 2.46 7.38 1.59 1.92 2.22 1.15 1.69 r 30 2.59 3.28 7.95 1.59 2.16 2.71 2.21 2.36 R 10 3.33 3.97 10.4 2.14 2.99 5.19 2.24 3.29 R 30 4.67 5.29 11.2 2.14 4.97 6.34 4,48 4.60

WD-XRF IS HP WD-XRF IS LP ED-XRF non polarized E D - X R polarized WD-XRF HP WD-XRF LP UV Fluorescence Microcoulometry

TABLE 6---Reproducibility r and the repeatability R for petrol samples at 10 and 30 mg&g sulfur using different analytical techniques. r r R R WD-XRF IS HP WD-XRF IS LP E D - X R non polarized ED-XRF polarized WD-XRF HP WD-XRF LP UV Fluorescence Microcoulometry 10 2.12 3.77 6.29 2.00 1.92 3.24 0.98 1.32 30 3.08 3.77 7.87 3.00 2.57 4.32 2.24 2.63 10 3.54 5.96 10.05 2.28 3.45 5.93 2.71 3.58 30 4.95 5.96 12.56 3.43 5.74 7.90 6.21 7.17

In one laboratory the long-term stability over a period of two weeks was tested with a diesel sample containing sulfur close to 10 mg/kg. The measurements were performed in duplicates on nine different days. An average concentration of 9.63 mg/kg was analyzed. The standard deviation is calculated as 0.55 mg/kg. The results are displayed in Fig. 10.

136

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. l O---Long-term repeatabilityfor a diesel sample with a sulfur content close to 10 mg/kg. Summary and Outlook In the last few years EDXRF using polarized and focused excitation has become a very powerful analytical technique for the analysis of low S content in automotive fuel. The technique shows the potential to achieve repeatability and reproducibility values, which are comparable to reference methods for this application, In preparation for a new ASTM and EI norm, a new round robin with a larger number of participants is under way. The future for this technique will involve improved crystal optics for improved sensitivity and new detection systems with improved peak-to-background ratios helping to achieve even lower detection limits. References [1] CEN TC 19 WG 27 2002, Test Methods for the Determination of Sulfur Content, Round Robin report 2001. [2] Grigorieva, I. G. and Antonov, A. A., "HOPG as Powerful X-Ray Optics," X-Ray Spectrometry, Vol. 32, No. 1, January/February 2003, pp. 64-68. [3] Heckel J., "Bragg and Barkla Polarization in EDXRF," EDXRS Conference, Lisboa, 1996. [4] Heekel J., "Bragg Polarization Optics in EDXRF," 49 th Annual Denver X-ray Conference, 2000. [5] Ryon, R.W. and Heckel, J., "Polarized Beam X-ray Fluorescence Analysis," Handbook of Xray Spectrometry, 2nd ed., R. E. Van Grieken and A. A. Markowicz, Eds., Marcel Dekker Inc., New York, 2002, pp. 603-630.

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JA112980 Available online at www.astm.org

Steve Tarkanic I and John Cl'nko 1

Rapid Determination of Sulfur in Liquid Hydrocarbons for AtLine Process Applications Using Combustion/Oxidation and UV-Fluorescence Detection
ABSTRACT: With the reductionin sulfurlevelsin gasolineand dieselproducts,itisnecessaryto have a method availablewhich can preciselyand accuratelymeasure the very low levelsof sulfurin thesefuels. A S T M testmethod D 5453, Sulfurby Combustion and Ultraviolet Fluorescence(UVF), is a laboratorybased technologythathas been widely used in the industryfor such purposes for more than a decade.An on-line/at-process versionof thistechniqueis describedwhich can complete a fastanalysisin lessthan one minute.Thus, near real-timedata can be obtainedwhen refineries terminalsneed to make rapid and determinationsof sulfurlevelsfor process controland pipelinedistribution decisionsinvolvingthe new generation"clean"hydrocarbon fuels.Particularly discussedisthe advantage gained for decisionmaking involvingdetectionof transportboundariesand relatedcustodytransferissues.

KEYWORDS: rapid sulfur analysis,gasoline analysis,diesel analysis,ultra low sulfur in diesel,


combustion/oxidation UV-fluoresccnccdetection, on-line/at fast process sulfurdetermination

Contemporary Challenges for Sulfur Determination


Sulfur levels in gasoline and diesel fuels have been decreasing steadily over the last several years, primarily in response to government regulations both in the U.S. and in Western Europe. For example, U.S. Environmental Protection Agency's (EPA) Tier 2 emission standards mandate sulfur level reduction from current 330 and 550 mg/kg sulfur in gasoline and diesel, respectively, to 30 and 15 mg/kg by 2006 and near zero by 2010 and later years [I-3]. Regulations in California are even more restrictive. Similarly, E N 590 European diesel specification mandates sulfur content of diesels from 350 mg/kg m a x i m u m in 2000 to 10 mg/kg m a x i m u m by January l, 2005 in European Union countries. These mandated sulfur reductions, coupled with other regulatory driven specification changes, such as fuel volatility and oxygen content, have spurred large capital and operational cost increases for a wide range of refinery and downstream (distribution)business models. In fact, a major time compression in the entire Well-to-Wheel dynamic is evolving as the oil industry invests in projects ranging from installation of sulfur reduction/removal processes to significant "remodeling" of final product blending, storage, and transportation schemes. Strong product demand, industry capacity limitations, crude oil price fluctuations, and the need to realize a timely return on investment has oil industry administrative, operational, and engineering personnel realizing the need for all types of instantaneous or near real-time data. The benefits of very rapid and accurate product quality data can now, more than ever, bring value to the entire petroleum and petrochemical industry.

Manuscriptreceived2 December 2004; acceptedforpublication21 March 2005; publishedOctober2005. PresentedatA S T M Symposium on ElementalAnalysisof Fuels and Lubricants:Recent Advances and Future Prospectson 6-8 December 2004 in Tampa, FL. iAntek Instntments/PAC,300 Bammel WestfieldRoad, Houston,TX 77090.
Copyright9

2005 by ASTM International, 100 BaITHarbor Drive, PO Box C700, West Conshohoekcn, PA 19428-2959. 137

138

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

As the new "clean" fuels enter the marketplace, both refinery process control and maintenance of product quality during distribution are crucial. Availability of near real time sulfur data, particularly in downstream applications, can have significant value when time critical decisions must be made for a custody transfer and for detecting transport boundaries of similar fuel streams flowing through a pipeline system. One of the parameters historically used for determining product transition in a pipeline is API gravity. However, there is not necessarily a correlation between the gravity of a fuel and its sulfur content. When an ultra-low sulfur fuel and "other" similar fuel streams with higher sulfur levels are being transported through the same pipeline, in order to minimize contamination and loss, it is critical to be able to identify a change of sulfur levels in the streams with as little delay as possible. Ideally the change in sulfur concentration is reported at the boundary between the materials. This interface between different product streams is called trans-mix region, and pipeline companies strive to keep this region to a minimum. Depending on factors such as pipeline flow rate and speed of analysis, the amount of cross-contamination and waste which can occur as the two streams are directed to separate destinations can vary. Often, in an effort to minimize the chance for contamination occurrence, larger volumes of material near the interface region of the two streams are diverted or cut to special trans-mix holding vessels. Test methods that might be used for sulfur determination applications in fuels include techniques such as lead acetate paper tape, oxidative combustion followed by gas chromatographic separation for flame photometric detection [4], and energy dispersive X-ray fluorescence with coaxial proportional counter detectors [5]. The first of these two methods was recently issued as a new standard D 7041 by ASTM. However, these and other commercially available analyzers, such conventional on-line/at process sulfur by combustion and UVF analyzers typically requirr analytical cycle times of 4-10 rain. This delay in reporting the sulfur concentration levels limits the real time detection capability of changes in the sulfur concentration of fast moving or rapidly changing transport or process streams. A frequent, very fast, and accurate determination of sulfur contamination contained in the hydrocarbon materials being transported in a pipeline or involved in a controlled blending situation can facilitate more precise material management and significant reductions in waste. This paper will describe such a rapid technique for on-line/at process determination of ultra-low levels of sulfur in hydrocarbon engine fuel streams based on the D 5453 oxidative combustion UV-fluorescence technique. The described technology can continuously determine and report ultra low levels of sulfur at a rate of once per minute.

Analytical Laboratory Techniques for Sulfur Determination


There are about 20 laboratory based ASTM standard test methods available for the determination of sulfur in various petroleum products and lubricant samples [6]. These utilize diverse analytical techniques and have applicability range spanning from m% to low mg/kg levels. However, at the very low end of sulfur analysis there are only three or four test methods which can adequately determine sulfur in such fuels. These lab-based standard test methods include ASTM D 2622 - wavelength dispersive X-ray fluorescence, D 3120 - oxidative microcoulometry, D 5453 - combustion UV-fluorescence, and D 6920 - oxidative combustion electrochemical detection methods. Without a doubt, the most widely used two methods out of these in oil industry laboratories are D 2622 and D 5453. Studies have shown that at truly ultralow levels of sulfur only D 5453 can deliver accurate and precise results. This conclusion has

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

139

been corroborated in a Southwest Research Institute study [7] and an extensive cross check undertaken by a CEC work group in Europe [8].
ASTM D 5453 Test Method

The D 5453 method "Standard Test Method for Determination of Total Sulfur in Light Hydrocarbons, Motor Fuels and Motor Oils by Ultraviolet Fluorescence" has been widely used in the industry laboratories for over 15 years with a great deal of success. It is one of the preferred methods for such analysis approved by the California Air Resources Board (CARB) and recognized by the U.S. EPA. In a major study conducted by ASTM, it produced the best repeatability, reproducibility, and pooled limit of quantitation compared with D 2622, D 3120, or D 6920 test methods [9]. In this standard test method, a sample is combusted in an oxygen-rich environment at about 1050~ All sample components are completely oxidized, with various sulfur species in the sample being converted to sulfur dioxide (SO2). Hydrocarbons are converted to water vapor and carbon dioxide. The water vapor produced during the combustion is removed with a membrane dryer, and the gases are exposed to short wavelength ultraviolet light. The sulfur dioxide in the gas absorbs the energy from the ultraviolet light and is converted to excited SO2. The fluorescence emitted from the excited SO2 as it returns to a stable state is detected by a photornultiplier tube. The resulting signal is a measure of the sulfur concentration in the original sample. R-S + 02 --" 1050~ ~ SO2 + Combustion Products S02 + Xv -,. S0"2 ---- S02 + Xv' The complete vigorous combustion (oxidation) of the entire sample and the elimination of the resulting water byproduct virtually eliminates interferences. The method has been used for the analysis of a wide variety of liquid petroleum products such as naphtha, distillates, oils, ethanol, fatty acid methyl ester, gasoline, oxygen enriched gasoline, reformulated gasoline, diesel, biodiesel, and jet fuel for sulfur levels as low as 250 gg/kg. A method based on the same technique and used for the analysis of gaseous hydrocarbons and liquefied petroleum gases has been designated ASTM D 6667.
The Lab to On-Line/At Process Environment Transition

On-line/at process instrumentation based on the aforementioned ASTM D 5453 technology has been widely used in oil industry and petrochemical processes for over 10 years. Applications have ranged from feedstock analysis and catalyst protection to quality control of finished fuels and other chemical products. Common problems encountered whenever a technology is moved from the laboratory into the process environment include development of effective sample collection, treatment, and handling mechanisms. Once the sample is effectively isolated and prepared, it must promptly and properly be placed into the analyzer. The apparatus needed to execute the analytical technology must be appropriately packaged and made safe for operation in a variety of environments. Overall instrument ruggedness and ease o f maintenance and downtime frequency issues are often constraints. Data are typically displayed locally and transmitted over large distances so they can be used for process product quality control purposes. Figure 1 shows a very basic block diagram applicable for most on-line analyzers. Sample

140

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

collection and conditioning, sample injection, packaging of the technology apparatus, and data handling are depicted.

Sample Collectio~n end , Conditioning

~[ Sample -1 Introduclion

Combuslion ~ and Sulfur Detection

Quantifcation and Reporting

FIG. 1--Basic block diagram of a typical on-line/at process analyzer.


On-Line/At Process Sulfur Determination by Combustion and UVF

As the laboratory based sulfur by combustion and UVF technology was placed in the at process application environment, the associated design constraints and operational concerns of any high temperature analysis technique were realized. Application specific solutions were developed for concerns ranging from specialized flammable sample handling of multiple streams to various explosion proof apparatus enclosure requirements. Figures 2 and 3 illustrate a typical at process sulfur by combustion and UVF analyzer.

FIG. 2---Conventional sulfur by combustion and UVF on-line/at process instrument: Note the man-machine-interface (computer screen), metal safety enclosure, and purge system.

FIG. 3----An internal view of a typical on-line/at process sulfur by combustion and UVF analyzer: Note separation oJ furnace section (top) and electronics and detector enclosure (bottom).

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

141

Operations of Typical On-Line~AtProcess Sulfur by Combustion and UVF Analyzer


As with most on-line analyzers, a non-static (flowing) material of interest is supplied from and returned to its source through a flow path loop designed for the sample conditioning system. Velocity or flow rate through this commonly called "fast" loop is set so that material presented to the sample conditioning panel is representative of the process being monitored and can be correlated to the time of sample aliquot acquisition (injection) and the corresponding result. Once the fast loop presents the material of interest, the sample conditioning panel filters a portion of this fast loop flow to an acceptable particle size. Pneumatic inputs are then used to select either the filtered process stream, a calibration standard, or a validation material. The selected sample from this sample conditioning system is then flowed through a sample injection valve. At the point of injection, a fixed volume of the sample is inserted into the cartier gas flow and introduced into the analyzer through a small volume injection tube. Once the sample enters the high temperature combustion zone (fumace), the sample is analyzed and sulfur content is measured in the same manner described earlier for the laboratory based D 5453 test method. Manual controls or another man-machine interface allows the user to operate the analyzer, manipulate analysis parameters, and input the concentration values for calibration materials. The detector output is integrated to produce a value that is proportional to the total amount of sulfur that flowed through the detection chamber during the analysis. This proportional value is compared to the value produced by a standard to determine its concentration. This value can be corrected for the density of the standard or sample being analyzed. The resultant concentration can then be locally displayed, written to a register for use on a serial bus, and/or used to set a representative value on a 4-20 milliamp (mA) current loop.

Design Challenges for a High Speed Sulfur Analyzer for Hydrocarbons


Significant advances in sample introduction, combustion technique, and the handling of combustion byproducts have been incorporated into the basic on-line Sulfur by UV Fluorescence analyzer. The implementation of these new techniques has yielded a continuous rapid sulfur measurement capability that is based upon the reliable and well-established D 5453 standard test method. Figure 4 describes a typical high speed sulfur analyzer.
F'dbdon
and

10Q0.1U60~C

UV Light

Density

Cultbn~n Slmdnnl

VdidJon C m ~ n r 91andml Ou

Cmnbus~n

Gae

High Vdlqe

FIG. 4---Block diagram of typical high speed sulfur by combustion and UVF in an on-line~at process application.

A Summary of Enabling Techniquesfor a High Speed Sulfur Determination


A central issue for enabling rapid response for any at process instrumentation is to ensure that the analyzer is promptly supplied with a sample that corresponds to the process being monitored. Two issues emerge: elimination of sample discrimination and lag time. Materials

142

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

obtained for analysis must represent a true random example of the substance of interest and must be presented in a prompt reproducible timeframe that can be correlated to the system being monitored and the frequency of result. A properly designed sample conditioning system (SCS) is critical for reliable high speed sulfur measurements. The main function of the SCS is to prepare a representative process sample for introduction into the analyzer with as little delay as possible and to select what is sent to the analyzer for analysis or calibration. A correctly designed SCS will remove particulates and water, regulate the sample pressure, and control the sample flow to ensure sample introduction under constant conditions. A fast action injection valve constructed of materials compatible with the material being analyzed is essential. This injection valve precisely inserts the sample material into a carrier gas stream that transports the sample into the analyzer. A suitable injection valve will also be compatible with system pressures. All gas flow rates must be of sufficient volume to ensure that the sample and its combustion products are passed through the system in a timeframe that allows the analyzer to promptly return to baseline conditions prior to the next injection. Favorable gas flow rates will eliminate or minimize sample valve "wetting," ensure sufficient sample combustion, provide sharp Gaussian peak characteristics, and enhance detector response. A specially designed combustion tube allows the sample to be injected at much higher carrier flow rates than can be achieved with the standard combustion tube design. The sample inlet and furnace mixing zone promote complete sample oxidation. A proper combustion tube is typically constructed of quartz. A controlled zone of heating is applied to the combustion product's flow path and to the sulfur detector reaction chamber. Water vapor produced by the high temperature oxidation of each hydrocarbon sample injection is maintained in the vaporous state throughout detection and measurement of SO2 fluorescence. Appropriate heating prevents any water vapor condensation, thereby eliminating SO2 solubility and any problems with water buildup in the analytical flow path.
Materials and Methods

Most work discussed here regarding standard Antek on-line/at process analyzers, laboratory based equipment, and the High Speed Sulfur (HSS) on-line analyzer was carried out in the PAC Houston R&D Laboratory and Production facility. Other data were collected from various industry and public sources and are so identified in the associated text and graphics. Various solvents, sulfur sources, and standard reference materials were used in association with both historical and new data. Solvents and chemicals, such as iso-octane, toluene, and dibutyl sulfide, were of typically reagent grade purity and were obtained from readily available commercial sources, such as Aldrich chemicals. When appropriate, these materials were analyzed before
use.

In all cases the various materials analyzed by both the standard and HSS on-line analyzers used injection valves and typical gas flows and pressures found in end-use applications. In some instances, sample flow rates to the injection valve were slowed in order to preserve limited supplies of costly calibration materials or limited volumes of sample. Validation criteria for the HSS instruments included meeting or surpassing the performance established by the standard Antek on-line analyzers (6000/6200 series). Where possible, comparisons were also made to D 5453 lab instrumentation results. Due to the large number of

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

143

data collected in some experiments (large number of individual injections), some results are documented graphically. The on-line/at process industry often describes precision and linearity as a percentage of the full scale concentration. Full scale, for example, is the maximum concentration that is expected to be analyzed or the concentration that is equivalent to 20 mA on the 4-20 mA current loop. When used in this work, the term "% RSD of full scale" is the standard deviation for the results of a sam_pie analysis compared to the full scale value of the range that the analyzer is set for. Standard deviation as a percent of full scale indicates the maximum deviation that can be expected on a sample of any concentration that is within the analysis range. Because % RSD is dependent on the concentration being analyzed, it would have to be determined for each concentration in the range. Percent of full scale is a convenient way to define the precision of an analyzer that allows the maximum anticipated RSD to be inferred for any concentration in the range. See Calculation 1: Calculation 1 - (% Full Scale) X (Full Scale Value) / (A Specific Concentration) =

Maximum Expected % RSD


Results and Discussion

Minimizing or Eliminating Instrument Lag


In on-line/at process analysis, rapid indication of changes in sulfur concentration depends upon proper "fast loop" / sample conditioning design and the ability of the analyzer to produce a near immediate shift in detector response when a material with a different sulfur concentration is first presented to the analyzer. A delay or lag in the shift to a new level of detector response caused by residua from previous samples or by poor detector response can bias results. A good benchmark for demonstrating that an analyzer system does not suffer from lag or memory effects is to determine the system's capability to report very different sulfur concentrations within two injections following a significant sulfur concentration range. Figure 5 shows a typical response to sudden large sulfur concentration changes for the standard 6000/6200 on-line analyzers.
Rapid Response of Sulfur in Diesel When Switching from 1800 to 400 ppm 2000 GasOil 2 ~ 1500 C~

E 1000" 500' 9
0 "111 IIIIIIIIIIIIIIIIIIIIII

,wwwo-r
GasOil 1
IIIIIIIIIIIIIIII1||

16

26

36

46

45

Injections - 1 minute Analysis

FIG. 5--Rapid response of sulfur in dieselfuel when switchingfrom 1800ppm to 400 ppm.

144

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Figure 5 is an example of the analyzer response for sulfur in gas oil/diesel samples using standard Model 6000/6200 analyzers when switching from -1800 mg/kg to -400 mg/kg sulfur levels. The circled data points indicate the first injection after the sample was switched from high to low concentration. Injection rates were at five-minute intervals. Successful execution of a continuous high speed sulfur determination at injection rates of once per minute reduces the time necessary to recover from memory effect. The memory effect when changing from a high concentration sample to a low concentration sample is shown in Fig. 5 above. Data exemplified in Fig. 6 demonstrate two important characteristics: the HSS analyzer exhibits no significant memory effects; and the recovery from the minor memory effect of the analyzer is dependent upon the number of injections made after a concentration change, not on the amount of time between the injections.
i

Response Time for Sulfur in Gasoline at 110, 225 and 15 ppm

,,,

/11_
,, ...... ,,, , ,, ,

jl
,, ,,

,J,

+.t

10
i

20 30 40 Minutes (1 analysis each minute)


ill ii ill ii ill i i i i i

50
i i i

60

FIG. 6---Response time of high speed analyzer. Figure 6 demonstrates analyzer response for sulfur in gasoline samples using the HSS analyzer when switching from 225 mg/kg to 15 mg/kg sulfur levels. The arrow points to data from the first injection after the sample was switched from high to low concentration. Injection rates were at one-minute intervals. Table 1 lists some of the HSS analyzer results illustrated in Fig. 6. These data were collected just before, during, and after a change in sulfur contamination from the 225 mg/kg level to the lower 15 mg/kg level. An evaluation of the data points at injection times 17-23 finds a mean concentration of 221.9 mg/kg of S. Average (mean) sulfur values for data points at times 26--31 are found to be 14.3 mg/kg. In this case where the initial concentration was about 12 times greater than the lower sulfur concentration, the second result after concentration change (Result # 25) was found to be accurate within the analyzer's precision limit for the lower concentration. See Calculation 2.

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

145

Calculation 2 - V25 - M = D M = Mean of samples 26-31 (mg/kg) D = difference in M & V25 (mg/kg) samples 25-30.

V25 - value of injection 25 (mg/kg) D is less than the standard deviation of

Therefore, if an injection is made every minute, it would take only three minutes for the analyzer to begin yielding results within precision for the lower concentration and only one minute to indicate accurately a significant step change in sulfur concentration at over 95 % change. TABLE 1--Data summary of HSS response to changes in sulfur concentration. Sample 1 mg/kg Sample 2 mg/kg Sample Sample 1 2 Injection # Inj ecti0n # 17 224.2 23 22.2 Average 221.9 14.3 18 222.5 24 17.2 Standard 2.13 0.75 Deviation 25 14.7 19 221.4 %RSD 0.96 5.20 26 15.3 20 221.7 27 14.7 21 223.4 28 13.2 22 218.1 29 13.8 30 14.7 31 14.3

Precision of Results
Users also need to know that the analyzer responsible for monitoring a fast changing material stream can shift to various concentrations and quickly generate data with good precision. Figure 6 also shows the excellent repeatability e r a 110 mg/kg gasoline sample repeatedly injected about 30 times in a high speed sulfur analyzer. It resulted in a standard deviation o f 1.26 mg/kg or a relative standard deviation (RSD) of 1.14 %. Then a sample o f - 2 2 5 mg/kg sulfur was injected 10 times, resulting in a standard deviation o f 1.91 with a RSD of 0.86 %. This was followed by about 30 injections of a -15 mg/kg sulfur sample, resulting in a standard deviation o f 0.79 mg/kg or a RSD of 5.7 %. The ~225 mg/kg sulfur material was then analyzed for about five cycles. High speed sulfur analyzer data can now be compared to historical online analyzer performance. Figure 7 shows repeatability of a 50.44 mg/kg gasoline sample repeatedly injected about 70 times in the Standard 6000 process analyzer. It resulted in a standard deviation of mg/kg or a relative standard deviation (RSD) of 0.93 %. Then a sample of 29.9 mg/kg sulfur was injected about 20 times, resulting in a standard deviation of~- 0.23 with a RSD of 0.75 %. This was followed by about 10 injections of a 4.95 mg/kg sulfur sample, resulting in a standard deviation of 4- 0.14 mg/kg or a RSD of 2.87 %. See Table 2. At this stage, the sample was spiked with ~103 mg/kg of sulfur for four cycles, and then returned to the lower range of 50 mg/kg sulfur. Through all these cycles in different ranges of sulfur concentrations both the HSS and standard combustion UVF analyzers maintained excellent response to drastic changes in sulfur concentration.

146

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Repeatibility of Sulfur in Gasoline at 50, 30 and 5 ppm 110 .


9O --- 7 0
m m

50

="=
B

_.,__A .....

=. . . . .

,-.__

.==

-..~

. . . . . . . . . .

30 10"" 0 0

20

40

60 Injections

80

100

120

FIG. 7--Repeatability data at different sulfur concentrations in gasoline for model 6000. TABLE 2--Precision summary for Fig. 7 gasoline. mg/kg S # of Injections Standard Deviation mg/kg % RSD 50.44 ~70 0.47 0.93 29.2 ~20 0,23 0,75 4.95 -10 0.14 2.87

Linear Response
Demonstration of linear response enables the use of single point calibration techniques, which can offer considerable cost savings regarding the number of calibration materials that must be maintained and the time it takes to calibrate the on-line analyzer. A 2000 ppmw standard was made with 7.88 g of pure dibutyl sulfide diluted in 1 L of toluene. This standard was further diluted sequentially 1:1 nine times until the expected sulfur concentration in the sample was down to 3.91 mg/kg. Each of these 10 samples were injected multiple times in a 6000 series analyzer that was set for a 2500 ppmw full scale range and calibrated with the 2000 ppmw sample (see Figs. 8 and 9). Data from the results of the 16.13 ppmw sample were lost due to a data acquisition problem. The data, shown in Table 3, indicate a % RSD of better than 1 % in the sulfur range 62-2000 mg/kg. At lower sulfur levels of 4-31 mg/kg, the %RSD is 11-2 %. To demonstrate the linear response of the High Speed Sulfur Analyzer, a 750.1 mg/kg reference diesel sample was diluted eight times sequentially on a 1"1 basis with a near-zero sulfur light hydrocarbon of similar density. Each of the nine samples was injected multiple times in a HSS series analyzer that was set for a 1000 ppmw full scale range and calibrated with the neat reference material. The data are summarized in Table 4, and the correlation is plotted in Fig. 10. The r2 correlation coefficient was 0.999. The data in Table 4 indicate a % RSD of better than 10.1% in the 4-28 mg/kg sulfur range. At the higher sulfur levels, between 50--750 mg/kg, the %RSD is between 4-0.6%. Figure 10 illustrates the correlated linearity data.

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

147

2000

lOOq

500 250 125 0


,62.5

~31.25

Injections
FIG. 8--Graph of standard model linearity data (Table 3).

~--7.81 \3.9

2000 1800 1600 1400 "8 1200

Correlation = 0.999

> 1000
800

ooo

0 ~

'~.

200

400

600

800

1000

1200

1400

1600

1800

20oo

CalculatedValue

FIG. 9--Standard model correlated linearity data (Table 3).

148

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Expected Sulfur Concentration, , mg/kg 2000.00 1000.00 500.00 250.00 125.00 62.50 31.25 7.81 3.91

TABLE 3--Linearity of sulfur measurements. Average Sulfur Number of Standard Concentration Measurements Deviation, Found, mg/kg ,, mg/k~ 1993.57 43 14.81 998.67 41 8.41 489.6992 162 8.12 248.20 84 1.86 123.68 195 1.31 62.58 40 0.42 34.01 15 0.69 7.05 10 0.25 3.72 12 0.39

Relative Standard Deviation 0.74 0.84 1.66 0.75 1.06 0.66 2.01 3.54 10.56

TABLE 4 Linearity of sulfur measurements for the..high speed sulfur Average Sulfur Number of Standard Deviation, Concentration Found, Measurements mg/kg mg/kg . . . . . . 752.1 7 4.3 383.0 9 4.2 197.4 10 3.4 100.47 12 2.3 48.91 12 2.0 28.0 11 2.3 15.3 12 1.5 8.2 10 0.7 4.8 10 0.4
High Speed Linearity 700 I Correlation = 0.999
600 . . . . . .'"

(HSS) analyzer.
% Relative Standard Deviation 0.57 1.10 1.72 2.31 4.01 8.13' 10.07 8.40 9.20

"a 500
:> "o
O N

400 300 200 100

:~
e,.

<

O~

100

200

300 400 Calculated Value

500

600

700

FIG. 10---HSS correlated linearity data (Table 4).

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

149

Stability of Results
The user also wants to know that the analyzer is stable over time and does not drift. Figure 11 shows the long range stability of measurements for the HSS analyzer. In this case, a gas oil (diesel) sample containing about 130 mg/kg of sulfur was measured repeatedly over a period of 24 h. An average value of 131.4 mg/kg with a 0.662 RSD was found. Figure 11 illustrates the long-term stability of the HSS

High Speed Sulfur Analyzer 24-hour Repeatability Sample: Gazole/Gasoil @ 131.4 mg/Kg 16o1 ~4o ! ....................................................................
120

I
I

8o!
60,

4o!
20,
0

Calibrated Average = 131.4 ppm, Standard Deviation = 0.869, %RSD = 0.662%

0 "1 2
Hours (1 minute Analysis)

FIG. 11--Long-term stability of HSS.

Accuracy of Results
Finally, Table 5 compares the results obtained on six ASTM Ultra-Low Sulfur in Diesel (ULSD) cross check samples. Data from a D 5453 laboratory analysis and data using a process analyzer 6200 and 6200 HSS are compared with ASTM D 5453 consensus values based on about 100 measurements from the participating industry laboratories. Both the standard Model 6200 process analyzer and the 6200 High Speed Sulfur analyzer results are in excellent agreement with the ASTM consensus values as well as the D 5453 results obtained in our laboratory.
ASTM Standardization of the New Method

Work is underway to standardize the test method described here as a new ASTM standard. An Inter-Laboratory Study (ILS) is proposed to include at least nine each of gasoline and diesel Tier 2 class materials containing sulfur levels between <10 and 80 mg/kg in gasoline and between <10 and 80 mg/kg in diesels. Some blind National Institute of Standards and Technology (NIST) Standard Reference Materials (SRM) and solvent blanks are also planned for the cross check. The ILS will follow a protocol of an initial analysis of a set of samples and solvent blanks, and after an elapsed time of at least 12 h but no more than 72 h, a second set of measurements on the same set of samples and blanks will be completed. We expect this ILS to be completed by the end of 2005.

150

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

ASTM ULSD # 0401 0402 0403 0404 0405 0406

TABLE 5--Sulfur results of ASTM D 5453 ASTM High Lab Consensus speed Results Values* sulfur (n = 2) analyzer 14.5 15.08 1.27 14.90 q(99) 0.43 (10) 1.6 1.58 q-0.07 2.21 :e (104) 0.25 (10) 1.6 1.41 0.07 1.30 :L (118) 0.20

ULSD cross check samples (mg/kg).


% RSD HSS 2.87 11.07 15.5 3.58 4.01 % Full Scale 0.43 0.25 0.20 0.52 0.40 Standard Online Analyzer 6200 15.63 ~0.16 (12) 1.46 + 0.12 (10) 1.51 0.08 (10) 15.2 -4-0.27 (7) 9.66 0.32 (7) % RSD % Full Scale 0.31 0.23 0.16 0.53 0.64

1.01 7.98 5.16 1.75 3.33

(lO)
16.3 7.9 0.6 14.92 q-0.18 (104) 9.83 0.13 (113) 14.52 q0.52 (10) 9.98 0.40 (10) 0.62 0.17 (10)

0.66 0.28 27.12 0.17 0.64 0.08 12.8 0.17 (101) (16) All values are in mg/kg sulfur. * Mean value :~ standarddeviation(numberof results). The 6200 was set for a full scale concentrationof 60 mg/kg and calibrated with a 60 mg/kg standard that that was produced from thiophenein a matrix of tolueneand iso-octane. The 6200 HSS analyzer was for a full scale concentration of 100 mg/kg and calibrated with a 13 mg/kg diesel reference material.

Conclusions
Rapid determination of sulfur levels in fuel streams is an important parameter of low sulfur fuels produced in refineries today. It can be even more critical to have such capability in blending, fuel transport, or pipeline installations. The instrument described here can deliver updated sulfur data every minute, which will allow a more accurate delineation 0fboundary lines in the trans-mix region of diverse fuel streams. Timely information in such situations is critical for making cut-point decisions which impact revenues and potential cost savings. The analyzer described in this paper is capable o f accurately ( ~ 1 % RSD) and very rapidly (~1 min) determining very low levels of sulfur (1 mg/kg) in gasoline and diesel streams with a minimal recovery time between successive sampling of different material streams, thus enabling a refinery or a pipeline terminal operation to achieve significant savings of time and costs. The authors wish to thank Dr. R.A. Kishore Nadkami for his editing assistance and Franek Olstowski for some data generation regarding the production of this paper.

References
[1] U.S. Environmental Protection Agency - Federal Register 62(133); 40 CFR Part 80 Part II, Page 37337 (July 11, 1997). [2] U.S. Environmental Protection Agency - Federal Register 64(1992); 40 CFR Parts 80, 85, and 86, Page 26055 (May 13, 1999). [3] U.S. Environmental Protection Agency - Federal Register 65(28); 40 CFR Part 80, Pages

TARKANIC AND CRNKO ON LIQUID HYDROCARBONS

151

6752-74 (February I0, 2000); and 66(12); 40 CFR Part 80, pp. 5002-5141(January 18, 2001). [4] Johnson, W., "Measuring Low Sulfur Levels," Hydrocarbon Engineering, Vol. 8, No. 3, 2003, pp. 49-52. [5] Hietala, M., "Sensitive ASTM D 4294 Sulfur Analysis," Hydrocarbon Engineering, Vol. 8, No. 3, 2003, pp. 53-54. [6] Nadkami, R. A., "Guide to ASTM Test Methods for the Analysis of Petroleurn Products and Lubricants," ASTM MNL 44, ASTM International, West Conshohocken, PA, 2000. [7] Kohl, K., "Fitness for Use of Sulfur Methods," ASTM RR-DO2-1456, ASTM International, West Conshohocken, PA, 1999. [8] Tittarelli, P., "Round Robin Exercise for Sulfur Test Methods for EN 228 and EN 590 Fuel Specifications," CEC/TC 19/WG 27, April 2000. [9] Nadkami, R. A. and Bly, K. J., "An Interlaboratory Study for the Evaluation of Test Method Capabilities for the Determination of Low Levels of Sulfur in Gasoline and Diesel," ASTM RR-DO2-1547, ASTM International, West Conshohocken, PA, 2003.

Journal of ASTM International, September2005, Vol. 2, No. 8 Paper ID JAI12983 Available online at www.astm.org

John R. Rhodes I

Pyro-Electrochemical On-Line Ultra Low Sulfur Analyzer


ABSTRACT: Current and upcoming regulations worldwide require maximum concentrationsof sulfur in fuels to be in the range 10--50mg/kg. This paper describes a new on-line analyzer for ultra low sulfur, the first to employ "Pyro-electrochemicar' technology, which is significantly simpler than other pyrolysis methods such as Pyro-microcoulometry and Pyro-UV fluorescence. The technology is based on ASTM D 6920. The instrument is described with special reference to the modifications needed to adapt the technique to on-line analysis. Experimental results are given for long term stability, sensitivity, precision, speed and linearity of response, temperature stability, and sensitivity to nitrogen. KEYWORDS: ultra low sulfur, fuels, on-line process analyzers, thermal oxidation, pyrolysis, electrochemical sensors, nitrogen converters

Introduction
The maximum sulfur content o f fuels has been reduced steadily over the last few decades to comply with ever-stricter air pollution regulations. Lower sulfur levels have also enabled the development and economic operation o f more efficient engines. Before 1990, sulfur concentrations above 3000 mg/kg were common. During the 1990s, maximum permitted sulfur contents in gasoline and diesel fuels were brought down to 500 mg/kg in most technologically advanced countries and as low as 50 mg/kg in some. "Ultra low sulfur" (ULS), usually defined as a sulfur content less than 50 mg/kg, is now mandated in the USA, EU, UK, and Japan. The goal b y 2010 is "sulfur free" fuels, that is, fuels containing less than 10 mg/kg sulfur. The obvious benefit to air quality will be a greater than 99 % reduction in sulfur dioxide (SO2) emissions. Other benefits, no less important, include improved engine efficiency and fuel economy, reduced particulate emissions from diesel engines, and longer-lived catalytic converters (many sulfur compounds are notorious catalyst poisons).

Analytical Techniques and Instrumentation


The reduction in sulfur levels b y over two orders o f magnitude has significantly impacted the analytical techniques and instruments employed for sulfur measurement both in laboratory analysis and process monitoring. For example, Energy Dispersive X-ray Fluorescence (EDXRF), probably the most widely used method up to now because o f its relative simplicity, low cost, reliability, and easy sample conditioning requirements, may not be sensitive enough to measure ultra low sulfur levels. All other techniques employ to a greater or lesser extent more complicated instrumentation and sample conditioning than EDXRF. Table 1 compares values o f precision at 20 mg/kg S in ULS diesel obtained b y the analyzer described in this paper with Manuscript received 27 September 2004; accepted for publication 25 January 2005; published September 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. 1Rhodes Consulting, P.O.Box 201784, Austin, TX 78720-1784.

Copyright 9 2005 by ASTM International, 100 Ban" Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

152

RHODES ON PYRO-ELECTROCHEMICAL

153

values published in two recent Interlaboratory Comparisons [1,2]. Precision is defined here as the 95 % confidence limit on the difference between two test readings of the same sample (or about 2.8 standard deviations). In the case of the Intercomparisons the precisions are the "pooled within laboratory precisions" as defined in paragraph 23.1.3 of ASTM Standard Practice for Use of the Terms Precision and Bias in ASTM Test Methods (ASTM E 177-90a, reapproved 2002), usually called repeatability, r. The precision value for the present analyzer was obtained using the method described in paragraph 23.1.1 of E 177-90a. That is, a single operator performed measurements on a succession of aliquots drawn from the same homogeneous bulk sample on the same day using the same instrument. The ASTM methods employed are as follows: ASTM D 6920, "Standard Test Method for Total Sulfur in Naphthas, Distillates, Reformulated Gasolines, Diesels, Biodiesels and Motor Fuels by Oxidative Combustion and Electrochemical Detection"; ASTM D 3120, "Standard Test Method for Trace Quantities of Sulfur in Light Liquid Petroleum Hydrocarbons by Oxidative Microcoulornetry"; ASTM D 5453, "Standard Test Method for Determination of Total Sulfur in Light Hydrocarbons, Motor Fuels and Oils by Ultraviolet Fluorescence"; ASTM D 2622, "Standard Test Method for Sulfur in Petroleum Products by Wavelength Dispersive X-ray Fluorescence Spectrometry"; and ASTM D 4294, "Standard Test Method for Sulfur in Petroleum Products by Energy-Dispersive X-ray Fluorescence Spectrometry". In Table 1, Pyro-EC (APS) and Pyro-EC (Antek) refer to results obtained by two different instruments (from APS Technologies, now Analytik-Jena AG, Jena, Germany and Antek Instrumehts, Houston, TX, respectively) which had significantly different repeatabilities. WDXRF(HP) means that a high power X-ray tube (greater than 3 kW) was employed for excitation, and WDXRF(LP) means that a low power tube (less than lkW) was used. The European Round Robin [2] called out European or International standard methods, but they are similar to the corresponding ASTM methods, so they are not detailed here. Since the precisions were measured by different procedures, exact comparisons are not justified. However, the data in Table 1 can be used as a guide to the performance of the various techniques. TABLE 1--Methods and precisions for ULS in diesel. Technique , Pyro-EC On-Line Analyzer(this paper) Pyro-EC(APS) Pyro-EC (Antek) Pyro-mieroC Pyro-microC Pyro-UVF Pyro-UVF WDXRF (HP) WDXRF (LP) EDXRF EDXRF (polarized) ASTM Precision at Method 20mg/kg, (m~Jkg) S D6920 0.39 D6920 D6920 D3120 D3120 95453 95453 92622 92622 D4294 In process 0.41 [1] 3.89 [1] 1.76 [l ] 2.02 [2] 1.34 [1] 1.66 [2] 2.04 [2] 2.46 [2] 7.67 [2] 1.59 [2]

It is clear that any method whose precision is worse than about 1 mg/kg S will have difficulty performing satisfactory analyses of ULS diesel at, say, 15 mg/kg S. The results in Table 1 indicate that for most methods significant improvement is necessary. In fact, the US

154

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Environmental Protection Agency has published performance specifications (Code of Federal regulations, 40CFR part 80.584(a)(1)) which include a required standard deviation of 0.72 ppm S or better, in the range 5-15 ppm sulfur in diesel when repeat measurements of the same sample are made over a time interval of 20 days. A new Interlaboratory Crosscheck Program is being conducted by ASTM in 2004, targeting the sulfur content range 0--15 mg/kg S in diesel [3]. The following ASTM methods are represented: D 2622, D 3120, D 5453, and D 6920. An on-line analyzer must be packaged much more robustly than a laboratory instrument to withstand the process environment which, for example, may have an explosive atmosphere and significantly variable ambient temperature. It must also be capable of continuous, unattended operation over long periods of time. Clearly, the simpler the instrument the better. Of the methods listed in Table 1, WDXRF, polarized EDXRF, and Pyro-mierocoulometry have not been adapted to on-line process instrumentation, whereas the other methods have. The relative simplicity of Pyro-EC makes it particularly suitable for adaptation to process instrumentation. The sulfur dioxide sensor is a small, plug-in, low cost electrochemical cell, easily replaceable and with an expected lifetime of over one year. The UV lamp, UV optics, and photomultiplier used in PYro-UVF are not required. The X-ray tube (or radioactive source), X-ray detector, and X-ray optics used in all the XRF instruments are not required. Description of the Instrument Figure 1 shows a simplified flow diagram of the Model 682P-EC (Spectro Analytical Instruments, Marble Falls, Texas) [4]. The essential subassemblies are: means for sampling from the process stream; sample injection into the pyrolysis furnace; thermal oxidation at 1050~ conversion of any nitrogen dioxide (NO2) to nitric oxide (NO); control of flow rate, relative humidity and temperature of the resulting gas, quantitative detection of SO2 concentration. In addition (not shown in the Figure), there is a means for providing calibration liquids, a series of diagnostic alarms, and a programmable logic controller (PLC) to control instrument operation, calculate the sulfur concentration output, and provide alarm signals. The instrument is housed in two NEMA 4X (IP-65) stainless steel enclosures purged to either Class 1 Div 2 Groups C&D (Type Z purge) or Type X purge for Class I Div 1 Groups C&D and CENELEC Eex p II T4 explosive atmosphere specifications. LEL1 in the figure is a combustible gas sensor to provide an alarm in the event of a leak of combustible fluid into the cabinet. The Scrubber shown in the figure is an activated charcoal-filled cartridge to absorb any noxious gases, such as oxides of sulfur, nitrogen, and carbon before exhausting the sample gas to atmosphere. The other components shown are parts of the essential subassemblies, described below.

Sampling
Liquid from the process stream is sampled by a two position, six port rotary valve with a 520 I.tL sample loop connected between two ports. The sample may be preconditioned to remove any gas bubbles, entrained particulates, and water. Since this part of the instrument is outside the enclosures and could be in an explosive atmosphere, all valves are pneumatically operated, and the sample is driven by compressed air or inert gas.

RHODES ON PYRO-ELECTROCHEMICAL

155

FURNACE SAUPLE ~
ROTARY VALVE

PROCESS RETURN PROCESS INPUT

EXHAUST EXHAUST

OXYGEN INPUT ARGON INPUT

FIG. 1-- Flow diagram of model 682P-EC. In one position of the rotary valve, the process sample flows through the sample loop, thus filling it (and continues back to the process), while the carrier gas flows through the valve and the injector, ensuring that the injection path is clear and ready for the next sample injection. In the second position, the carrier gas is switched to push the sample out of the sample loop, through the injector, and into the pyrolysis tube, while the process fluid continues to flow though the valve and back to the process (bypassing the sample loop). In this way, the rotary valve and sample loop combination "cuts" and dispenses a sample of fixed volume from the process stream for analysis. The sample volume is usually chosen to be 10 IxL and is constant to better than one percent. The sampling frequency is usually selected to be once every 5 min. Since the sample has a constant volume, the analysis must be corrected for sample density if results are reported on a wt/wt basis. Note that the flow rate between the process and the rotary valve is not limited by the analyzer and may be quite high, so each analyzed sample is contemporaneous with the process.

Injection
The purpose of the injection subassembly is to inject the sample at an average rate of about 1~tL/s and an instantaneous rate that never exceeds that which would starve the thermal oxidation process of oxygen and so cause incomplete combustion. Also, any carbonization of the sample within the injector must be avoided. The sample is driven by a flow of inert carrier gas (argon) that is controlled by a combination of a flow restricter (MMV1) and a pressure regulator (LPR1). The injector, which is a stainless steel tube, penetrates the pyrolysis tube to a depth where the temperature is high enough to ensure sample volatilization within the injector while avoiding

156

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

carbonization. The pyrolysis tube is supplied with oxygen flows to carry the injected sample and completely oxidize it.

Pyrolysis
Pyrolysis takes place in a standard quartz tube inside a tube furnace controlled at a temperature of 1050~ as specified in the ASTM method. The output gases are passed through a quartz wool plug and a membrane filter to capture any soot particles.

Conversion
The sample gas stream then passes through a catalytic converter that converts any NO2 to NO without affecting the SOz content. Nitrogen in the fuel sample is theoretically completely oxidized to NO at the pyrolysis temperature of 1050~ However, some NOz is formed due possibly to reconversion of NO to NO2 in the cooler parts of the furnace downstream. The electrochemical sensor has a negative response to NO2 but is insensitive to NO. Hence, it is important to convert any NO2 to NO before it reaches the sensor. Another reason for making sure that all nitrogen is in the form of NO is that the instrument can potentially be used for total nitrogen determination by simply adding an NO sensor. The effective conversion efficiency (that is the conversion efficiency allowing for the fact that only some of the nitrogen ends up as NO2 in the first place) has been found to exceed 99 %. Long term tests show that the converter lifetime is at least one year. The converter is an easily replaceable cartridge.

Sample Conditioning
The relative humidity (RH) of the sample gas stream is such that the dew point is about 55~ during pyrolysis of the fuel sample but fails to very low levels when no liquid is being thermally oxidized. The dryer shown in Fig. 1 is a permeation dryer (Perma Pure Inc., 8 Executive Drive, Toms River, NJ 08754), which removes water vapor without affecting the gases of interest and so prevents water condensation when the sample has cooled down. It employs a purge gas of dry argon that is controlled through a metering valve (MV1) and monitored by an electronic flowmeter (FM1). It is important that the RH of gas incident on the electrochemical sensor is in the range 10-90 % to prevent the sensor getting too dry or too wet. To achieve these conditions, the sample gas is passed through a permeation humidifier (Perma Pure Inc., 8 Executive Drive, Toms River, NJ 08754)just upstream of the sensor. The flow rate of the sample gas stream must be constant, otherwise the SO2 content will vary due to changes in gas dilution. The flow rates are controlled to better than 1 % by metering valves (MV2 and MV3) and pressure regulators (not shown). Electronic flowmeters (FM2 and FM3) provide signals to the PLC that are used for alarm and diagnostic purposes. They can also be used to correct the SO2 concentration for the above dilution effect, if necessary. All parts "wetted" by the sample gas stream must be inert to SO2. This is achieved by the use of AISI 316 stainless steel or fluoropolymer plumbing where necessary.

RHODES ON PYRO-ELECTROCHEMICAL

157

Electrochemical Sensor and Response


The electrochemical sensor is an off-the-shelf SO2 sensor (City Technology Ltd., City Technology Centre, Walton Road, Portsmouth, PO6 1SZ, England). Typical sensor response profiles, or "pyrolysis spectra," are shown in Figs. 2 and 3. Each peak represents the response, as a function of time, to a single constant volume sample injection. The liquid injection rate rises rapidly from zero at the beginning of the injection cycle, remains more or less constant during the 10-s injection, then falls as the last of the sample is cleared out of the injector. Thermal oxidation takes place almost immediately since the residence time in the pyrolysis tube is only about 10 s. The gas transit time from pyrolysis furnace to sensor is only a few seconds, so there is not much dispersion of the SO2 concentration profile. It follows that the volume concentration of SO2 is proportional to the peak integral. It is also seen that the sensor time-dependent response is quite rapid.
Performance

This section describes the analyzer performance to ULS in diesel and gasoline.

Nitrogen Interference
Nitrogen is present in some fuels in concentrations up to, but rarely exceeding, the sulfur content. The approximate nitrogen content of a given fuel is often known, so that small interferences can be minimized by calibrating the instrument with samples that contain approximately the same amount of nitrogen as the fuel to be analyzed. However, it would be preferable to reduce any nitrogen interference to negligible proportions. Initial tests were performed using samples of diesel and gasoline containing zero or 20 mg/kg sulfur and either zero or 50 mg/kg nitrogen. Figure 2 shows the effect of nitrogen when the converter is not installed. The nitrogen content was changed from zero to 50 mg/kg and then back to zero, keeping the sulfur content at 20 mg/kg. The effect of the nitrogen is to considerably reduce the electrochemical cell response over a period of several sample injections. When the nitrogen content goes to zero, the cell response recovers slowly back to its original level. Figure 3 shows the effect when the converter is installed. Any nitrogen interference is now not observable. Table 2 gives results of quantitative measurements made on similar gasoline samples. Further tests have been made to determine the long term stability and lifetime of the converter. An example of these results is shown in Fig. 4, where the sample has an extremely high nitrogen content compared to the sulfur content. Experience over the past two years of operation with samples containing varying amounts of nitrogen have verified that in practice nitrogen interference does not occur, and the converter lifetime is in excess of one year.

158

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

to

FIG. 2--Effect of nitrogen at 20 mg/Kg sulfur in diesel, without converter.

~000

2000 Saeont~

3000

4000

5OOO

FIG. 3--Effect of nitrogen at 20 mg/Kg sulfur in diesel, with converter. TABLE 2--Test resultsfor nitrogen in gasoline. mg/kg S 20 20 blank blank mg/kg N 50 zero 50 zero mean S reading 20.2 20.5 1.2 1.5 two sigma 0.5 0.6 0.5 0.4

RHODES ON PYRO-ELECTROCHEMICAL

159

~ma~s
Sb:l [ ] e v = O,

50

100

150

2B) Msam~.~; Number

25O

3]O

3~0

FIG. 4

On-line analysis of approximately 7 mg/Kg S, 120 mg/Kg N in diesel (31-h run).

Precision and Sensitivity


Measurements o f precision at different concentrations o f sulfur in diesel yielded the following value as a function of concentration: p = 0.016088"X + 0.068238 where X is the sulfur content in mg/kg, and p is defined as the 95 % confidence limit on the difference between two test results. An accepted measure o f sensitivity is the minimum detectable limit (MDL). This is equal to the value o f p at zero concentration, which is seen to be 0.068 mg/kg S. The above values o f precision and sensitivity were obtained by one operator on a single instrument during a time interval o f a few days. However, many similar measurements have been made at different times over the past two years with comparable results.

Long Term Stability


A continuous 26-day unattended run with diesel containing 20 mg/kg sulfur was performed to test the long term stability. No calibration checks or recalibrations were done. Figure 5 shows the plot obtained o f sulfur readout versus measurement number. It is seen that the long term stability is excellent, with no significant up or down drift. Since drift is negligibly small, the standard deviation o f all the measurements (sigma = 0.17 mg/kg S) is another indicator o f the precision at 20 mg/kg sulfur.

Linearity
Linearity tests were performed using diesel samples with known sulfur content in the range zero to 500 mg/kg. Typical results are shown in Fig. 6. It is possible that the responses o f different electrochemical cells will be slightly different from each other. However, nonlinearities of the degree shown here are negligible over the smaller concentration ranges used in practice. Table 3 gives examples of non-linearity error to be expected over practical concentration ranges. The maximum errors are well below the normal instrument precision.

160

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

25

20!

~wmlm
Tot~ r

m m
us run wilhc t caJibratior update. ents at 5 rri ~tes per rnE :=+0.13%

~ P

15

10

~verage re~ :ling: 20.6 "g/Kg S Precision = ).83% O" 0.1 ' rrg/Kg S

1000

2000

30OO

4OOO

5O0O

6OOO

70OO

8OO0

Mea~-=T=t

FIG. 5 - - L o n g term stability.


80OO OO O 7OOO OOO 6OOO OOO ,O O0 5 O 0O
m

J
Net :~eak = 5. 1917"X^2 + 1227ff X - 19421 std clev = 2.8 L~JK.q

"4000000
Z

3OOO OOO

jf
J
50
100 150 200 250

,J

z /

2OOOOOO 1000000

300

350

4OO

450

5C0

S~Xmg~g

FIG. 6 - - N e t p e a k versus sulfur content.

RHODES ON PYRO-ELECTROCHEMICAL

161

TABLE 3--Examples of non-linearity error. Range (mg/kg S) 0-20 0-30 10-50 50-100 100-200 200-300 300--400 400-500 Maximum Error (at mid-range) mg/kg % -0.04 -0.41 -0.09 -0.61 -0.16 -0.54 -0.24 -0.33 -0.92 -0.61 -0.86 -0.34 -0.80 -0.23 -0.75 -0.17

Speed of Response
The speed of response is generally measured by plotting the instrument output as a function of time in response to step changes in sample concentration. Such a plot is shown in Fig. 7.
3~)

250

200

15o
E
100

l ~OnVKg

6Omoj~
50
10 rn, I / K g

ZSrrg<g [
15 10

~eeoe,~
9
30 35 40 10 rrg/l~

9149149149149149
20 25

0
0

eeee~, 9149149
5

45

50

FIG. 7--Response to step changes in concentration. The response is nearly but not quite "instantaneous," where instantaneous is defined here as a 100 % response to a step change by the next measurement after the change. The response speed is influenced mainly by the degree to which, during the injection of the previous sample, the next sample flushes out the previous one in the Rotary Valve and associated plumbing. This, in turn, is a function of the sample flow rate and the magnitude of the step change in concentration. So

162

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

the response can be made faster by increasing the sample flow rate, which is a choice the operator can make. The flow rate used in this test was about 40 ml/min.

Temperature Dependence
There are three main temperature effects. One is due to the variation of sample gas density with temperature, according to the gas law. This causes a change of sulfur concentration in the sample gas stream since the volume of liquid sample injected does not change significantly with temperature. This effect is common to all the pyrolysis techniques. The second effect is the temperature sensitivity of the electrochemical sensor, which is small but not negligible. The third effect is the influence of changes in total water vapor content of the gas sample stream. All three effects have been minimized by proprietary design. Figure 8 shows the results of varying the ambient temperature over a range greater than that likely to be experienced in the instrument shelter of a refinery or other plant. It is seen that the concentration output does not vary when the temperature is changed over the range 17-29~ I f a temperature dependence is ever observed, the instrument has a built-in algorithm that can be used to correct the output for measured cabinet temperature.

FIG. 8--Response to temperature changes.


Conclusions

9 9 9 9 9

Pyro-electrochemical technology is viable for ultra low. sulfur determination. Suitable instrumentation is available for on-line or at-line monitoring of petroleum fuels. The sensor is a small, low-cost plug-in component, leading to a greatly simplified detection subsystem compared to those of other techniques. Nitrogen interference has been eliminated. Sensitivity, repeatability, stability, linearity, and speed of response are equal to or better than those for other on-line ULS analyzers.

RHODES ON PYRO-ELECTROCHEMICAL

163

References

[1] ASTM Lnterlaboratory Crosscheck Program, Low Level Sulfur, ASTM International, West Conshohocken, PA, 2002. [2] CEN TC19 WG27, "Test Methods for the Determination of Sulfur Content. Report of the 2001 Round Robin," European Committee for Standardisation, 36 rue de Stassart, B-1050, Brussels, Belgium. [3] ASTM Interlaboratory Crosscheck Program, Ultra-Low Sulfur Diesel Fuel, ASTM International, West Conshohocken, PA, 2004. [4] U.S. Patent #6,830,730, Dec 14, 2004, and U.S. Patent #6,867,047, March 15, 2005.

Joumal of ASTM International, July/August 2005, Vol. 2, No. 7

Paper ID JA113090 Available online at www.astm.org

Ronda L. Gras, 1 B.S., Jim C. Luong, 1B.S., Robert V. Mustacich, 2 Ph.D., and Randy L. Shearer, 3 Ph.D.

DP-SCD and LTMGC for Determination of Low Sulfur Levels in Hydrocarbons


ABSTRACT: A novel measurement technique has been developed for the rapid determination of low levels of sulfur in hydrocarbon matrices. The technique employs low thermal mass gas chromatography (LTMGC) and a dual plasma sulfur chemiluminescence detector (DP-SCD). Highly sensitive total volatile sulfur measurement can be made in less than 30 s with detection limits in the 20-30 parts per billion range. Response is linear over at least three orders of magintude sulfur with excellent repeatability. By taking advantage of LTMGC's rapid temperature programming and cool down time, it is also possible to measure individual sulfur species, analogous to ASTM D 5623. Conditions have also been developed using the same analytical apparatus to detect the major elemental sulfur allotropes ($2, $3.$6, $7, and $8) in gasoline at levels that can cause silver corrosion.
KEYWORDS: sulfur, elemental, sulfur chemiluminescence detection (SCD), low thermal mass gas

chromatography(LTMGC), dual plasma SCD


Introduction

The United States Environmental Protection Agency (U.S. EPA) and other environmental regulatory agencies, particularly those in Europe, Japan, and other industrialized nations, have developed regulations aimed at further reducing air pollution from mobile sources. Sulfur plays a key role in this since it is known to poison catalytic converters that are used to reduce concentrations o f carbon monoxide, hydrocarbons, nitrogen oxides, and particulates in engine exhausts. The allowable levels o f sulfur in liquid fuels, such as gasoline and diesel fuels, are continuously being driven to lower levels. For example, in California the total average allowable level o f sulfur in gasoline has been 30 ppm total sulfur since 1996 [1], and the U. S. federal limits for sulfur in on-road diesel fuel will be 15 ppm (mg/kg) starting in 2010 [2]. This is a significant development, considering that in the past average sulfur levels for gasoline were higher, often b y a factor o f 10 times or more. Numerous techniques and instrumental methods can be used to measure sulfur in gasoline and diesel fuels [3]. It is beyond the scope o f this paper to cover any o f the standard techniques in detail. Unfortunately, many o f the traditionally used techniques are out o f consideration for analyzing fuels at low levels because o f inadequate sensitivity or lack o f precision at those levels. For instance, a study group comprised o f 69 industry participants has recommended that Energy Dispersive X-ray Fluorescence (EDXRF) and some other older technologies be dropped from consideration for analyzing fuels at the 30 ppm level [3].

Manuscript received 7 October 2004; accepted for publication 14 December 2004; published July 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. l Dow Chemical, Hwy 15, P.O. Bag, Bldg. 158, Fort Saskatchewan, Alberta, Canada, T8L 2P4. 2 RVM Scientific, 5511 Ekwill St. # A, Santa Barbara, CA 93111-2398, USA. 3Ionics Instruments, 6060 Spine Rd., Boulder, CO 80301, USA.

Copyright 9 2005 by ASTM lntornational, 100 Barr Harbor Drive, PO Box C700, West Consbohockcn, PA 19428-2959.

164

GRAS ET AL. ON DP-SCD AND LTMGC

165

Sulfur measurements from process and product quality standpoints are extremely important and necessary. Nevertheless, total sulfur measurements alone may provide little information about the best manner to process a feed or even to determine whether a process unit is functioning as designed. For example, in a process that converts mercaptans to disulfides, total sulfur measurements would show little or no difference between the feed and the product, while a sulfur distribution measurement would show if the intended chemical conversion is taking place. Furthermore, there may be unexpected or unintended consequences from process changes that affect product quality that are related to individual trace sulfur species. For example, it has recently become apparent that low ppm levels o f elemental sulfur can contribute to a product quality issue o f corrosion of automobile fuel gauge sensors [4]. This is especially surprising and ironic considering that sulfur in fuel levels are lower than ever before. There is speculation that some other sulfur species may contribute synergistic or even inhibitive effects on elemental sulfur formation and subsequent silver corrosion. Nevertheless, at least 45 000 vehicles have been affected by this issue to date, and the financial impact caused is apparent. Repair costs alone are on the order of at least $200-600 per vehicle. In this paper, the use of LTMGC to attain high speed heating and unprecedented cool down time and a new commercially available detection system, Dual PlasmaT M technology, based on sulfur chemiluminescence detection (SCD) is described. The system is able to provide quantitative measurements of individual sulfur species, such as the allotropes responsible for silver corrosion, and also can be operated in various modes to provide a rapid total sulfur measurement by itself or a total sulfur measurement in conjunction with speciated sulfur analysis. The new Dual Plasma technology for the SCD (DP-SCD) has been developed to improve performance in several areas, including stability and selectivity [5]. The SCD has become a commonly used sulfur selective detector in petroleum and petrochemical laboratories and has been described in detail [6,7,8,9,10,11,12]. As of this writing, there are three ASTM methods currently being considered that utilize SCD technology. These are ASTM D 5504 (Standard Test Method for Determination of Sulfur Compounds in Natural Gas and Gaseous Fuels by Gas Chromatography and Chemiluminescence), D 5623 (Standard Test Method for Sulfur Compounds in Light Petroleum Liquids by Gas Chromatography and Sulfur Selective Detection), and D 7011 (Test Method for Determination of Trace Thiophene in Refined Benzene by Gas Chromatography and Sulfur Selective Detection) [13]. The high sensitivity, selectivity, absence from hydrocarbon interferences, linea~rity of the SCD, and equi-molar response are well established. It may be operated in series with a flame ionization detector (FID), or the column may be connected directly to a combustion furnace. Briefly, detection of sulfur compounds in the SCD is based upon the two-step mechanism shown below: S-compound + 02 --> SO + Other Products

(1)

SO + O 3 ~ SO2" + 02 -9 SO2 + hv (2) In Eq 1, sulfur compounds are combusted to sulfur monoxide (SO) and other products. In Eq 2, the second step of the mechanism, light energy (hv) in the blue region of the spectrum is emitted from the excited species resulting fi:om the ozone reaction. The basic mechanism of the DP-SCD is the same as that described above, but two plasmas (flames) instead of one are provided to improve selectivity and the ability to measure lower sulfur levels without hydrocarbon interferences. A conceptual drawing of the flow dynamics used in the Dual Plasma burner is shown (Fig. 1).

166

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 1--Conceptual flow diagram of Dual Plasma burner. A schematic block diagram illustrating an entire DP-SCD detection system is shown in Fig. 2. An analytical system consists of a gas chromatograph equipped with a split/splitless injector with the option of a Pressurized Liquid Injection System (PLIS), with or without low thermal mass gas chromatography apparatus, for sample introduction and sulfur speciation (if required); an electrically heated burner with an interface that controls the burner gas flows and temperature; and a detector that contains a chemiluminescent reaction cell, ozone generator, optical filter, amplifier, and electronics. Lastly, a vacuum pump is used to keep the reaction cell under low pressure conditions to prevent loss of chemiluminescent species and to reduce collisional quenching. Gas chromatography ( G C ) a s an analytical technique is accepted and widely used, particularly in the petroleum industry. Its application is amenable to gaseous, liquid, and even solid samples using a variety of sample introduction techniques. In the case of gaseous and liquid samples, introduction of the sample is usually based on a volumetric injection using a valve or syringe, either of which can be automated. Because of the volumetric nature of sampling, it is important that density be taken into account when calculating results on a mass basis. In some cases, density values will not affect calculated results, for instance when an instrument is calibrated on a mass per unit volume basis, or if the sample matrix and calibration matrix are similar or have the same density. In addition, there are intricacies with regard to the injector

GRAS ET AL. ON DP-SCD AND LTMGC

167

technique that is used for sample introduction, such as split and on-column injection [14]. Even though GC could be considered to be a mature technique, significant improvements are being made in several areas besides detection, such as approaches to improve separation or analysis speed.

25 g psimax~
25 psig ma

F.C.
Dual Plasma Controller

Dual Plasma r-"-IInlctor Burner Injeet


Transfer

Detector

Une

Column

Gas Chromatograph

Vacuum ~ent Pump


FIG. 2--Block diagram of GC system with DP-SCD. The process of increasing column temperature during a GC run is referred to as temperature programming gas chromatography (TPGC). For a particular solute, the increase of temperature during temperature programming leads to a decrease of retention factor, retention volume, and retention time. Key benefits of TPGC include better separation of wide boiling point range solutes, improved detection limits, peak shapes, and precisions, especially for late eluting peaks. It also provides an excellent means of column cleaning. A drawback for temperature programming is that it requires considerable cool down time (several minutes or longer), which lengthens the analysis cycle time compared to isothermal chromatography. Cryogenic fluids can be used to lower oven temperatures and speed cooling, but this adds additional expense.

168

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TPGC of capillary columns is typically conducted by electronic control of the temperature of an oven containing the GC column. The winding of the columns on the wire frame support provides extensive surface contact of the capillary gas chromatography column with the heated air in the air-circulation oven for fast temperature equilibrium. Using this technique, depending on the size of the oven, reproducible temperature programming rates ranging from 30-70~ can be attained. Resistive heating elements combined with GC columns have been demonstrated to provide effective and rapid temperature control, but higher thermal effieiencies, accommodation of a wide range of column dimensions, and immunity to electrical shorts at high temperatures were still needed. New technologies that address these needs have been recently developed and implemented. Recently, RVM Scientific of Santa Barbara, California introduced an advanced low thermal mass gas chromatographic accessory (LTMGC) that can be easily integrated to Agilent GCs 4 for high throughput chromatography. A typical LTMGC module is comprised of: an analytical column, a temperature sensor, a nickel alloy heating wire, and a metal tray to support the module and its interface. Mounted under the tray is an electric fan for heat removal during the cooling down cycle of a run. The low thermal mass design, coupled with high surface area provides rapid temperature programming rates in excess of 1800~ with unprecedented cool down time. Power consumption for temperature programming is only about 1 % of that required for conventional gas chromatographs. Fewer than 60 s were required to cool a 5 m column in an LTMGC module from 280~ to 40~ LTMGC takes advantage of many of the key fundamental principles of high-speed gas chromatography [15]. LTMGC provides the capability to either collapse the separation power of the analytical column or the capability to maximize its separation in the case where speciation is required. This technique, coupled with a specialized pressurized liquid injection system (PLIS) [16] and conventional injection, with DP-SCD are discussed. In GC, highly volatile liquefied hydrocarbons are commonly injected using devices such as high-pressure syringes, piston valves, liquid rotary valves, or vaporizing regulators. Although these techniques are adequate in some cases, there are known deficiencies. The PLIS is a small injector that can be coupled directly to an injector port without the need of pre-injection vaporization and transfer lines and sample sizes range from 0.1-2 uL for liquid, with the valve capability to be helium leak-free of up to 1200 psig. These characteristics make the PLIS attractive for introduction of sulfur containing samples into a GC.
Materials and Methods

The work covered in this paper was carried out in two laboratories with two different sets of instruments. At the Dow Chemical Canada, Western Canada Operations, Analytical Sciences Laboratory, an Agilent GC, equipped with a split/splitless injector, a Transcendent Pressurized Liquid Injection System (PLISrU), a RVM 5 Scientific LMTGC system, and a Sievers6 SCD with a prototype dual plasma burner was used. Data collection was done using PerkinElmer AccessChrom Chromatography System. Various columns were used including an 18 m, 0.25 mm id, 0.25 micro Varian VF-1 column; a 50 m, 0.32 mm id, 5 micron CP-Sil 5 CB column; and a 5 m, 0.25 mm id, 0.25 micron DB-1 column. 4Agilent 6890 and HP 5890 GCs, AgilentTechnologies,Litflefalls,DE, USA. 5RVM Scientific, SantaBarbara, CA USA. 6SieversModel 355 SCD, IonicsInstruments,Boulder, CO USA.

GRAS ET AL. ON DP-SCD AND LTMGC

169

In the Ionics Instruments R&D laboratory, an Agilent GC was used with a Sievers SCD with Dual Plasma Burners. Agilent ChemStation software was used for data acquisition and instrument control for most experiments. Both on-column and split injectors with electronic pressure control were used. Work covering the use o f several GC columns, such as a CP-Wax from Varian and HP-5 for Agilent will be described in greater detail. The typical sample size was 0.1-1 pL, split 1:10 to 1:50, and chromatographic separations were done with a variety o f capillary columns. A PLIS was used for some liquid injections. The oven was typically operated in the range o f 40-340~ at programming rates o f around 10~ for conventional GC and about 200~ for the RVM module. In some cases, the run was ended at a lower temperature when complete elution of the sample was ensured. For fast chromatographic analysis, an SPB-1, 30-m x 0.32 mm ID capillary cohnnn with a 4/~m film thickness or an equivalent SPB-1 of 5-m length was used with a helium cartier gas at a constant flow rate o f 6 mL/min. An uncoated 30m, 0.32 mm ID deactivated fused silica capillary acted as a transfer line for single-peak totalsulfur determinations in conjunction with a Dual Plasma burner for some experiments. Pertinent experimental details are given with the appropriate results. Solvents used in the work, such as hexane, iso-octane, toluene, and benzene, were obtained from Aldrich Chemical and Fisher Chemical Canada and were analyzed prior to use. Benzene was treated to remove thiophene as per ASTM D 7011 [12]. Sulfur compounds and elemental sulfur (flake) were obtained from Aldrich Chemical and used as received for preparation of gravimetric blends. Silver wool (0.05 mm OD fibers) was obtained from Alfa Aesar. A California Phase II reference fuel analyzed by D 5453 (Standard Test Method for Determination o f Total Sulfur in Light Hydrocarbons, Motor Fuels and Oils by Ultraviolet Fluorescence) [12] and found to contain 28 mg/kg sulfur was obtained from Spectrum Quality Standards. Elemental sulfur was first dissolved in benzene as a stock solution, and this stock solution was volumetrically added to the reference gasoline fuel to generate low ppm elemental sulfur levels. In addition to the instrumental sulfur analysis, approximately 10 mg o f silver wool was placed in about 1:5 g o f the reference gasoline with and without elemental sulfur. The fuel sample was then placed in a small high-pressure all-glass (except for aluminum foil seal) reaction vessel in a secondary container and oven and heated to 150~ for about 1 h. After the samples were cooled, the silver wool was removed and inspected with a photomicroscope at 10x and 40x magnifications.
Results

The detection o f trace levels o f sulfur in complex hydrocarbon matrices, such as fuels, is dependent on sensitivity (the ability to detect trace sulfur levels) and selectivity (the ability to detect sulfur without interference from potentially interfering hydrocarbons). Hydrocarbon interference can manifest itself positively or negatively or even both in some instances. It is most desirable to have no interference whatsoever, however, if the interference is known and repeatable, then calibration can compensate for it. In most analytical techniques designed to measure trace levels, interferences are likely to exist, and selectivity is most likely to be the ultimate dominant factor in determining detection limits. It is relatively straightforward to determine selectivity by using GC because the hydrocarbon effect can be decoupled from the sulfur response via chromatographic separation. Figure 3 is illustrative o f both the sensitivity and selectivity of the GC Dual Plasma SCD for trace sulfur species when used in GC. In this example, the matrix consisted o f about 99.9 % toluene with

170

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

sulfur species added gravimetrically to a stock solution, which was further diluted. The chromatogram shows that the toluene solvent does not cause any significant deviation of the baseline, either positive or negative. The ratio of sulfur response to hydrocarbon response indicates a selectivity sulfur: carbon selectivity of > 2 x 10 7. This, coupled with the sensitivity of < 0.5 pg S/s, allows detection of low ppb levels of sulfur compounds.

180140120100-

Carbon disulfide Toluene Solvent Virtually No Response

9.% 99
30 ppb Sulfur 3-methiophene

Ethylmethyl sulfide

00o I;E Bo-

/
Thiophene

O0

0.5

t ,0

1.5

2.0

25

3.0 Time - Minutes

:3.5

4.0

4.5

5.0

5,5

I~.0

FIG. 3--GC Dual Plasma SCD chromatogram of trace sulfur species in toluene. Injection volume of 1 ILLsplit 1:10 using a 0.32 mm ID CP-Wax column. High sensitivity and selectivity are required to measure trace sulfur species in a complex fuel sample like gasoline. This is especially so for elemental sulfur, which can form several allotropes and which at ppm levels can contribute to the fuel gauge sensor issue described previously [4]. Using the DP-SCD, chromatographic conditions were chosen to elute all of the volatile sulfur species in gasoline. Analysis of the California reference fuel (28 mg/kg total sulfur) showed no evidence of containing the major elemental sulfur allotropes, $6, $7, and $8. Chromatographic separation of the minor allotropes, $2 and $3, from naturally occurring sulfides (typically present at low ppm levels) found in the gasoline was not possible using the conditions described here, but their addition results in detectable differences in the chromatograms. The procedure is readily capable of detecting $6 and $8 as shown in Fig. 4, however, even at the low level that can contribute to fuel gauge sensor corrosion. Though not shown here, it is possible to deduce which peaks of a sulfur chromatogram contribute to silver corrosion by comparing chromatograms of a sample prior to and after reaction of the sample with silver wool. In this type of procedure, the reaction vessel must be capable of withstanding the high pressure that the gasoline will produce when heated, and nonreactive volatile sulfur compounds must not be allowed to escape. It was hoped that standard laboratory GC auto-sampler vials could serve this purpose, but they were found to leak when heated above 100~ and a temperature of 150~ was necessary to react with the elemental sulfur in a reasonably short time, such as 1 h. Some researchers have noted that elemental

GRAS ET AL. ON DP-SCD AND LTMGC

171

mercury rapidly reacts with elemental sulfur in gasoline at room temperature, and this could be used for the same purpose [4]. However, use of silver is more representative of the materials used in the fuel gauge sensors that corrode. Furthermore, mercury is more toxic and is subject to greater concerns with regard to material disposal and industrial hygiene issues. 15uV ~1 AIB1B, (JUN04~SULFUR13.D) I I 2.6 ppmelementalsulfurin CA Refgasoline1:12.5split

20O

" r

"1

AIB1B, (JUN04\SULFUR16.D) 15~1 1 I 10 pprnelementalsulfurin CA Refgasoline1:12.5sptit

400 2O0

, 10

lfi

?fl

.A

FIG. 4---Sulfur chromatograms of a California reference gasoline that had 2. 6 ppm (top) and 10 ppm (bottom) of elemental sulfur added. (GC Conditions: 1 IzL split1:25; 9.8 psig headpressure and 280~ injection temperature; column: HP-5, 30 m, 0.32 mm ID with a 0.25 pm film; 50 ~ for I min programmed to 260 ~ at 8 ~ and held at the upper temperature for 4 min.) Formation of black silver sulfide is visible on silver wool when 10 mg of silver wool was placed in about 1.5 g of the reference gasoline containing 2.6 ppm elemental sulfur and heated as described above. No visible formation of silver sulfide occurred with silver wool in the reference fuel prior to the addition of elemental sulfur. This is illustrated in Fig. 5, where "A" represents silver wool heated in the reference fuel versus "B," which represents silver wool heated in the reference fuel that contained 2.6 ppm elemental sulfur. Even though the pictures in Fig. 5 were obtained under magnification, the silver corrosion was readily apparent to visual inspection without magnification. The surface of the silver wool became even darker when exposed to higher levels of elemental sulfur as one would expect, and it appeared that the entire surface became coated at about the 10-15 ppm level. Further experimentation was not pursued in this area, as it was demonstrated that the more qualitative silver wool test and quantitative GC analysis appear to be correlated. One could expect to find an optimum ratio between the amount of sample and silver wool to maximize the visible contrast that is observed, however. It is apparent that elemental sulfur species that are suspected to cause silver corrosion can be detected at the appropriate levels that contribute to the fuel gauge issue. Of course, the ability to make a rapid total sulfur measurement is extremely useful in many circumstances. For instance, in a process plant upset, speciation of sulfur may be desirable only when a total sulfur specification is exceeded. In those circumstances, it is desirable to be able to simply change operating conditions back and forth between total sulfur and speciated sulfur analysis, and this is achievable using LTMGC.

172

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 5--Silver wool corrosion; control sample ",4 "" (left) heated with reference fuel without elemental sulfur and sample "B'" (righ 0 reference fuel containing 2.6 ppm elemental sulfur. (Top ~ l Ox magnification; bottom approximately 40x magnification). A standard containing 13 sulfur compounds, each in the range of 170-300 ppm, was used to determine a number of analytical figures of merit including short-term precision and response as shown in Table 1. The system demonstrated good short-term precision with most compounds yielding a relative standard deviation of response less than 4 % RSD (Relative Standard Deviation) for number of injections, n=4. The results also indicate that the system responds to sulfur on an equi-molar basis within experimental error (:t:5 %), meaning that sulfur response is equivalent for different sulfur compounds on a molar or mass sulfur basis. This is a desirable property because it enables quantitative analysis of unknowns, and it allows for easier calibration. A chromatogram for the 13 sulfur compounds that were used is shown in Fig. 6. The results are particularly good considering that mercaptans and disulfides are very reactive. Figure 7 shows a chromatogram of an unleaded gasoline sample obtained with the above LTMGC DP-SCD system. This demonstrates that the system is amenable to sulfur analysis of petroleum products, described by ASTM D 5623 [12]. By operating the LTM GC module isothermally at a high temperature, e.g., 250~ instead of temperature programming the module, it is possible to measure total sulfur as a single unresolved peak. It should be noted that peak shape will be dependent on several factors, such as sample size, sample matrix volatility, the column used, and the temperature of the module. In this part of the study, the focus was on light hydrocarbons used in a chemical plant, for which the PLIS injection valve was a good choice. The effect of split ratio of the GC injector was adjusted so that no hydrocarbon interference was observed (at ca. 1:5). Figure 8 illustrates sensitivity, peak shape, and repeatability for a butane standard that contained about 1.5 ppm total sulfur. The detection limits obtained with the LTM GC in the total sulfur mode were between 20-30 ppb S for a signalto noise ratio of 3:1. Even though variation of the peak height is readily apparent, peak area was used to calculate repeatability. For the 15 injections, the area repeatability was 3.3 % RSD.

GRAS ET AL. ON DP-SCD AND LTMGC TABLE 1--Short-term precision and response for sulfur compounds.

173

400000-

11

== o

12

13

10

5771

A__JU U ~L._.J)U UL__/t y~ 2' 4' 0' Time (min)

L
i

h..__..,,,.,_, I0'

12'

FIG. 6---Standard chromatogram of 13 sulfur compounds in the range of 170-300 ppm sulfur. The column was an 18 m, 0.25 mmid, 0.25 micron PDMS, programmed at 30~ for 2 rain to 200~ at 30~ ~rain; Split rate: 10:1, Helium carrier, 30 cm/sec [Peak identifications: (1) Dimethyl Sulfide; (2) Isopropyl mercaptan; (3) Tert-butyl mercaptan; (4) Propyl mercaptan; (5) Sec-butyl mercaptan; (6) lsobutyl mercaptan; (7) Diethyl sulfide; (8) Butyl mercaptan; (9)Dimethyl disulfide; (10) Dipropyl sulfide; (11) Diethyl disulfide; (12) Diisopropyl disulfide; (13) Dipropyl disulfide].

174

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Benzothiophene

2&3-Methylthiophene=
t-0

Methylbenzothiophenes
Thiophene

5804

4
Time (rain)

lo

FIG. 7--Sulfur chromatogram of an unleaded gasoline sample obtained with the LTM GC SCD system operated in a fast chromatographic mode.
6500

5400. 0.~

~.5' 3.0' Time (min)

3.5'

4.0'

4.B'

~.O'

FIG. 8--Sulfur chromatogram of an unleaded gasoline sample obtained with the LTMGC SCD system operated in afast chromatographic mode. Figure 9 demonstrates that a linear response to total sulfur is obtained with the system over a concentration range of ca. 1.2 ppm to about 1200 ppm. The linear least squares analysis produced a correlation coefficient > 0.9999, and the y-intercept was essentially zero. Lack of a yintercept is a good indicator that there is no positive or negative interference. Still, caution should be exercised when making measurements close to the detection limit because of the difficulty of the potential for interferences and contamination at such low levels. If one is going to make measurements in that region, then it would be advisable to check for interferences and contaminants using the standard addition technique, for example. Keep in mind that not only is it possible for "pure" solvents to contain trace levels of sulfur, but the possibility of sulfur

GRAS ET AL. ON DP-SCD AND LTMGC

175

adsorption and even sulfur release from surfaces could affect analytical results. Several injections were made at each concentration level. Nevertheless, in this experiment good repeatability was observed over the entire concentration range, for example 2.2 % RSD at the 120 ppm level. These results are particularly good considering the volatile nature and difficulty of sampling this liquefied butane matrix.

FIG. 9--Plot of sulfur response as a function of concentration for a liquefied butane matrix over the range of approximately 1.2-1200 ppm PLIS injection valve and Dual Plasma SCD with LTMGC module at 250~
Discussion

The main advantage of using the LTMGC approach is that the LTM can easily give speciated or total sulfur (single peak) results without changing hardware. That is, only the temperature programming profile of the LTMGC is changed for the particular type of analysis. This does not preclude the use of a standard GC for total sulfur or speciated sulfur analysis. In fact, obviously a standard GC is designed for speciated analysis, but it can also be used for dedicated total sulfur analysis as well. This can be accomplished with an uncoated fused silica tube, as described previously, attached to a conventional split/splitless injector. Again, no interference to hydrocarbons was observed with the dual plasma burner (injection size, 1 ~tL; split 1:10), and detection limits and speed of analysis are similar to those obtained with LTM GC. Using an auto-sampler, short-term precision for ppm levels of sulfin: is similar to that obtained for a typical GC analysis, 1-2 % RSD. These results are acceptable for a typical laboratory analyzer. Stability and longer-term precision were evaluated using a gasoline fuel that contained ca. 32 ppm sulfur, making an analysis approximately every 20 rain over a period of 100 h. Individual data points collected over this period are plotted in Fig. 10. Stability of response and precision for the above gasoline was about 1.7 % RSD over the entire test period. These results compare favorably with other accepted methodologies, such as combustion UV fluorescence, ASTM D 5453 [12]. This approach is attractive, especially because of the wide availability of GC instrumentation and its relative low cost.

176

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. lO--Stability of response to ca. 32 ppm sulfur gasoline fuel over 100 h using a conventional GC configured for total sulfur analysis. A fast sulfur speciation can also be achieved with a conventional GC by sacrificing some resolution using a short, e.g., 5 m, 0.32 mm ID 4 #In PDMS column. This is illustrated in Fig. 11 as an isometric plot of a six replicates of a gasoline sample that contained approximately 32 ppm total sulfur. Benzothiophene is eluted at about 5.2 min in the "fast" mode compared to about 23 min using conventional conditions. Of course, this gain in speed results in some loss in chromatographic efficiency as readily seen by the loss of resolution of the methylbenzothiophene isomers. The precision of total sulfur values obtained by this technique was about 1.6 % RSD for number of injections, n = 6. These results illustrate the applicability of the technique for low sulfur gasoline that meets most current regulations.
150

.n o 400

j~t

10X Dilution

Oveday of 6 Replicate Injections

'

0 .O

1.0

2.0

3.0
Time

4.0
- Minutes

5.0

8.0

7.0

8.0

FIG. 11--Isometric plot of six replicates of the California reference gasoline sample that contained ca. 32 ppm total sulfur. The small insert is the response to ca. 3 ppm total sulfur obtained by 10 x dilution of the reference sample.

GRAS ET AL. ON DP-SCD AND LTMGC

177

To demonstrate the potential for application at even lower level sulfur fuels, the California reference gasoline was diluted by 10 times in a "sulfur-free" toluene solvent and re-analyzed. Results from this test are illustrated in the small insert of Fig. 11. The results show that this technique is capable of detecting and speciating sulfur at levels even lower than the 5 ppm total sulfur level. On the other hand, it is likely that regulations may mandate sulfur concentrations in fuels at even lower levels. The insert also illustrates that the dilution of the sample to ~90 % toluene has little effect on the sulfur distribution, and clearly no hydrocarbon interference is observed. Loss of resolution of the 2 and 3-methylthiophene isomers (eluted just prior to 2 min) was observed due to a chromatographic "solvent effect," but otherwise this did not significantly impact the results.

Summary
Gas chromatography with sulfur chemiluminescence detection has become a useful technique to help refiners meet their sulfur challenges by providing sulfur speciation and analysis. This work shows the potential for extending a new version of the SCD (Dual Plasma technology) to chromatographic speciation and total sulfur detection in at least two modes using a LTM GC or conventional GC as a single sulfur peak measurement approach. In addition, this work shows that both of these approaches are able to measure the low sulfur levels required today, and they are capable of meeting even lower sulfur measurement requirements that are anticipated. Future work involving this technology will be directed toward lower concentration ranges, near the detection limits of 20-30 ppb to about 100 ppb, and examining the technology for heavier sample matrices, such as diesel fuel. Significant findings include: The DP-SCD demonstrates excellent sensitivity and selectivity (which is frequently more important) for measurement of sub-ppm levels of sulfur. 9 This technique is able to detect the major elemental sulfur allotropes at levels that contribute to silver corrosion involved in the fuel gauge issue. 9 A LTMGC module can be successfully used with this sulfur detection technology in both a chromatographic and non-chromatographic mode for trace sulfur detection. The use of LTMGC is attractive because, at least for volatile matrices, the same hardware without physical modification can be used to make chromatographic and non-chromatographic sulfur measurements. 9 The technique for a non-chromatographic total sulfur measurement using a conventional gas chromatograph has also been demonstrated. 9

Acknowledgments
The authors wish to acknowledge the support of Dr. Mary Fairhurst, contributions and helpful discussions from Myron Hawryluk, Dr. Joaquin Lubkowitz, Paul Cecil, Dr. Bruce Quimby, and others.

References
[1] Title 13, California Code of Regulations, Motor Vehicles Division, Section 2281 and 2282, 2 Rules.

178

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

[2] Title 40, CFR, Part 80, Compilation of Federal Regulations of Fuel and Fuel Additives, Including Diesel Fuel. [3] Nadkarni, R. A. K., "Determination of Sulfur in Petroleum Products and Lubricants: A Critical Review of Test Performance," American Laboratory, Vol. 32, 2000, pp. 16-22. [4] Special session to address reactive sulfur in gasoline, ASTM D2 Meeting, June 2004, Salt Lake City, UT. [5] Gras R. L., Luong, J. C., Shearer, R. L., and Mustacich, R.V., "Dual Plasma Sulfur Chemiluminescence Detector and Low Thermal Mass Gas Chromatography (LTMGC) for ppb Level of Sulfur Analysis in Hydrocarbons," Poster 1.17, June 2, 2004, the 27 th ISCC Conference Proceedings, Rival Del Garda, Italy. [6] Benner, R. L. and Stedman, D. H., "Universal Sulfur Detection by Chemiluminescence," Analytical Chemistry, Vol. 61, 1989, pp. 1268-1271. [7] Hutte, R. S., Johansen, N. J., Plato, M., and Legier, M. F., "Column Selection and Optimization for Sulfur Compound Analyses by Gas Chromatography," Journal of High Resolution Chromatography, Vol. 13, 1990, pp. 421-426. [8] Shearer, R. L., "Development of Flameless Sulfur Chemiluminescence Detection: Application to Gas Chromatography," Analytical Chemistry, Vol. 64, 1992, pp. 2192-2196. [9] Shearer, R. L., Poole, E. B., and Nowalk, J. B., "Application of Gas Chromatography and Flameless Sulfur Chemiluminescence Detection to the Analysis of Petroleum Products," Journal of Chromatographic Science, Vol. 31, 1993, pp. 82-87. [10] Chawla, B. and Di Sanzo, F., "Determination of Sulfur Components in Light Petroleum Streams by High-Resolution Gas Chromatography," Journal of Chromatography, Vol. 589, 1992, pp. 271-279. [11] Di Sanzo, F. P., Bray W., and Chawla, B., "Determination of the Sulfur Components in Gasoline Streams by Capillary Gas Chromatography with Sulfur Chemiluminescence Detection," Journal of High Resolution Chromatography, Vol. 17, 1994, pp. 255-258. [12] Shearer, R. L. and Meyers, L. M., "Simultaneous Measurement of Hydrocarbons and Sulfur Compounds using Flame Ionization and Sulfur Chemiluminescence Detection for Sulfur Simulated Distillation," Journal of High Resolution Chromatography, Vol. 22, 1999, pp. 386-390. [13] Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA. USA. [14] Modern Practice of Gas Chromatography, 4th Addition, John Wiley and Sons, Grob, R. L. and Barry, E. F., Ed., New York, USA, 2004. [15] United States Patent 6,217,829, Mustacich, R. V., et al., "Reduced Power Consumption Gas Chromatograph System," April 17, 2001. [16] Luong, J., Gras, R., and Tymko, R., "Innovations in High Pressure Liquid Injection Techniques for Gas Chromatography: Pressurized Liquid Injection System," Journal of Chromatographic Sciences, Vol. 41, 2003, pp. 550-559.

MERCURY DETERMINATION

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JAI12985 Available online at www.astm.org

S. Mark Wilhelm, Ph.D.,1 David A. Kirchgessner, Ph.D., 2 Lian Liang, Ph.D., 3 and Peter 1t. Kariher 4

Sampling and Analysis of Mercury in Crude Oil


ABSTRACT: Sampling and analytical procedures used to determine total mercury content in crude oils were examined. Three analytical methods were compared with respect to accuracy, precision, and detection limit. The combustion method CLT.S.EPA 7473 hybrid) and a commercial extraction method (non-standard) were found adequate to provide a good combination of sensitivity and accuracy, while instrumental neutron activation analysis was found to suffer from interferences from elements other than mercury (Ni, Se) that are typically found in crude oil. Aliquot removal and processing steps were found to be important to accuracy and precision of the combustion method. Potential errors caused by incomplete homogenization and loss of volatile mercury were investigated in detail. Homogenization of oils that contain suspended mercury compounds was found important to allow retrieval of representative aliquots for analysis. Ultrasonication at slightly elevated temperature yielded the more reproducible and higher magnitude results. It was discovered that volatile mercury can be lost from sample containers when they are opened to remove aliquots. Alternative sample containers and aliquot removal techniques were developed that minimized loss of volatile mercury. Speciation experiments were utilized to determine the identity of volatile species and the rates at which they are lost from containers, Both elemental mercury and dimethylmercury were identified as volatile mercury components of one freshly sampled crude oil.

KEYWORDS: petroleum, mercury, oil, volatile mercury, speciation, refining


Introduction

The United States Environmental Protection A g e n c y (U.S. EPA), in cooperation with the American Petroleum Institute (API) and the National Petrochemical and Refiners Association (NPRA), has initiated a project to determine the mercury content o f crude oil processed in the U n i t e d States. The focus o f the project is to determine the m e a n concentration and range o f concentrations o f total mercury in crude oil accurately and with statistical justification. Data generated in the project will be used to calculate the m a x i m u m potential contribution o f crude oil mercury to anthropogenic mercury emissions in the United States. Samples o f crude oil are acquired at locations immediately upstream o f refinery tank farms and analyzed for total mercury concentration (THg). Data are generated in the form o f total m e r c u r y concentration in liquid aliquots (THg, gg/kg) o f market-named oil streams obtained at particular points in time. The project is divided into three parts. Part 1 involved comparison o f analytical methods and has been reported previously [1]. Part 2 involved investigation o f sampling and analysis procedures and is reported here. Part 3 is in progress and involves sampling and analysis o f the majority o f crude oils processed in the U.S. in years 2003-2005.

Manuscript received 1 October 2004; accepted for publication 9 March 2005; published October 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. i Principal Scientist, Mercury Technology Services, 23014 Lutheran Church Rd., Tomball, TX 77377. 2 Senior Research Scientist, U.S. EPA, Mail Drop E305-02, Research Triangle Park, NC 27711. 3 Principal Scientist, Cebam Analytical, Inc., 3927 Aurora Ave. N, Seattle WA 98103. 4 Research Scientist, ARCADIS, 4915 Prospect Dr., Durham, NC 27713.
Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

181

182

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Data for the oil ensemble and their statistical treatment will be reported at the completion of the project.

Technical Background
One goal of the API/EPA/NPRA project is to obtain an estimate of the amount of mercury that may be emitted to the environment from crude oil processed in the United States. The analogous question for coal has been answered [2]. The average concentration of mercury in U.S. coal is approximately 100 gg/kg. U.S. coal combustion contributes on the order of 70 tons of mercury to the atmosphere annually or about 60 % of the total amount from all U.S. combustion sources [3]. It is equally important to understand the concentration of mercury in crude oil because the mass of petroleum processed in the U.S. annually is roughly the same as the mass of combusted coal (1012 kg) [4]. Most mercury in coal is substitutional in pyrites found in the coal matrix, and thus coal presents mostly non-volatile species to the analyst [2]. ASTM Method D 6722-01 (Standard Test Method for Total Mercury in Coal and Coal Combustion Residues by Direct Combustion Analysis) and U.S. EPA Method 7473 (Mercury In Solids and Solutions by Thermal Decomposition, Amalgamation and Atomic Absorption Spectrophotometry) were developed primarily for analysis of mercury in coal and are now widely practiced. A number of mercury species have been identified in crude oil (Table 1) [5]. Some evidence exists that other species may also be present in crude oil, but those listed in the table are thought to encompass all the major species. Of those listed, elemental mercury and mercuric sulfide dominate the others. Mercuric sulfide is a generic term that really means mercury bonded to sulfur such that the resultant compound is thermally stable and non-volatile in the primary distillation of crude oil. Actual (meta-) HgS has been identified in crude oil conclusively, but other Hg-SR species, where R is an organic substituent, are likely present as well. Aside from mercuric sulfide, most of the mercury species are volatile to one degree or another. Ionic mercury is not thought to be a major species in the reservoir because elevated mercury concentrations are not measured in produced water associated with either crude oil or gas condensates having high concentrations of total mercury. The prevalence of dialkylmercury (DMHg) in oil is a matter of considerable debate, but DMHg is thought to be a very low percentage of the total mercury content of crude oil and gas condensates based on existing data. Historically, a variety of methods has been applied to measuring the total mercury concentration in crude oil. Notable in the evolution of technologies is the early work of Filby and colleagues [6,7] with instrumental neutron activation analysis (INAA), Bloom [8] with extraction/cold vapor atomic fluorescence (CVAF), Liang [9] with combustion - CVAF, and Kelly [10] with isotope dilution. The only standard method for mercury in crude oil is a method developed by Universal Oil Products identified in industry as UOP 938. UOP 938 is a combustion-amalgamation method specific to a particular instrument (Nippon Instruments Corporation). Mercury in crude oil historically has been routinely determined using some form of digestion followed by cold vapor atomic absorption (CVAA), but mercury losses in the digestions and extractions have been reported [I 1]. More recently, many industry analysts have adopted some variation of the combustion/amalgamation method because of its relative simplicity and demonstrated accuracy for mercury in a variety of hydrocarbon matrices. The initial experiments conducted in the EPA/API/NPRA project examined analytical method and compared INAA, combustion/CVAF, and extraction/CVAF [1]. Three separate laboratories participated, each analyzing a series of constructed single-species standards using

WILHELM ET AL. ON CRUDE OIL

183

dodecane or toluene as the oil surrogate. It was found that certain mercury species were problematic to routine analysis. Suspended HgS, for example, was difficult to homogenize in samples, and thus the analysis o f a liquid suspension of HgS in mineral oil was not reproducible. Likewise, problems were encountered when attempting to measure solutions o f elemental mercury in dodecane and toluene leading to the experimental work reported here. TABLE 1--Properties of mercury and mercury compounds found in petroleum. _ Property Hg~ HgC12 HgS CH3HgC1 (CH3)2H~__ Melting -39 277 584 167 Point (~ (sublimes) (sublimes) "" Boiling 357 303 96 Point (~ Vapor 0.180 0.899 n.d. 1.76 8.3 x 103 Pressure (Pa) (20~ (20~ (25~ (25~ Water Solubility 49.4 x 10~ 66 2 x 10-24 5-6 2.95 (gL") (20~ (20~ (25~ (25~ (25~ Alkane 2 x 10-3 -2.5 x 10-z Solubility (gkg"l) (20~ (25~ miscible Henry's Law 729 3.69 x 10-5 1.6 x 10"5 646 Coefficient ... (Pa mamol~) (20~ (20~ (15~ (25~ Octanol - water 4.2 0.5 n.d. 2.5 180 partition ratio Crude oil is known to contain volatile mercury because mercury is found concentrated in the liquefied petroleum gas (LPG) and naphtha fractions of the atmospheric distillation when oil is refined. It is known conclusively that elemental mercury is at least one o f the volatile mercury species in crude oil because it sometimes is found condensed in trays in refinery distillation towers and condensed in cryogenic heat exchangers that liquefy petroleum gases. What is not known is whether elemental mercury is the only volatile species. The rate at which volatile mercury evaporates from samples o f crude oil is much greater than most analysts realize, leading to the possibility o f a prevalent low bias in oils routinely analyzed for total mercury. The logical path that leads to this conclusion is the principal subject o f the experimental work to be discussed.

Experimental
Three independent laboratories, each using a different method, were selected to analyze crude oil samples. One objective o f the project was to identify methods and procedures that could be used by industrial or academic laboratories to analyze crude oil for mercury in a practical fashion. The practicality of a method is a function o f sample processing steps, reagent availability, instrumental analysis time, instrument expense, detection limit, and accuracy o f result. Crude oils exhibiting total mercury concentrations (THg) below 1-2 ~tg/kg are not problematic to refining, but some specifications for refined products (naphtha) are set close to 1 ~tg/kg; thus a method detection limit (MDL) o f 0.1-0.3 ~tg/kg was deemed acceptable for routine quality assurance (QA) purposes. Differentiating mercury concentrations below 1 ~tg/kg may have academic utility, but an MDL o f 10 ng/kg (parts per trillion) is not necessary for oil quality determinations or for assessment of atmospheric emissions attributable to petroleum.

184

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

LAB 1 employed the extractive method developed by Bloom [8] in which an aqueous oxidant, BrC1, is used to oxidize and thus extract all mercury species in an aliquot without chemical decomposition of the oil matrix. Mild heating (100~ was required to promote complete oxidation o f mercury species. The mercuric ion in the aqueous extract was treated in the same manner as acid digests that are measured using U.S. EPA Method 1631. The sequence o f steps involved reduction of aqueous phase mercuric ion using stannous chloride (SnC12), sparging o f elemental mercury produced by the reduction step, and trapping o f the mercury vapor from sparging on gold (Au). The amalgamated mercury was then thermally desorbed from the gold amalgam and detected using cold vapor atomic fluorescence (CVAF). Bloom developed these extractive techniques for operational speciation of mercury in crude oils and condensates [8]. Data quality parameters obtained separately from this project are shown in Table 2. References [1] and [8] provide details on the operational speciation procedures also utilized in this project (see Table 5 and discussion). TABLE

2--Data qualityparametersfor Hg extractions in petroleum samples [8].


CH3Hgt 0.001 0.005 4 0.02 0.58 0.47 0.09 8 0.25 16.87 4 105.5 99-109

Mercury Concenlrations, lag/kg(ppb) Parameter Total* Hg(II) Hg~ Mean method blank 0.10 0.007 0.16 SD blank 0.02 0.003 0.01 N blanks 4 4 4 eMDL at blank 0.06 0.01 0.04 Low spike level 0.30 0.12 0.26 Mean low spikes 0.24 0.07 0.24 SD low spikes 0.05 0.03 0.03 N low spikes 7 8 8 Measured MDL 0.17 0.10 0.09 Matrix spike level 487 14.3 93.8 N matrix spikes 4 4 4 Mean spike recovery 100.3 93.1 98.7 % recovery range 98-102 76-101 94-103 * Same figures apply to total dissolvedHg. t Results are matrix spike recovery averagecorrected (63.7 % recovery).

LAB 2 used a combustion method developed and constructed in-house [9,12]. The combustion-trap system was composed o f a quartz temperature controlled combustion column, a gas scrubbing section, and gas flow apparatus. The combustion column was divided into two independently controlled heating zones. The initial segment vaporized the sample, and the second segment combusted the hydrocarbon. Liquid aliquots were injected directly into vaporization section o f the instrument as opposed to using a ceramic boat as in UOP 938 or U.S. EPA 7473. The carrier gas was air purified to remove any ambient mercury using a gold-coated sand trap. Mercury vapor exiting the combustion chamber was scrubbed to remove partially combusted hydrocarbon and then collected on gold traps (double amalgamation). Mercury vapor detection was by cold vapor atomic fluorescence (CVAF). LAB 3 employed INAA and used procedures developed by Filby and Musa [6,13,14]. Oil aliquots were irradiated in a uranium fission reactor, and the flux and energy spectrum of emitted gamma radiation was measured. Mercury concentrations were determined by comparison to standards that were irradiated at the same time as the samples. Oil aliquots were removed from sample bottles by pipette and placed in quartz vials that were then torch-sealed prior to irradiation.

WILHELM ET AL. ON CRUDE OIL

185

When analyzing actual crude oil samples, the INAA method suffers from interferences caused by nickel and selenium. Total mercury was determined by correction for the selenium interference as discussed by Filby and Shah [7]. The principal 7 photopeak of 80Hg2~ overlaps the 7 photopeak of selenium (345e75). The quantification of 80Hg2~ was accomplished by subtracting the 279.6 keV 348e 75 overlap of the 279.2 keV 80Hg2~ line with knowledge of the 348e 75 264.6/279.2 keV intensity ratio. The 34Se75 264.6 keV gamma line is not subject to interference and was determined independently. The 34Se75 264.6/279.2 keV intensity ratio is fixed by isotopic abundance and known independently. This correction procedure provided reproducible results in samples containing both Hg and Se, but the resultant detection limits were elevated substantially. Nickel in oil was likewise problematic to mercury determination INAA due to a general increase in background 7 radiation levels. The combination of the Se and Ni interferences produced high detection levels for oils that contained either or both nickel and selenium. Three crude oils were selected for analysis. The oils were selected to demonstrate a range of density (20-50 API gravity), sulfur content (0.05-3.5 wt. %), and metals content (Ni + V; 0.05300 ppm). Grab samples of crude oil were acquired from sampling stations at the terminus of pipelines to refineries or at stations adjacent to tanker unloading facilities. Initially, oil samples were obtained in the field by mostly filling 125 ml glass bottles. The headspace in each bottle varied, but was generally in the range of 15-30 ml (10-25 % of sample bottle volume). Experimental work conducted separately from the project identified borosilicate glass as the most suitable sample container material. Somewhat ironically, the industry standard for crude oil samples historically has been tin-plated metal cans because metal cans are necessary for compliance with air transport regulations. Because mercury amalgamates with tin, the concentration of mercury measured in oil contained in a tin-plated metal can is biased (very) low. The loss of elemental mercury in any type of polymeric container has been observed in several studies [8].
Results and Discussion

Issues related to variations in measured THg associated with actual crude oil samples acquired using field sampling techniques from actual pipelines and from tankers during discharge to refinery tanks were examined. The items of concern were organized by task with each task designed to answer specific questions. Tasks were added and analytical procedures changed as data were acquired in the course of the experimental work.

Method Comparison
Total mercury MDLs, as defined by U.S. EPA [15], were determined by analysis of reagent dodecane using each laboratory's standard procedures, which differed because of instrumental differences. Data are recorded in Table 3. LAB l ' s extractive method was blanked using extraction reagents and reflects reagent background. Subsequent analyses of dodecane by LAB 1 were blank corrected. LAB 2's blanking method was the detector response from a combustion cycle and reflected instrument and carder gas cleanliness. LAB 2's MDL determinations using dodecane are not blank corrected. LAB 3's MDL was determined from the background emission of irradiated vials. Because of the volatile nature of crude oil and the associated safety restrictions on transportation of samples, significant time intervals can exist between when oil is sampled and

186

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

when samples are analyzed in laboratories. One hypothesis concerning sample stability was that certain species o f mercury could partition to bottle walls or sediment irreversibly over time. This would result in lower than actual measured concentrations, if the adsorbed mercury could not be redistributed to the liquid phase by a simple homogenization procedure (heat and shake). Secondly, it was thought that volatile mercury could be lost when sample bottles were opened. Lastly, it was thought possible that species' conversion by oxidation due to exposure to air might occur, which would produce a redistribution o f species being more inclined to attach to bottle wails or sediment and thus become elusive to aliquot removal. All of these hypotheses would translate into measured amounts o f mercury in aliquots that would be sensitive to homogenization procedure or frequency of bottle openings, and thus concentrations could diminish over time or in successive analyses. TABLE 3--MDL determination, LAB 1 LAB 2 (/ag/kg) ____(Egz/kg) _ ( ~ g / k g ) (~g/kg)_ Method Method Dodecane* Dodeeane* Blank Blank 0.22 0.01 <0.11 0.095 0.47 0.03 <0.11 0.182 0.15 -0.01 <0.11 0.153 0.18 0.00 <0.11 0.100 0.22 0.00 <0.11 0.158 0.11 -0.03 <0.11 0.172 0.11 -0.07 0.172 -0.03 0.206 -0.04 0.195 0.21 -0.02 0.159 Meall 0.12 0.03 0.039 SD 0.14 0.11 MDL *Corrected for method blank. tNot corrected for methodblank.

LAB 3 (~tg/kgQ____ Method <2.3 <2.4 <1.3 <1.6 <1.6 <1.1 <1.1 <1.3 <1.1 ~2

Data for the method comparison study are reported in Table 4. Method performance and time-dependant variation o f measured total mercury concentrations were compared using three field-sampled crude oils (US4, US19, and MX4) that were selected based on crude oil assays of gravity, metals, and sulfur content. Crude oils were sampled in the field according to a uniform sampling protocol (125 ml glass bottles with perfluorocarbon lined lids; three grab samples not composited; 25 mL headspace in sample bottles). Individual sample containers were distributed to selected laboratories within one week o f when the samples were taken. Replicate analyses of oil from a single sample container from the first sample event were performed at each of the three laboratories (LABs 1, 2, 3) repetitively at approximately one month intervals. For each analysis, triplicate aliquots were removed from samples that had been warmed (30-40~ and manually shaken prior to bottle opening. The first sample event was repeated three additional times. In Table 4, the sample event is indicated by the number after the oil code. Sample events did not follow a time sequence but were independent. Independence was defined as samples taken fi'om different tankers or at least two weeks apart for oil delivered by pipeline.

WILHELM ET AL. ON CRUDE OIL

187

TABLE 4--Method comparison-total mercury in commercial crude oils. Sample ID LAB 1 (SD) 0tg/kg) LAB 2 LAB 3 (SD) (~g/kg) (SD) (pg/kg) Mean (~g/kg) At (days from sampling) LAB 1/2/3 46/41/43 73/67/78 99/97/121 _ 42/45/46 69/71/81 78/101/124 16/11/5 42/37/40 68/67/53 27/23/32 27/23/35 19/19/14 27/23/12 23/29/41 67/67/56

US4-1* 1.8 (0.7)t 2.4 (0.4) <10s 2 (mean labs 1, 2) US4-1 0.7 (0.2) 2.4 (0.1) 10 4 US4-1 2.2 (0.2) 2.0 (0.I) <10 2 (mean labs 1, 2) US19-1 33.9 (0.6) 31.4 (0.6) 41 (2) 35 USI9-1 36.1 (2.5) 32.8 (2.8) 44 (5) 38 US19-1 33.6 (3.3) 33.3 (1.0)__ 50 (3) 39 MX4-1 0.1 (0.5) 4.5 (0.5) <20 2 (mean labs 1, 2) MX4-1 1.1 (1.2) 1.7 (0.1) <20 1 (mean labs 1, 2) MX4-1 1.1 (0.5) 1.7 (0.1) <20 1 (mean labs 1, 2) US19-2 8.2 (0.6) 11.2 (0.6) 9 (4) 9 US19-3 31.3 (5.6) 35.9 (1.1) 34 (3) 34 US19-4 32.5 (2.6) 34.3 (1.7) 48_(_3) 38 MX4-2 -0.3 (0.6) 9.8 (1.3) <20 5 (mean labs 1, 2) MX4-3 3.0 (0.7) 5.1 (0.4) <20 4 (mean labs 1, 2) MX4-4 1.4 (0._5) 2.1 (0.4) <20 2 (mean labs 1, 2) Homogenization 40 40 40 T (~ Aliquot Taken Hot Warm Warm * US4 is the code name for the oil. Numbers after the dash refer to sampling events. For example, US4-1 refers to oil acquired in the first sampling event. Three grab samples were obtained at each sampling event. t Reported data are the mean of triplicate analyses. Standard deviations are in parenthesis. < indicates less than MDL due to interference. The 1NAA method suffered from interferences caused b y nickel and selenium discussed previously. The selenium correction procedure provided reproducible results in samples containing both Hg and Se, but the resultant detection limits were substantially elevated. Nickel in oil was likewise problematic due to a general increase in background ~ radiation levels. The combination o f the Se and Ni interferences produced high detection levels for oils that contained either or both nickel and selenium. For these reasons the I N A A method was not used in any additional tasks. The variation in mercury content o f oils having the same trade name was examined. Samples were obtained from independent sampling events. Independence was defined as different tankers or at different times from pipelines (weeks apart). Two o f the three oils examined were found to exhibit significant variation in mercury concentration from one sample event to another. The US19 oil exhibited concentrations o f roughly 38, 9, 34, and 38 Ixg/kg (average of 3 laboratories) in 4 separate sampling events and the MX4 values o f 2, 5, 4, and 2 ~tg/kg (average o f 2 laboratories). The variation in mercury concentration o f market-named oils over time was important to project objectives, in that, i f the temporal variation o f mercury content in oils was found to be large, the number o f individual samples required to define an exact mean and range for a particular market-named oil likewise would be large. It was not surprising that oil concentrations were found to vary from tanker to tanker and in pipelines given the blending o f oils that typically takes place in production systems.

188

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Speciation
Speciation and bottle wall determinations (Table 5) were performed using procedures developed by Bloom [8]. Fresh aliquots were removed from the same sample bottles used for method comparison experiments (Table 4). Data in Table 5 are independent. Separate aliquots were used for each determination. TABLE 5--Speciation data (LAB 1). Dissolved Suspended Hg2 Hg~ Walls Sample ID (~tg/k~) (btg/kg) 0-tg/kg) (lag/kg) (~tg/kg)US4-2 rep 1 1.4" 0.0 1.0 0.2 0.06 US4-2 rep 2 1.6 -1.2 1.0 -0.4 mean 1.5 -0.6 1.0 -0.1 SD 0.1 0.8 0.0 0.4 0.04 US19-2 rep 1 3.3 4.6 2.4 -0.4 US19-2 rep 2 2.9 5.1 2.6 -0.2 mean 3.1 4.9 2.5 -0.3 SD 0.3 0.4 0.1 0.1 MX4-2 rep 1 0.6 -1.1 -0.2 -0.5 0.07 MX4-2 rep 2 1.2 -0.9 -0.2 0.2 mean 0.9 -1.0 4).2 -0.2 SD 0.5 0.1 0.0 0.5 * All data corrected for method blank. The speciation measurements did not find volatile mercury (assumed to be Hg ~ [8]) in any of the samples. Mercury in samples consisted o f ionic and suspended forms. The fact that low THg concentrations were found in the oils examined was unfortunate, given the goal of understanding species effects. Partition o f mercury to bottle walls was not discovered to any significant degree for any of the three oils examined. The partition of mercury to sample bottle walls is a known complicating effect to measurement o f THg in some crude oil samples based on published investigations [8]. It certainly may be the case that some oils demonstrate container wall adsorption to a much greater extent than that discovered in the small number o f oils examined in these experiments. Likewise, it was suspected that any volatile mercury may have been lost or oxidized by the time the speciation experiments were performed. The samples used for the speciation experiments had been opened several times for analysis and density measurements prior to speciation experiments.

Sample Homogenization
For the data reported in Table 4, samples ofoil in 125 ml bottles were warmed, shaken, and then aliquots were removed using a syringe (LAB 2) or micro-pipette (LABs 1, 3). The temperature at which the aliquot was removed varied between hot (40~ LAB 1) to an intermediate temperature between 40~ and ambient (warm, LABs 2 and 3). It was thought that procedures used to remove hot aliquots may have provided opportunities for loss o f volatile mercury or may have allowed reactions to take place possibly with oxygen in air when bottles were opened for aliquot removal.

WILHELM ET AL. ON CRUDE OIL

189

Variations in measured THg were examined as a function of the manner in which samples were homogenized prior to removal of aliquots. LAB 2 examined the homogenization procedure by measuring concentrations in unheated samples followed by homogenization at 40~ and vice versa. As opposed to the situation associated with data in Table 4, the sequence of experiments started with a sample bottle that had not been opened previously. Care was taken to minimize the opportunity for evaporation by removing aliquots quickly and by closing containers between aliquot removals. Data are reported in Table 6. Samples analyzed first exhibited higher concentrations, irrespective of whether the samples were heated (40~ or not. Close examination of quality assurance data and standard deviations obtained from replicate analyses suggested that the downward trend in measured concentrations was real and not the result of random variation. ..
Sample

TABLE 6--Homogenization procedure (LAB 2).

.....
_

Homogenization Temperature (~ 25~ 80/40~ 40~ 40oc (lSt _ g _ ~ 0 / 0 2 _ ) _ _ _ _ ( 2 "d opening - 7 / ~ o E e _ n i n g o~ - 8/16/~10/15/02) Otg/kg) (~tg/kg) (g/kg) (gg/kg) US4-1 5.6 (0.2)* 2.4 (0.4) 2.4 (0.1) 2.0 (0.1) MX4-1 12.4 (1.5) 4.5 (0.5) 1.7 (0.1) 1.7 (0.1) US19-1 51.0 3~.4) 31.4 (0.6) 32.8 (2.8) 33.3 ( 2 . 9 ) 40~ 25~ . _ _ _ _ _ ( 1 ~topem~g 8/16/02) (2~do~_nin~ 8/16/02) MX4-2 9.8 (1.3) 2.5 US19-2 11.2 (0.6) 7.5 * Reported data are the mean of triplicate THg analyses. Standard deviations are in parenthesis.

The reason for the observed decline in measured mercury concentration in successive measurements is thought to lie in the fact that volatile (elemental) mercury in the headspace (25 ml) of the sample bottle is lost when the bottle is opened. When the bottle is closed, volatile mercury in the liquid phase redistributes to the headspace of the bottle, and thus the first aliquot reflects a higher concentration (in the liquid phase) than aliquots obtained in successive removals. If volatile mercury is lost when bottles are opened, one would expect the measured concentration in liquid aliquots to diminish the more times a bottle was opened until most of the volatile mercury had escaped and only non-volatile mercury forms remained. The data suggest that only a relatively small number of bottle openings is sufficient to deplete a significant portion of the volatile mercury in the sample bottle. The homogenization procedure was further examined using an oil sample of HgS suspended in paraffin oil. This sample suffered .from inconsistent results when analyzed previously [1]. Lack of reproducibility was attributed to the inability to produce a homogenous HgS suspension prior to aliquot removal, at least using simple techniques (heat and shake). Table 7 demonstrates that ultrasonic treatment is much more effective to homogenize samples than a heat and shake approach.

Headspace Mercury
Loss of volatile mercury as might occur due to partitioning of volatile mercury to the headspace of sample containers was examined. The term "headspace" refers to the portion of the sample bottle not occupied by liquid. The hypothesis was that the total initial measured

190

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

concentration o f mercury in a freshly taken and homogenized sample was closer to the actual sum of volatile and non-volatile mercury components in the sample and that volatile mercury components can escape when bottles are opened. TABLE Aliquot Conditions 20~ shake 40~ shake 30~ ultrasonic 30~ ultrasonic THg n g / g (lag/kg) 16.9 611 1312 1470

7--Homogenization procedure (LAB 2).


Duplicate (~tg/kg) 7.9 397 1361 1414 Mean (I.tg/kg) 12.4 504 1338 1442 RPD 72 42 2 2 Target (~tg/kg) 1659 1659 1659 1659 Reanalysis Recovery (%) 1 30 81 87

Elemental mercury (Hg ~ spiked dodecane (C 12) was added to a crude oil sample that had been previously analyzed repetitively (numerous bottle openings). Forty mL of US19-4 and 40 mL of C12/Hg~ were mixed in a 125 mL glass bottle, and the mixed sample was identified as US19-C12/Hg~ Each of three bottles, C12/Hg ~ US19-4, and US19-C12/Hg ~ had about 80 mL of sample (~68 g) and 45 mL o f headspace. In addition, a H~ ~ spiked toluene sample was also constructed. Initial concentrations are shown in Table 8, HgU (liquid phase) was measured by sparging an aliquot and trapping mercury in the sparge gas [16]. In Table 8, the amount of mercury (ng) in the liquid phase of the constructed samples was measured (THg in liquid phase). The concentration o f mercury in US 19-C12/Hg ~ was calculated from added mercury (oil density 0.85). TABLE

S--Mercury in samples usedfor headspace experiments (LAB 2).

Hg~ THg Hg~in the sample bottle (liquid phase) (liquidphase) (~tg/kg) (~tg/kg) (ng) Hg~spiked C12 124.4 (n = 2) 145.4(n = 2) 8000 Hg~spiked toluene 14.5 14.8 (n = 2) 1000 US19-4 ND 40.0 0 US19-C12/Hg~ 60 (calculated) 90 (calculated) 5000 Date - 1/13/03 The experimental procedure for headspace Hg measurements involved removing headspace vapor using a volumetric syringe (20 mL) without opening the sample bottle. The syringe was discharged onto a gold trap to capture the mercury in the gas volume extracted by the syringe. Liquid phase samples (same bottle as headspace samples) were also removed with a syringe and analyzed for THg. The volume o f the syringe used to measure headspace mercury concentration was thought to approximate a bottle opening. Headspace mercury was sequentially analyzed as a function of time, temperature, and bottle equilibration procedure (Table 9). The 40 ~tg/kg THg (liquid) concentration exhibited by oil US 19-4 represented its residual non-volatile mercury concentration as demonstrated by the low values o f mercury found in headspace measurements. In constructed samples, the mercury that partitioned to the bottle headspace and then retrieved in successive headspace analyses originated from elemental mercury added to the sample. The initial attempt to measure mercury in the headspace o f the US19-C12/Hg ~ sample bottle produced an off scale response by the detector, and thus the amount lost in the first attempt at measurement is not known but is certainly high. Leaving the bottle open for 24 h was sufficient to deplete the elemental mercury originally added to the bottle.

WILHELM ET AL. ON CRUDE OIL

191

TABLE 9--Sequential measurements of heads ace and liquid phase THg in spiked oil (LAB2). US19US19US19-4 C12/Hg0 C12/Hg0 Toluene/Hg~ Order of Headspace Headspace Headspace Liquid Date Conditions Experiments Hg Hg Hg THg (rig/20 rot) (ng/20 ml) (ng/20ml) (p.g/kg) Shake/settle/no heat ...* > 100 ... 9 1/13/63 1st 5th Shake/settle/no heat 18.8 28.9 1/14/03 Ist No shake/no heat 01i 25.4 73.'9 50~ 25.6 72.9 4th 70.9 No shake/go heat NDt 16.1 1/15/03 1st 40.8 2nd Shake/no heat 0.2 19.6 53.8 3rd 50~ 0.5 43.7 67.9 115.2 1/16/03 1st No shake / no heat ND 15.8 44.7 70~ heat/shake/settle 1.5 44.6 2nd 117.6 Bottle left open with 1/17/03 1st stirring (20~ for 25 h 4.3 34.2 70~ shake/settle *not measured. t not detected. temperature of vapor aliquotremoval. The amount of mercury in the syringe should be proportional to the concentration o f mercury in the liquid phase, if equilibrium was established. While the toluene sample was constructed at a lower concentration (14.5 ~tg/kg) than the US19-C12/Hg ~ sample (124 lag&g), it consistently provided more mercury to the vapor phase. This demonstrates the effect o f liquid phase composition on mercury volatility. It is not known if complete equilibrium was established between mercury in the vapor and liquid phase, but the qualitative trends are apparent. Volatile mercury partitions to the vapor space o f sample containers at a significant rate. Figure 1 shows a compendium o f THg measurements for both constructed and real crude oils on a (natural) log scale for clarity of trend in time dependence. Real, as opposed to elemental mercury spiked crude oils, decay to a seemingly constant value that is thought to represent the non-volatile fraction. The elemental mercury-spiked dodecane decreased in mercury concentration from approximately 170 ppb originally to approximately 7 ppb after numerous bottle openings for various purposes.

Identification of Volatile Species


The procedure to identify volatile mercury species followed the methods o f Liang [16] for determination o f dimethylmercury and elemental mercury in vapor. An open oil sample vial (40 mL) was placed in a Teflon chamber (250 mL, see Fig. 2). The chamber was then purged with nitrogen for varying periods of time at 165 mL/min. The purge gas exiting the chamber flowed sequentially through a Tenax trap and gold trap. The Tenax trap adsorbs organic or organometallic (DMHg) compounds but not elemental mercury. The gold trap amalgamates both DMHg and Hg ~ After purging the chamber, the gold trap was analyzed by thermal desorption/CVAF. The Tenax trap was heated under flowing argon to desorb mercury species onto a gas chromatography column with CVAF detector (column 1 m; 85~ carrier gas flow rate 50 mL/min.; DMHg peak retention time 2.7-2.8 min.; Hg ~ peak retention time 1.7 min). The particular crude oil selected for the experiment had a natural concentration of total mercury close to 600 ppb and was selected so as to have a sufficient quantity o f evaporated mercury for

192

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

measurement. Data (Table 10) show the amount of mercury collected on the traps as a function of purge time. Both Hg~ and DMHg are clearly identified. Hg~ is dominant in the purge gas, but the relative amounts of Hg~ and DMHg in the oil cannot be inferred because the rate of evaporation of each species is unknown.

FIG. 1--THg decay in solvent and crude oil samples.

FIG. 2--Apparatus for volatile mercury species determination.

WILHELM ET AL. ON CRUDE OIL

193

TABLE l O---Vo(atile mercury species (LAB 2)" Purging Time . . . . . . . 2.5 rain 10 rain DMHg, ~tg(Tenax) N.D. 0.0023 H~~ ~tg(sold) . . . . . . . 0.048 0.045
Conclusions

~30 rain 0.0050 0.092

Mercury concentrations in constructed and actual crude oil samples were measured using three analytical methods that were compared with respect to accuracy, precision, and detection limit. The combustion method (U.S. EPA 7473 hybrid) and a commercial extraction method (non-standard) were found adequate to provide a good combination of sensitivity and accuracy, while instrumental neutron activation analysis was found to suffer from interferences from elements other than mercury but typically in crude oil. In the combustion method, direct syringe injection of aliquots to the combustion chamber was found advantageous in that it minimized opportunities for loss of volatile mercury. Mercury in crude oil consists of both volatile and non-volatile forms. The study identified two volatile species (Hg~ and DMHg) that partition to sample bottle headspace. The volatile forms of mercury are difficult because they can be lost from sample containers when containers are opened, lost from aliquots that are removed from containers, and lost in aliquot processing steps prior to element detection. The non-volatile forms were not investigated for identity but could include suspended HgS, mercury associated with asphaltenes, and mercury chemisorbed to normal sediment. Homogenization of oils that contain suspended mercury was found important to allow retrieval of representative aliquots for analysis. Ultrasonic agitation at slightly elevated temperature yielded the more reproducible and higher magnitude results. Dilution of some viscous oils may be necessary to assist homogenization or to assist aliquot removal with a syringe, but dilution could allow the possibility of loss of volatile mercury when bottles are opened for dilution. Speciation experiments on oils in sample bottles previously opened numerous times did not discover any elemental mercury. The lack of elemental mercury in speciated oils likely was due to the loss of this component during prior bottle openings as evidenced by the lower total mercury concentrations measured in the speciation experiments compared to those obtained previously upon first bottle opening. The volatility of elemental mercury was demonstrated by analysis of headspace vapor in sample bottles containing oil and solvents spiked with elemental mercury. The amount of mercury that partitions to bottle headspace depended on temperature and on liquid phase composition, as expected. Based on these results, one would expect significant variation in the rate of mercury evaporation depending on oil composition. Oils that contain a volatile mercury component typically do not analyze reproducibly unless precautions are taken to prevent escape of the volatile mercury. Oils that analyze reproducibly irrespective of bottle openings or other precautions likely have lost their volatile component and may be unrepresentative. Such oils demonstrate reproducible concentrations over time. The ongoing project did not investigate sampling procedures per se, but the principles that have been elucidated no doubt apply to sampling as well. Sampling hot oil or using methods that are slow and allow prolonged contact of oil with the atmosphere while filling containers provide opportunities for the volatile mercury component to escape. The sampling methods now employed in the project entail filling 40 mL glass containers with oil flowing from a conditioned

194

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

sample port as quickly as possible (a few seconds), leaving a very small portion of the sample container as headspace. We have adopted a 40 mL glass sample vial with a Teflon septum as an improvement to standard glass bottles utilized previously. The vial is filled almost full (1-2 mL headspace) and is not refrigerated. Samples are homogenized ultrasonically. The ultrasonic energy warms the sample slightly. Aliquots are removed using a pre-conditioned volumetric micro-syringe and injected directly into the vaporization/combustion chamber of the apparatus for determination. Using these techniques, reproducible and higher magnitude concentration values are found for the majority of actual crude oil samples so far investigated in the project. To meet the objectives of the EPA/APUNPRA project, measuring the sum of dissolved volatile, dissolved non-volatile, and suspended mercury compounds is important as each type enters the refinery in crude oil. The newly developed procedures for sample homogenization, for retrieving representative liquid aliquots, and for preservation of volatile components are thought to eliminate a major portion of the negative bias that can exist in routine sampling and analysis of crude oil for determination of total mercury content.

Acknowledgments
The U.S. Environmental Protection Agency through its Office of Research and Development funded and collaborated in the research described here under contract 68C 02 088. The material in this document has been subjected to Agency editorial and policy review and approved for publication. The views expressed by individual authors, however, are their own, and do not necessarily reflect those of the U.S. Environmental Protection Agency. Mention of trade names, products, or services does not convey, and should not be interpreted as conveying, official EPA approval, endorsement, or recommendation. The participation and cooperation of the American Petroleum Institute (Karin Ritter) and the National Petrochemical and Refiners Association (Norbert Dee) are gratefully acknowledged. The donation of crude oil samples by API and NPRA member companies is sincerely appreciated.
References

[1] Wilhelm, S. M. and Kirchgessner, D. A., "Mercury in U.S. Crude Oil: A Study by U.S. EPA,

API and NPRA," SPE/EPA/DOE Exploration and Production Environmental Conference, San Antonio, TX, Society of Petroleum Engineers Paper 80573, 2003. [2] Brown, T. D., Smith, D. N., Hargis, R. A., and O'Dowd W. J., "Mercury Measurement and Its Control: What We Know, Have Learned, and Need to Further Investigate," J. Air & Waste Manage. Assoc., June 1999, pp. 1-97. [3] Mercury Study Report to Congress, U.S. EPA, Washington, DC, EPA-452R/R-97-004, 1998. [4] Wilhelm, S. M., "An Estimate Mercury Emissions to the Atmosphere from Petroleum," Environ. Sei. Tech., Vol. 35, No. 24, 2001, p. 4704. [5] Wilhelm, S. M. and Bloom, N. S., "Mercury in Petroleum," Fuel Processing Technology, Vol. 63, 2000, pp. 1-22. [6] Shah, K. R., Filby, R. H., and Hailer, W. A., "Metals in Crude Oil by Neutron Activation Analysis,",/.. Radioanal. Chem., Vol. 6, 1970, p. 413. [7] Filby, R. H. and Shah, K. R., "Neutron Activation Methods for Trace Metal Analysis of Crude Oil," The Role of Trace Metals in Petroleum, T. F. Yen, Ed., Ann Arbor Science

WILHELM ET AL. ON CRUDE OIL

195

Publishers, Ann Arbor, MI, 1975. [8] Bloom, N. S., "Analysis and Stability of Mercury Speciation in Petroleum Hydrocarbons," Fresenius'J. Anal, Chem., Vol. 366, No. 5, 2000, p. 438. [9] Liang, L., Horvat, M., and Danilchik, P., "Novel Analytical Methods for Determination of Low Level of Total Mercury in Petroleum and Its Products by Gold Amalgamation CVAFS," Proceedings of the 4 th International Conference on Mercury as a Global Pollutant, Hamburg, Germany, August 4-8, 1996. [10] Kelly, W. R., Long, S. E., and Mann, J. L., "Determination of Mercury in SRM Crude Oils and Refined Products by Isotope Dilution Cold Vapor ICP-MS Using Closed-System Combustion," Anal, Bioanal. Chem., Vol. 376, 2003, pp. 753-758. [11] Liang, L., Horvat, M., Fajon, V., Prosenc, N., Li, H., and Pang, P., "Comparison of Improved Combustion/Trap Technique to Wet Extraction Methods for Determination of Mercury in Crude Oil and Related Products by Atomic Fluorescence," Energy & Fuels, Vol. 17, No. 5, 2003, pp. 1175-1179. [12] Liang, L., Lazoff, S., Horvat, M., Swain, E., and Gilkeson, J., "One Step Determination of Mercury in Crude Oil and Related Products Using a Simple Lab Built Thermal Decomposition System," Fresenius' J. of Anal, Chem., Vol. 367, 2000, pp. 8-11. [13] Hitchon, B., Filby, R. H., and Shah, K. R., "Geochemistry of Trace Elements in Crude Oils, Alberta, Canada," The Role of Trace Metals in Petroleum, T.F. Yen, Ed., Ann Arbor Science Publishers, Ann Arbor, MI, 1975. [14] Musa, M., Markus, W., Elghondi, A., Etwir, R., Hannan, A., and Arafa, E., "Neutron Activation Analysis of Major and Trace Elements in Crude Petroleum," J. Radioanal. Nucl. Chem., Vol. 198, No. 1, 1995, p. 17. [15] 40 CFR Part 136, Appendix B. [16] Liang, L., Kichgessner, D. A., Wilhelm, S. M., Kariher, P. H., and Pang, P., "Occurrence of Dimethylmercury in Organic Solvents," Materials and Geoenvironment, Vol. 51, No. 2, 2004, pp. 1968-1971.

Journal of ASTM International, June 2005, Vol. 2, No. 6 Paper ID JAI13089 Available online at www.astrn.org

Brian S. Fox, 1 Kelly John Mason, Ph.D., 2 and Frank C. McElroy, Ph.D. 2

Determination of Total Mercury in Crude Oil by Combustion Cold Vapor Atomic Absorption Spectrometry (CVAAS)
ABSTRACT: Unlike many heavy metals, mercuryis a long-rangeair and water pollutantwhose load on the environmentis increasing. Accordingto publishedreports, a major source of atmosphericmercuryis the burningof coal for electricalpower generation. However,certain crude oils may euntainup to several hundred ng/g of total mercury. Accurate analysis of total mercury in crude oil is the first step in determiningthe value of a crude oil and one's ability to process the material within the refinery,while generating finishedproducts. This paper details the use of CombustionCold Vapor Atomic Absorption Spectrometry (CVAAS) technology for the analysis of crude samples and distillation cuts. The instrument described in the report is one of several instrument methods currently being evaluatedby groups in industry and government. The use of a boat-inlet sample introductionsystem allows crude oil and other types of material to be analyzed without acid digestion. The strengths and weaknesses of the combustion-CVAAStechnique,along with a year's worth of operating experience, will be presented from the user's perspective. KEYWORDS: mercury, crude oil, combustionCVAAS

Introduction
World consumption of hydrocarbons is steadily increasing. To meet current and future demand, crude oils that possess less desirable characteristics from a processing standpoint must be utilized. These are sometimes referred to as challenged crudes. Certain crude oils may contain significant concentrations o f arsenic, selenium, and mercury. These elements can affect how a crude oil is processed and what steps must be taken to mitigate its presence in product and waste streams. Mercury is a particular challenge, given its nature as a high profile pollutant with so many negative effects. When present in liquefied natural gas (LNG) product, elemental mercury may cause embrittlement of aluminum vessels and heat exchangers [1]. The potential for catastrophic equipment failure could be a direct outcome of processing mercury-containing LNG. Even when corrosion does not occur, elemental mercury can pose an exposure hazard to refinery workers during turnaround operations o f process units. Mercury may also find its way into fuel gas and wastewater streams with subsequent release to the environment. Once in the environment, mercury may be transformed by bacteria into organic forms, moving up the aquatic food chain where it reportedly causes neurological damage to fish-eating adults and birth defects and developmental problems in children [2,3]. Based on its serious environmental impact, the determination o f total mercury in crude oil is a priority. Researchers have used numerous detection techniques in the pursuit of total mercury Manuscriptreceived 15 October 2004; accepted for publication2 December2004; publishedJune 2005. Presented at ASTM Symposiumon ElementalAnalysis of Fuels and Lubricants:Recent Advancesand FutureProspects on 6-8 December 2004 in Tampa, FL. 1AdvancedResearch Associate,ExxonMobilResearch and Engineering,600 BillingsportRoad, Panlsboro,New Jersey 08062. 2 ExxonMobilResearch and Engineering.

Copyright 9 2005 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

196

FOX ET AL. ON ATOMIC ABSORPTION

197

measurement. They include Cold Vapor Atomic Absorption Spectrometry (CVAAS), Cold Vapor Atomic Fluorescence Spectrometry (CVAFS), and Isotope Dilution Cold Vapor Inductively Coupled Plasma Mass Spectrometry (ID-CV-ICP-MS) [4]. To accurately measure total mercury concentration in crude oil, several criteria were deemed important. Since both the analyte and the matrix are volatile, minimal sample handling and preparation are key attributes. Researchers have previously developed acid digestion schemes, using both open and closed vessel microwave systems to prepare food and biological samples for mercury analysis [5]. Petroleum products are often more difficult to digest than these sample types, requiring even more rigorous conditions. The hot acid digestions may contribute to inaccuracy from either thermal evolution of the volatile analyte or contamination. Extraction techniques have also been used to determine total mercury and its species in petroleum products [6]. The technique is effective, but it appears to be more labor-intensive than what one may desire for a refinery laboratory. An instrument that provided a direct determination of mercury in crude oil by thermal decomposition followed by CVAFS was built [7]. This approach appears to hold the promise of uniquely low detection limits, but unfortunately a fully automated instrument based on this design has not been commercialized. An instrument that uses thermal sample decomposition followed by gold amalgamation and CVAAS was specified in EPA Method 7473 [8]. A few commercial instruments based upon this technique have been used to analyze crude oil samples [9,10]. Another important attribute of the candidate instrument is that its response to mercury should be independent of the form in which it is found in the sample. The NIC instrument thermally decomposes the sample and converts all of the mercury to the elemental form, thereby ensuring that instrument response is independent of the mercury species. This enables the analyst to use a stable ionic form of mercury in an acidified aqueous medium for the purpose of instrument standardization. The focus of this paper is on the technique of Combustion CVAAS and its applicability to some of the samples encountered in the petroleum industry. It describes the equipment and test procedure adopted by our laboratory for the determination of total mercury in crude oil, condensate, and other process stream samples. Validation of this technique using standard reference materials (SRM), certified standards, and spiked samples is addressed. Instrument response to various organic mercury species and precision data derived from actual crude oil samples is presented.
Instrumentation

The instrument used to measure total mercury by Combustion-CVAAS was the SP-3D mercury analyzer manufactured by the Nippon Instruments Corporation (NIC), Osaka, Japan. The boat inlet design of the instrument allows for analysis of either solid or liquid samples. With a separate attachment, LPG and gas samples may be analyzed. The instrument uses air that is supplied by an on board pump as the carrier gas. Before use, the air is dehumidified and passed through an activated carbon filter to scrub away background mercury. A computer-controlled furnace in the ceramic tube thermally decomposes the material in the boat. A choice of six different heating cycles is available for combusting various sample matrices. Mercury released by the sample passes through a second 850~ decomposition furnace that has a copper catalyst insert. This step converts all of the mercury to elemental form. Mercury vapor and combustion gas pass through a scrubber filled with an aqueous phosphate buffer solution, then through a dehumidifier before reaching the first mercury collector. Mercury

198

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

is selectively retained by gold-coated Chromosorb, while combustion by-products pass through a multi-port valve and onto vent. The first mercury collector is heated, and the mercury vapor passes through a six-port Teflon valve and onto a second gold-coated Chromosorb trap. This second step concentrates mercury again and minimizes the presence of interfering species. This trap is heated, and the mercury vapor is swept into the cell of a CVAAS, where measurement is performed at 253.7 nanometers. The amount of energy absorbed in the cell increases proportionally to the number of ground state mercury atoms present. The concentration of mercury in the sample is quantified by comparison to the signal from a previously established calibration curve. The exhaust from the CVAAS cell passes through an activated carbon filter, where it is scrubbed of mercury before release into the atmosphere. Data from the analyses are collected by a PC running "Mercury MD-I" data analysis software. This software can store multiple calibration curves and raw data from the samples. The mercury response can be measured using either peak height or integration over time. The PC software is strictly for data acquisition and does not control the instrument start or any operating parameters.

Reagents and Materials


The cleanliness of equipment and reagents used in the test is of paramount concern. Mercury is a ubiquitous contaminant that can show up in surprising ways, especially at the range of concentrations in the samples. An example of potential contamination is the mercury compounds that may be used as preservatives in certain phosphate buffer solutions. The analyst must be continuously vigilant when choosing reagents and supplies. All water used in the analysis is deionized and polished with a Millipore Simplicity system, available from Millipore Corporation, Billerica, MA, USA. Two reagents are used in the sample boats to assist the combustion of organic materials. Additive B is a coarse aluminum oxide. Additive M is a 35:50 combination of calcium hydroxide and sodium carbonate. The additives, specially prepared by Wako Pure Chemical Industries, Ltd., Osaka, Japan, are available for purchase through NIC. The phosphate buffer in the scrubber vessel was prepared from a powdered concentrate (Wako). A stock standard aqueous solution of mercury chloride, 997 mg I"1 (Wako) was used to prepare a series of working standard solutions. The working standards are prepared in an aqueous L-Cysteine, 10 mg 1-1 in 0.2 % nitric acid. The L-Cysteine, 97 % was obtained from Aldrich Chemical Company, Inc. Milwaukee, W-I, USA. The nitric acid, concentrated, Trace Metal Grade was obtained from Fisher Scientific, Pittsburgh, PA, USA. Mercury free solvents were used to dilute heavy hydrocarbon samples, when necessary. HPLC Grade toluene, Optima Grade hexanes, and low odor kerosene (Fisher Scientific) were all found to have suitably low mercury concentrations for most work. Samples were caught in precleaned, 40-mL Volatile Organic Analysis (VOA) vials with Teflon lined septa (Fisher Scientific). The ceramic sample boats and high-form crucibles are available from NIC or other suppliers. A Hamilton Gastight Model 81000 100-#1 syringe (Fisher Scientific) was used to withdraw crude oil sample from the VOA vials. To perform the instrument validation and determine analytical figures of merit, stock standard solutions of dimethyl mercury, methyl mercury chloride, and ethyl mercury chloride were prepared by dissolving the pure compounds in HPLC Grade toluene. Each solution was prepared to approximately 100 ~tg/g of the mercury compound. These solutions were obtained

FOX ET AL. ON ATOMIC ABSORPTION

199

from Spectrum Quality Standards, Sugarland, TX, USA. To prepare sample spikes, 10 /xg/g certified mercury in oil standard was obtained from Conostan Division, Conoco Specialty Products, Ponca City, OK, USA. Aqueous mercury check standards, prepared in 10 % nitric acid were also obtained from SCP Science, Champlain, NY, USA. Finally, standardized reference materials SRMs and CRM were obtained from NIST, Gaithersburg, MD, USA and from the Canada Centre for Mineral and Energy Technology, Ottawa, Ontario, Canada.
Test Procedure

A general test procedure that describes instrument standardization and analysis of various sample matrices, including liquid hydrocarbons, is found in the NIC instruction manual for the SP-3D [11]. A second, nearly identical procedure that is specifically designed for the analysis of mercury in liquid hydrocarbons may be found in UOP Method 938-00 [10].

Instrument Preparation and Standardization


Ceramic crucibles that contain the necessary amount of reagent for the expected analyses are placed in a muffle furnace at 750~ the prior day. Ceramic sample boats are similarly stored in the muffle furnace to keep them free of ambient mercury contamination. The sample boats and reagents are allowed to cool before use in a covered pan that is over-pressured by dry, mercuryfree, filtered air from the instrument. It is important to allow the surfaces that the sample contacts to reach ambient temperature to minimize loss of elemental mercury. After choosing the appropriate instrument range and working standard, as shown in Table 1, the analyst measures aliquots of standard into the cooled boats using an adjustable pipettor with disposable tips. Instrument heating mode "1" is chosen for the aqueous standards or samples of a similar nature. Several heating modes are available, and examples of the samples for which each mode may be used are shown in Table 2. A calibration curve is established from the instrument response to various known mercury concentrations. TABLE 1--Working standards and injection sizes. MeasuringRange, n g . 2 20 200 ..... 1000 Hg Standard, mg/L 0.01 0.1 1.0 5.0 InjectionVolumes,.~L 50,100,150,200 50, 100, 150, 200 50,100,150,200 50,100,150,.200

...

TABLE 2--NIC SP-3D Heating Cycles and Sample Types Mode 1 2 3 4 5 6 ...... Sample Heating Cycle . 1 min - 350~ 4 min - 700~ 4 min - 350~ 6 min - 700~ 10 rain - 350~ 6 min - 700~ 2 min - 350~ 4 min - 700~ 4 rain - 600~ 6 min - 900~ Decomposition heating element off Sample Types Aqueous standards, drinking water Coal, fly ash, wastewater, soil Crude oil, heavy naphtha, lube oil Light naphtha, gasoline Sulfuric acid LPG, LNG, air

200

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Sample Boat Preparation


A reagent blank value must be determined, since the process of filling the sample boats leads to trace contamination by ambient mercury sources. The effect of the reagent blank value is most pronounced for samples with mercury concentrations at, or close to the minimum detection limit. Since the sample boats vary in shape and capacity, there is no hard rule for the amount of reagent to add. Approximately 2-3 g of Reagent B will fill the lower third of the sample boat. On top of this layer, about 3-5 g of Reagent M is used to bring the level of the reagents up to the top of the boat. The sample boat with reagents is placed at the entrance of the combustion tube and slid into the heated zone with a long fork. The cap at the entrance to the combustion tube is screwed into place, heating mode "3" is selected, and the start button is pressed. The mercury response values from this measuring cycle are discarded. While the first boat is being heated, a second sample boat is prepared with reagents in a similar manner. At the end of the cycle, the first boat is removed and allowed to cool in a covered tray. The second boat is also subjected to one heating cycle. After the second boat is removed, the first boat is placed into the heated zone of the combustion tube. The cap is screwed into place, heating mode "3" is selected, and start is pressed. The mercury response from this run is captured by the data acquisition software and used to calculate a reagent blank. The second boat is treated in a similar manner. It is important to allow sample boats to cool to ambient temperature before proceeding with sample measurements.

Sample Preparation and Analysis


Crude samples normally arrive for testing in 40-mL VOA vials with Teflon-coated silicone septa. Mercury loss needs to be minimized by opening containers as briefly as possible for sample transfer. If the crude oil sample flows into a syringe at ambient temperature, then all sampling may be done through the septum. For heavy oil samples, the containers must be opened briefly to withdraw an aliquot. Samples are placed in an ultrasonic bath for 1 h immediately prior to analysis. This step was recommended to us by the NIC representative to increase mercury recovery in a sample. Ice chips are added to the ultrasonic bath to prevent the temperature from rising more than 3~ above ambient laboratory temperature. The crude oil aliquot is obtained by piercing the septum on the container and slowly drawing the sample into a 100- #1 syringe. The container is inverted to dispel bubbles from the syringe. A cooled sample boat, loaded with reagent, is placed at the entrance of the combustion tube. The crude sample is injected beneath the surface of the reagent, and the boat is quickly pushed into the combustion tube. The cap is replaced, heating mode "3" has been pre-selected, and the start button is quickly pressed.

Heavy Crude Sample Preparation


Heavy crude samples do not flow at room temperature. Their highly viscous nature makes them impossible to withdraw from a sample vial with a syringe. These samples must be diluted before injection. The vial of heavy crude is placed in the ultrasonic bath for 1 h. Allowing the bath temperature to rise above ambient temperature often is enough to cause the heavy oil to

FOX ET AL. ON ATOMIC ABSORPTION

201

flow. The vial is quickly opened, and a weighed aliquot is transferred to an empty, tared 40-mL VOA vial. Toluene is added, a total weight is recorded, and the vial is re-capped. Typically, a 1:2 mass/mass ratio of sample to toluene is enough to lower the viscosity for withdrawal using a syringe. From this point, diluted samples are treated similarly to light crude oil samples.

Sample Mass Determination


The mass of the crude sample that was injected into the boat is determined by measuring a second aliquot of the crude oil sample. A capped 2-mL GC vial is tared on a balance. A 100-#1 aliquot of the crude oil sample is slowly drawn into the same syringe that is used to inject the sample into a boat. Bubbles are carefully excluded from the syringe. The sample is then injected into the tared GC vial and weighed again. This procedure allows the mass of an injection to be measured directly, rather than calculated based upon a separate density determination.
Results and Discussion

The response of the instrument is linear using the recommended standard solutions prepared in nitric acid and L-cysteine. Figure 1 shows a calibration curve for the low range. The correlation coefficient is typically 0.9995 or higher. Several organic mercury compounds were tested, and the instrument response was found to be independent of the mercury species in the samples. This allows the analyst to choose more stable inorganic mercury standards in acid solution. Inorganic standards stored in tightly capped, FEP Teflon bottles appear to last for a few months.

FIG. 1--Typical low range standardization curve o f the instrument.

202

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Certified materials were used to evaluate the accuracy o f the technique. These included coal, coal fly ash, soil, and crude oil. Table 3 shows that measured concentrations of mercury in the SRMs agree well with certified values. With only one exception, the Relative Percent Difference (RPD) is within 5 % of the certified values. The one exception is for a value that is less than half o f the claimed detection limit for the published method. Six repetitions were made for each material. The expanded uncertainty o f each measurement is calculated by multiplying the value for o f one experimentally derived standard deviation o f the data times the critical value. The critical value o f 2.57 is determined from a Student's t-distribution [12] with 5 degrees of freedom and a confidence level o f 95 %. TABLE 3--Total mercury results of standard reference materials. SRM Name Certified Hg, ng/g Found Hg, ng/g ..... RPD,% 10.9+ 1 11.1+ 1 2 NIST 1635 Subbituminous Coal NIST 2683b Bituminous Coal 90.0 + 3.6 94.5 + 4.7 5 NIST 2685b Bituminous Coal 146.2 + 10.6 152.2 + 1.4 4 0.0417 + 0.0057 0.047 + 0.031 12 NIST 2721 Light-Sour Crude Oil 0.129 + 0.013 0.135 + 0.06 5 NIST 2722 Heavy-Sweet Crude Oil 160 158 + 13 1 NBS 1633a Coal Fly Ash CRM SO-1 Reference Soil 2 22 21 + 2 5 i This is an inactive SRM obtained from NBS. The mercury value was obtained from the containeriabel. 2Current status of CRM is unknown. The CRM was obtained from CANMET, Ottawa Ontario, Canada. The mercury value was obtained from the container label. The NIC operating instructions call for weighing about 100 #g of sample onto a layer o f Reagent B in the sample boat and then covering that with a layer o f Reagent M. Other researchers had determined that this step causes the loss of volatile analyte when compared to sampling the crude oil through the Volatile Organic Analysis (VOA) vial septum using a syringe [13]. To minimize volatile analyte loss, we inject a volume o f sample below the layered reagents in the ceramic boat. The mass is obtained by measuring a separate 100-~1 aliquot o f the sample on a balance. After repeatedly testing a material of known density, it was observed that uncertainty around using a mass determined by this procedure was less than 0.5 % of the value. Some of this observed uncertainty may have been generated by temperature change, due to continuous handling o f the syringe. There are two factors that influence the minimum detection limit (MDL) for this method. The first is the maximum sample size for a crude oil sample. In the interest of combustion efficiency, it is recommended that no more than 100 mg o f an oil sample be used. The second limiting factor is the magnitude and variability o f the reagent blank signal. By carrying out a pre-heat cycle o f the reagents after they are loaded into the boat, the blank response is reduced along with its variability. Figure 2 shows how the blank response varies before and after a preheat cycle. With multiple heating cycles immediately followed by the reagent blank measurement, it is possible to drive the reagent blank down to the level o f the instrument blank. However, this will result in a low bias for the reagent blank signal. The best procedure is to allow the sample boat to cool to ambient temperature, thereby causing the analyst to treat the reagent blank measurement in exactly the same way as a sample measurement.

FOX ET AL. ON ATOMIC ABSORPTION

203

FIG. 2--Effect of a pre-heat cycle on the response of the reagent blank. The positive outcome of taking this extra step is that the mercury concentrations in the two available NIST crude oil SRMs that straddle the instrument detection limit are practically resolved. However, the uncertainty around the measured mercury values is about five times greater than what NIST obtained with ICP-MS. Many of the crude oils samples that have been tested to date in our laboratory have less than 2 ng/g of mercury. The use of a syringe to inject a volume of sample below the level ofpre-treated reagent has resulted in less variability between repeat injections being observed, along with the possibility of determining mercury concentrations at or near the claimed detection limit. The sensitivity and precision of this Combustion-CVAAS technique is compatible with published results of Liang et al. [7], who used CVAFS after thermal decomposition of crude oil samples. Crude oil samples that had very low measured mercury concentrations were spiked with a variety of organic mercury species to test the recovery. The results of spiked samples are shown in Table 4. Elemental mercury was not used to spike crude oil samples, due to the difficulty in preparing a known, stable concentration in an organic matrix. TABLE 4--Total mercury results of spiked samples. Sample ID Crude 1 Crude 2 Crude 3 Crude 4 Crude 5 Conostan Base Oil Conostan Base Oil Mercury Spike Compound Methyl Mercury Chloride Methyl Mercury Chloride Dimethyl Mercury Ethyl Mercury Chloride Conostan Hg Standard Conostan Hg Standard Conostan Hg Standard Total Hg, ng/g 15.9 + 0.4 32.3 + 1.1 13.2 + 1.1 8.7 + 0.7 33.7 + 1.1 11.0 + 1.6 104.5 + 2.1 Recovery, % 105 105 93 104 97 100 104

204

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Three analyses were made of each spike sample. The uncertainty was determined in a similar manner as before. The critical value of 4.3 is determined from a Student's t-distribution [12] with 2 degrees of freedom and a confidence level of 95 %. With just one exception, the recoveries were within 5 % of the expected concentrations. The fact that dimethyl mercury was the most volatile of the analyte species used in the study may have contributed to the slightly lower recovery of its spike sample. Our laboratory has tested over 80 crude oil and condensate samples to date. Three external laboratories also tested some of the samples as part of a cooperative effort to develop standard procedures for handling and analysis. The laboratories used techniques other than combustionCVAAS. Details of the outside labs' testing statistics are not known. Table 5 shows a representative subset of these results for comparison. Each analysis by our laboratory was done three times and averaged. The same statistics apply as those discussed in the preceding paragraph. The critical value of 4.3 with 2 degrees of freedom from the Student's t distribution table [12] exaggerates the small differences that were actually observed between the three mercury determinations. TABLE 5--Total mercury results of actual crude oil and condensate samples. Sample Type Crude 1 Crude 2 Crude 3 Crude 4 Crude 5 Crude 6 Crude 7 Crude 8 Condensate 1 Condensate 2 Total Hg, ng/g In-house Laboratory 0.3 + 0.4 3.9 + 1.3 82 + 2 1.5 + 0.4 0.5 + 0.4 0.6 + 0.4 1.4 + 0.4 10.1 + 1.7 450 + 15 51 + 3 Average Total Hg, ng/g Outside Laboratories 0.2 4.5 . . . . 1.6 0.2 0.6 1.3 . . . . . . . . . . . . RPD, % 20 14 . . 6 86 0 7 . . . . . .

Precision is good, and agreement with the outside laboratories is acceptable, considering the fact that many of the samples in the comparison have sub-ppb mercury concentrations that are just above the stated minimum detection limit of 0.1 ng/g for the technique. Since these samples were rtm over the course of a year, some did not benefit from accumulated experience with reagent blanks and the special handling required for low ppb mercury samples.
Observations

Several observations have been gleaned from more than a year of using Combusfion-CVAAS to analyze total mercury in crude oil samples and several other matrices. The instrument flow path must be kept clean. Carbon residue, no matter how slight, can cause dramatic precision and recovery problems. Teflon connectors and valves must be handled with care to prevent damage. Regular leak checking also helps maintain mercury recovery and precision. Additionally, laboratory cleanliness is imperative for accurate trace mercury

FOX ET AL. ON ATOMIC ABSORPTION

205

measurements. All sources of contamination must be controlled. The utensils used to dispense reagent and handle the boats must be kept covered. The boats themselves should be treated in a muffle furnace, before use. Their condition must be monitored, since the caustic Reagent M will erode the glazing from ceramic after a period of use. This may lead to precision problems with aqueous standards, since water has been observed being wicked over the side of the boat. Scrupulous flushing of the syringe with clean solvent minimizes analyte carryover from high mercury concentration samples. Running a blank between the high calibration standards and the samples also helps to flush residual mercury out of the instrument flow path. Container integrity can be a problem with certain samples. The Teflon surface of the septum on the 40-mL VOA vial serves as a barrier between hydrocarbons and the soft silicone rubber of the septum. Just a few punctures of this barrier by a syringe needle may allow the vapor to swell the septum. Both light crude oil and condensate samples can swell a septum to a point where it splits and releases material. These problems of container integrity and the volatility of both the matrix and the analyte make it a challenge to find a suitable Statistical Quality Control (SQC) sample for this test. Once it had been determined that instranaent response is independent of mercury species or matrix, any stable mercury sample, even a coal SRM, can be used to verify instrument performance on a daily basis. The solid samples seem to be far more stable than any mercury standard in a hydrocarbon matrix that we have prepared to date. The boat inlet design of the NIC Combustion-CVAAS instrument system precludes the addition of an auto-sampler. Due to the necessity of minimizing the residence time of the sample in the boat prior to the analysis, an auto-sampler may not be a practical accessory for the analysis of crude oil. Conclusion A mercury analyzer based upon complete sample combustion followed by CVAAS has proven to be a relatively simple instrument to operate and maintain. The technique requires practically no sample preparation for a light crude oil; therefore loss of analyte or contamination by mercury sources in the laboratory is minimized. By testing a variety of mercury species and sample types, we have demonstrated that instrument response is independent of the form of mercury and the sample matrix. This makes standardization with stable mercury species an easy and reliable procedure. The improvements in techniques for the trace level determination of difficult analytes, such as mercury in crude oil, serve to focus future efforts on proper sampling and preservation procedures. The inherent simplicity of a total combustion procedure for hydrocarbon samples removes many of the analytical variables that would otherwise confuse the issue.

Acknowledgements
The authors would like to thank Amy S. Brown of EMRE for providing the variety of samples used in the study.

206

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

References

[1] Edmonds, R., Moorwood, A. S., and Szczepanski, R., "Mercury Partitioning in Natural Gases and Condensates," Paper Presented to the GPA European Chapter Meeting, London, March 21, 1996. [2] Weise, E. and Watson, T., "Mercury in Many Lakes, Rivers," USA Today, Wednesday, August 25, 2004, p.1. [3] Raines, B., "Mercury Testing Planned in Gulf," Mobile Register, Friday, March 22, 2002, p. 4A. [4] Wilhelm, S. M. and Bloom, N., "Review - Mercury in Petroleum," Fuel Processing Technology, Vol. 63, 2000, pp. 1-27. [5] Microwave Enhanced Chemistry, Kingston, H. M. and Haswell, S. J., Eds., American Chemical Society, 1997, pp. 380-381. [6] Bloom, N. S., "Analysis and Stability of Mercury Speciation in Petroleum Hydrocarbons," Fresnius J. Anal Chem, 2000, Vol. 366, pp. 438-443. [7] Liang, L., Lazoff, S., Horvat, M., Swain, E., and Gilkeson, J., "Determination of Mercury in Crude Oil by In-Situ Thermal Decomposition Using a Simple Lab Built System," Fresnius J. Anal Chem, 2000, Vol. 367, pp. 8-11. [8] US EPA Method 7473 - Mercury in Solids and Solutions by Thermal Decomposition, Amalgamation, and Atomic Absorption Spectrometry. [9] Mercury Application Note #6 Rev. 01/20/04 - "Determination of Mercury in Crude Oil Using A Direct Mercury Analyzer," Milestone Incorporated, Monroe, CT 06468. [10]Total Mercury and Mercury Species in Liquid Hydrocarbons, UOP Method 938-00, UOP LLC, Des Plains, IL, 2004. [11] Instruction Manual for SP-3D, Nippon Instruments Corporation, Takatuki-shi, Osaka, Japan. [12]Anderson, R. L., Practical Statistics for Analytical Chemists, Van Nostrand Reinhold Company Inc., 1987, p. 299. [13] Teleconference with Dr. S. Mark Wilhelm.

"Journal of ASTM International, January 2006, Vol. 3, No. 1 Paper ID JA112988 Available online at www.astm.org Professor Peter B. Stockwell, 1 Dr. Warren T. Corns, 2 and Dr. Derek W. Bryce 3

Mercury Measurements in Fossil Fuels, Particularly Petrochemicals


ABSTRACT: The dangers of mercury and its derivatives, especially organomercury compounds have been well documented for almost half a century. Concerns remain regarding mercury contamination in aqueous ecosystems. Significant quantities of mercury are present in fossil fuels, particularly petrochemicals. We still manage to pollute the atmosphere and environment by the use of coal and petrochemicals. The impact of mercury in natural gas has been such that there have been at least six experiences of plant failure which have been traced to corrosion due to mercury attacks on aluminium rotors in the plant fabric. Very little attention is being given to mercury removal strategies, and in order for these to be successful there is an urgent requirement to determine the speciation profile of the natural gas or liquid hydrocarbon condensates. Atomic fluorescence spectrometry is an ideal analytical technique to determine the levels of mercury both before and after removal systems. The authors will outline achievements of both laboratory and online measurement systems for this important area of research. These systems have been operating at the required levels of performance and in the case of some of the online systems, have been in place operating 24 hours per day 7 days per week for more than 2 years. This has encouraged other companies to install similar systems. KEYWORDS: mercury, fossil fuels, petrochemicals
Introduction

The dangers o f m e r c u r y (Hg) and its derivatives, especially organomercury compounds, have been well documented for almost h a l f a century. Significant concerns remain regarding mercury contamination in aqueous ecosystems. The U S E P A has advanced water quality criteria [1] for the protection o f organisms native to water environments. The criterion for mercury in fresh water ecosystems is 12 nanogram/liter (ng/1), and the mercury chronic criterion for salt water is 25 ng/1. These extremely low criteria present significant demands for the analyst, and atomic fluorescence spectroscopy provides a viable option for the measurements. W e still manage to pollute the atmosphere and environment b y the continual use o f coal and petrochemicals as our m a i n sources o f energy. The impact o f m e r c u r y in natural gas has been such that there have b e e n at least six experiences o f plant failure which have been traced to corrosion due to m e r c u r y attacks on aluminium rotors in the plant fabric. The six experiences o f plant failure have had significant financial impact on the petrochemical companies. Since very Manuscript received 8 October 2004; accepted for publication 14 June 2005; published January 2006. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. i Managing Director, P S Analytical Ltd, Arthur House, Crayfields Industrial Estate, Main Road, St Paul's Cray, Orpington, Kent, BR5 3HP, U.K. z Research and Development Manager, P S Analytical Ltd, Arthur House, Craytields Industrial Estate, Main Road, St Paul's Cray, Orpington, Kent, BR5 3HP, U.K. 3 Application & Technical Support Specialist, P S Analytical Ltd, Arthur House, Crayfields Industrial Estate, Main Road, St Paul's Cray, Orpington, Kent, BR5 3HP, U.K.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

207

208

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

little attention is being given to mercury removal strategies and in order for these to be successful, there is an urgent requirement to determine the speeiation profile of the natural gas or liquid hydrocarbon condensates. Burning fossil fuels as a source of electricity also creates a source of mercury pollution. Hg (~ and Hg 2+ are both produced from electricity generation, but the fate of these species is totally different. Atomic fluorescence has long been available as a measurement technique, however, it suffers from matrix interference effects, which have limited its use. The measurement of mercury at the levels of interest and legislation mandate the use of vapor generation techniques as a measure of sample introduction. These techniques overcome the majority of the matrix interference effects and exploit the sensitivity and linearity of atomic fluorescence. The sensitivity gained is equivalent to or better than that achieved by the more expensive ICP-MS instrumentation and is two orders of magnitude better than those obtained by atomic absorption. It is therefore also important to know the speciation profile of the coal-fired stack emissions. The US Environmental Protection Agency is actively pursuing an ambition to regulate the levels of mercury that can be emitted from the chimneystacks. Therefore, there is a requirement for fast, reliable measurements, firstly to determine the levels of mercury species so that removal strategies can be determined. Both of these requirements are important for the control of environmental pollution. However, the measurement of mercury and its products in these complex matrices is extremely difficult and requires considerable attention to detail. The reliability of these measurements can be facilitated by the atomic fluorescence technique; it provides a simple but sensitive solution with linearity over several orders of magnitude.
Measurement

The use of atomic fluorescence for the determination of mercury was first reported by Thomson and Reynolds in 1971 [2]. Since then, several authors [3-6] have described enhancements to the technique that have reduced formal instrument detection limits (IDL) for the fluorescence technique to the 1-10 ng/1 range. Knox et al. [7] report on the use of atomic fluorescence detection limit for mercury to less than 1 ng/1. A European standard EN 13506 was published in 2001. This uses vapor generation coupled to direct atomic fluorescence measurement. The most recent version of the US EPA standard 1631 utilizes an additional gold amalgamation step. The amalgamation provides the potential for an additional order of sensitivity but also requires considerable attention to detail and cleanliness to avoid contamination. Although many improvements have been made to the measurement techniques over the last 30 years, there has been little attention paid to the Achilles heel of analysis, that of the sampling procedures. The regulatory requirements put a heavy burden on the analyst even for water analysis, and proper sampling strategies and quality reagents are required. Extending the procedures for the measurement of mercury levels in natural gas and coal fired stack emissions places far more difficult demands on the skills of the analyst. Both present very complex matrices, which although different in many respects, provide similar requirements to overcome the effects of the make up within the matrix. Measurement of the mercury levels is not in itself the main issue; firstly, it is important to obtain a sample, which is fully representative of the stream being examined without modifying it in any way. Secondly, the mercury present should be removed from its complex matrix and transferred into a stream of argon carrier gas. Passing the sample over a gold substrate, which collects all forms o f mercury prior to revaporizing into a stream of argon, has been well

STOCKWELL ET AL. ON FOSSIL FUELS

209

documented for these matrices. Requirements for an instrttment to achieve this are shown in Table 1 and are represented schematically in Fig. 1. The development of offline and online systems for measurement of mercury in process gas streams are discussed here. The applications are focused in the petrochemical industry and in the coal fired utilities operation. TABLE 1--Analyzer spec~cation. ..... 0.1 pg detection limit Easy to use Suitable for both on and offline operation 3000 ~tg/m3 range Typical 5 min per.sample

REMOTE "~g~ I ::::

PERMANENT TR/P

ToAFS

FIG. 1--Schematic diagram for samplinj mercury levels in gaseous samples. It is important that all forms of mercury are trapped and measured because the form of mercury in natural gas and liquid hydrocarbons changes in relation to the geographical location of the product. In South East Asia, for example, organomercury compounds are present. Not recognizing these as mercury can cause the mercury embrittlement referred to previously, since the processing arrangement can transfer the organic mercury into metallic mercury, which creates the problem and causes a plant shutdown. Table 2 shows that the Amasil trapping arrangement is efficient at trapping all forms of mercury and that it is ideal for control and measurement purposes, confirming the work ofDumarey et al. [8]
. . . . . . TABLE .

2--Trapping e,,ffilcien c y [8] .


(CH3)2Hg 98.9 99.7~0.5 99.5~0.4 CH3HgC1 99.9 99.1~0.6 100.7~0.8 (C2Hs)2Hg 99.7~0.4 100.3 99.9~0.4 C2HsHgC1 98.8 99.1 99.9~0.3

F10w Rate 1/min 1.0 2.5 5.0

HgO 99.8~0.3 101.0 99.9

HgC12 99.3~0.4 99.7 99.8

Natural gas is often pressurized above 3000 psig, and to take a sample at atmospheric pressure it is necessary to design a suitable pressure let down system, which avoids liquefaction of the hydrocarbon gases. For coal fired gas sampling, it is important to avoid any condensation of moisture present in the stack gas, since with the matrix containing sulfur dioxide, nitrous oxide, and carbon dioxide, these may result in acidic solutions. The reason for the measurement of mercury is often for regulatory requirements, but basically if you cannot measure the levels correctly, then you cannot control the processing or

210

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

produce a reliable product, nor can you sell it. The latter is particularly important in the petrochemical industry where mercury embrittlement has caused several plant shutdowns over the years, the latest being in Australia in January 2004. Natural Gas Figure 2 shows a schematic diagram o f a pressure let down system, which conforms to the European Standard ISO 6978 (ASTM 6350) for natural gas sampling. By careful choice o f components and temperature of operation it is possible to provide reliable data on the mercury levels present in the natural gas samples. By injecting defined samples of mercury into the sample line prior to the sample collection, it is possible to check the validity o f the measurements by testing the spike recovery. Recoveries in the region o f 95-105 % will show that there is little, if any, loss with the system, and the analyst can be confident of the levels found in the gas streams. Typical data for Hg in export natural gas are shown in Table 3. The reproducibility for triplicate measurements was found to be less than 4 %. For these measurements two traps are operated in series, and the total mercury is the sum of that collected on each of the traps. Since the levels are very low it is common practice to measure the levels from a trap without sample collection, i.e., a blank level, and to correct measurements for these blank levels.
PRIMARY BY-PASS RV 1 L RV1 VENT FIR 1 FIR 2 V E N T VENT

PIPELINE

ENCLOSURE

I
r

25 PSIG "~" RELIEF VALVE L P1 PRV 1 SAMPLE


i
4~

P2 FIR 2

(OPTIONAL) PROBE

ELECTF ICALLY HEATED REGULATOR

--O

~ i~J
:

I
INJECTION PORT

uJ

FIR I

ONE ADSORBER LINKS: 3 SAMPLEPORT 1 TO FIR 1 iN.


5 SERIESBYPASSTO FIR 2 iN.

I 1i

ill i

'~t~APII= ~ ~IESBYP FIR1 IN FIR21N ./~$$ PORT2 PORT1

FIG.

2--Schematic diagram for commercial pressure let down system.

STOCKWELL ET AL. ON FOSSIL FUELS

211

....... Sample Reference Export Gas (V4500) Export Gas (V4500) Export Gas (V4500) Export Gas (V4100) Export Gas (V4100) Export Gas (V4100)

TABLE Gas Volume (I)

3--Determination of mercury in export naturalgas.


Tube No 1 2 1 2 1 2 1 2 1 2 1 2 ng Hg/tube 41.88 2.897 42.59 2.056 38.79 2.738 45.20 2.182 42.63 2.103 44.98 1.438 ng Hg/tube blank corrected 41.38 2.293 42.04 1.644 38.43 2.255 44.83 1.305 42.29 1.465 44.108 0.873 ng Hg/sample 43.673 43.684 40.685 46.135 43.755 44.981 #g Hg/ra3 28.54 28.55 26.59 30.35 28.98 29.99 Mean #g/rn3 27.89 + 1.12 RSD = 4 % (n -- 3)

1.53 (T = 19~ 1.53 (T = 19~ 1.53 (T= 19~ 1.52 (T=22~ 1.51 (T = 23~ 1.50 (T = 25~

29.77 + 0.71 RSD = 2.4 % (n = 3)

Challenges for Online Natural Gas Measurements


Online systems can also be devised, but these provide additional challenges due to the safety aspects required for process stream measurements at petrochemical facilities. Sampling systems based on carefully designed and tested speed loops will be required, as will explosion-proof enclosures. A speed loop is a configuration where sample is taken out o f the process line at one point and returned to the process line at a lower pressure down the process line. This provides a sample that is continuously updated and representative o f the sample being processed. The online measurement o f mercury in natural gas presents additional problems: operating on a continuous basis put instruments on the design of the instrumentation, and the safety issues involved put an additional requirement on the suppliers. The mercury level in natural gas streams can be quite varied and the problems caused by high levels quite catastrophic. As such, online measurement can qae essential in spotting potential problems allowing a measured response to increased Hg levels, potentially saving millions o f dollars. A schematic block diagram o f the online system is shown in Fig. 3. Practical experience from a number o f operating companies has greatly assisted the process o f designing fully functional online systems that operate 24 hours a day on a continuous basis. Collecting a representative sample is a major challenge. Fundamental to the design is the ability to collect a fully representative sample and to do this from sampling streams which have pressures o f up to 3500 psig. The sampling must be carried out onto the traps at atmospheric pressure, so it is necessary to avoid any liquefaction in the traps. A series of speed loops is used to present the sample to a selection system, which introduces the relevant samples to the analyzer for measurement. The requirements for the sampling system are set out in Table 4.

212

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Industrial PC

O~m~rto (Can b e ~ p ~

~oams)
$k
Probe PSA 1O64O Filter

s~m

--

Stream Selector

Genenl~
FIG.

3--Block diagram of typical system.

9 9 9 9 9

TABLE 4 Sampling requirements. Sample is delivered using speed loop. A normally closed high pressure solenoid valve is included in the speed loop to stop the sample flow under alarm conditions. Heated regulator reduces pressure of sample to 6 psig. Pressure switch and relief valve are used to prevent high pressure entering enclosure. Pressure switch activates solenoid in alarm conditions. All components are explosion proof.

Calibration Accessory
In order to ensure that the measurements are valid, the authors have developed a simple calibration accessory that is traceable to NIST standards [9]. This allows a set value of mercury concentration to be introduced at different stages of the measurement/sampling setup to check the integrity of the measurements. All of the components in the sampling system must be tolerant to mercury being present and not absorb or desorb any mercury which might distort the measurements. It is essential to ensure that measurements are correct, and that reliable data are provided. This requirement is facilitated by specifically designed process control software which has the following features: 9 9 9 The online process control software provides a simple interface between the analyzer, the user, and the control room. The software allows the user to define not only the sampling and calibration frequency but also actions to be performed under alarm conditions The software connects to other systems either by 4-20mA, RS232, RS485, RS422, MODBUS, TCP/IP, or contact closure.

STOCKWELL ET AL. ON FOSSIL FUELS

213

The software is fully password protected to allow the user the ability to define who can do what. Figure 4 shows a typical installation on a plant where mercury removal beds have been installed. The mercury level has to be measured both at the inlet and outlet of the system. This requires that the measurement system can cover a wide concentration range, from a few ng/m3 to parts per trillion/m 3, typically up to three orders of magnitude. The Sir Galahad Online Natural Gas Analyser handles this reliably using only one detector. Icier S ~ ,

MweuW Removal Beds

4D"

41'-

FIG. 4--Schematic of typical plant installation.

Summary
Using the above criteria, a number of online and offline explosion proof sampling systems for Hg in natural gas applications have been designed, and these consist of a multi-stream process analyzer, which performs inlet and outlet measurements across Hg removal beds. To validate these measurements, an automatic calibration system was used and this has been tested by NIST [9] and found to give values within 5 % of theoretical concentrations.

Online Hg Monitor for Stack Emissions


Mercury has become a major concern with respect to the emissions from coal-fired power stations. Systems have been operated in the field since 2001 and have provided consistent and reliable data. Oxidized Hg is water soluble and is generally washed out of the atmosphere or adsorbed onto particles which are deposited locally with precipitation. In contrast to this, elemental Hg has an atmospheric lifetime between six months to two years [10]. Mercury emissions are therefore a global issue, since Hg can travel long distances with no distinction of boundaries. As a result of health concerns, there is an increasing number of international programs concerned with reducing mercury emissions to the environment. One of the most recent

214

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

estimates [11] for global emissions of Hg from human activities was 5000 tons/year, two-thirds of which are due to the emissions from the combustion sources worldwide. Other models predict differing levels, but all provide the urgency to solve problems [12]. The complex behavior of mercury species in the atmosphere and the continuous cycling of mercury between land, water, and air complicate the development of accurate models and estimates. Without reliable measurement data, research into understanding the fate and transportation of mercury within the environment and its potential control from combustion sources will be severely hindered. Durham et al. [13] have summarized the numerous problems of manual sampling: 9 9 9 9 9 9 9 They are slow, complex and elaborate They characterize emissions during a sampling period on processes that may be variable over time. They are man power intensive. They are prone to interferences. Speciation is difficult. They are not real time or online and are therefore limited in their use for process control or compliant monitoring. They are expensive.

Current methods for sampling and analysis are known to be problematic and prone to error. Several wet chemical impinger-based methods have been approved for the determination of total Hg in flue gas. These include US EPA method 29 and 101A. More recently, these methods have been modified to enable the speciation of particulate, oxidized, and elemental mercury. The most commonly used procedure is the ASTM Standard Test Method for Elemental, Oxidized, ParticleBound and Total Mercury in Flue Gas Generated from Coal-Fired Stationary Sources (D 678402) known as the "Ontario Hydro Method." This method was developed by public consultation with ASTM membership. There is a need, therefore, to develop online continuous emission monitors to enable accurate mercury speciation measurements in flue gas. Real time continuous monitoring of mercury is essential for numerous reasons. Control of mercury emissions from combustion sources is currently being considered, and when implemented it is likely to cost billions of pounds per year. Most of the mercury control strategies being proposed include some sort of sorbent injection. Prior to installing a Hg control device, long-term mercury speciation data will be required by the legislators and utilities to make more informed decisions concerning their needs and control options. Real time continuous monitoring of mercury would also provide options for advanced process control feedback, as well as monitoring the performance of control technology. This would minimize the sorbent usage and thus lower the operating costs. The development of Hg emission monitors would also benefit other stationary sources such as municipal and sewage waste incinerators, commercial and industrial boilers, medical waste incinerators, crematoria and smelters. In 1996, the authors' research team was invited by the Energy and Environmental Research Center (EERC), University of North Dakota to participate in a comparison study of mercury speciation measurements in flue gas. This proved to be a very difficult application given the low mercury concentration (1-4 ppbv) and difficult flue gas matrix. During these tests we discovered that the Amasil tube could be used to collect Hg species at elevated temperatures so that the condensation of acid gases could be avoided. Using a manual collection procedure, we achieved

STOCKWELL ET AL. ON FOSSIL FUELS

215

reasonable agreement (+/-20 %) with the wet chemical methods. However, we discovered later from Benchscale studies performed at the EERC that this approach had some limitations. The combination of nitrogen oxides (NOx) with hydrogen chloride (HCI) to form nitrosyl chloride (NOC1) or the presence of chlorine (C12) at low ppm concentrations in the flue gas severely lowered the collection efficiency of rig on the gold substrate. With the increasing demand for mercury speeiation measurements, we developed an online wet chemical speciation module. In this case, the hot flue gas was separated into two streams performing different chemistries. The first stream represents total Hg, since an alkaline starmous chloride (SnC12) reagent was used to convert divalent mercury to elemental mercury. The second channel utilized a reagent of potassium chloride (KC1) to complex divalent mercury, whereas the elemental fraction was un-retained. The difference between the two streams was attributed to oxidised Hg. Figure 5 summarizes the process in the form of a schematic diagram. The system was used to develop the first fully automated online speciated measurements of mercury in flue gas. As a research tool in the development and optimization of mercury control technologies, the system has proved invaluable.

I
LU

Prim~l~Q~s

3
U.

ap V~

FIG. 5--Schematic diagram of wet chemical speciation module and amalgamation - AFS. The majority of research to date has been completed on bench and pilot scale facilities. Testing in the field for long-term duration has proved to be somewhat more difficult. Numerous measurement issues have come to our attention, which need to be addressed before the full potential of rig emission monitors can be realized. One of the major disadvantages of the mercury measurement is the industry standard Ontario Hydro method. Whereas most mercury CEMs provide data every few minutes, the results from the Ontario Hydro method are often not available for 1-3 months. Experience from our research shows that the system described above will operate reliably over long periods of time and will provide data that compare well with the industry standard method. Figure 6 shows results of the series tests comparing the wet impinger based system with the Ontario Hydro method. The mercury CEM data agrees to within 10 % of the Ontario Hydro method, and it is available within a few minutes of the measurement being taken.

216

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 6--Comparison of Ontario Hydro and PSA Hg CEM.

Recent Developments for Continuous Emissions Monitors for Mercury in Stack Emissions
Regulations relating to the levels of mercury emissions are considered in the USA and elsewhere. Whereas the problems of measurement are similar to those for mercury in natural gas, the matrix is quite different and often more challenging. On the one hand, you have a hydrocarbon matrix, and in stack emissions you have a set acidic matrix which contains sulfur dioxide, sulfur trioxide, and other by-products from burning fossil fuels. The standard methods using the Ontario Hydro method and some commercial systems based on wet chemistry provide data that can correlate well. These are demanding on the involvement of staff to ensure that the instruments remain operational. A system that requires less intervention, essentially a dry-based system, is desirable. The challenges of this requirement are being address by several groups, but the inherent problems of a suitable reference method that can be easily operated is hampering this achievement. A schematic diagram of the dry based system developed by the authors is shown in Fig. 7. This allows the measurement of both speciated and total mercury levels in a continuous manner. Comparative results for the wet and dry based systems are shown in Fig. 8. This shows promise, and the next priority is to challenge the instrurnental system with a range of coal types and to ensure that comparative results are obtained across the full spectrum of available fuels.

STOCKWELL ET AL ON FOSSIL FUELS

217

FIG. 7--Schematic diagram for a dry based mercury speciation module.

FIG. 8--Comparison of data provided by both wet and dry based speciation module using a continuous emissions monitor for mercury.
Conclusions

Atomic fluorescence, coupled to innovative sample conditioning arrangements, offers a significant advantage over other approaches. In the petrochemical industry, understanding the fate of mercury is vital to the efficient production of the range of products. The presence of mercury will create financial implications, and therefore there is the motivation both to measure and control the mercury levels. In the coal fired utilities area, where the possibility of regulatory control is an issue, the commitment for the samples' potential to provide solutions at a fast time scale is not really evident, and this is delaying developments in this area. Wherever we look, mercury is around us; its measurement and removal from the environment are goals that will

218

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

only be beneficial if we continue to monitor the use of the fossil fuels. Atomic fluorescence spectrometry offers an affordable approach to the measurement of mercury, and several procedures to provide accurate methods have been described above.
References

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

EPA, Quality Criteria for Water, US Environmental Protection Agency, Office of Water Regulations and Standards, EPA 440/5-86-001, 1986. Thompson, K. C. and Reynolds, G. C., Analyst, Vol. 96, 1971, p. 771. Bloom, N. S., Fresenius JAnal Chem, Vol. 355, 2000, pp. 438-443. Stockwell, P. and Godden, R., Journal of Analytical Atomic Spectroscopy, Vol. 4, 1989, p. 301. Kammin, W. and Knox, R., Environmental Laboratory, August/September 1992. Potter, B. et al., US EPA Method 245-7, Revision 1-1, Environmental Monitoring and Systems Laboratory, Cincinnati, Ohio, 1994. Knox, R., Kammin, W. R. and Thomson, D., Journal of Automatic Chemistry, Vol. 17, No. 2, pp. 65-71. Dnmarey, et al., Anal. Chem., Vol. 57, 1985, pp. 2638-2643. "Mercury in Nitrogen Gas Research Gas Mixture," EPA Reference: DW13939860-01-0, G. D. Mitchell and W. D. Dorko, Anal. Chem. Div., Chemical Science & Technology Lab, National Institute of Standards and Technology. Ohnez, I., Ames, M. R., Gullu, G., Environmental Science and Technology, Vol. 32, 1998, pp. 3048-3054. Schindler, D. W., Proceedings for USEPA/NERL/NACEC Science experts" workshop on Hg, Oct 1998, Las Vegas, pp. 20-30. Mason, R. P., Fitzgerald, W. F., Morel, F. M. M., Geochimica et Cosmochimica Acta, Vol. 58, No. 15, August 1994, pp. 3191-3198. Durham, M. D., Schlager, R. J., Hyatt, D. E., Fuel Processing Technology, Vol. 39, 1994, pp. 285-299.

OTHER HETEROATOMS

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JAI12989 Available online at www.astm.org

Frank P. Di Sanzo I and John W. Diehl 1

Recent Advances in Gas Chromatographic/Atomic Emission Hetero-Atom Selective Detection for Characterization of Petroleum Streams and Products
ABSTRACT: The selective detection of organo-compounds containing hetero-atoms is used extensively in the petroleum industry for process control (sulfur, nitrogen, etc.) and additives in fuels (sulfur, nitrogen, oxygen, phosphorous, etc.). The most common selective atom detection is performed with atomic emission (AED), sulfur/nitrogen chemiluminescence (SCD, NCD), and photometric (FPD, pFPD). Atomic emission detection has an advantage over the other selective detectors since in an automated analysis scheme several hetero-elements may be detected for rapid sample screening. This paper describes the unique capabilities of atomic emission detection of fuels and petroleum fractious using several applications. In addition, a simplified version of comprehensive GCxGC that enhances hetero-atom selectivity by significantly reducing the matrix interference from the large hydrocarbon background is coupled with a microwave induced helium plasma atomic emission detector (GCxGC/AED). The GCxGC uses simple valve based flow modulation optimized for samples with a boiling point in the gasoline and jet fuel range. The technique is useful for screening fuels additives, contaminants, etc. KEYWORDS: Gas chromatography, atomic emission detection, betero-atom speciation, sulfur, comprehensive two-dimensional gas chromatography, comprehensive two-dimensional gas chromatography/AED, GCxGC/AED, GCxGC, fuels, gasolines, jet fuels

Introduction
The selective detection o f hetero-atom containing organo-compounds is used extensively in the petroleum industry for process control (sulfur, nitrogen, etc.) and additives in fuels (sulfur, nitrogen, oxygen, phosphorous etc.). The most c o m m o n selective atom detection is performed b y gas chromatography with atomic emission (AED) [1-3], sulfur and/or nitrogen chemiluminescence [4-8] and flame photometric (FPD, p-FPD) [9]. Furthermore, in some cases, these detectors m a y be used in conjunction with a higher chromatographic resolution/selectivity to m i n i m i z e matrix interference. Comprehensive twodimensional gas chromatography (GCxGC) has b e e n under development for more than ten years [10(and references contained therein)] as an enhanced resolution technique. The technique involves continuously transferring or modulating short time periods from a primary column, the first dimension, to a second c o l u m n o f different polarity or selectivity, which serves as the second dimension. The modulation period is typically less than the peak widths observed on the Manuscript received 15 October 2004; accepted for publication 11 April 2005; published October 2005. Presented at ASTM Symposium on ]~lementalAnalysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. l ExxonMobil Research and Engineering, Analytical Science Laboratory, Paulsboro, NJ 08066. This paper was also part of an invited presentation at the Second International Conference on GCxGC, Atlanta, GA, Sept. 1-3, 2004. Copyright9 2005by ASTMInternational, Ban"HarborDrive,POBoxC700,WestConshohonkon, 19428-2959. 100 PA 221

222

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

primary column, and the total elution time from the second dimension is also less than or equal to the modulation period. Because of the narrow peak widths of the compounds eluting from the second dimension, detectors for the technique have been limited to those with very fast acquisition rates such as flame ionization and time of flight mass spectrometry. In this paper the GCxGC process is 'slowed down' by use of 0.53 mm i.d. Megabore columns and relatively long modulation times using valve flow modulation [11], thus potentially allowing a wider variety of detectors. In addition, this paper describes the unique capabilities of atomic emission detection of fuels and petroleum fractions. A simplified version of comprehensive two-dimensional GCxGC to enhance hetero-atom selectivity is coupled with a microwave induced helium plasma atomic emission detector (GCxGC/AED) using a total sample transfer valve based flow modulation for samples boiling in the gasoline and jet fuel range. The technique is useful for screening fuels additives, contaminants, etc.

Experimental
An Agilent Technologies 6890N gas chromatograph/G2350 atomic emission detector (note: a current supplier of GC/AED is Joint Alytical Systems, Moer, Germany) was configured as follows: On-Column Injector Injector Temperature Program 50~ (0.1 min) 30~ to 260~ (until end of oven program). 0.1 #L with a 7683 autosampler. Injection Volume

Columns
First Dimension Second Dimension Carrier first dimension Carrier second dimension Oven Temperature Program Agilent Technologies 60 m x 0.53 mm i.d. 5.0 micron film HP-1. Agilent Technologies 60 m x 0.53 mm i.d. 1.0 micron film DB-WAX Helium at 2 mL/min Helium at 50 mL/min 40~ (0 min) 2~ to 250~ (15 min or until all compounds of interest eluted). 260~ 260~ 20 ml/min Research grade hydrogen, oxygen, and 10 volume % methane in nitrogen. Flows set per factory specifications.

Detector (AED)
Cavity temperature Transfer Line Temperature Helium Make-up flow Reagent Gases

Elements Detected
Carbon Sulfur Nitrogen Oxygen Phosphorous Data Acquisition Rate 179 181 174 171 186 10 points/s (10 Hz)

Wavelength (rim)

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

223

GCxGC Modulator General Set-Up--A Valco DC8WT 1/16 in. 8 port valve (maximum temperature = 350~ was configured with two 5 mL gas loops as shown in Figs. la and lb. Turning the valve on and off allowed effluent from the first dimension to be collected alternately in one of the two gas loops, while the other loop was being swept into the second dimension. The valve was pulsed on and off with an Agilent valve actuator controlled by a Valco digital valve interface and a Valco digital valve sequence programmer. The modulation interval was controlled by the sequence programmer, which was started and stopped by timed contact closures from the gas chromatograph. Columns were connected to the valve with Restek fused silica lined stainless steel tubing, Swagelok 1/16 in. low-dead volume unions and Supelco M2-A 1/16 in. graphitevespel ferrules. The modulator was synchronized by adjusting the start signal (in this case to 1.70 min) until pentane, relatively unretained on Carbowax column, eluted near the beginning of its respective second dimension interval. Modulator discrimination was tested with n-paraffins from C5 through C12. As expected with first dimension column flow, loop sizes, and modulation interval used, no loss was found and no sample was vented during modulation. The valve was mounted at 330~ in a standard valve oven on the 6890 GC. Gas lines from the actuator on this oven were connected to the Valco digital valve interface. Data System
A Hewlett Packard Vectra Pentium IV personal computer operating under Microsoft Windows NT 4.0 and Agilent Technologies A.07.01 ChemStation software was used for data acquisition and instrument control. Two dimension plots were obtained with Fortner Software Transform Version 3.4 after the signal files were exported in CSV file format by the ChemStation software. Pure compounds were purchased from Aldrich Chemical Co., Milwaukee, WI.
Results and Discussion

The system described above was used to screen a number .of fuel samples for bulk composition, additives, and potential contaminants. The following describes qualitatively a number of these applications. Figure 2 shows the carbon specific C-GCxGC chromatogram o f a fluidized catalytic cracking (FCC) gasoline. Note that the compounds present were resolved into distinctive bands for nonaromatics, mono-aromatics, and di-aromatics. This can also be seen in Fig. 3, which shows the C-GCxGC separation of a steam-cracked naphtha (SCN). Because of the process used to produce this gasoline, additional aromatic compound classes such as styrenes and indenes were identified. Figures 4 and 5 show the sulfur specific chromatograms (S-GCxGC) of the two above FCC and SCN samples. The sulfur compounds were also resolved by the GCxGC process into sulfur classes with distinct bands for mercaptans + sulfides, thiophenes, and benzothiophenes. Note that there was a sulfur compound in the steam cracked naphtha at approximately X=55 min, Y=110 s. This unknown compound was also found to contain carbon, hydrogen, and oxygen by further GCxGC/AED analysis. Further characterization of the compound may be possible by GC/Mass Spectrometry (GC/MS) using a comprehensive GCxGC/GC/MS set up very similarly to that described in this paper.

224

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIGURE 1A. Flo~ Modulation Valve Diagram Second Dimension (DB-~AX) First Dimension (HP-1)

Loop 1 - -

Loop 2 FIGURE lB. Flow Hodutation Valve Diagram ~


"I ~ _________+ /

Second Dimension (DB-iIAX)

First Dimension (HP-1)-ON

Laap 1

~
Loop 2

> I

FIG. 1--Flow modulation valve diagrams.

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

225

FIG. 2--Carbon GCxGC/AED of a fluid catalytic cracked (FCC) naphtha.

FIG. 3--Carbon GCxGC/AED of a steam cracked naphtha.

226

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 4--Sulfur GCxGC/AED of a FCC naphtha.

FIG. 5--Sulfur GCxGC/AED of a steam cracked naphtha.

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

227

Figure 6 shows the C-GCxGC separation of a racing fuel. Note that this mixture blended for high engine performance had a very distinct fingerprint compared to the full range naphthas shown in Figs. 2 and 3. Benzene was absent, but toluene was very abundant. Figure 7 shows the oxygen specific separation (O-GCxGC) of this same fuel and the presence of methyl-tbutylether. Figure 8 shows the C-GCxGC of a jet fuel. In this separation, the oven temperature program rate was reduced from 2~ to I~ As with gasoline, the hydrocarbon aromatic classes were readily resolved, and their carbon number distributions are readily distinguishable from C7 to C13. As expected, this sample did not contain benzene. Note that cleanly separated compounds at approximately X = 85 min, Y = 90 s (labeled 'phenolics') were almost as retained as the di-aromatics (naphthalenes). The O-GCxGC separation (Fig. 9) revealed the presence of oxygen in these compounds, which were identified as phenolics. Figure 10 shows the C-GCxGC of a hydraulic fluid. Phosphorous (Fig. 11) P-GCxGC and oxygen O-GCxGC (Fig. 12) specific detections were consistent with the fact that the most abundant compound at approximately X = 105 rain, Y = 30 s was a phosphate ester. Two additional compounds at approximately X = 115 rain, Y = 80 s contained only carbon and oxygen, which was consistent with their identification as epoxides. The effect of the increased selectivity and detection limits using GCxGC over single dimensional GC/AED [1-3 and references therein] for the elements in such samples was not determined in this study. However, it is expected that the detectability of many of the elements is enhanced by the removal of a significant amount of the hydrocarbon matrix background, such as the saturated, olefinic, and much of the aromatic hydrocarbon compounds, by using the second dimension separation of GCxGC. Therefore, it is expected that in many cases the detection limits of the elements will approach those determined with pure compounds using such a system.

FIG. 6--Carbon GCxGC/AED of a racing fueL

228

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 7--Oxygen GCxGC/AED of a racing fuel.

FIG. 8--Carbon GCxGC/AED of a jet fuel.

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

229

FIG. 9--Oxygen GCxGC/AED of a jet fuel.

FIG. 1O---Carbon GCxGC/AED of a hydraulic fluid.

230

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

FIG. 11--Phosphorous GCxGC/AED of a hydraulic fluid.

FIG. 12--Oxygen GCxGC/AED of a hydraulic fluid,

DISANZO AND DIEHL ON GAS CHROMATOGRAPHIC

231

Conclusion

Atomic emission detection coupled with a simplified version of comprehensive GCxGC to reduce hydrocarbon matrix interference using simple and rugged modulation, together with wide bore capillary columns, may be used successfully for the analyses and screening of light boiling fuels, such as gasolines and jet fuels. The technique, together with other spectroscopic techniques such as GC/MS, can provide information on many selective elements in compounds that may be added to fuels as additives or be present as a result of unwanted contaminations.
References

[1] [2] [3] [4]

[5] [6] [7] [8] [9] [10]

[11]

Albro, P. A., Dreifuss, P. A., and Worrnsbecher, R. F., "Quantitative Determination of Sulfur Compounds in FCC Gasoline by GC/AED," J. High ResoL Chromatog. Chromatog. Commun, Vol. 16, 1993, pp. 13-17. Quimby, B. D., Grudoski, D. A., and Giarrocco, V., "Improved Measurement of Sulfur and Nitrogen Compounds in Refinery Liquids Using Gas Chromatography-Atomic Emission Detection," J. ChromatographicSci., Vol. 16, 1998, pp. 435--443. Diehl, J. W., Finkbeiner, J. W., and Di Sanzo, F. P., "Determination of Ethers and Alcohols in Reformulated Gasolines by Gas Chromatography/Atomic Emission Detection," J. High Resol. Chromatog. Chromatog. Commun,Vol. 18, 1995, pp. 108-110. Di Sanzo, F. P., Bray, W., and Chawla, B., "Determination of Sulfur Components of Gasoline Streams by Capillary Column Gas Chromatography with Sulfur Chemiluminescence Detection," J. High Resol. Chromatog. Chromatog. Commun,Vol. 17, 1994, pp. 255-258. Chawla, B. and Di Sanzo, F. P., "Determination of Sulfur Components in Light Petroleum Streams by High-Resolution Gas Chromatography with Chemiluminescence Detection," J. of Chromatography, Vol. 589, 1992, pp. 271-279. Wang, F. C., Robbins, W. K., and Greaney, M. A., "Speciation of Nitrogen-Containing Compounds in Diesel Fuel by Comprehensive Two-Dimensional Gas Chromatography,"J.. Separation Science, Vol. 27, 2004, pp. 468-472. Chawla, B., "Speciation of Nitrogen Compounds in Gasoline and Diesel Range Process Streams by Capillary Gas Chromatography with Chemiluminescence Detection," J. Chromatographic Science, Vol. 35, 1997, pp, 97-104. Wang, F. C., Robbins, W. K., DiSanzo, F. P., and McElroy, F. C., "Speciation of SulfurContaining Compounds in Diesel by Comprehensive Two-Dimensional Gas Chromatography,"J. ChromatographicScience, Vol. 41, 2003, pp. 519-523. Chambers,L. and Duffy, M. L., "Determination of Total and Speciated Sulfur Content in Petrochemical Samples Using a Pulsed Flame Photometric Detector," ./. Chromatographic Sci., Vol. 41, 2003, pp. 528--534. Gorecki, T., Harynuk, J., and Panic, O., "The Evolution of Comprehensive TwoDimensional Gas Chromatography (GCxGC)," J. Separation Sci., Vol. 27, Nos. 5-6, 2004, pp. 359-379. Gorecki, T., Harynuk, J., and Panic, O., "The Evolution of Comprehensive TwoDimensional Gas Chromatography (GCxGC)," d. Separation Sci., Vol. 27, Nos. 5-6, 2004, p. 373.

Journal of ASTM International, November/December 2005, Vol. 2, No. 10 Paper ID JAIl 2990 Available online at www.astm.org
L i s a J. Nash l

Improvements in the Determination of Fluorine in Fuel and Lubricants by Oxidative Combustion and Ion-Selective Electrode Detection
ABSTRACT: Oxidative combustion combined with ion-selective electrode detection is one of several techniques available for the analysis of fluorine in fuels and lubricants to the petroleum and petrochemical industry. This technique was first developed in the mid-1990s to provide an alternative to the widely used Wickbold method, which is very time consuming and extremely hazardous. The early designs were marginally successful due to the lack of sensitivity and stability, the use of large volumes of solvents, inconsistent delivery of fluids, plugging of catch and hold valves, numerous sources of contamination, and long analysis times. The design has been refined to enhance the sensitivity, stability, and accuracy of the fluoride determination. Enhancements include improved ion-specific electrodes with increased sensitivity, the use of circulating baths to stabilize temperature fluctuations, improved sample introductionand flow systems, and redesigned electronic circuitry. KEYWORDS: fluorine, fluorine, fluoride, ion-selective electrode, fuel analysis, oxidative combustion, elemental analysis
Introduction

The determination o f fluorine in various liquid and gaseous hydrocarbons is vital at many points in the refining process primarily in any blend component that has been sourced from the hydrogen fluoride (HF) Alkylation Unit. Fluorinated compounds poison process catalysts; therefore, it is essential that process feeds be as free o f fluorine as possible. As an example, butane is used to produce methyl tertiary-butyl ether (MTBE). The butane must be fluorine free prior to butane isomerization to prevent the poisoning o f the process catalyst. In addition, any H F acid or its combustion products m a y be extremely destructive in any environment. Therefore, any finished hydrocarbon product or synthesized material that is utilized in the presence o f sufficient heat (i.e., car engine), such as fuel and lubricating oils, must be flee o f fluoride. Classical measurement o f fluorine in inorganic and organic materials included sample collection, preparation, and determination. For the reliable determination o f fluorine in any type o f sample, the fluorine: 1) was separated from the interfering substances, or the interferences are suppressed; 2) the experimental blank was low, and the final concentration o f fluorine was adequate for measurement with the method used; and 3) the results were corroborated b y an independent method o f analysis [1]. Preparation involved one or more o f the following stages: washing, drying, ashing, fusion, acid extraction, distillation, or diffusion. Ashing and fusion were sometimes necessary to oxidize organic matrices and to release fluorine from refractory compounds, respectively. Separation was used to avoid interference and/or as a means o f concentration. Separation techniques included distillation, diffusion, solvent extraction, Manuscript received 8 October 2004; accepted for publication 23 May 2005; published November 2005. Presented at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. l Applications Laboratory Manager, Antek/PAC, 300 Bammel Westfield Road, Houston, TX 77090. Copyright9 2005byASTMInternational,100Ban"HarborDrive,POBoxC700,WestConshohocken,PA 19428-2959. 232

NASH ON FLUORINE IN FUEL AND LUBRICANTS

233

chromatography, adsorption, pyrolysis, and pyrohydrolysis. Once the fluorine was released, it was typically measured in the form of fluoride ions (F). Many methods were available for the determination of fluoride in suitably prepared samples including titrimetry, spectrophotometry, fluorometry, gas chromatography, ion chromatography, mass spectrometry, and potentiometry. However, the advent of the fluoride specific electrode, developed by Frant and Ross in 1966 [2], greatly decreases the use of spectrophotometric and fluorometric techniques because it offers speed and relative freedom from interference; and in some circumstances, separation may not be necessary. The ion-selective electrode (ISE) is a variation on the pH electrode with a different sensing membrane and internal solution. Figure 1 illustrates the measurement of potential difference between an ISE and a reference electrode. The membrane of a fluoride specific electrode consists of a single lanthanum crystal which has been doped with europium fluoride. The electrode responds to the activity of fluoride ions in aqueous solutions and is directly proportional to concentration. It is very selective, and its only interferent is the hydroxide ion (OH-). The pH is adjusted to >5.5 to eliminate this interference. The electrode can determine fluoride ion activity over a wide concentration range and down to 0.1 #g/g F" in the presence of several other ions.

FIG. 1--Measurement with an ion-selective electrode. The fluoride selective electrode is routinely used for the determination of fluoride in drinking water, industrial effluents, sea water, air, aerosols, flue gases, soils and minerals, urine, serum, plasma, plants, and other biological materials. The following discussion focuses on the development of analogous techniques for the determination of fluorine in hydrocarbons.
Early Methods

Several methods were developed prior to 1990 for the combustion of organic fluorine compounds like those found in hydrocarbons. The Parr bomb method was first developed in the

234

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

early 1950s. The Parr combustion apparatus consists of a sealed steel reactor that is pressurized to 40 atm (4053 kPa) with oxygen. The sample is ignited and the combustion process releases covalently bound fluoride. The fluoride is converted to its anion and analyzed by titration or other wet chemical techniques. This method works for most organic fluorines except in the case of highly volatile compounds and compounds with high fluorine content where there may be difficulty in the decomposition step due to incomplete combustion and the formation of salt residue. The oxy-hydrogen flame combustion method was developed in the late 1950s by R. Wickbold [3,4,5] to address the critical decomposition step during combustion. The method, commonly referred to as the Wickbold method, is used for the analysis of liquids and liquefied petroleum gases but may be used for viscous liquids and solids if the sample can be dissolved in an appropriate combustible solvent. In the present method, UOP Method 619-83 Fluorine in Hydrocarbons by Oxy-Hydrogen Combustion, the sample is passed through an oxy-hydrogen burner in combination with a Wickbold quartz-tube combustion apparatus which decomposes the sample to carbon dioxide and HF. The combustion products are absorbed in 1 % aqueous boric acid and treated with formaldehyde to remove peroxides. The fluoride content can be determined by several means including measuring the extent of bleaching action by the fluoride on an intensely red zirconium-SPADNS (4,5-dihydroxy-3-(p-sulfophenylazo)-2,7-naphthalenedisulfonic acid trisodium salt) complex or potentiometric titration. The elapsed time for a single analysis is given as 2.8 h in the written method but is often closer to 4 h in actual practice. After burning a sample of high fluoride concentration, the Wickbold system must be purged free of entrapped fluoride by running consecutive blank determinations until a constant low level of fluoride is obtained before a sample of low fluoride can be accurately determined. The procedure can be hazardous due to the use of hydrogen and oxygen and the potential for peroxide formation. The oxygen flask method (also called the Schrniger flask combustion method) involves the ashing of the organic material in a closed flask containing oxygen and an absorbing solution in which the fluoride is subsequently determined by ISE or titration. This method is relatively simple; however, the size of the combustion flask limits the amount of the sample that can be used, occasionally causing incomplete combustion. Also, due to the absence of a concentration step, this method has a higher limit of detection.

Oxidative Combustion/ISE Method Development


All of the previously described methods have a separate combustion step prior to detection. A combination of combustion and detection in one instrument would greatly enhance the determination of fluorine, making it simple and rapid to operate and suitable for automation. The goal was to develop a rapid, sensitive, and accurate method with safe operation for the analysis of fluorine in liquids, solids, and gases with little to no sample preparation. In the mid-1990s, development of a technique began which combined oxidative combustion with ISE detection in a flow through system to provide an instrumental alternative to the aforementioned methods. In this method, the organic fluoride-containing sample is delivered to a quartz pyrotube in an 800-1100~ furnace. The sample is burned in the presence of excess oxygen and steam (pyrohydrolysis). Excess oxygen and high temperature are needed to ensure complete combustion of the sample and to form free fluoride ions. The steam provides sufficient hydrogen ions to react with the released fluoride ions rapidly to form HF, thus preventing the fluoride from reacting with the quartz walls of the pyrotube. The combustion gases and the HF

NASH ON FLUORINE IN FUEL AND LUBRICANTS

235

are bubbled through a solution of Total Ionic Strength Adjustment Buffer (TISAB). The TISAB serves to dissociate the hydrogen from HF, to prevent the formation of H E and HF2", which cannot be detected by ISE, and to adjust the pH to >5.5 to prevent any interference from hydroxide ions. The dissociated fluoride ion concentration in the solution is measured directly by ISE. The reaction chemistry is shown below.

R-F + H20 + O~ . . ~ . ~ HF T ~

HF + CO 2 + 1t20 + oxides F"

The first fluoride instrument design employed catch and hold valves which served to capture and concentrate the combustion products prior to detection. The results achieved with this oxygen combustion/ISE method were compared with the commonly used Wickbold method. It was shown that the results corroborated and, furthermore, the relative standard deviations were decreased with the oxygen combustion/ISE method (see Table 1). TABLE 1--Com arison of Wickbold and oxidative combustion/ISE. Method Wickbold Oxidative Combustion Sample Conc. RSD Conc. RSD Variance (#g/g) (%) (ppm) (%) Between Methods LPG 1 6.1 10.7 6 3.8 2% LPG 2 18.3 8.3 17.1 2.6 7% LPG 3 0.3 8.8 0.3 7.6 0% Fluorobenzene 116 10.4 114 3.5 2% in Toluene Std. Fluorobenzene 1048 3.8 1013 0.1 4% in Toluene Std. The initial design proved to be marginally successful due to the use of large volumes of solvents, inconsistent delivery of fluids, plugging of catch and hold valves, numerous sources of contamination, and long analysis times. The catch and hold valves were eliminated in favor of a direct delivery of the solution containing the dissociated fluoride ions and TISAB to the ISE. This arrangement required the solution to be cooled prior to the ISE to prevent damage to the electrode and an attempt to introduce the sample to the ISE at a constant temperature. A receiving vessel/condenser downstream o f the pyrotube was added to accept the combustion gases and also served to mix the gases with the TISAB. A cooling block and fan were installed that served to reduce the temperature of the mixture prior to the ISE. This design was more successful than the first, but temperature fluctuations were experienced with the air-cooled condenser, resulting in varying potential readings from the temperature sensitive electrode. Beginning in 2000, numerous modifications were made to the fluoride instrument design to greatly enhance the sensitivity, stability, and accuracy of the fluoride determination. Several electrodes were evaluated to determine the most sensitive and durable fluoride electrode available. The ELIT 8221TM, a solid state sensor with a mono-crystalline membrane, was chosen.

236

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

The electrode specifications include a working temperature range of 5-80~ a linear measuring range of 0.1-1 900 ppm F, and a detection limit of 0.02 mg/kg F. Use of this electrode greatly increased the sensitivity of the instrument as can be seen in Fig. 2. Figure 2 shows the comparison of results of a fluoride determination of 5 mg/kg in isooctane between the previously used electrode and the new electrode. The existing air-cooled condenser was replaced with a vertical, thermostatically controlled water condenser located directly at the exit of the pyrotube where the TISAB is added to mix with the exiting combustion gases. This arrangement greatly reduced the temperature of the effluent of the pyrotube and effectively eliminated one meter of heated transfer tubing and potential fluoride losses. Furthermore, another thermostatically controlled condenser was added after the vertical condenser to maintain the condensate temperature at 40~ while flowing through the electrode cell. In this way, the temperature fluctuations were alleviated, and there is significantly less drift in the electrode readings (see Fig. 3). Additionally, the time required to stabilize the instrument before use and between analyses was significantly reduced.

FIG. 2--Comparison of sensitivity achieved with initial (top) and current (bottom) design.

FIG. 3--Comparison of resulting baseline from initial (top) and current (bottom) design.

NASH ON FLUORINE IN FUEL AND LUBRICANTS

237

Other upgrades included a redesigned fluoride detector board with three attenuation levels and improved sample introduction techniques. Currently, the software is being updated to incorporate automation and to use a variety of integration techniques. Although the initial oxidative combustion/ISE designs were able to achieve comparable results to existing techniques, the culmination of recent improvements to the instrument design has greatly enhanced the sensitivity and stability of the fluoride determination and decreased the overall analysis time.
Results and Discussion

The oxidative combustion/ISE instrumental method is used routinely for the analysis of fuels and liquefied petroleum gases (LPG). A standard method, UOP Method 975-02 Fluorine in Liquid Hydrocarbons and LPG by Oxidative Combustion with 1on-Selective Electrode, is available through ASTM International. The method is applicable to liquid hydrocarbons with fluorine in the range of 1-500 ng/~d and LPG in the range of 0.5-500 ng/~tl with the appropriate calibration standards and materials. Higher concentrations can be analyzed by dilution in an appropriate solvent or higher concentration standards can be used as the calibration is linear through several orders of magnitude (see Fig. 4).

FIG. 4 Calibration with Fluorohexane in lsooctane. Currently, the method is used predominantly for the analysis of liquid and gaseous process feeds and automotive lubricants. Results are shown below for an automatic transmission fluid (ATF) and butane (see Figs. 5 and 6, respectively). These results serve to illustrate the sensitivity and repeatability of the fluoride measurement that can be achieved with the current design. With the improved detector response, the level of detection has been shown to be less than that stated in the UOP Method 975-02. Work has started to develop the method as an ASTM standard for the measurement of fluoride in LPG and to establish the lower limit of detection, repeatability, and reproducibility.

238

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 5--Analysis of 30 mg/kg F in automatic transmission fluid (in duplicate).

FIG. 6--Analysis of l Omg/kg F in Butane (in triplicate).


Conclusions

The oxidative combustion with ion-selective electrode method described here can be used to accurately determine fluorine in liquid hydrocarbons and liquefied petroleum gases at levels down to 0.5 ng/Ixl. Results obtained are in good agreement with those achieved with the Wickbold Oxy-Hydrogen Combustion method. The oxidative combustion with ion-selective electrode method is less time consuming, less hazardous, and more sensitive than the Wickbold method. The analysis is also applicable to a range of sample types not discussed here, including aqueous based systems and solids.
References

[1] Venkateswarlu, P., "Evaluation of Analytical Methods for Fluorine in Biological and Related Materials," Journal of Dental Research, Vol. 69, Special Issue, 1990, pp. 514-521. [2] Ross, J. W. and Frant, M. S., Science, Vol. 154, t966, pp. 1553. [3] Wickbold, R., Angew. Chemistry, Vol. 69, 1957, pp. 530-533. [4] Wickbold, R., Angew. Chemistry, Vol. 64, 1952, pp. 133. [5] Wickbold, R., Angew. Chemsitry, Vol. 66, 1954, pp. 173.

Journal of ASTM International, October 2005, Vol. 2, No. 9 Paper ID JAI12977 Available online at www.astm.org

T. W. Selby, 1 R. J. Bosch, 2 and D. C. Fee 2

Phosphorus Additive Chemistry and Its Effects on the Phosphorus Volatility of Engine Oils
ABSTRACT: The most widely used and effective anti-wear/anti-oxidation additives in engine oil contain phosphorus that can partially volatilize during engine operation. Unfortunately, volatile phosphorus in the exhaust stream degrades the fimction of the exhaust catalyst in reducing air pollution. Earlier studies in a special volatility bench test using the Phosphorus Emission Index have shown that phosphorus volatility is not related to engine oil volatility or to phosphorus content in the unused engine oil. At the time, it had been speculated that this unexpected lack of correlation with initial phosphorus concentration could be explained by 1) the effects of other engine oil additives and/or 2) variations in the phosphorus additive chemistry. The first speculation was relatively recently confn-med by taxi fleet studies of catalyst degradation by phosphorus-containing oils by the Ford Motor Company. It remained to determine if differences in phosphorus additive chemistry were also a factor, and this is the subject of the present study. KEYWORDS: phosphorus volatility, Phosphorus Emission Index, PEI, ZDDP, zinc dithiophosphates, exhaust catalyst, exhaust emissions

Nomenclature
ZDDP PEI API ILSAC EGR zinc di(organo) di(thio) phosphate Phosphorus Emission Index American Petroleum Institute International Lubricant Standards and Approval Committee exhaust gas recirculation system

Introduction and Background


Importance o f ZDDPs
The class o f anti-wear, anti-oxidation additives c o m m o n l y k n o w n as ZDDPs has been a mainstay o f engine oil formulation and performance for more than 60 years [1]. Engine designers and metallurgical engineers have found significant help in overcoming problems b y utilizing the protection brought b y the chemists and lubrication engineers forming and formulating with these additives. Concentrations o f Z D D P phosphorus in the formulated engine oil have often been above 0 . 1 % to ensure the desired level o f protection, even though somewhat lesser amounts have been shown to suffice. More recently, with emphasis on smaller, more powerful engines (carrying ever longer warranties), the circumstances would seem ideal for increasing use o f ZDDPs. Manuscript received 9 November 2004; accepted for publication 11 April 2005; published October 2005. Presented at ASTM Symposium on Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. 1Director of R&D, Savant, Inc., Midland, Michigan. 2 Technical Service Fellow and Process Improvement Specialist, respectively, Astaris LLC, Webster Groves, Missouri. Copyright9 2005byASTMInternational, Ban"HarborDrive,POBoxC700,WcstConshohoeken, 19428-2959. 100 PA 239

240

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Phosphorus and Catalyst Interaction


A problem demanding creative solutions has arisen over the last few decades. Automotive exhaust pollution control requires expensive exhaust catalysts, and as a result of government mandates today, these catalysts must remain highly effective for up to 120 000 miles of service. The problem is that it became evident that these catalysts were degraded by phosphorus in the exhaust stream [1-8]. Association of Phosphorus Emission with Oil Consumption and Volatility--Oil consumption caused by oil volatility has always been a concern for automotive engineers because of the associated deposits often formed on piston rings, combustion chambers, and valves. Concern that oil volatility could cause phosphorus in the exhaust stream adds to the desire to reduce oil volatility. Consequently, progressively lower formulated engine oil volatility (as measured by the classic Noack volatility test [9-12]) was required for oils meeting API and ILSAC GF-2 (22 % loss) and GF-3 (15 % loss) specifications. Phosphorus Concentration Reduction and Concerns--A more challenging and debated direction in limiting phosphorus volatility was to limit the concentration of ZDDPs in engine oil. Understandably, this brought intense discussion among automotive and lubrication engineers reagarding: 1. lowering levels of ZDDP in engine oils without considerable field experience in modem high-performance engine durability, at a time when 2. engine oil operating viscosities and levels of hydrodynamic lubrication were being reduced to gain fuel efficiency, as well as concern about 3. the impact of such changes on older, used, and presumably more vulnerable automobiles. Despite these concerns, phosphorus levels were reduced first to a maximum of 0.10 % in lower viscosity grade passenger car engine oils designated API SJ and ILSAC GF-2 in 1995 and then in 2004, to a level of 0.08 % for API SM and ILSAC GF-4.

A New Noack Testfor Determining Phosphorus Volatility Selby-Noack Volatility Test--The toxic nature of the Woods Metal used for heat transfer in the classic Noack volatility test [9-12] led to the development of a new approach [13-17]. This approach, while retaining all Noack test conditions, 3 was designed to heat the test oil safely using noble metal resistive heating and to collect 99+ % of all the material volatilized. Figures la and b show a picture of the special Noack instntment and a sketch of the arrangement of the essential components, respectively. As indicated in Fig. 2, early studies with the special Noack showed that more than 99 % of the oil mass lost from the volatilization chamber was accounted for either in the collector or the ~1 g on the walls of the tubing leading to the collector [16]. Such efficiency is important to be sure that the volatilized material collected is essentially identical to all material volatilized instead of forming a questionable composition. Analyses of phosphorus in the collected volatiles from the special Noack in these early studies [13-17] showed that phosphorus was present. Moreover, the limited data also suggested that phosphorus volatility was not closely related to oil volatility.
3The Noack volatilitytest determinespercent loss of volatile materialfrom 65 gramsof the test oil when exposedto a small vacuumof 20 mm water at 250~ for one hour.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

241

FIG. 1--(a) Selby-Noack instrument and (b) essential components of Selby-Noack.

Mass Balance of Amount Collected and AmountVola'dlized


Grams Volatiles Recovered 20 Rz = 0.998 Slope : 0.993 g/g Intercept .1.224g : j ~ J j ~ r " JJ"

t5

~ j ~ r

10

Ideal ~

~/__ ~d~'~Actual

10

16

2O

26

GramsEn~line Volatilized Oil


FIG. 2--Mass balance of loss and collection of volatilized material.

Effects of Oil Volatility and Concentration on Phosphorus Volatility


This special Noack approach - permitting collection of all volatiles - was opportune. Among other applications, it presented the ability to evaluate the relationship of phosphorus volatility to oil volatility and/or phosphorus concentration in the fresh oil. A study of data published by the Institute of Materials (IOM) [ 18] for the years 1999 and 2000 of over 1200 oils was used. Phosphorus volatilized from fresh engine oils was compared to: 1) their oil volatilities and 2) initial phosphorus concentrations. Phosphorus volatility was first shown no__At be dependent on engine oil volatility, as can be seen in Fig. 3 where phosphorus to volatility varies independently of the associated engine oil volatility. In a further analysis of the IOM data, it was surprisingly found that phosphorus volatility is unrelated to phosphorus concentration in the fresh oil [19,20] as shown in Fig. 4.

242

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 3 Showing lack of relationship between the grams of an engine oil volatilized in the special Noack test (from an initial charge of 65 g for the tesO and the grams of phosphorus volatilized during the same test.

FIG. 4---Showing lack of relationship between the concentration of phosphorus in a fresh engine oil and the concentration of phosphorus volatilized from the particular engine oil in the special Noack test. Phosphorus Emission lndex Considering the lack of relationship of volatilized phosphorus with either oil volatility or initial phosphorus concentration in the oil, it seemed important to have a measure of the comparative phosphorus volatility of engine oils. This led to a concept of ranking engine oils based on the volatility of their phosphorus-containing additive(s) [20]. The ranking method was called the Phosphorus Emission Index, or PEI. PEI is defined as milligrams of phosphorus volatilized per liter of fresh oil based on results obtained in the special Noaek test. Repeatability was tested and shown in past papers [20,21].

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

243

Considering the surprising lack of correlation with phosphorus concentration in the fresh oil, it was thought to be likely that phosphorus volatility might be dependent on ZDDP variations of chemistry and/or effects of other additives in the formulation. However, at the time, there was no evidence whether the Phosphorus Emission Index and the volatile phosphorus to which it was related had any correlation with the degeneration of exhaust stream catalysts.

Correlation of PEI with Field Study


Further work testing the significance of the PEI concept was undertaken as a consequence of taxi fleet tests of phosphorus volatility and catalyst degradation conducted by Ford Motor Company [22,23]. Interest in the PEI concept led Ford engineers to submit three blind-coded, fresh engine oils of the several that had been used in the taxi fleet tests. On completion and submission of this blind-sample PEI study to Ford, it was revealed that the ZDDP, concentration of ZDDP, and the base oil used in each engine oil were the same. Moreover, one of the oils had no additives present other than the ZDDP. Results are shown in Table 1 and Figs. 5 and 6, in which four of the Ford test oils are listed. (Oil X, which contained no phosphorus, was not submitted for PEI evaluation by Ford but has value in the comparisons made.) The figures show the additives present as well as the level of phosphorus deposits and the NOx emissions for each oil.

TABLE 1--Comparison of PEI to phosphorus found on exhaust catalyst and NOx emitted at the end of Ford Motor field taxi tests.

FIG. 5--Correlation of PEI and catalyst.

FIG. 6---Correlation of PEI and NOx.

244

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

This study of field test oils resulted in three important findings on the questions raised in the earlier work [19-21] regarding the significance of PEI and the effects of other oil additives: 1. The PEI level was found to correlate with the phosphorus found on the catalyst. 2. The presence of other additives reduced the PEI value considerably. 3. PEI also correlated with the grams/mile of NOx emitted. Thus, the PEI analysis of the Ford Motor Company field study resolved the previous posed question [20] of whether other additives in the oil formulation can influence the volatility of phosphorus. More specifically and importantly, it seems apparent that engine oil formulation can be tailored to control phosphorus volatility. Interestingly, Ueda and his associates [8] found much earlier that both calcium and magnesium diminished phosphorus effects on the catalyst. (If these two types of additives were capable of forming volatiles, they might curtail adverse phosphorus effects.) The major remaining question was whether the various chemistries of ZDDPs also affect the volatility of phosphorus, and answering this question was the focus of the present paper.
Studies of the Chemistry and PEI of ZDDPs

Instruments and Techniques Noack Test Procedures--The Selby-Noack is the only instrument capable of generating the Phosphorus Emission Index and was consequently used to obtain the PEI data of this study. Volatile and residual materials were also used to provide samples for the associated NMR determinations presented in a conjoined paper [24]. As briefly described in the previous section, the routine special Noack technique applies the same conditions as the original Noack procedure. (For greater detail, this procedure is fully described in ASTM Method D 5800c [12].) However, in generating data for this study, in some experiments the technique was modified to be run either for longer temperature ramps to the operating temperature of 250~ or for shorter intervals than one hour at 250~ In these modified tests, other than collecting volatiles for longer or shorter time periods, the overall technique was identical to the routine special Noack test. Elemental Analyses--A Leeman Model PS 1000 inductively coupled plasma (ICP) spectrometer was used to determine all elemental concentrations of phosphorus required in these studies. ASTM Method D 4951 [25] was followed using cobalt at 200 PPM as an internal standard. Nuclear Magnetic Resonance Procedure--Small samples of the original oils and the volatile materials and residues obtained during the special Noack tests were analyzed by 31p NMR at Washington University (St. Louis, MO). The spectra were obtained either on a 500 MHz Varian NMR equipped with a 10 mm probe or on a 600 MHz Varian NMR equipped with a 5 mm probe. Samples were diluted with 10-15 % chloroform-d (CDC13), which also served as an internal reference for establishing spectral positions. Average data accumulation time for these spectra was one hour. As mentioned, a conjoined paper [24] covers this aspect of the study.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

245

Initial Studies

Reference Oils Formulated and Reference Oils--The fully formulated oil, RO 780, was classified as SAE
5W-30 engine oil known to be in APUILSAC categories SL/GF-3. This oil - minus any ZDDP was the primary Reference Oil, RO 779, providing blending stock for specially prepared ZDDPs. PEIs of the Comparator and Reference Oils--Volatilities and PEIs using the special Noack were obtained on these two oils, and results are shown in Table 2. As expected, the formulation without ZDDP, RO 779, shows no phosphorus in either the formulation or the volatiles. In comparison, RO 780 has a PEI of 4.0 - considered relatively low among the engine oils collected around the world for the international engine oil database of the IOM. TABLE 2--0il and phosphorus volatilities of fully formulated Oil RO 780 and ZDDPdeficient Reference Oil 779.

PEI Analysis of Engine Oils Selected from the Market


For purposes of determining the effect of large differences in PEI on the NMR signature of the volatile and residual material, four oils having similar oil volatilities but widely different phosphorus volatilities were requested from the North American IOM engine oil database. These samples were kindly supplied, and results are shown in Table 3 and Fig. 7 (also including RO 780 for comparison). Compared to phosphorus concentrations of the fresh engine oils, Table 3 and Fig. 7 show that the volatilized material of the four oils has widely different concentrations of phosphorus and, thus, considerably different levels of associated PEIs. It was considered reasonable that these oils would provide good opportunity in later NMR studies [24] to determine the cause(s) of differences in phosphorus volatility and the related PEI. TABLE 3--Comparative oil and phosphorus volatilities of four marketed engine oils of considerably different PEIs.

246

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

FIG. 7--Variation of PEI among four marketed North American engine oils of different formulations with somewhat similar overall volatility loss and initial phosphorus concentrations.

Preliminary Study of Temperature Ramping on Phosphorus Volatility


As a preliminary test to determine when phosphorus volatiles were generated during the special Noack test, the ZDDP-containing RO 780 was given a one-hour ramping time to the temperature of 250~ followed by the normal hour duration at that temperature. Results are shown in Table 4, contrasted with the normal ramp time of about nine minutes to 250~ TABLE 4 Effects on phosphorus volatilization using one-hour ramp versus normal, 9-rain ramp.

The data show that, although the total volatile material increased by about 13 % because of the longer test time, the phosphorus volatilized increased by slightly more than 100 %. From these results, it would seem that the longer test encouraged the volatilization of more phosphorus and/or the slower, longer ramping interval affected the nature of the phosphorus being volatilized. The questions raised by these considerations led to further tests reported in the next section of the paper.

Second Stage of Studies

General Considerations
The questions of when phosphorus volatilizes, as well as what volatile compounds are produced and what compounds are left in the residual oil during the special Noack test were of considerable interest. It was recognized by the authors that the study of these various

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

247

relationships would constitute a sizable effort, and it was decided to resolve the questions progressively. Some of these efforts are reported in this paper, others in the conjoined second paper [24].

Dependence of PEI on Residence Time in the Special Noack Test


One of the first studies was a modified special Noack test on RO 780 (having a PEI of 4) and a formulated engine oil having much higher PEI. Modified special Noack tests were run in four progressive time intervals of 10, 20, 30, and 50 min exposure after the normal 9-min ramp to volatility-inducing conditions. A special vial-filling drain for collecting small amounts of the volatilized oil was attached to the bottom of the collection vessel. Normal 65-g samples were used for the test. Following the test, the collected samples were analyzed by ICP to determine their phosphorus content and the related PEI. Table 5 and Figs. 8 and 9 show the data obtained (including the 60-min data given earlier in Table 2). TABLE 5--First sequence of exposure-modified special Noack results on the low-PE1 formulated Reference Oil 780 compared to the results from a marketed oil having a relatively high value of PEI.

FIG. 8--Dependency of oil volatility on exposure.

FIG. 9--Dependency of PEI on exposure.

248

ELEMENTALANALYSIS OF FUELS AND LUBRICANTS

As would be expected, the oil volatilities rise with increasing time of exposure, as shown in Fig. 8. Oil 780 may be coming to equilibrium in about 50-60 min, while the volatility of EO 7450 is still rising, although slowing. Effects of the forgoing study on the Phosphorus Emissions Index are shown in Fig. 9. With Reference Oil 780, the PEI remains essentially constant and may even be interpreted as falling somewhat. In contrast, the ZDDP producing the much higher volatilizing Oil 7450 increases during the exposure. These dissimilar results raise further questions regarding the nature of the chemical composition of the phosphorus additives as well as effects of the other additives in the two oils - questions appropriate for the NMR study in the conjoined paper [24].

Modification of the Chemistry of ZDDPs on PEI Chemistry and Test Setup---Four ZDDP samples were prepared in the Astaris labs. The alcohols employed were 2-ethylhexanol, a primary alcohol, and 4-methyl-2-pentanol, a secondary alcohol. Two P2S5 samples were employed, one that consists of slightly lower (at 27.77 %) and the other of slightly higher (at 28.07 %) levels of phosphorus than the stoichiometric value of 27.87 % phosphorus theoretically calculated for P2S5. These four ZDDPs were blended into a typical GF-3 oil at Savant using the non-ZDDP-containing RO-779. These blends, identified as Test Oils A through D, were again exposed to the time-interval approach. In this portion of the work, exposure times of 10, 20, 30, and 50 min in the special Noack instrument were used with the routine 9-min ramp to an operating temperature of 250~ Test Results--Results are shown Table 6 and in Fig. 10. As evident from Fig. 10, PEI values of each of the Test Oils A through D show major differences between the primary and secondary ZDDPs when added in similar concentrations to the Stock Oil RO 779. That is, primary ZDDPs A and B increase in PEI with increasing exposure time in the test, while secondary ZDDPs C and D show little change in their PEIs.
TABLE 6--Effect of ZDDP chemistry and reaction stoichiometry on the level of phosphorus volatility in exposure-modified special Noack tests.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

249

FIG. l O--Plot of PEI data from Table 6 showing the clear differences of type of alcohol chosen and the chemistry of preparation of ZDDP. A response showing little or no increase of volatile phosphorus indicates that ZDDPs C and D emit all volatile phosphorus components in less than 10 min exposure in the special Noack test. More important, however, is the fact that the one-hour PEI values of the secondary ZDDPcontaining oils are comparatively low and do not increase, whereas the primary forms of ZDDP investigated in this study are considerably higher and increase further in PEI values with exposure time. For reasons offered earlier, increase in PEI implies continuing generation of phosphorus volatiles during the test. It will also be noticed in Table 6 that the primary and secondary alcohol preparations are made with two samples of P2S5 that contain different levels of phosphorus. Of these two pairs of ZDDPs, little difference is shown between the primary ZDDPs A and B. However, significant differences are shown between the secondary ZDDPs C and D. That is, at closely similar concentrations (991 versus 1064 PPM, respectively), ZDDP C has about half the value of PEI shown by ZDDP D. Table 7, obtained from the Institute of Materials Engine Oil Database, shows that both of these secondary ZDDPs are at relatively low levels of PEI compared to other oils around the world.
Discussion

Relevance of the PEI to Observed Field Catalyst Deterioration


On the basis of the Ford Motor Company field studies, the correlation of the PEI generated using the special Noack shown in Table 1 and Figs. 5 and 6 is evident. The degree of correlation suggests that the volatility of phosphorus is perhaps the most important source of such catalyst contamination, whether or not it is generated from the crankcase and passed to the exhaust stream through the EGR valve or from oil leaking directly into the combustion zone from piston ring bypass or down the valve stems.

250

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE 7--Comparison of PEI of specially prepared primary and secondary alcohol ZDDPs to marketed oils from around the world. (Data from IOM Engine Oil Databases, used with permission.)

Moreover, the work clearly shows that engine oil formulation is a critical aspect o f controlling phosphorus volatility and raises the question o f whether certain additives are more effective than others or whether some o f the other additives actually enhance phosphorus volatility.

General Comments on PEI Technique Used in These Studies


In this work a special technique was used to drain the contents of the collection cup o f the special Noack instrument into a small vial after a shortened run where it was desirable to catch small amounts o f volatile material. This was successful in obtaining PEIs over the series of time intervals. Considering the results obtained on those samples continuing to generate phosphorus to the end of the 60-min test, it is evident that longer exposure periods should be even more revealing.

Comparison of Two Special Reference Oils


As previously noted, two special reference oils, RO 779 and RO 780, were provided by Chevron Oronite Company LLC. RO 779 had no ZDDP added but otherwise was identical in base stock and additive content to RO 780. Table 1 showed that the responses o f the two reference oils were as expected, with RO 779 evidencing no volatile phosphorus and RO 780 showing a PEI o f 4 - a comparatively low value. RO 779 was thus considered to be acceptable as a stock for the subsequent blending studies o f specially made ZDDPs.

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

251

Comparison of Four Marketed Engine Oils


In an effort to determine the differences between engine oils having similar oil volatilities and fresh oil phosphorus concentrations, but markedly different phosphorus volatilities shown by the PEI, four oils were chosen from the IOM Engine Oil Database. Sufficient samples were received through the courtesy of IOM to run comparative special Noack tests, from which the volatile material and the residual oil for NMR studies were collected. These data, shown in Table 3 and Fig. 7, show the similarities of both oil volatilities and fresh oil phosphorus concentrations and are in clear contrast to the different levels of PEI, which are as great as 3700 %.

Effect of Different Temperature Ramps


A preliminary effort was made to determine whether ramping time from ambient to 250~ for the special Noack test would show differences in the phosphorus volatiles collected. Data presented in Table 4 using RO 780 indicate that significant differences in ramping time do have an effect and raised the questions of whether the effect is related to a difference in phosphorus decomposition products or merely the extended time of exposure of the sample. In view of the later experiments shown in Table 5 and Fig. 9 conducted by varying exposure time for Oil 780 (in which the full content of volatilizable phosphorus was delivered to the special Noack collector in less than 10 min), it is more likely that there is a difference in the path of phosphorus decomposition. That is, for Oil 780, the difference in oil volatility increases with and without extended ramping time - about 13 % - for the longer ramping test is considerably less than the 125 % increase in volatilized phosphorus. This preliminary finding emphasized the value of investigation using 31p NMR spectrometry to determine what differences were present.

Effects of Exposure Time on Otherwise Normal Test Technique First Evaluation of Technique--Desire to determine the progression of the breakdown and volatilization of ZDDP in the fresh oil led to the technique of varying the length of exposure of the oil to 10, 20, 30, 50, and 60 min in the special Noack test. In this initial study, RO 780 was compared to IOM's EO 7450, a formulated oil from the market having relatively high PEI. First results were shown in Table 5 and Figs. 8 and 9. Not surprisingly, increasing exposure showed progressive increase in oil volatilization. The technique, however, showed dissimilar results in progressive values of the PEI since RO 780 gave results indicating that all volatile forms of phosphorus from the ZDDP additive in this oil were generated within the first 10 min of exposure at 250~ Volatile Phosphorus, a Small Part of Available Phosphorus--As noted previously, the PEI data of RO 780 showed no further generation of phosphorus with more lengthy exposures (and, in fact indicated some decrease in PEI at 50 and 60 rain exposure). Thus, whatever volatile decomposition products were produced before 10 rain exposure comprised only a small fraction of the phosphorus-containing additive(s) present in the fresh oil. That is, 850 g of the fresh RO 780 contains 966 PPM phosphorus or approximately 820 mg/L, whereas the PEI is 9 mg/L about 1% of the phosphorus in the fresh RO 780 volatilized in this test. In contrast, the IOM EO 7450 contains 1198 PPM or approximately 1020 mg/L. In this case, the PEI rises to 60 at 60 min of exposure, and thus, about 4 % of the phosphorus material in the fresh oil is capable of being volatilized at this length of exposure. Greater length of exposure might increase this level of phosphorus volatility of IOM EO 7450, although it does seem to be

252

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

reaching a plateau.

Application of Progressive Exposure Technique to Special ZDDPs--The four special ZDDPs made by Astaris to generate further understanding of the causes and forms of phosphorus volatilization were added to the stock oil, RO 779, at levels of about 1000 PPM. Using the technique of progressive time exposure, the results are shown in Table 6 and Fig. 10. The PEI values obtained demonstrate obvious differences in phosphorus volatility between the particular primary- and secondary-alcohol ZDDPs. Specifically, both higher and lower stoichiometric variants of this primary alcohol's ZDDPs show continuing release of volatile phosphorus during the one-hour special Noack test with the rate of release decreasing with increasing time of exposure. At 50 min, about 2 % of the original phosphorus in the oil has been volatilized. Both preparations of the primary ZDDP at slightly higher and lower P2S5 stoichiometries are closely similar in their PEI response to increasing exposure. In contrast, the particular secondary alcohol ZDDPs show little or no change in PEI with exposure time indicating that all the volatile phosphorus entities that will be formed are volatilized early in the test. However, in this case there is a difference in whether the ZDDP is formed from P2S5 that contains a higher or lower than stoichiometric level of phosphorus. The higher-than-stoichiometric ZDDP C gives phosphorus release of about 1 % of the original phosphorus in the oil. In contrast, the lower-than-stoichiometric ZDDP D gives phosphorus release of about 0.6 % of the original phosphorus - a considerable difference in phosphorus release. Resolution of the Question of the Volatility Role Played by ZDDP Formation--Most
important, these latter data answer the question of whether the chemical form of ZDDP can markedly affect its tendency to volatilize. As to how well this statement applies to other forms of ZDDP, this is a matter of fiarther study. Certainly, from the viewpoint of controlling phosphorus volatility, there is a strong need for much more information regarding the relationship of the chemical form of ZDDP and its dependence on the other additives used in formulating engine oils.
Conclusions

This series of studies of the variation of the Phosphorus Emission Index generated from several applications of the special Noack instrument has been effective in revealing some of the characteristics of the phosphorus volatilization from ZDDPs made from a primary and a secondary alcohol. Thus, the primary objective of this study has been met. That is, it has been shown that phosphorus volatilization and the associated PEI are affected strongly by the chemistry of the ZDDP. Whether or not these differences are caused by thermodynamic effects and/or some form of steric interference remains to be determined, as well as the range of differences that may exist among the different alcohols. Moreover, using the PEI as an investigative tool for this initial study, differences between the primary and the secondary alcohols chosen are also found in the immediacy and ultimate levels of forming volatile forms of phosphorus species. Variation in PEI of four marketed engine oils in North America has shown marked differences of 3700 %. Even comparing the two alcohols used in this study, the differences in PEI were about 400 %. To be sure, at this point, the latter observations are limited to the two alcohols used to make the ZDDPs for this study, but the technique is obviously readily applicable to all other variations of ZDDPs. The study also generated sufficient volatilized samples to permit the generation of 31p NMR spectra, and this work is reported in a conjoined paper [24].

SELBY ET AL. ON PHOSPHORUS ADDITIVE CHEMISTRY

253

One area of question that was not part of this study but should be considered is the impact of phosphorus volatilization on the residual oil from the special Noack test. That is, to what extent is the anti-wear and anti-oxidation performance of a ZDDP compromised by phosphorus volatilization? This information is obviously of high interest and importance given the primary purpose of ZDDPs in engine oils.

Acknowledgments The authors would like to thank those who dedicated their efforts in preparing, blending, and analyzing the various oils used in this study. Among these were Skip Ramsey of Astaris for his assistance in preparing the ZDDPs, George Schaller of Savant Labs for running the special Noack tests, and both he and Mike Habitz of Savant Labs for blending the test samples and analyzing the results. For the marshalling of samples and data, the help of Amy Parker and Joyce Gee were indispensable. Special thanks are extended to Chevron Oronite Company LLC for providing typical GF-3 oil samples with and without ZDDP.
References

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

Spikes, H., "The History and Mechanisms of ZDDP," Tribology Letters, Vol. 17, No.3, October 2004, pp. 469--489. Minamitani, A., et al., "Effects of Engine Oil Additives on the Activity of Exhaust Gas Oxidation Catalysts, J. Japan Petroleum Institute, Vol. 21, 1978, p. 116. Caracciolo, F. and Spearot, J. A., "Engine Oil Additive Effects on the Deterioration of a Stoichiometric Emissions Control System," SAE Paper #79094, 1979. Monroe, D. R., "Phosphorus and Lead Poisoning of Pelleted Three Way Catalysts," SAE Paper #800859, 1980. Miyoshi, N., "Poisoning Deactivation of Automotive Catalysts by Lead and Phosphorus," SAE Paper #852219, 1985. Brett, P. S., et al., "An Investigation into Lubricant Related Poisoning of Automotive Three Way Catalysts and Lamda Sensors," SAE Paper #890490, 1989. Dmry, C., et al., "The Effect of Lubricant Phosphorus Level on Exhaust Emissions in a Field Trial of Gasoline Engine Vehicles," SAEPaper #94075, 1994. Ueda, F., et al., "Engine Oil Additive Effects on Deactivation of Monolithic Three-Way Catalysts and Oxygen Sensors," SAE Paper #940746, 1994. Noack, K., Angewandt Chemic, Vol. 49, 1936, p. 385. Determination of Evaporation Loss of Lubricating Oils, (Noack Method), DIN 51-581, 1981. Evaporation Loss of Lubricating Oils, CEC L-40-T-87, 1987. ASTM Standard D 5800, "Evaporation Loss of Lubricating Oils by the Noack Method," Annual Book of ASTM Standards, Vol. 5.03, ASTM International, West Conshohocken, PA, 2004, pp. 352-366. Selby, T. W., et al., "A New Approach to the Noack Test for Volatility Measurement," SAE International Fuels & Lubricants Meeting and Exposition, Philadelphia, PA, 1993. Reichenbach, E., et al., "A New Approach to the Noack Volatility Test," Proceedings of Ninth International Colloquium, Esslingen, Germany, 1994; Tribotest Journal, Vol. 1, No. 1, Leaf Coppin Publishing, September 1994, pp. 3-29.

254

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

[15] Selby, T. W., et al., "Base Oil Characterization Techniques Using a New Approach to the Noack Volatility Test," ACS Symposium on "The Processing, Characterization, and Application of Lubricant Base Oils: Part 1I, San Diego, CA, March 1994. [16] Selby, T. W. and Reichenbach, E. A., "Engine Oil Volatility Studies - Generation of Phosphorus," Proceedings of the International Tribology Conference, Yokohama, Japan, 1995, pp. 813-816. [17] Cluff, B., et al., "A Laboratory Method for Measuring Bulk Volatility of Engine Oils Comparative Results," SAE Paper #961227, International Spring Fuels and Lubricants Meeting, Dearborn, MI, May 1996. [18] Institute of Materials Engine Oil Database, Institute of Materials, Midland, MI, issued yearly from 1984. [19] Selby, T. W., "Analysis of Engine Oil and Phosphorus Volatility - Development and Use of the Selby-Noack Apparatus to Recover and Study Phosphorus Volatiles," 11th International Colloquium, Ecological and Economic Aspects of Tribology, Esslingen, Germany, 1998. [20] Selby, T. W., "Development and Significance of the Phosphorus Emission Index of Engine Oils," Proceedings of 13th International Colloquium - Lubricants, Materials, and Lubrication Engineering, Esslingen, Germany, 2002, pp. 93-102. [21] Selby, T. W., "Phosphorus Volatility of Lubricants - Use of the Phosphorus Emission Index of Engine Oils," 9th F&L Asia Conference, Singapore, January 21-24, 2003. (available on CD from F&L Asia, Inc., Manila, Philippines). [22] Johnson, M., "Catalyst Contamination by Phosphorus from Engine Oils - Emission Effects and Transport Processes," ILSAC-Oil Meeting, Romulus, MI, July 10, 2002. [23] Johnson, M., "Effects of Engine Oil Formulation Variables on Exhaust Emissions in Taxi Fleet Service," SAE Paper #2002-01-2680, SAE Powertrain Meeting, 2002. [24] Bosch, R. J., et al., Analysis of the Volatiles Generated during the Selby-Noack Test by 31p NMR Spectroscopy, to be published by ASTM International, West Conshohocken, PA. [25] "Determination of Additive Elements in Lubricating Oils by Inductively Coupled Plasma Atomic Emission Spectrometry," Annual Book of ASTM Standards, Vol. 5.02, ASTM International, West Conshohocken, PA, 2004, pp. 974-980.

Journal of ASTM International, November/December 2005, Vol. 2, No. 10 Paper ID JAI 12984 Available online at www.astm.org

R. J. Bosch, 1 D. C. Fee, 1 and T. W. Selby 2

Analysis of the Volatiles Generated During the Selby-Noack Test by 31p NMR Spectroscopy
ABSTRACT: A series of engine oil samples collected during a study of their Phosphorus Emission Index

values were analyzed by 31p Nuclear Magnetic Resonance Spectroscopy. (The PEI analyses themselves were presented in an associated paper [1]). NMR spectra were generated to obtain and explain the mode of formation and identity of the phosphorus-containing species in the volatiles generated during the Selby-Noack volatility test and to compare these species to those found in both the fresh oil and the residual oil remaining after the volatility test. KEYWORDS: phosphorus volatility, exhaust catalyst, exhaust emissions, ZDDP, zinc dithiophosphates, Phosphorus Emission Index, PEI, 31p NMR, NMR spectroscopy
Nomenclature

ZDDP PEI NMR


Introduction

Zinc Di(organo) Di(thio) Phosphate Phosphorus Emission Index Nuclear Magnetic Resonance

Zinc Di(organo) Di(thio) Phosphates, commonly referred to as ZDDPs, are the most widely used and effective anti-wear/anti-oxidation additives in engine oil. Phosphorus contained in the ZDDP molecules has been shown to partially volatize during engine operation [2]. Unfortunately, volatile phosphorus in the exhaust stream degrades the function o f the exhaust catalytic converter, and as a consequence, there has been pressure to reduce the amount o f ZDDP in engine oil. One o f the new analytical test procedures developed during the 1990s is the Selby-Noack instrument [3]. This instrument uses the earlier protocol developed b y Noack [4-6] to measure the volatility o f motor oil but, in addition, collects all o f the volatile material generated in the test. In Noack-protocol tests, 65 g o f o i l are heated in air in a metal or glass cup for 1 h at 250~ under a slight vacuum o f 20 mmw~t~, and the percent loss o f sample mass is measured. During the development and early use o f the Selby-Noack test, it was found that phosphoruscontaining compounds were present in the volatiles obtained from the procedure [7]. With the growing interest in the subject o f volatile phosphorus, the Institute o f Materials incorporated the Selby-Noack test as part o f the database they maintain on worldwide oil analyses [8] and also determined the concentration o f phosphorus in the volatile material generated (initial phosphorus concentration o f the fresh oil was already a part o f the database). Manuscript received 15 January 2005; accepted for publication 28 April 2005; published November 2005. Presented at ASTM Symposiumon Elemental Analysis of Fuels and Lubricants: Recent Advances and Future Prospects on 6-8 December 2004 in Tampa, FL. J Technical Service Fellow and Technology Fellow, respectively, Astaris LLC, Webster Groves, Missouri. 2 Director of R&D, Savant, Inc., Midland, Michigan.

Copyright 9 2005 by ASTM International, 100 Ban"Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

255

256

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

In 2002, a paper was published on the phosphorus volatility results that appeared in the IOM database from 1999-2001 [9]. Contrary to expectations, it was found that phosphorus volatility was neither related to engine oil volatility nor to the phosphorus content in the unused engine oil. It was speculated that this lack of correlation with initial phosphorus additive concentration could be explained by either or both: 1) effects of other engine oil additives and 2) variations in the chemistry of the phosphorus additives. A 2002 field study conducted by Ford Motor Company of catalyst degradation by phosphorus-containing oils [10] was subsequently shown to correlate with the PEI data generated from the Selby-Noack bench test and, further, showed that phosphorus volatilization was strongly affected by other oil additives [11]. One of the objectives of the present studies is to examine the second premise concerning whether variation in the chemistry of phosphorus-containing ZDDP additives also affects phosphorus volatility from the formulated engine oil.

Experimental
Small samples of the original oils and the volatile materials and residues obtained during the Selby-Noack tests in the associated paper [1] were analyzed by 31p NMR at Washington University (St. Louis, MO). The spectra were obtained either on a 500 MHz Varian NMR equipped with a 10 mm probe or on a 600 MHz Varian NMR equipped with a 5 mm probe. Samples were diluted with 10-15 % chloroform-d (CDC13), which also served as an internal reference for establishing spectral positions. The individual phosphorus-containing species are identified by a "chemical shift" in the phosphorus-31 peak because of the different chemical environment around the phosphorus atom in phosphorus-containing oil components (or decomposition fragments) compared to the phosphorus atom in phosphoric acid (H3PO4). This shift is reported in units of ppm (which term has no relation to concentration). By definition, the shift for the phosphoric acid standard is zero ppm. Average data accumulation time for these spectra was 1 h.

Results First Study---Comparison of Formulated Oils with Different PEIs


A series of five motor oil samples, four from the Institute of Materials database (samples EO7450, EO-7459, EO-7538, and EO-7551) and one of a typical GF-3 motor oil supplied by Chevron Oronite Co., L.L.C (Sample RO-780) was subjected to the Selby-Noack test. The first four engine oils were chosen from the database because of similar volatilities but dissimilar phosphorus contents in the volatilized oil. The fifth oil was matched to a second sample of oil that contained no ZDDP (Sample RO-779). A series of 31p NMR spectra was taken of each of the fresh oils and of their volatiles and residue generated by the Selby-Noack test (see Tables 1-5 and Figs. 1-5). Also included in the Tables are the Selby-Noack oil volatility results and the PEI values for each of the starting oils. TABLE 1--Summary of 31 P NMR Results for E0-7450. Selby- Noack PEI, Sample: EO-7450 Volatility mg/L Chemicalshifts(ppm) EO-7450 Fresh Oil 19.4 % 46 104.3, 99.5, 86.4, 77.8, 75.6 EO-7450 Volatiles 97.1, 69.4, 69.3, 29.6, 29.4, 29.3, 12.4 EO-7450 Residue 97.3, 3 (broad)

BOSCH ET AL. ON PNMR SPECTROSCOPY

257

Sample 7450

Volatlllm

1
/

Original 011

Residue

...........................
' x~, d.."

............................

~.." 'V ~ '~: ~" :~'~;: ~ ~IV-:'..:Z~:~,,'~.,'~'~":.IT '~/":i

-"~
~.~'

"*-', .... ,',

~'a

~',

*,

s'*

*'~

" ~', .... z', .... ~',

-' ' -i,

FIG. 1--E0-7450, Oil. TABLE 2--Summary O f 31 P NMR Results for E0-7459.


Sample: EO-7459 Selby Noack Volatility 18.05 % PEI, mg/L 1 Chemical shifts (ppm)

EO-7459 Fresh Oil EO-7459 Volatiles EO 7459 Residue

93.2, 83.9, 83.5, 81.2, 74.6, 74.2 125.6, 94.8, 94.6, 94.3, 94.0, 93.6, 83.2, 48.2 2 (broad)

Sample

7459

Volatlhm

II

Odglnll Oil

l~Iduo

lZO

1i|

lee

se

8|

7e

oo

se

~e

3e

zo

le

-It

ppe

FIG 2--E0-7459, Oil

258

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

TABLE Sample: EO-7538 EO-7538 Fresh Oil EO-7538 Volatiles EO-7538 Residue

3--Summary of ~1P NMR Results for E0-7538.


Selby - Noack Volatility 22.55 % PEI, mg/L 31 Chemical shifts (ppm) 93.3, 83.7, 82.6, 82.5 97.1, 96.9, 96.3, 95.8, 95.7, 95.5, 69.3, 69.2, 29.5, 29.4, 29.3, 28.0 4 (broad)

Sample
7538 Volatiles

J.

Original Oil
i..,,L., .u.,, l..~Jh~,llkl~ll~

|[
i ~ a l ~ J j l l ~ t L , ld.. L I . J ~ a . a . , L i . , l . L J a d t ~ . , . . l . . i , h , * l d ~ A t a ~ . - - a . ~ . ~ / h . . l l ~ , ~ , . l A Ii,Ia...

~'"'"';~ ""'11~" 7;';...... ,';'"" "j," *'1~;;'1' " "k'0 j, ...... ~',.... '3';...... W ' " ';, . . . . : . . . . -;i" "';;-"' ........ FIG.

3--E0-7538, Oil.

TABLE Sample: EO-7551 EO-7551 Fresh Oil EO-7551 Volatiles EO-7551 Residue

4--Summary of 31P NMR Results for E0-7551.


-

Selby Noack Volatility 17.34

PEI, mg/L 8

Chemical shifts (ppm) 103.6, 98.1, 95.8, 93.3, 86.7, 86.6, 78.0, 77.8 93.7, 93.3, 92.6, 25.7, 25.6 0 (broad)

BOSCH

ET

AL.

ON

PNMR

SPECTROSCOPY

~5~

Sample

7551

Volatlles

iResidue

Prll

~ o

" ~;o " " ~ o

eb . . . . 7b . . . .

~ . . . . ~ . . . . 4~ . . . .

....

~b ....

~b ....

6'

" -;b-~

/~IG. 4---E0-7551, Oil.


TABLE 5--Summary of 3t P NMR Resultsfor R0-780. Sample: Typical GF-3 Selby - Noack PEI, Chemical shifts (ppm) oil (RO-780) Volatility mg/L RO-780 Fresh Oil 14.23 % 4 100 (broad), 93.8, 84.3, 83.7, 83.6, 83.5, 82.8, 74.6, 67.9 RO-780 Volatiles 101.7, 95.1, 94.6, 94.5, 93.9, 72.4 RO-780 Residue 2 (broad)

Sample

RO-780

Volatlies

, . . . . ,

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

, . . . . ,

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

Residue

FIG. 5--R0-780, Oil.

260

ELEMENTAL ANALYSIS OF FUELS AND LUBRIGANTS

Second Study---Rate of Heating Dependence


As a preliminary test to determine when phosphorus volatiles were generated during the Selby-Noack test, RO 780 was given a 1-h ramping time to the temperature of 250~ followed by the normal hour duration at that temperature. The volatiles generated during this study were subjected to 31p NMR analysis. The results o f this study are presented in Fig. 6 and are summarized 1'n Table 6 . The 3 1 P N M R analysis o f the volatiles generated during a typical SelbyNoack test (normal ramp time o f about 9 rain to 250~ followed by a 60-min hold) are also presented in Fig. 6 and are also summarized in Table 6. Both residue samples were also analyzed by 31p NMR and were found to consist o f a large broad peak centered at approximately 2 ppm.
:Sample 780 Standard vs t Hour R a m p

1- I
Ramp Standard
_.,....Jl ...........

~'P

tl
ij

,~2"-T'~,'?T ~"?T 7:~'7 2 V T ; 7 S4"IL'T" 7 " t : 1 : ' T T T 2 T T T - T


log

t ......................................................................... "TT" 7 : " : " : T ' 2 T Jl ~g ig ]1 41 34 ii IJ i

? "'T2TTT', -11 p}

FIG. 6--R0-780, 1-h ramp versus normal ramp, volatiles. TABLE 6--Summary of the 31p NMR results of a 1 h ramp versus the standard ramp in the Selby-Noack Test. Sample: Typical GF-3 oil (RO-780) 1Hr Ramp volatiles Standard Ran~-volatiles Selby - Noack PEI, Volatility rng/L 16.03 % 9 14.23 % .... 4 Chemical shifts (ppm) 102.5, 95.0, 94.5, 93.8, 93.7, 92.7 102.5, 95.0, 94.5, 93.8, 83.2

Third Study--Time Dependence


Modified Selby-Noack tests were performed on EO-7450 and RO-780. These experiments were performed using a normal ramp heat-up time o f about 9 min to 250~ followed by a variable hold o f 10, 20, 30, or 50 min for each progressive test. The volatiles and residues were all subjected to 31p NMR analyses, however only the results of the 10-min and 50-rain runs were included in this paper. They are presented in Figs. 7 and 8 and are summarized in Table 7 for EO-7450 and Fig. 9 and are summarized in Table 8 for RO-780.

BOSCH ET AL. ON PNMR SPECTROSCOPY

261

Sample 74r~, 10 rain.

VolaUles
. . . . . I

J
' I ~ " [ ' l 1 . . . . .

Elm

FIG. 7--E0-7450, t = 10 min, volatiles and residue.


Sample 7450, t:50 rain.

VolaU~s

' " ' " t~t

ill

"

1-'1

Ii

a,

Ti

,i

sl

tl

st

z*

t,

-xl

-~l

pl..

FIG. 8 - - E 0 - 7 4 5 0 , t = 50 min, volatiles and residue. TABLE 7--Summary o f 31 P NMR results f o r E0-7450, 10 and 50 min runs. Chemical shifts (ppm) Selby - Noack PEI, Sample: EO-7450 Volatility mg,/L 104.3, 99.5, 86.4, 77.8, 75.6 19.4 % 46 Original oil (from Table 1) 97.0, 91.1, 77.8, 75.6, 69.3, 65i3, 29.6, 29.4, 6.12 % 39 Volatiles, T=10min 29.3, 28.4, 14.5, 14.1, 12.5, 12.4, 97.1, 29.4, 29.2, 0 (broad) Residue T=10 rain. 97.2 91.1, 75.5, 69.5, 69.4, 61.0, 29.5, 29.3, 12.97 % 53 Volatiles, T=50min 28.4, 12.4 97.3, 4 (broad) Residue T=50 rain.

262

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Sample750,
10 and ISO mln. SO min. Volatiles

-I/'~
~W,.~] ..................
r ] . . . . . . . . i . . . . . . . . . 1,,~

....... lat,~ ..........


......

~,-L,, ..................................
. . . . . . . . ,. . . . . . . . . . , . . . . . . . . . . .

A.L, L..Ld.,.,la ,, i[i xalR I I


, ,1 . . . . . . . ~,.r,. e

I. . . . . . . . . . . . SO mill. Residue

...... ill ....... iii ....... :, ......... ,; ......... g,. . . . . . . . ,',......... ; ........ ;i, ' p,'

t0 rain. Volatlles

[[
.t . . . . . . . . r ,

,L ~ j ~]L&,.I t. . . . . . . . . . . . . . . . . . . . . . . . . . ...... ;;":. . . . . . ."; . ~ . . . . . . . . . . .7 Y'7!'~"Y~, ?V.??L"~" . 'Y.?. . .."?T[.tlZ"Y77.'.7.77": "T',Z"." : Y;" Y. .';", ~ 77 ~ 7: T ~ "?TE" ~"l', Y,~ .' "~'. . .'~ ,'7: ' ': '! 10mln.Res due FIG. 9---R0-780, t = 10 min volatiles and residue and t = 50 min volatiles and residue.

TABLE 8--Summary of 31 P NMR results for R0-780, 10 and 50 min runs. Sample: RO-780 oil Selby - Noaek PEI, Chemical shifts (ppm) Volatility mg/L Original oil (from Table 5) 100 (broad), 93.8, 84.3, 83.7, 83.6, 83.5, 82.8, 74.6, 67.9 Volatiles, T=10 min 4.03 % 9 103.0, 102.4, 94.4, 94.0, 93.8, 93.3, 88.9, 83.2, 67.6, 54.1 Residue, T=10 min. 54.1, 0 (broad) 13.62 % 5 95.1, 94.6, 93.9, 83.0, 54.1 Volatiles, T=50 rain. Residue, T=50 min. 4 (broad)

Fourth Study--Effects of ZDDP Chemistry and Formation


Four Z D D P samples were prepared in the Astaris labs. The alcohols employed were 2Ethylhexanol, a primary alcohol, and 4-Methyl-2-pentanol, a secondary alcohol. Two P2S5 samples were employed, one with a phosphorus content o f 27.77 % (Low Phos. P2S5) and the second with a phosphorus contend o f 28.07 % (High Phos. PzSs). These four Z D D P s were blended into a typical GF-3 oil at Savant using a non-ZDDP-containing version o f RO 780 provided b y Chevron Oronite C o m p a n y LLC. ZDDP ZDDP ZDDP ZDDP A B C D 2-Ethylhexanol 2-Ethylhexanol 4-Methyl-2-pentanol 4-Methyl-2-pentanol Low Phos. P2S5 High Phos. P2S5 Low Phos. P2S5 High Phos. P2S5

These samples were subjected to a series o f Selby-Noack tests in which the runs were terminated at 10, 20, 30, and 50 m i n progressively. The resulting volatiles and residues, along with the starting oil samples, were subjected to 3~p N M R analyses.

BOSCH ET AL. ON PNMR SPECTROSCOPY

263

Figures 10-14 contain the most relevant 31p NMR spectra; however, Tables 9-12 contain summaries of the results of all of the NMR spectra. The 31p NMR spectra for the experiments involving the two 2-Ethylhexyl ZDDPs were integrated, which are summarized in Tables 13 and 14.

Original OII

FIG. 10---ZDDP A, Original Oil and t = 10 volatiles and residue.

!
Sample ZDDP A 50 mln. Volatltes
l

,.

.~

Is

-1.

-t,

p,.

FIG. 11

ZDDP A, Original Oil, and t = 50 volatiles and residue.

264

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

T A B L E 9--Summary Of 31 P NMR results for Sample: ZDDP A Selby - Noack .... Volatility Oil Volatiles, T=10 min. 3.31% Volatiles, T=20 rain. 7.09 % Volatiles, T=30 min. 8.43 % Volatiles, T=50 min. 11.82 % Residue, T=10 min. Residue, T=20 min. Residue, T=30 n'fin. Residue, T=50 rain.

ZDDP A (2-Ethylhexanol + Low Phos. P2S5).


PEI, mg/L 6 13 14 19 Chemical shifts (ppm) 104, 98.1, 69.5, 65 104, 96.7, 84.4, 69.0, 29.6 103.3, 97.2, 84.2, 69.5, 29.9 103.2, 97.2, 69.6, 30.0 103.5, 97.2, 69.6, 30.0 97.3, 69.6, 29.7, 2 97.2, 69.6, 29.7, 2 97.3, 69.7, 29.7, 2 97.3, 69.7, 65, 2

Sample

ZDDP

1
I

,,

. . . . . . . . . . . . . . . .

. . . . . . . . . .

9 .

50 MIn. Residue

7*

le ,

Wl

,J

* ~ ' ~ ' Zl

" "

~X ,

- i*

" " - .~l

Pm

'

FIG. 12--ZDDP t?, Original Oil, and t = 50 volatiles and residue.

T .ABLE lO--Summary of 31P NMR results for ZDDP B (2-Ethylhexanol + High Phos. P2S5). Sample: ZDDP - B Selby - Noack PEI, Chemical shifts (ppm) Volatility mg/L Oil 104.1, 103.5, 98.5, 69.7, 65 Volatiles, T=10 rain. 3.59 % 6 103.9, 96.8, 84.3, 69, 29.6 Volatiles, T=20 rain. 5.97 % 10 103.3, 97.2, 84.2, 70, 29.9 Volatiles, T=30 min. 8.55 % 13 103.5, 97.1, 69.4, 29.8 Volatiles, T=50 rain. 13.72 % 18 103, 97.3, 69.6, 30.0 Residue, T=10 rain. 97.2, 69.6, 65.3, 29.6, 2 Residue, T--20 min. 97.3, 69.7, 65.2, 29.7, 3 Residue, "1"--30rain. 97.2, 69.6, 65.3, I Residue, "I"--50min. 97.3, 69.7, 65.4, 2

BOSCH ET AL. ON PNMR SPECTROSCOPY


SampleZDDPC mln.VolMlkll

265

20

.... ' ' " " "

1 ~ .........

''

F~'I rl'''

'r] ~-

", ~ " , ' " r r -

' ,']l'

,, i,p,~ H I ' , - -1[~1 I, , r I ~ I' 1' '~"1,'

FIG. 13--ZDDP C Original Oil and t = 20 volatiles and t = 10 residue. TABLE, 1 X--Summary of3'P NMR results for ZDDP C (4-Methyl-2-1aentanol+Low Phos. PeSs). Sample: ZDDP-C Selby - Noack PEI, Chemical shitts (ppm) Volatility m~/L 99.2, 93.1, 77.3 Oil Volatiles, T=10 rain. 4.28 % 5 103, 97, 94.1, 89.1, 83.2, 66.4 Volatiles, T=20 rain. 6.23 % 5 102.7, 101.6, 95.0, 94.5, 94.2, 93.5, 90.2, 89.2, 89.1, 88.1, 85.3, 83.1, 66.6, 26.5 Volatiles, T=30 rnin. 9.80 % 5 102.5, 101.5, 94.3, 93.5, 89.3, 88.1, 83.1, 66.7, Volatiles, T=50 rain. 15.29% 4 102.3, 101.2, 95.3, 94.7, 94.4, 93.7, 93.5, 83.1, 66.8, 26.8 Residue, T=10 rain. 3 (broad) Residue, T=20 rain. 4 (broad) Residue, T=30 rain. 3 (broad) Residue, T=50 rain. 3 Ibroad)
SIl~pleZDDPD /

,ir.l~-ir Hi ~ "r ['" .i,i- ~ . 1~-. U,, .,.[r,T ...... i...1.~.,rni..,p~.. ,], ~..,n ,n , l~li?vl'" T

Orlgln~O~OII

FIG. 14---ZDDP D - Original Oil and t = 10 volatiles and residue.

266

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

T A B L E 12--Summary of31P NMR results for ZDDP D(4-Methyl-2-pentanol+High phos. PeSs). Sample: ZDDP D Selby - Noack PEI, Chemical shiRs (ppm) Volatility mg/L 100.4, 93.1, 77.3, 5.5 Oil Volafiles, T=10 rain. 3.98 % 102.9, 101.9, 94.2, 93.9, 83.2, 66, 26,5, 5.9, 5.5 Volatiles, T=20 rain. 5.83 % 102.4, 101.3, 83.1, 67.5, 66.7, 26.6, 5.9, 5.4, 4.8 Volatiles, T=30 rain. 8.65 % 102.6, 94.5, 83.1, 67.8, 66.7, 5.5, 4.8, Volatiles, T=50 min. 13.99% 102.2, 95.4, 94.5, 83.1, 66.8, 5.9, 5.4, 4.8 Residue, T=10 rnin. 4 (broad) Residue, "1"--20rain. 4 (broad) Residue, T=30 rain. 3 (broad) 6 (broad) Residue, T=50 rain. . ,7 TABLE 1 3 - - I n t e g r a l s of the 31p NMR results for ZDDP A (2-Ethylhexanol + Low Phos. PeSs).

(Reported as relative mole %.)


ZDDPA
Sample V, t=10 V, t=20 V, t=30 V, t=50 R, t=10 R, t=20 R, t=30 R, t=50 104pprn 97ppm 7 5 5 trace 54 57 64 70 38 26 28 25 84pprn 6 3 69ppm 28 30 27 28 6 4 5 5 30ppm 5 5 4 3 2 1 1 1 2ppm

55 69 67 69

T A B L E 14---Integrals of the 3Jp NMR results for ZDDP B (2-Ethylhexanol + High Phos. P2S5).

(Reported as relative mole %.)


zDDPB Sample V, t=10 V, t=20 V, t=30 V, t=50 R, t=10 R, t=20 R, t=30 R, t=50 104ppm 97ppm 9 4 3 2 53 62 67 67 32 24 23 19 84ppm 6 2 69ppm 26 28 25 27 5 4 4 7 65ppm 30pprn 6 5 5 5 2 1 2ppm

2 2 2 1

59 69 71 73

Discussion

General
From the foregoing studies engine oils decompose during species, some portion o f which ZDDPs also break down under in an engine. it is evident that the ZDDPs in all the engine oils and simulated the Selby-Noack test and are converted to different phosphate are volatile. Similarly, it has been shown in other studies [2] that analogous conditions o f temperature, time, and oxygen exposure

BOSCH ET AL. ON PNMR SPECTROSCOPY

267

First Study
The NMR analyses shown in Figs. 1-5 and Tables 1-5 indicated that the various samples of volatiles from these Selby-Noack volatilization tests contain multiple phosphorus species and that most of these species are different from those present in the original motor oil. Perhaps more interesting, the first study gave clear evidence that different formulations of engine oils with different values of PEI gave considerably different phosphorus breakdown products. The residue oil left in the heated cup after completion of the Selby-Noack volatility test also contains totally different phosphorus species than those in the original motor oil. More specifically, all of the residue samples were observed to consist of a broad peak centered at 0-4 ppm, which is indicative of inorganic and/or simpler organophosphate compounds than ZDDP. In these typical GF-3 oils, a PEI of 8 corresponds to approximately 1 % of the phosphorus volatilizing from the engine oil sample. In three of the five motor oils tested, the ZDDPs completely decomposed under test conditions. In samples 7450 and 7538, small amounts of ZDDP remained in the residual oil at the end of the test.

Second Study
In the experiments on RO-780 (typical GF-3 oil) in which the effect of extending the time of the heat up ramp from 9 min to 60 min was examined, it was found that although the PEI increased when the time to operating temperature was considerably extended, the NMR spectra did not change significantly. That is, the same species were volatilized from the oil. This is illustrated Fig. 6 and summarized in Table 6.

Third Study
The experiments performed on samples EO-7450 and RO-780 (see Figs. 7-gand Tables 7 and 8) were, as noted, run to determine what and when different volatile phosphorus species were generated. The tests were run in 10-, 20-, 30-, 50-, and 60-rain (from the first and second studies) exposure times to the Selby-Noack operating condition. From these samples the NMR analyses of the residue and volatile samples collected from the samples at 10 and 50 min were compared Unexpectedly, in the RO-780, the decomposition of the motor oil was essentially complete after 10 min. In addition, it was noted that the PEI of the 50-min sample was lower than that of the 10-min sample. Although this might be associated with repeatability of the PEI at low values, it may also indicate that some of the phosphorus-containing species in the collected volatiles could be lost because of even further volatilization from the collection vessel during the SelbyNoaek test. This requires further investigation. The NMR spectra of the volatiles obtained in these two runs were compared with Figs. 5, 6, and 9. Qualitatively, the respective spectra were found to be very similar, which illustrates the repeatability of the volatilization and collection technique as an analytical method. As observed before, in sample 7450, the decomposition of the phosphorus containing species is not complete after 50 rain. It is very interesting to note that both the 10-rain and 50-rain residue spectra are significantly different from that of the original oil sample. It appears that this oil first decomposes to an intermediate state, which then undergoes further reaction to generate additional phosphorus volatile species. It is also noted that the NMR spectra obtained on the 10and 50-mi volatile samples qualitatively consisted of the same peaks. However, the proportionate size of the peaks varied between the two samples.

268

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

Fourth Study---ZDDP Composition Effects


Evolution of volatile components was found to depend strongly on the alcohol used in ZDDP manufacture, as illustrated in Figs. 10-14 and Tables 9-12. ZDDPs C and D made from the secondary alcohols, 4-methyl-2-pentanol, appear to break down readily under the conditions of the Selby-Noack test. Within the first 10 min, whatever volatile species were formed were given off, and the only phosphorus-containing species in the residual oil were fully oxidized phosphates. Interestingly, the PEI values were comparatively low and constant. In contrast, ZDDPs A and B made from the primary alcohols, 2-ethylhexanol, have a longer life. Volatile phosphorus-containing species were given off for up to 60 min. Concomitantly, the residual oil contained some of the original ZDDPs, but consisted primarily of other phosphoruscontaining components in addition to fully oxidized phosphates. However, in contrast to aforementioned results with the primary alcohol, the PEI values increased with time and became comparatively high. These differences in rate and degree of breakdown between ZDDPs made with primary and secondary alcohols reflect other findings in the literature [2].

Fourth Study--Stoichiometric Effects Results with ZDDPs A, B, C, and D---In the limited testing described in this paper, the phosphorus/sulfur molar ratio of the PzS5 does not appear to have a very significant effect on the composition of the volatile species generated from the ZDDP made from 2-ethylhexanol. Similar results were obtained from ZDDP made from both High and Low Phos P2S5. However the PEI values obtained on the ZDDPs manufactured form High and Low Phos PzS5 and 4-methyl-2pentanol differed significantly. Close examination indicates that some of the unique impurities in ZDDP D are most likely volatilizing and thereby increasing the PEI value of this oil. These impurities are observed in both the original oil and the volatiles at 3-8 and from 65-70 ppm. Unpublished studies at Astaris have found that typical yields of dithioacids (the ZDDP precursor prepared from the reaction of P2S5 and alcohol) are generally between 85 and 95 %. These reactions are typically run with an excess of alcohol to drive the reaction to completion. The phosphoms/sulfiar molar ratio in the P2S5 has been found to affect the trace components generated during the reaction of PzS5 and alcohols. Use of P2S5 o f a slightly lower-thanstoichiometric phosphorus level generates some impurities that are rich in sulfur, while some of the impurities generated in reactions performed with PzSs at a slightly higher-than-stoichiometric level of phosphorus consist of p+3 compounds but do not include sulfur-rich impurities. Similarly, the P+3 impurities of higher than stoichiometric reactions are not generally found in the products generated from P2S5 with a lower than stoichiometric level of phosphorus. This was one of the primary points of interest in performing the stoichiometric experiments with P2S5, i.e., to see if these different impurity mixtures lead to ZDDPs with different PEIs, which, in fact, appears to be the case. To generate more detailed information regarding the composition of the volatile products from the PEI studies, the relative ratios of the various peaks found in Tables 13 and 14, which were taken over the time interval of 10-50 rain of the Selby-Noack test, were multiplied by the PEI values of these samples to calculate the PEIs of the individual peaks. That is, the total PEI value for the sample was separated into the individual PEI components in the same relative ratio as the 31p NMR peaks. The results o f this comparison are presented in Figs. 15 and 16 and show interesting trends.

BOSCH ET AL. ON PNMR SPECTROSCOPY

269

FIG. 15--ZDDP A, PEI values by NMR peak.

FIG. 16--ZDDP B, PEI values by NMR peak. Some peaks grow while some stay the same or slightly shrink during the duration of the test. Two peaks, those at 104 and 84 ppm, remain the same or shrink. The peaks at 97 and 69 ppm clearly grow, and the peak at 30 ppm appears to grow somewhat, but not dramatically. Before the test, Oil A has peaks at 104, 98.1, 69.5, and 65 ppm, and Oil B has peaks at 104.1, 103.5, 98.5, 69.7, and 65 ppm. Therefore, it is possible that the peak at 104 ppm is an impurity that volatilizes and the small peak at 69 ppm in the before-test oils is generated during the thermal decomposition of the oils during the PEI test and subsequently is volatilized. The large peak at approximately 98 ppm in the before-test oil is perhaps represented as the dithioacid in the volatilized oil samples. Further study is required.

General Observations
Volatile components from ZDDPs may arise from the manufacturing process. Zinc dithiophosphates are typically a mixture of phosphorus containing species. In addition to the desired ZDDP, [(RO)2P(S)S]2Zn and basic ZDP salt, there are a number of other trace components containing phosphorus, as shown in Table 15. Some of the trace components in the unused oil were also observed in the volatiles.

Table 15 - Literature Survey of 31p NMR Analyses of ZDDPs


",.4 0

Species Observed.,
m rm m z

((RO)2P(S)S)sZr~C ((RO)2P(S)S)~.r RO)2P(S)SnR(RO)2P(S)SH (RO)zP(S)OR (RO)zP(S)O- [(RO)2P(O)S]2 (ROhP(O)SR (RO)3P(O) (OR)P(S)(SR)2 (S)P-(SRh sulfide n = basic ZDDP neutral ZDDP I r2r3 thio acid thiophosphate :hiophosphodcsai' thiophosphate phosphate

Ref.

Expedment

'shift, ppm]
(RO)zP(S)OR [60 to 72] [37 to 551 -,, (RO)2P(O)SR [24 to 30] [13 to-13] I

15000km 12 ertgine field test ((RO)2P(S)S)eZn~O((RO)2P(S)S)2Zr (RO)2P(S)SR

I
I '~"~"~

t'---

Z
ffJ

13

lab oven 180*C, in air

t
I ~~

t I
I (RO)sP(O)

t I I
[0.5 to .12]

""14,'1 lab4-ballin air rigs

o -rl
C m r"

15 I

lab; 150" to 200~ in air *

16

lab; 200=to 260~ in air *~

I
I
17o]
J'40TO 45] I

I
(RO)2P(S)O" [(RO)zP(O)S]~ [49l [22]
[691

I
Feto8o] I
(RohP(s)o"
I(RO)zP(O)SR [20to26]

I I

I I I
I (RO}~P(O) I [13t0-13] I

ZDOP+ cumene I ] 17 ~ hydropemxide ((RO)2P(S)S)eZn,O ((RO)2P(S)S~Zn (RO)zP(S)SR

,]

40~

[,o3] i

m]

i r~toss]

i1> z 0 FC 0 Z

1817-..ODPsyrlthesis

((RO)=P(S)S)eZn40((RO)2P(S),S)~Zn~ (RO)zP(S)SR (RO)2P(S)SH[ (RO)2P(S)OR

~101to103]

[96to100)

*ZDDP gone in 1 hr at 200~ or 24 hr at 150~

(RO)zP(S)O- I ~ all ZDDP gone in 4 hr at 200" and 260~

The volatile components from ZDDPs may also arise from thermal degradation and/or oxidation. The mechanisms proposed for ZDDP anti-wear and anti-oxidation performance in motor oil have included both the effects of ZDDP degradation and oxidation [2]. In addition, 3~p NMR studies at other Laboratories have identified ZDDP breakdown fragments in the used motor oil [12], as shown in Table 15. Moreover, ZDDP breakdown fragments have been observed using 3~p NMR in lab tests elsewhere [13-16]. These lab tests were also done in the presence of air. So the relative contribution of thermal degradation versus oxidation is not clear. The direct oxidation era ZDDP with hydroperoxide [17], as observed by 3~p NMR, also gives fragments that may be volatile.

BOSCH ET AL. ON PNMR SPECTROSCOPY

271

The evolution of volatile components has been found to depend strongly on the alcohol used in ZDDP manufacture. ZDDPs made from the secondary alcohol 4-methyl-2-pentanol appear to break down readily under the conditions of the Selby-Noack test, but only a small percentage of the phosphorus break-down products were volatile. Within the first 10 min, all the volatile species were given off, and the only phosphorus-containing species in the residual oil were fully oxidized phosphates. In contrast, ZDDPs made from the primary alcohol 2-ethylhexanol appear to have a longer life but produce more volatile phosphorus over their breakdown life. Volatile phosphorus-containing species were generated for at least 60 min. In concert, the residual oil contained ZDDPs and other phosphorus-containing components in addition to fully oxidized phosphates. The rapid breakdown of ZDDPs has also been observed elsewhere: Lab tests in air using 31p NMR show that all the ZDDP is gone in 1 h at 200~ [15], 24 h at 150~ [15], 4 h at 200~ [16], or 4 h at 260~ [16]. In the latter two tests, the only phosphorus containing species was a fully oxidized phosphate. A recently published study by scientists at MIT [20] has indicated how to quantify the amount of oil lost by three possible mechanisms. These include thermal decomposition of the oil in the engine, aerosol generation, and oil volatilization. The information gathered in the present paper using the Selby-Noack and the related NMR spectroscopy testing methods can be used to quantify phosphorus volatility in the same manner. Measurement of zinc in the volatiles can be used to determine the amount of phosphorus generated via aerosol generation. In addition, it appears that careful 31p NM-R analyses of the volatiles could be used to measure the phosphorus generated by volatilization of the components of the motor oil versus the phosphorus homologs generated by thermal decomposition. Earlier studies by Savant showed that there were trace levels of zinc in the phosphorus volatiles obtained from the Selby-Noack test. The combination of varying the length of the Selby-Noack test, followed by 3lp NMR analysis of the volatiles and residues, has the potential to be a very useful tool to further understand the mechanism of phosphorus volatility of motor oil.
Conclusions

The 31p ~ data are reproducible using the Selby-Noack procedure. For the same conditions of time and temperature, the information, though limited, shows that the composition data are consistent when the experiments are repeated. Different oils formulated with different ZDDPs and other additives have different volatile components. Moreover, the composition of the volatile phosphorus containing components depends on the specific ZDDP. In most cases, the phosphorus forms found in the volatiles appeared to contain some of the ZDDP additive. But the 31P NMR spectrum also indicated the presence of more species in the volatiles than in the untested oil. That is, the volatile components appear to be generated primarily by the decomposition of ZDDP. However, volatile phosphorus containing trace components arising from the ZDDP manufacturing process may be significant in some cases. It appears possible to meet engine oil performance standards using additives having low phosphorus volatility. Although all the oils in the IOM Database for North America are presumed to have met minimal standards for engine wear and oil oxidation, there is a large difference in Phosphorus Emission Index (PEI) among these oils. By using PEI as an additional criterion for additive selection, it would seem reasonable to provide protection for both the

272

ELEMENTAL ANALYSIS OF FUELS AND LUBRICANTS

engine and the emissions control system. Certainly, approaching the control of phosphorus volatility by actual measurement is highly preferable to attempting to control phosphorus volatility by limiting the phosphorus content of fresh engine oil. Acknowledgments The authors would like to thank Dr. Andre' d'Avignon at Washington University for performing the 31p NMR studies, Mr. Skip Ramsey for synthesizing the ZDDP samples, and Chevron Oronite Company, LLC for providing typical GF-3 oil samples with and without ZDDP.
References

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

Selby, T. W., Bosch, R. J., and Fee, D. C., "Phosphorus Additive Chemistry and Its Effects on the Phosphorus Volatility of Engine Oils," submitted to ASTM International. Spikes, H., "The History and Mechanisms of ZDDP," Tribology Letters, Vol. 17, No. 3, 2004, pp. 469-489. Selby, T.W., et al., "A New Approach to the Noack Test for Volatility Measurement," SAE International Fuels & Lubricants Meeting and Exposition, Philadelphia, PA, USA, 1993. K. Noack, Angewandt Chemie, Vol. 49, 1936, p. 385. ASTM Standard D 5800, "Evaporation Loss of Lubricating Oils by the Noack Method," Annual Book of ASTM Standards, Vol. 5.03, ASTM International, West Conshohocken, PA, 2004, pp. 352-366. Determination of Evaporation Loss of Lubricating Oils, (Noack Method), DIN 51-581, 1981. Selby, T. W. and Reichenbach, E. A., "Engine Oil Volatility Studies - Generation of Phosphorus," Proceedings of the International Tribology Conference, Yokohama, Japan, 1995, pp. 813-816. Institute of Materials Engine Oil Database, Published by the Institute of Materials, Midland, Michigan, issued yearly from 1984. Selby, T. W., Development and Significance of the Phosphorus Emission Index of Engine Oils, Proceedings of 13th International Colloquium - Lubricants, Materials, and Lubrication Engineering, Esslingen, Germany, 2002, pp. 93-102. Johnson, M. D., McCabe, R. W., Hubbard, C. P., Riley, M. E., Kirby, C. W., Ball, D.J., et al., "Effects of Engine Oil Formulation Variables on Exhaust Emissions in Taxi Fleet Service," SAE Paper #2002-01-2680, SAE Powertrain Meeting, 2002. Selby, T. W., "Phosphorus Volatility of Lubricants - Use of the Phosphorus Emission Index of Engine Oils," 9th F&L Asia Conference, Singapore, January 21-24, 2003. Peng, P., Hong, S. Z., and Lu, W. Z., "The Degradation of Zinc Dialkyldithiophosphate Additives in Fully Formulated Engine Oil as Studied by P-31 NMR Spectroscopy," Lubrication Engineering, Vol. 50, No. 3, 1994, pp. 230-5. Coy, R. C. and Jones, R. B., "The Thermal Degradation and EP Performance of Zinc Dialkyldithiophosphate Additives in White Oil," Internationales Jahrbuch tier Tribologie, Vol. 1, 1982, pp. 345-50. Coy, R. C. and Jones, R. B., "The Degradation of Zinc Dialkyldithiophosphate Additives in Rigs and Engines," 1Mech E Conference Publications, Vol. 1, 1982, pp. 17-22.

BOSCH ET AL. ON PNMR SPECTROSCOPY

273

[15] Fuller, M. L. S., Kasrai, M., Bancroft, G. M., F3ffe, K., and Tan, K. H., "Solution Decomposition of Zinc Dialkyl Dithiophosphate and Its Effect on Antiwear and Thermal Film Formation Studied by X-Ray Absorption Spectroscopy," Tribology International, Volume Date 1998, Vol. 31, No. 10, 1999, pp. 627-644. [16] Harrison, P. G. and Brown, P., "External Reflection FTIR, Phosphorus-31 MAS NMR and SEM Study of the Thermal Decomposition of Zinc(II) Bis(O,O'-dialkyldithiophosphates) on 316 Stainless Steel," Wear, Vol. 148, No. 1, 1991, pp. 123-34. [17] Yagishita, K. and Igarashi, J., "31P NMR and Mass Spectrometric Studies of the Reaction of Zinc Dialkyldithiophosphates with Cumene Hydroperoxide. (Part 1). Kinetics and Mechanisms of the Initial Homolytic Reaction," Sela'yu Gakkaishi, Vol. 38, No. 6, 1995, pp. 374-83. [18] Zimmermann, V., Jaeger, G., and Meyer, H., "Characterization of Reaction Mixtures in Zinc Bis(O,O'-dialkyl dithiophosphate) Synthesis by Phosphorus-31 NMR Spectroscopy," Chemische Technik (Leipzig, Germany), Vol. 38, No. 4, 1986, pp. 169-72. [19] Oehler, R., Zimmermann, V., and Jager, G., "Preparation of Zinc Dialkyl Dithiophosphates. Reaction and Reaction Products," Erdoel & Kohle, Erdgas, Petrochemie, Vol. 40, No. 2, 1987, p. 87. [20] Yilmaz, E., Tian, T. Wong, V. W., and Heywood, J. B., "The Contribution of Different Oil Consumption Sources to Total Oil Consumption in a Spark Ignition Engine," SAE Paper, 2004-01-2909.

STP1468-EB/Sep. 2005

Author Index
A Anderson, Daniel E, 71
B

Mason, Kelly John, 42, 196 McEIroy, Frank C., 42, 196 Mennito, Anthony S., 42 Mustacich, Robert V., 164
N

Beumer, Berry, 116 Bosch, R. J., 239, 255 Brouwer, P., 98 Bryce, Derek W., 207 C Chen, Z. W., 116 Corns, Warren T., 207 Cruko, John, i 37
D

Nash, Lisa J., 232 Ngobeni, Prince, 59 O Onyeso, Chris C., 17


P

Diehl, John W., 221 Di Sanzo, Frank P., 221 Donard, O. E X., 51 Dreyfus, S., 51
F

Paszek, Joseph, 42 P6cheyran, C., 51 Prinzhofer, A., 51 Priola, Marco, 59


R

Fee, D. C., 239, 255 Fess, Scott, 108 Fox, Brian S., 24, 196 G Gras, Rhonda L., 164 I-I Horton, Meaghan, 33 Hwang, J. David, 33
K

Radley, Ian, I 16 Rhodes, John R., 152 Ricchiuto, Silvia, 59

Selby, T. W., 239, 255 Shearer, Randy L., 164 Stockwell, Peter B., 207 T Tarkanic, Steve, 137 Tittarelli, Paolo, 59 V Vrebos, B., 98 W Wei, Fuzhong, 116 Wilhelm, S. Mark, 181 Wissmann, Dirk, 128 Wolska, J., 98

Kariher, Peter H., 181 Katskov, Dmitri A., 59 Kirchgessner, David A., 18 I Kishore Nadkarni, R. A., I, 85 L Leong, David, 33 Liang, Lian, 181 Lienemann, C. P., 51 Lukas, Malte, 71 Luong, Jim C., 164
M

u Yurko, Robert J., 71


275

Magnier, C., 51

Copyright9 2005by ASTMInternational www.astm.org

STP1468-EB/Sep. 2005

Subject Index
A Additives, 17, 24, 221,239 Analysis, perfect, 1 Analytical laboratory, I Asphaltenes, 51 ASTM D 5453, 137 ASTM D 5623, 164 ASTM D 6920, 152 ASTM D 7039, 116 Atomic emission detection, 221 Atomic fluorescence spectrometry, 207 C Combustion cold vapor atomic absorption spectrometry, 196 Combustion/oxidation UV-fluorescence detection, 137 Comprehensive two-dimensional gas chromatography, 22 I Contaminant particles, analysis, 71 Crude oil, 51, 108, 181, 196
D

Fluorine, 232 Fossil fuels, 207 Fuels, 108, 221 analysis, 98, 232 G Gas chromatography, 221 Gasoline, 17, 59, 85, 137, 223 low sulfur content, 128 ultra-low sulfur, 152 Gasolines, 221 l-I Hetero-atom speciation, 221 Hydrocarbons, sulfur determination, 164

Diesel fuel, 17, 59, 85, 116, 137 ultra-low sulfur, 116, 137, 152 Doubly curved crystal optics, 116 Dual plasma sulfur chemiluminescence detector, 164
E

Inductively coupled plasma atomic emission spectrometry, 17, 24, 33, 71 Inductively coupled plasma-mass spectrometry, 33, 42, 51 Interlaboratory crosscheck programs, 85 Ion chromatography, 42 Ion-selective electrode, 232

Jet fuel, 59, 221 L Liquid chromatography, 42 Lower limits of detection, 98 Low thermal mass gas chromatography, 164 Lubricants, 232 Lubricating grease, 24, 98
M

EDXRF, 128 Electrochemical sensors, 152 Electrothermal atomic absorption, 59 Elemental analysis, 98, 232 Energy dispersive x-ray spectrometer, 98 Engine oils, 239 Exhaust catalyst, 239, 255 Exhaust emissions, 239, 255
F

Fast on-line/at process sulfur determination, 137 Filter furnace electrothermal atomic absorption spectrometry, 59 Fluoride, 232

Maltenes, 51 Mercury, 181, 196, 207 volatile, 181 Microflow nebulizer, 51 Microwave digestion, 33 Monochromatic wavelength dispersive x-ray fluorescence, 116

276

INDEX

277

Neutron activation analysis, 181 NIST 1084a, 51 NIST 1085b, 51 NIST 1634c, 51 Nitrogen converters, 152 Nuclear magnetic resonance spectroscopy, 255 O On-line process analyzers, 152 Organic solvents, 51 Oxidative combustion, 232

Sample preparation, 33 Selby-Noack volatility test, 255 Selenate, 42 Selenite, 42 Selenium, 42 Selenocyanate, 42 Speciation, 181 Spray chamber, 17, 51 Sulfur, 108, 116, 128, 164, 221 analysis, 85 ultralow, in diesels, 85 Sulfur chemiluminescence detection, 164 Sulfur/nitrogen chemiluminescence, 232 T

3zp Nuclear magnetic resonance spectroscopy, 255 Petroleum and petrochemical products, 33, 183, 207 Phosphorus emission index, 239, 255 Phosphorus volatility, 239, 255 Pipeline, 108 Pneumatic nebulizer, 17, 51 Polarization, ! 28 Polarized energy dispersive x-ray spectrometry, 128 Pooled limit of quantification, I 16 Pyro-electrochemical on-line ultra low sulfur analyzer, 152

Thermal oxidation, 152 Trace metals, 51, 59 Transverse heated filter atomizer, 59 U Ultrasonication, 181 Ultra-trace metals, 42, 51 U. S. EPA 7473, 181 Used oil analysis, 71 W Wavelength dispersive x-ray spectrometer, 98 Wear metals, 17 Wear particles, analysis, 71 X

Q
Quality management, 1
R

Rapid sulfur analysis, 137 Refinery, 108 Refining, 181 Rotating disc electrode spectrometer, 71 Rotrode filter spectroscopy, 71 Round robin test, 128

X-ray fluorescence spectrometry, 98 X-ray spectrometry, 98 X-ray transmission, 108

Zinc dithiophosphates, 239, 255

You might also like