Dynamic shear rheometers

J Carswell BP Oil International, Sunbury M J Claxton BP Oil International, Sunbury P J Green BP Oil International, Sunbury

Introduction
There is increasing interest in the use of fundamental rheological parameters to specify binders. In the USA, the Strategic Highways Research Programme (SHRP) has developed binder specifications which classify the grade of a binder on the basis of its performance in a number of rheological tests. Parameters specified are based on both the binder complex shear modulus (|G*|) and the phase angle (*). In the UK, new specifications for modified binders in both hot rolled asphalt (HRA) wearing course and surface dressing also require rheological measurements. It is proposed that rheological data will be required for modified binders for any road application. The specification (Highways Agency Clause 943) is still in draft but some contracts already request rheological information for both applications. The draft specifications, once finalised, will eventually be adopted in the Specification for Highways Works. Data required include either isotherms of |G*| as a function of frequency at a number of temperatures or binder master curves. A Dynamic Shear Rheometer (DSR) may be used to determine the rheological properties of bituminous binders and it is generally assumed that a DSR can accurately measure binder properties over a wide range of conditions. However, the use of DSRs to characterise bitumen properties is still in its infancy when compared with well established measurements such as penetration, softening point and viscosity. These conventional tests recognise the importance of closely controlled conditions and appropriate calibration to obtain reproducible results, particularly for specification purposes. In contrast, for rheological measurements using DSRs, little work has been reported to date to identify the factors which may affect the accuracy of results. This paper discusses some of these factors and quantifies the effects of two important sources of error : temperature control and sample test geometry.

Factors affecting results on Bitumens

A number of factors have been found to affect the results of rheological measurements on bitumens using a DSR :

Temperature control
Bitumen is a good thermal insulator and is also highly temperature susceptible. In order to make accurate measurements it is important firstly, that the whole sample is at the same temperature and secondly, that the temperature is accurately controlled. Temperature calibration is sometimes carried out using a standard viscosity oil which is much less temperature susceptible than bitumen. In such cases, inaccurate temperature control is much less evident than when making measurements on bitumens.

Sample geometry
Bitumen stiffness varies over several orders of magnitude over a relatively narrow temperature range. DSR manufacturers offer a number of different parallel plate and cone/plate geometries to measure the wide range of stiffnesses (and viscosities) encountered. However, in the UK there seem to be no clear guidelines on the choice of suitable sample geometry (plate diameter and gap setting) and it is left up to the individual to determine the most appropriate for the binder under test. As a result, there are no

most appropriate for the binder under test. As a result, there are no standard geometries which have been identified as suitable for specific ranges of binder stiffness. This is regarded by the authors as a major potential source of error.

Sample preparation
As mentioned earlier, conventional bitumen tests such as penetration, softening point and viscosity have long recognised the importance of sample conditioning and thermal history in obtaining repeatable and reproducible results. At present in the UK, no standard procedure exists for sample preparation for rheological measurements using dynamic shear rheometers. The authors recognise the importance of sample preparation and there is an urgent need to standardise on a procedure. It is however recognised that, as a result of the SHRP program, a sample preparation method has been proposed by AASHTO for use with bitumens (1). In the work reported in this paper a standard in-house procedure for sample preparation was used for all testing.

Equipment calibration
For calibration purposes, most DSR manufacturers carry out measurements of displacement and applied torque and confirm such calibrations with measurements in steady shear on a standard fluid. Unfortunately, these calibration fluids are often low viscosity, Newtonian oils which are very different in consistency from bitumen, where the range of stiffnesses encountered can be much higher. Consequently, such calibrations can be misleading, particularly if subsequent measurements are to be made on high stiffness binders, as discussed later.

Linear viscoelasticity
Many viscoelastic materials such as bitumens do not behave linearly in terms of their stiffness as a f unction of strain. Unless measurements are made at strains where the properties are independent of the level of strain, then comparison with data from different instruments can be very difficult. Although the importance of this is recognised by the authors, it is not discussed further in this paper, except to say that all results quoted were obtained in the linear viscoelastic region for the binders tested, to enable valid comparisons to be made between instruments and under different conditions. The effects of temperature control and sample geometry will now be discussed in detail.

Binders
The effect of temperature control was studied using an unmodified binder. Physical properties of the binder are given in Table 1. Table 1 Binder Physical Properties (Temperature Control Study) Property Test Method Value

Penetration @ 25*C Ring & Ball Softening Point Dynamic Viscosity at 60*C Dynamic Viscosity at

62 mm/10 IP58 IP 222 Brookfield

IP49 49.2*C 322 Pa.s 1070 mPa.s

Dynamic Viscosity at 150*C

Brookfield

295 mPa.s

The temperature dependency of the viscosity of bitumens may be described by the function: C(4- log(*)) * K + Te = T (1) (C-4+log(*)) Where * is the viscosity in Pa.s, Te is the temperature at which the viscosity is 104 Pa.s, K is a constant, T is the temperature in *C and C=8.86 (T*Te) or 12.5 (T*Te). For a bitumen of known viscosities at two or more temperatures the viscosity at any temperature may be calculated. Alternatively, the temperature for a given viscosity may be determined. (Equation 1 is preferred to the reduced Walther equation since it gives a linear plot of viscosity over a wider temperature range.) For the study of different geometries, unmodified 15 pen and 200 pen bitumens were used to represent the hard and soft extremes of normal paving grades. Test methods and conditions All work was carried out using two commercially available rheometers DSR 1 and DSR 2. For ease of use, DSR 1 was used with a 20mm parallel plate (PP) geometry and DSR 2 was used with 8mm and 25mm PP geometries.

Effect of temperature control

Rheological testing was carried out using DSR. Two temperature control systems were used. The first employed a standard Peltier system in which only the temperature of the lower plate was controlled. This may present a source of error when making measurements on bitu mens which are poor thermal conductors. To try to quantify the effect of thermal gradients across the bitumen sample, results obtained with the Peltier temperature control system were compared using a standard stainless steel upper plate and also a 'Tufnell' upper plate of lower thermal conductivity. In addition, the effect of using an annular aluminium block around the sample to try to minimise heat losses from the upper plate was also evaluated. Finally, the effect of water around the sample as a means of improving heat transfer from the Peltier device to the upper plate was assessed, in an attempt to obtain a more uniform sample temperature (Figure 1). Figure 1 Schematic of DSR 1 with water jacket and aluminium block.

The second method of temperature control comprised an Extended Temperature Module (ETM), which used electromagnetic induction heating of both the upper and lower plates. To determine the effec ts of the different methods of temperature control, the viscosity of the standard binder was measured in steady shear at 60*C using both temperature control systems and compared with that obtained using the vacuum capillary method (IP 222). Both systems were precalibrated to determine the settings required to obtain 60.0*C. A parallel plate (PP) geometry was used for all testing with two gap settings, 0.5 and 1.0 mm.

Effect of sample geometry

To assess the effect of sample geometry, both DSRs were used i n oscillation mode. In this mode of operation, a sinusoidal stress is applied to the sample at different frequencies and a resultant strain is measured. As mentioned previously, the level of strain was selected to be within the linear visco-elastic region for the binders used. The complex shear modulus, |G*|, and visco-elastic phase angle, *, were measured under different temperature and frequency conditions as shown in Table 2, for different plate diameters and gaps.

different plate diameters and gaps.

Table 2 DSR Test Conditions (Sample Geometry Study)

DSR 1 20mm PP DSR 2 8mm PP DSR 2 25mm PP

Strain Gap (mm) Temperature Frequency

0.02 0.01 0.5 5-60C 0.1-10Hz

0.02 0.01 0.5 5-60C 0.1-10Hz

0.02 0.01 1.0 5-60C 0.1-10Hz

Results
Temperature Control Table 3 summarises the viscosity results obtained. Acceptable reproducibility of the vacuum capillary method is defined as results from two laboratories having a difference of less than 10% of their mean value. Although this definition of reproducibility applies to results from a single test method, it was used to assess agreement between the results using different methods of temperature control in this study. On this basis, for a viscosity of 322 Pa.s, a result in the range 291-356 Pa.s from the DSR 1 would be acceptable. The results which meet this criterion are shown underlined in Table 3. Gap (mm)

System

Viscosity at 60*C (Pa.s) No modification Al block Al block + water jacket

ETM Peltier (steel plate) Peltier (Tufnell plate)

0.5 1.0 0.5 1.0 0.5 1.0

345 354 540 720 413 500

460 564 365 383

420 426 387 383

Results using the ETM to heat the upper and lower plates show good agreement with the IP 222 method. In all other cases, the results do not meet the reproducibility criteria. The effect of heating only one side of the sample is shown clearly in the results using the steel plate with the Peltier system and to a lesser degree using t he Tufnell plate. As expected, increasing the gap resulted in a higher measured viscosity further away from that obtained using IP 222, indicating the problem of temperature gradients across the sample. Heating the air space around the sample or use of a w ater jacket (particularly for the steel plate) improved the agreement. The viscosities and stiffnesses of bitumen are highly temperature susceptible. From equation (1) a change of just * 1*C at 60*C will change the viscosity of this bitumen by 45Pa.s i.e. a variation from 282-367 Pa.s. Therefore, it is not surprising that IP 222 specifies close temperature control of the water bath in which the sample is immersed. Although the control of temperature to this level of accuracy is well established for conventional tests on bitumen, it is only relatively recently that such accuracy and precision have been available with modern rheometers. The largest deviation in Table 1 (720Pa.s) corresponds to an average sample temperature of just 54.2*C. The best measurement with the Peltier system (365Pa.s) corresponds to an average sample temperature of 59.0*C and again indicates the problem of heating the sample from one side only. The ETM gives better agreement and shows the benefit of heating the sample both top and bottom. However, there were disadvantages found with using this ETM for routine testing including difficulty in trimming the sample.

this ETM for routine testing including difficulty in trimming the sample. Developments within the industry have led to improvements and simplifications being made to rheometers in order to make them mor e readily useable for routine testing. In order to achieve accurate temperature control of the sample, a number of DSRs are now being marketed which incorporate a water bath in which the sample is fully immersed. Provided sufficient time is allowed for the sample to reach thermal equilibrium, accurate temperature control can be achieved. Plate Geometry The results obtained for both the 15 pen and 200 pen bitumens are shown in Figures 2 to 8. Figure 2 shows the data for the 15 pen bitumen as often presented in research papers. It is also the current format required by the UKs Highways Agency. Initially, results appear as would be expected with increases in |G*| and a gradual reduction in phase angle (d) as the binder becomes more elastic at higher frequencies and/or lower temperatures. However, closer inspection of Figure 2 suggests that the limiting elastic stiffness for the 15 pen binder is not around 109 Pa, the approximate value found for all bitumens, but a value closer to 108 Pa. Figure 2 DSR 2 results for 15 pen bitumen (8mm PP, 0.5mm gap)

An alternative way of presenting these results is to plot |G*| and delta against temperature at a single f requency. Figure 3 shows |G*| and delta at 0.4Hz for the 200 pen bitumen plotted against temperature using the 8mm and 25mm parallel plates with DSR 2, using 0.5mm and 1mm gaps respectively. Figure 4 shows similar results for the 15 pen bitumen at 10Hz. From Figures 3 and 4, it is clear that there are differences between the results obtained using the two sample geometries. Moreover, the differences become greater as the stiffness of the binder increases. At 0.4Hz with the 200 pen bitumen, there are large differences in the values of delta with the two plate geometries at all temperatures below about 40C while significant differences in the values obtained for |G*| are seen below about 25C. At 10Hz with the 15 pen bitumen, significant differences in |G*| are obtained below about 35C, while the values measured for the phase angle, *, are very different over nearly the entire temperature range. Figure 3 DSR 2 results for 200 pen bitumen (0.4 Hz)

To eliminate the effect of temperature on the results obtained, |G*| may be plotted against * over the temperature and frequency range used. In such a plot, * decreases as |G*| increases and the value of |G*| when * is 0 should be the limiting stiffness of the bitumen i.e. 109 Pa. Figures 5 and 6 show these plots for the 200 and 15 pen bitumens respectively. The following conclusions may be drawn from the data : The limiting value of |G*| (109 Pa) is never attained with either DSR for all the geometries (plate diameter and gap) used.

characterised using a number of standard empirical tests (penetration, Ring & Ball softening point, Fraass point and ductility recovery) and also rheological measurements including absolute viscosity (vacuum capillary method), and using CarriMed CSR500 and Metravib rheometers. Oscillatory testing was carried out on the binders using both rheometers. In addition, using the CarriMed rheometer, creep tests were also conducted. All temperature and loading conditions for the rheological tests will be discussed in the results.

Asphalt testing
For all asphalt tests, the mix was a continuously graded 0/10 asphaltic concrete (dense mix). Binder contents (by weight) were 5.3 % for wheel-tracking and fatigue tests and two binder contents of 4.75 % and 6.0 % for low temperature asphalt testing. The high temperature performance of an Hot Rolled Asphalt (HRA) mix was also evaluated to determine the applicability of rheological parameters to different mixes. HRA is a gap-graded, dense asphalt and is the preferred surfacing material in the UK for motorways and trunk roads. It typically contains approximately 7% wt binder, 10% wt filler, 53% fine aggregate and 30% coarse aggregate. The following tests were used to characterise the bind er performance in asphalt mixes: Figure 7 shows the effect of different gap settings for each plate diameter on results for the 15 pen bitumen obtained using DSRs 1 and 2 (test frequency of 0.4Hz). Figure 7 Effect of plate diameter and gap on d and |G*| (15 pen bitumen)

It is clear that the gap setting affects the results. In the extreme case, with the gap set to 10 microns, surprisingly low and constant stiffness values of about 3-4x104 Pa were obtained for the 20mm (DSR 1) and 25mm (DSR 2) parallel plate geometries. Under these conditions of low temperature and high binder stiffness, the displacement measured is believed to be not solely due to the binder but also due to compliance of the measuring system. Increasing the gap (sample height) reduces the problem but the effect is still significant, causing lower than expected |G*| values as * ten ds to zero. Reducing the diameter of the plate to 8mm improves the situation but, even with a combination of reduced plate diameter and increased sample height (2mm), it does not appear possible to measure the limiting elastic stiffness of 109 Pa. (This can be seen by extrapolation of the d vs |G*| plot for this geometry in Figure 7.) Use of an even larger gap may extend the measurement range further. Clearly however, there is a practical limit to the maximum gap which can be used. The foregoing plots of * vs |G*| show how it is possible to estimate a maximum stiffness which can be measured for a given sample geometry. However, it is not easy from such plots to estimate an upper limit of stiffness beyond which results may be inaccurate. Research carried by BP out in the late 1960s found a linear relationship between the logarithm of (1+tan *) and the logarithm of |G*| for penetration grade bitumens, within the range of 0 to 1 for log(1+tan*) (2,3). When plotted in this form the limiting stiffness may be more readily extrapolated and it is also easier to identify the point of departure from linearity where lower than expected values of |G*| are obtained. Figures 8 and 9 show plots of log(1+tan*) versus log|G*| for both bitumens used in this study. Figure 8 Log(1+tan*) vs log|G*| for 200 pen bitumen

From Figure 8 it can be seen that disagreement in the values of |G*| measured using the two plates occurs at about 105 Pa. For the 25mm

measured using the two plates occurs at about 105 Pa. For the 25mm plate (1mm gap), the extrapolated value for the limiting stiffness is about 107 and inaccurate values of stiffness would be obtained in the range 105-107 Pa. However, for the 8mm PP geometry (0.5mm gap), extrapolation of the line gives a value for the limiting stiffness which is close to 109 Pa, although departure from linearity may occur above 2x107 Pa. Figure 9 Log(1+tan*) vs log|G*| for 15 pen bitumen

Figure 8 also shows that, for values of log(1+ tan *) greater than about 1, the linear relationship does not hold, because tan * tends to infinity and consequently there is a lot of scatter in the data. A similar plot for the 15 pen bitumen (Figure 9) using the 8mm PP geometry (0.5mm gap) gives a similar value for the limiting stiffness. Departure from linearity appears to occur at a similar stiffness to that observed in Figure 8, i.e. 2x107 Pa.

Discussion
Work presented in this paper indicates that care is required to obtain accurate rheological measurements on bitumens. Small variations between the set temperature and the actual temperature of the binder can cause large variations in the results obta ined. Accurate temperature control of the whole environment around the bitumen, for example by a fluid bath, is essential. In addition, the choice of sample geometry can have a large influence on the results. This influence is considered to be due to machine compliance, the extent of which depends on both the rheometer and the plate diameter and gap. The problem is measuring small displacements which contain contributions from both the binder and the machine using a particular measuring geometry. As the stiffness of the binder increases, the strain (or displacement) in the binder decreases until a point is reached where torsional movements in the machine itself become significant. The instrument and software assumes the torsional movement is entirely due to the binder and, where this is not the case, this will result in a lower measured stiffness, as shown in this study. The results indicate that, for practical purposes, the 25mm PP and 20mm PP geometries with gaps of 0.5-2 mm are not suitable for the accurate measurement of binder stiffnesses above about 105 Pa without some correction for machine compliance. For the 8mm PP geometry, the upper limit of binder stiffness appears to be approximately 108 Pa for a gap of 2mm. These limiting stiffness values occur over the temperature ranges normally selected for characterising the rheological behaviour of bitumens. In order to measure higher stiffnesses, it is necessary to consider either an alternative geometry, for example tension-compression, or an alternative instrument, such as the Bending Beam Rheometer. The former of these approaches has been reported by researchers using a Rheometrics Mechanical Spectrometer (4) to overcome a similar problem of low limiting stiffness values with parallel plate geometry. To put these data into context, it is interesting to consider some of the SHRP binder specifications. Two parameters have been proposed which require the measurement of |G*|, which are related to deformation and fatigue. The fatigue criterion is defined as the temperature at which the binder has a value of |G*|.sine d equal to 5000 kPa. Since sine * is always less than 1, the value of |G*| will be more than 5000 kPa (or 5x106Pa). The proposed American Association of State Highway and Transportation Officials (AASHTO) Provisional Standard for binder measurements using a DSR (1), specifies the use of two parallel plate geometries (25mm and 8mm diameter), but does not recommend which plate to use for specific ranges of binder stiffness. This study has shown that, w ithout careful selection of sample geometry, inaccurate results could easily be obtained.

obtained.

Conclusions
The major conclusions from this study are as follows :

A difference of just 1*C in temperature can lead to significant variations in the viscosity of a binder measured using a DSR. In order to ensure accurate temperature control it is recommended that the sample is in a suitable temperature-controlled environment. e.g. a water bath. In order to check the accuracy of DSR measurements, a plot of log(1+tan*) versus log|G*| may be made and the limiting stiffness of the binder (|G*|) estimated. The limiting stiffness for unmodified bitumens should be about 109 Pa. With the two DS Rs used with 20mm and 25 mm diameter PP geometry, a practical stiffness limit appears to be about 105Pa for gaps in the range 0.5-2 mm. For the 8mm PP geometry, an upper limit of 108 Pa seems to apply for a gap of 2mm. This may increase for an increased gap, although clearly there are practical working limits to be considered. Stiffness measurements which exceed these values will be affected by machine compliance which is both DSR and geometry (plate diameter and gap) dependent. The lower operational limits were not specifically addressed in this study but, for a given DSR, there was good agreement between the geometries at stiffness values between 103 and 105Pa. The current AASHTO Provisional Standard does not recommend specific plate geometries for different ranges of binder stiffness. This may lead to inaccurate stiffness results. Modified binders were not tested as part of this study but the measurable stiffness range in order to guarantee accurate results remains the same. However, the linear relationship of log(1+tan *) against log(|G*|) will not be valid for modified materials which exhibit highly elastic behaviour over the in-service temperature range. Since empirical relationships between binder properties and the in-service performance of asphalt mixes may not always be valid for modified binders, it is important that future specifications are developed which are valid for these binders. The use of fundamental rheological measurements, as required by the Highw ays Agency and some Highway Authorities in the UK, is of increasing importance and is fully supported by the authors. It is essential that the methods for making such measurements, including accurate temperature control and selection of an appropriate geo metry are carefully defined in order to ensure accurate and reproducible results. In addition, the effects of equipment calibration, sample preparation and nonlinear visco-elastic behaviour should not be overlooked. It is the authors view that all of these factors should be fully addressed, and the precision of the tests established, before such measurements are used for product compliance purposes, particularly in contractual situations.

References
1. American Association of State Highway and Transportation Officials (1995). AASHTO provisional standard TP5-93: standard test method for determining the rheological properties of asphalt binder using a dynamic shear rheometer. AASHTO, Washington DC 2. Dobson G R (1967). An apparatus for measuring the dynamic elastic properties of bitumens. Journal of Scientific Instrumentation, Vol 44, p375, 1967. 3. Dobson G R (1969). The dynamic mechanical properties of bitumen. Proceedings of the Association of Asphalt Paving Technologists, Vol 38, p123-139, 1969. AAPT, Minnesota.

4. Pink H S, Merz R E and D S Bosniak (1980). Asphalt rheology: experimental determination of dynamic moduli at low temperature. Proceedings of the Association of Asphalt Paving Technologists, Vol 49, p64-94, 1980. AAPT, Minnesota.