ANALELE TIINIFICE ALE UNIVERSITII AL. I.

CUZA IAI
Tomul II, s. Biofizic, Fizic medical i Fizica mediului 2006

MEMBRANE COVERED POLAROGRAPHIC OXYGEN SENSOR

MANUFACTURING THEORETICAL CONSIDERATIONS
Alina Manole1, I. Neacsu2, M-O. Apostu3, V. Melnig1

KEYWORDS: Clark electrode, polarography, diffusion oxygen models

The operating principle of the Clark type polarographic oxygen electrode is based on the
use of an electrolyte solution contained within the electrode assembly to transport oxygen
from an oxygen-permeable membrane to the metal cathode. In this paper are reconsidered
theoretical assumption of one layer and two layer diffusion models and the evaluation of
optimal calibration and error measurements.

1.

INTRODUCTION

Electrochemical oxygen analysers are based on electrochemical reduction of O2

at a negatively polarized electrode. This principle lies in the bases of the Clark-type
oxygen-sensitive electrode which was proposed in 1956 by Leland Clark [1, 2]. Clark
studied the electrochemistry of oxygen gas reduction at platinum (Pt) metal electrodes,
pioneering the use of Pt electrodes as an oxygen sensor. (Pt electrodes used to detect
oxygen electrochemically are often referred as Clark electrodes.)
An electrode is an electrical conductor used to make contact with a metallic
part of a circuit (e.g. a semiconductor, an electrolyte or ionised gases). The word was
coined by the scientist Michael Faraday from the Greek words elektron (meaning
amber, from which the word electricity is derived) and hodos, a way. An electrode in
an electrochemical cell is referred to as either an anode or a cathode, words that were
also coined by Faraday. The anode is defined as the electrode at which electrons leave
the cell and oxidation occurs, and the cathode as the electrode at which electrons enter
the cell and reduction occurs. Each electrode may become either the anode or the
cathode depending on the voltage applied to the cell. A bipolar electrode is an
electrode that functions as the anode of one cell and the cathode of another cell.

1
2
3

Faculty of Physics, Al. I. Cuza University, Bd. Carol I, Nr. 11A, 700506, Iasi
Faculty of Biology, Al. I. Cuza University, Bd. Carol I, Nr. 20A, 700506, Iasi
Faculty of Chemistry, Al. I. Cuza University, Bd. Carol I, Nr. 11A, 700506, Iasi

ALINA MANOLE, I. NEACSU, M-O. APOSTU, V. MELNIG

2.

26

THEORETICAL CONSIDERATIONS

If an electrode of noble metal such as platinum or gold is made 0.6 to 0.8 V

negative with respect to a suitable reference electrode such as Ag/AgCl or an calomel
electrode in a neutral KCl solution (Fig. 1), the oxygen dissolved in the liquid is
reduced at the surface of the noble metal following the reactions [3]:
Cathodic reaction: O2 + 2H2O + 2e- H2O2 + 2OH
H2O2 + 2e- 2OH
Anodic reaction: Ag + Cl- AgCl + eOverall reaction: 4 Ag + O2 + 2H2O + 4 Cl- 4 AgCl + 4 OH-

Fig. 1 Setup for polarography.

The oxygen reduction at the noble metal surface can follow two pathways:
4 electron pathway - the oxygen in the bulk diffuses to the surface of the
cathode and is converted to OH- via H2O2 (path a in Fig. 2)
2 electron pathway - the intermediate H2O2 diffuses directly out of the
cathode surface into the bulk liquid (path b in Fig. 2).
The oxygen reduction path may change depending on surface condition of the noble
metal. This is probably the cause for time-dependent current drift of polarographic
probes. Since the hydroxyl ions are constantly being substituted for chloride ions as
the reaction starts, KCl or NaCl has to be used as the electrolyte. When the electrolyte
is depleted of Cl-, it has to be replenished.

Fig. 2 Alternative pathways of oxygen reduction at cathode surface.

27

THEORETICAL AND PRACTICAL

If we study the current-voltage diagram of the electrode (called a

polarogram) we can observe that the current output of the electrode increases, in the
beginning, with the increase of the negative voltage applied to the noble metal
electrode (the cathode) and because the reaction of oxygen at the cathode is very fast
and the rate of reaction is limited by the diffusion of oxygen to the cathode surface,
the current becomes saturated (Fig. 3a). When the negative voltage is further
increased, the current output of the electrode increases rapidly due to other reactions,
mainly, the reduction of water to hydrogen. If a fixed voltage in the plateau region (for
example, - 0.6V) is applied to the cathode, the current output of the electrode can be
linearly calibrated to the dissolved oxygen (Fig. 3b). It has to be noted that the current
is proportional not to the actual concentration but to the activity or equivalent partial
pressure of dissolved oxygen, which is often referred to as oxygen tension. Originally
developed for measuring oxygen gas, it is only a matter of polarity, whether the
electrode senses hydrogen or oxygen gas. For hydrogen measurements +600 mV (vs.
Ag/AgCl) are supplied. The electrode is calibrated using two gas mixtures of known
oxygen (or hydrogen) concentration.

(a)
(b)
Fig. 3 (a) Current-to-voltage diagram at different oxygen tensions;
(b) Calibration obtained at a fixed polarization voltage of -0.6V.

Sensor Output and Design Variables. One Layer Model

A first-order diffusion model of the Clark electrode is illustrated in Fig. 4 [2].
The membraneelectrolyteelectrode system is considered to act as a one-dimensional
diffusion system with the partial pressure at the membrane surface equal to the
equilibrium partial pressure p0 and that at the cathode equal to zero.

28

ALINA MANOLE, I. NEACSU, M-O. APOSTU, V. MELNIG

The following assumptions are made [4]:

1. The cathode is well polished and the membrane is tightly fit over the cathode
surface such that the thickness of the electrolyte layer between the membrane and the
cathode is negligible.
2. The liquid around the sensor is well agitated that the partial pressure of oxygen at
the membrane surface is the same as that of the bulk liquid.
3. Oxygen diffusion occurs only in one direction, perpendicular to the cathode surface.
Ficks 2nd law describes the unsteady-state diffusion in the membrane:

p
2p
= Dm
t
x 2

(1)

Dm - the oxygen diffusivity in the membrane; x - the distance from the cathode
surface.
The initial and boundary conditions are:
p = 0 at t = 0
(2)
p = 0 at x = 0
(3)

p = p o at x = d m

(4)

dm - the membrane thickness; po - the partial pressure of oxygen in the bulk liquid
The solution of eq. (1) with the boundary conditions is:

p
x
=
+
p0 dm

n ( 1)

sin

n =1

nx
2
exp n 2 2 D m t d m
dm

(5)

The current output of the electrode is proportional to the oxygen flux at the
cathode surface:

I = NFAD m (C/x) x =0

(6)

= NFAPm (p/x) x =0

N - number of electrons per mole of oxygen reduced; F - Faraday constant (= 96,500

C/mol); A - surface area of the cathode; Pm - oxygen permeability of the membrane
The permeability Pm is related to diffusivity by:
Pm = D m S m
(7)
Sm - the oxygen solubility of the membrane
From eqs. (5) and (6), the current output It of the electrode is:

I t = NFA

Pm
n
2
p 0 1 + 2 ( 1) exp n 2 2 D m t d m
dm
n =1

(8)

The pressure profile within the membrane and the current output under
steady-state conditions can be obtained from eqs. (5) and (8), respectively:

p
d
=
p0 dm

(9)

and

I s = NFA

pm
p0
dm

(10)

29

THEORETICAL AND PRACTICAL

At steady state, the pressure profile is linear and the current output is
proportional to the oxygen partial pressure in the bulk liquid. Eq. (10) forms the basis
for oxygen diffusion measurement by the sensor. Eq. (8) shows that the rapidness of
the sensor response is characterized by the response time parameter:
2

dm
Dm

(11)

For a rapid response (small ) the membrane must be thin and/or diffusion
coefficient, Dm, high. Adjusting dm (rather than Dm) is more effective in adjusting
(because depends on the square of dm). Eqs. (10) and (11) indicate that the design
variables for a diffusion oxygen sensor (DO sensor) are Pm, dm, Dm and A.

Fig. 4 One layer electrode model.

Fig. 5 Two layer electrodemodel

Two Layer Model

In reality, the assumption 2 made earlier is not entirely satisfactory, because
when a DO sensor is used to measure DO concentration in two different liquids at the
same partial pressure of oxygen, the readings are not the same. This indicates that the
sensor output depends, to a certain extent, on the properties of the liquid. The effect of
liquid layer on sensor current output can be estimated using the two layer model witch
considers the membrane and the liquid film as shown in Fig. 5.
At steady state, the oxygen flux J through each layer in Fig. 5 should be the
same:

J = Kp 0

= k L (p 0 p m )

(12)

= k mpm
K - the overall mass transfer coefficient; kL- mass transfer coefficient for liquid film;
km - mass transfer coefficient for membrane
The inverse of the mass transfer coefficient can be termed as the mass
transfer resistance:

ALINA MANOLE, I. NEACSU, M-O. APOSTU, V. MELNIG

1
1
1
=
+
K kL km

30

(13)

Equation (13) says that the overall mass transfer resistance, 1/K, is the sum of the
liquid phase mass transfer resistance, 1/kL, and the membrane phase mass transfer
resistance, 1/km. The derivation is based on Ohms law analogy; J is considered the
current and p the voltage. The individual resistances can be replaced by:

1 dL dm
=
+
K PL Pm

(14)

dL- liquid film thickness; PL - the oxygen permeability of the liquid film
When the individual mass transfer resistances are considered the steady state
sensor output becomes:

I s = NFA

pm
d

p0

(15)

where d is defined by:

d = d m + Pm

dL
PL

(16)

Alternatively, Is can be written as:

I s = NFA

dm

p0
Pm + d L PL

(17)

The time constant of eq. (11) can be modified to:

d2
Dm

(18)

where d is defined by:

d = dm +

Dm
dL
DL

(19)

When placed in a stagnant liquid, the DO sensor produces a diffusion

gradient extending outside the membrane and into the liquid. When the liquid is
stirred, the diffusion gradient can no longer be extended beyond the liquid film around
the membrane. Since the diffusion gradient becomes steeper with decreasing liquid
film thickness, the current output of the sensor increases with increase in liquid
velocity as shown in Fig. 6. Also, the response time of the sensor increases as the
liquid velocity decreases. This so-called "flow sensitivity" is greater for a sensor with
a larger cathode because the size of the stagnant diffusion field is proportionally with
the size of the cathode. For proper operation of the sensor, the liquid has to be stirred
beyond a certain level to maintain membrane control of oxygen diffusion. The critical
velocity, Vc, of the liquid is the velocity where the probe output reaches 95% (95-99%
depending on the definition) of the steady state value. For a given liquid, Vc is smaller
for smaller cathodes. For example, with a 25 m Teflon membrane, a cathode of 5 mm
diameter requires Vc of 70 cm/s in water, whereas only 5 cm/s is required for 25 m

31

THEORETICAL AND PRACTICAL

cathode. When the cathode diameter is less than 1 m, the sensor becomes insensitive
to liquid flow even without the membrane because, the diffusion field of the cathode is
so small that it is always contained inside the minimum liquid boundary layer around
the cathode.
One obvious effect of the cathode size is the area effect. Eq. (17) shows that
the current output is directly proportional to the sensor area A. When the current
produced by the sensor is too small (when the cathode area is small), the sensor signal
tends to be more susceptible to noise. The current output should be greater than 10-6 to
10-7 range for ease of signal amplification.
From eq. (17), the condition for a membrane-controlled diffusion becomes:
d m /Pm >> d L /PL
(20)
To achieve this condition, a relatively thick membrane with low oxygen
permeability has to be used (this contradicts the requirement for a fast sensor
response). In this case the oxygen sensor output depends only on membrane properties
as given by eq. (10) and the sensor calibrated in one liquid can be used in other liquids
without recalibration. In reality, however, there is always a liquid film (however thin it
may be) and this causes variations in calibration in different liquids.

Fig. 6 Effect of cathode diameter on flow dependency of DO sensor.

Running experiments it has been observed that there is 1 to 5% increase in
sensor output current per Celsius degree increase in temperature (Fig. 7a). From eq.
(17) we can see that the temperature effect comes from Pm and PL because they are
functions of temperature. In the case of membrane diffusion control, the temperature
dependency of Is should come entirely from Pm and when the liquid film resistance is
not negligible, both Pm and PL contribute to the temperature dependency. Generally,
Pm is expressed as:
Pm = Pm * exp(-E/RT)
(21)
E - the activation energy for the permeation (8.8 kcal/g-mol for polyethylene and 7.89.6 kcal/g-mol for polypropylene membranes) [5].

ALINA MANOLE, I. NEACSU, M-O. APOSTU, V. MELNIG

32

(a)
(b)
Fig. 7 Temperature dependency of DO sensor.
From Fig. 7b we can observe that as the temperature increases, the diffusivity
increases and this makes the sensor respond faster.
Calibration and kLa Measurement
For liquid phase calibration an air saturated and also a nitrogen saturated
water is prepared by passing air bubbles into a small volume (100 mL) of water, then
the voltage output for both water solutions is measured. To obtain proper calibration
the liquids have to be stirred at high speed. This is so-called a two point calibration.
To perform gas phase calibration the sensor is exposed to air and also to nitrogen, as
the gas phase, and then the two calibrations are compared [6].
For measuring the response time a step change in oxygen partial pressure in
the measurement medium has to be done and then the sensor response is determined.
The sensor can be approximated as a first order system:
c c p = p (dc p / dt )
(22)
c - the oxygen concentration in the measurement sample; cp - the oxygen concentration
measured by the sensor; p - the sensor time constant
When a step change is made in c (by transferring the sensor from air into
nitrogen saturated, stirred water), the sensor output decreases roughly exponentially
(not exactly exponentially because the sensor may not be a true first order system).
The time constant p is the time when the sensor response reaches 63.7% of the
ultimate response (Fig. 8a). The solution to eq. (22) with the following boundary
condition is an exponential function:
c = 1 at
t=0
(23)

c c p = 1 exp t / p

(24)

A normalized concentration is used: c of 1 means 100% air saturation and 0 means

nitrogen saturation. Eq. (24) indicates that when t = p, c/cp will be 0.64. The time
constant p can also be determined conveniently by using an integral method - the area
above the response curve is equal to p (see Fig. 8b). This method is especially useful
when there is a lot of noise in the measured signal.

33

THEORETICAL AND PRACTICAL

Fig. 8 (a) Sensor response time measurement; (b) integral method for measuring the
sensor time constant.
Measurement of kLa
kLa, the liquid phase overall volumetric mass transfer coefficient, represents
the oxygen absorption capability of a bioreactor [4]. A DO sensor is used frequently to
measure kLa. The reactor is first sparged with nitrogen and at time zero, the nitrogen is
switched to air. The oxygen mass balance in the reactor yields:

dc dt = k L a c * c

(25)
c - the oxygen concentration in the reactor; c* - the oxygen concentration at the gasliquid interface. This equation can be rearranged to:

c * c = k (dc dt )

(26)

where

k = 1 k La
(27)
Eqs. (22) and (26) can be solved simultaneously to obtain an expression for kLa and k
can be obtained graphically as shown in Fig. 9 when p is known.

Fig. 9 Measurement of kLa by integral method.

It is very important to know that the magnitude of p depends on the liquid
velocity in the vicinity of the sensor. Therefore, if a p measured at one agitation rate is
used for measuring kLa for different agitation rates, the results will be in error. A safe

ALINA MANOLE, I. NEACSU, M-O. APOSTU, V. MELNIG

34

way is to used the same agitation rate for both p and k measurements. However, if k
is much greater than p, such a precaution is not necessary.
3.CONCLUSIONS
As the models reveals, there is the possibility of constructing an oxygen
sensor, based on oxygen diffusion limited process (the Clark electrode), that consists
of an anode and cathode in contact with an electrolyte solution (usually potassium
chloride, 1 M). It is covered at the tip by a semi-permeable polymer membrane, which
is permeable to gases but not contaminants and reducible ions of the sample. The
cathode is in a glass envelope in the body of the electrode. The anode has a larger
surface that provides stability and guards against drift due to concentration of the pO2
electrolyte. This silver/silver chloride (Ag/AgCl) anode provides electrons for the
cathode reaction. When most of the mass transfer resistance is confined in the
membrane, the electrode system can measure oxygen tension in various liquids. This
is the basic operating principle of the membrane covered polarographic DO probe
(DO sensor) (Fig. 10) [7].

Fig. 10 A Clark-type oxygen-sensitive electrode.

REFERENCES
1. T.C. Tan, and C.C. Liu, 1991, Principles and fabrication materials of electrochemical sensors. In:
Chemical Sensor Technology. Kodansha Ltd., Vol. 3.
2. L.C. Clark, 1956. Monitor and control of blood and tissue oxygen tension. Trans. Am.Soc. Artif. Internal
Org. 2, 41 46.
3. Magda Aflori, I. Neacu, V. Melnig, 2006. Analysis methods of oxygen consumption in biological systems,
Scientific Annals of University of Agricultural Sciences and Veterinary MedicineIon Ionescu de la Brad IA I,
Tom XLIX.
4. Young H. Lee, 1999. Biosensors, Engineering Biotechnology Gateway Project, Drexel University,
5. Park, J.H., N.O. Kaplan, E.P. Kennedy, 1966.Current Aspects of Biochemical Energetics, Academic Press, New
York, 299.
6. David R. Caprette, 1996. Experimental Biosciences, Introductory Laboratory Bios 211, Rice University.
7. Trinity College Dublin, 2005-2006. Biochemistry Laboratory: Manual for Senior Freshman Science.