XVIII IMEKO WORLD CONGRESS

Metrology for a Sustainable Development

September, 17 – 22, 2006, Rio de Janeiro, Brazil

CONSIDERATIONS ON THE ACCURACY OF MEASUREMENTS

OF ELECTRICAL CONDUCTIVITY OF LIQUIDS

Zbigniew Moroń 1
1
Department of Biomedical Engineering and Instrumentation, Wroclaw University of Technology, Wroclaw, Poland,
Zbigniew.Moron@pwr.wroc.pl

Abstract: The purpose of this paper is the presentation of for new designs and new technologies for manufacturing the
new requirements created by contemporary applications of sensors. Therefore a new look at the problem of accuracy of
electrolytic conductivity measurements, the main factors measurements of electrical conductivity of liquids seems to
limiting accuracy of such measurements and also discussion be necessary. Because of the author’s present interests the
of the means enabling minimization of effects caused by considerations presented are oriented towards biomedical
these factors. Recently observed demand for accurate applications.
microscale conductivity measurements, mainly in
biomedical engineering, involves new features required 2. PRINCIPLES OF MEASUREMENT
from the conductivity sensors and co-operating electronics:
miniaturization, robustness, low price (mass produced, Electrical conductivity κ of a conductor characterizes its
disposable devices). Consequently, a new approach in ability to conduct electric current and is defined as the ratio
designing such instrumentation is required. of the density J of the current flowing through the conductor
to the electric field strength E generating this current, i.e. κ
Keywords: electrolytic conductivity, cell constant, electrode = J/E. It is determined indirectly, from the solution
polarization, microscale measurement, miniature sensor resistance or conductance, using a suitable conductivity
sensor – the principle is illustrated in fig. 1. [1].
1. INTRODUCTION
Most of the liquids conducting electrical current are
electrolytic solutions, i.e. ionic conductors. They are also the
main component of all live organisms, hence the great
practical importance of conductometry. The area of its
traditional applications has recently considerably extended
over many new applications: biological, medical, industrial,
objects of daily use, etc. Many new conductometry based
devices, like chemical micro-analyzers, electrical diagnostic Fig. 1. Illustration of the principle of measurement
devices, biosensors and electrical resistance tomography of electrical conductivity
systems have been developed.
The basic parameter of all conductivity sensors is the so
New applications create new requirements, especially for called “cell constant” K – a scale factor dependent on the
much better spatial resolution, smaller sample volumes and geometry of the sensor and the current paths (often affected
wider range of measuring frequencies [1,2,3] (although by the overall geometry and volume of the sample),
basic conductivity measurement is carried out usually at one necessary for calculating the conductivity from the
frequency, investigations are often performed over a wide resistance or conductance of the sensor. Considering a
range of frequency – to minimize the influence of the volume of solution bounded by two parallel cross-sections, 1
electrode polarization or to distinguish specific features of and 2 (fig. 1), its resistance R12 determined as the ratio of the
matter, e.g. in impedance spectroscopy of biological voltage drop V to the current I can be calculated from
tissues), etc. This involves new features required from the equation (1) where κ is the conductivity and L and S are,
sensors: miniaturization, robustness, low price (mass respectively, the length and the cross-section of the currents
produced, disposable devices), and new conditions of their paths:
use – multi-frequency and multi-point measurements in V 1 L K (1)
variable environments, very low level of measuring signals, R = = ⋅ =
12
I κ S κ
etc. Very often absolute values of the conductivity are
required (not relative changes only) for distinguishing Hence the cell constant is
particular objects (type of electrolyte dissolved, type of L (2)
tissue examined, etc.). That indicates an increasing demand K = κ ⋅ R12 =
S
If the cross-sections of the current paths are not uniform
along their lengths (fig. 2), integral expressions are required.
Then:
L
1 (2b)
K=∫ dl
0
S (l )

The conductivity determined in the described way is an

average value over the whole volume of the specimen
measured by the sensor. Local conductivity measurements
require applying sensors of appropriately small size.
Miniature sensors of size of the order of tens to hundreds
micrometers are available now.

Fig. 3. Illustration of two- and four-electrode method of measurement

and distribution of the potential along the conductivity sensor
(electric field close to uniform assumed) [1]

It is assumed that the cell constant has, for a given

sensor, at a given temperature, a certain and invariable
value. In reality it can exhibit variations depending on may
factors: occurrence of current paths spreading out of the
sensor (stray electric field), the value of the measured
Fig. 2. Illustration of the integral definition of the cell constant
of conductivity sensor [1] conductivity of the liquid and the measuring frequency
(effects of electrode polarization at lower frequencies and
The most popular is the two-electrode method of parasitic stray capacitances at higher frequencies) [1,5], etc.
measurement – the same pair of electrodes, 1 and 4 in fig. 3, Consequently, the majority of usable conductivity sensors
is used for exciting the current flow I and for measuring the require experimental determination of their cell constant
voltage drop V1. The ratio V1/I is identified then with the value at the conditions close to the expected application [1].
resistance Re of the solution. In reality the voltage V1 The recommended method of calibration is based on the
contains also the voltage drops Vp across the electrode measurement of electrical resistance of the cell filled up
impedances (polarization impedances) – as shown in fig. 3. with the standard solution [6]. Such calibration has a serious
Therefore the resistance measured in this way is always disadvantage – it is strongly disturbed by temperature
greater than the bulk solution resistance. This is the so called changes, similar to all measurements of electrolytic
“polarization error”. Analyzing the potential distribution conductivity. Free of this disadvantage are absolute
along the current paths (fig. 3) one can observe that the measurements of conductivity, performed by cells of
regions adjacent to the electrodes exhibit much greater calculable constant [5,7]. Until now such sensors are rather
potential gradient than the bulk of the solution. Polarization seldom applied as usable ones and they have not been
error can be eliminated (in theory) or minimized (in realized yet in a miniature form.
practice) applying a differential [4] or four-electrode method
of measurement [1,2]. The second one is more effective and 2. MAIN FACTORS LIMITING ACCURACY
hence much more popular. In the four-electrode method an OF CONDUCTOMETRIC MEASUREMENTS
additional pair of electrodes, 2 and 3 – in fig. 3, is inserted The main sources of errors and uncertainties in
into the bulk solution – to the region of linear potential measurements of electrical conductivity of liquids
gradient. The potential electrodes should not disturb the (electrolytic conductivity) are: inaccurate determination of
original distribution of the potential, hence they must be the cell constant value, strong influence of temperature on
sufficiently small and the voltage drop V2 should be the measured solution conductivity, electrode polarization
measured with negligible (in theory zero) current. The ratio and also instrumental errors, including residual and parasitic
V2/I is then accurately equal to the bulk solution resistance, effects [1,5].
free of polarization effects.
2.1. Influence of temperature
The cell constant defined in the way illustrated in fig. 2
has a general meaning and can characterize all types of Electrical conductivity of all electrolytic solutions is
conductivity sensors, two-electrode and four-electrode ones strongly dependent on temperature (fig. 4) – its value
as well as electrodeless sensors, e.g. inductive ones in which increases 1%–3% per degree celsius, or even more. Usually
the electric field is excited in the volume bounded by the it is necessary to convert results obtained at different
cross-sections 1 and 2 using electromagnetic coupling. temperatures to a standard reference temperature (usually
25oC, this is so called “temperature compensation”) – only
then are the conductivity values comparable. Such a
procedure is not easy because various solutions have
different temperature characteristics. Lack of or inexact
temperature compensation can be a source of very serious
errors [1].
For instance, skin temperature can vary in the order of
10oC from the core body temperature and is strongly
dependent on the depth, and therefore the temperature at the
point of measurement needs to be known. This requires the
application of a miniature temperature sensor integrated
with the conductivity sensor [2].
A significant source of the temperature error can be self-
heating caused by excessive power dissipation in the sensor,
resulting from the measuring current applied, essential
particularly when the sensor has a closed vessel and heat
exchange with surroundings is poor. Fig. 5. Equivalent circuit of a two-electrode conductivity sensor (a)
ω) and Cp(ω
and its transformation to the serial form (b); Rp(ω ω) are the
equivalent serial resistance and capacitance of polarization [1]

To avoid chemical reactions at the electrodes and to

minimize the polarization impedances, conductivity
measurements are usually performed using nearly ideal
polarized electrodes (such as electrodes made from platinum
or gold), excited with alternating current (AC) of
sufficiently high frequency. But even then the electrode
polarization can considerably effect the measured sensor
resistance and hence worsen the accuracy of determining the
electrolytic conductivity. Exemplary values of the
polarization resistance rp per unit of the area of one
electrode, measured by the author [1], are presented in fig. 6
and 7. These results confirm that the polarization resistance
strongly depends on the material, porosity of the electrode
surface, the measuring frequency and the measured
conductivity (it can depend also on the kind of the measured
electrolyte, which is not presented there). Porous electrodes
Fig. 4. Influence of temperature on the conductivity have low polarization impedance but they are unacceptable
of some aqueous solutions (the vertical axis presents the ratio
of conductivities at actual and reference temperature) [1]
in most of biological, medical and food industry applications
whereas suitable for such applications are smooth metal
electrodes that have higher values for this impedance. The
2.2. Influences of electrode polarization results obtained by the author are in good agreement with
The total impedance of the two-electrode conductivity those presented later in [9] – fig. 8 (this figure presents also
sensor can be presented as a sum of the resistance of the the results of measurements of the polarization capacitance
cp per unit of the area of one electrode, which determines
bulk electrolyte solution Re (measurand) and two electrode the sensor reactance).
impedances ZE consisting of double layer capacitances CDL,
Warburg impedances (diffusion impedances) ZW and
Faraday resistances (chemical reactions resistances) RF,
where ZW and RF are frequency dependent – a suitable
equivalent circuit is shown fig. 5a [8,1] (electrode
impedances can be considered as linear ones only in the
range of small electrical signals). Transforming this circuit
to a serial form (fig. 5b) one can show that the measured
sensor resistance RS is a sum of the solution resistance Re
and two frequency dependent polarization resistances Rp.
The major problem is how to arrange the measurement
to extract accurately the measurand Re from the measured
resistance RS. It can be achieved minimizing the
contribution of the polarization resistances or applying the
four-electrode method of measurement. Often both these
approaches are necessary. Fig. 6. Polarization resistance per unit for one electrode surface
made up of different materials, as a function of conductivity [1]
 potential difference depending on the measuring frequency
and on properties (type, concentration) of the measured
solution – and therefore the cell constant of such sensor is a
function of these factors. This effect was modelled by the
author for the four-electrode sensor having stainless steel
potential electrodes of width w = 0.5 cm and 3.6 cm
interelectrode distance, using FEM (Finite Element Method)
[1]. The results of this modelling are presented in fig. 9, in
comparison with the results of experiments obtained for two
identically manufactured copies of the sensor (A and B in
fig. 9). Two methods of FEM modelling were applied: the
relative method and the combined one (energy method plus
relative one). Modelling was performed as a function of the
ratio of the polarization resistance to the solution resistance,
i.e. Rp/Re, for different widths w of the potential electrodes.
Different values of the Rp/Re ratio were obtained by
Fig. 7. Polarization resistance per unit for one electrode surface changing the measuring frequency. The results presented in
made up of different materials, as a function of frequency [1]
fig. 9 demonstrate clearly the previously mentioned
influence of the finite size of the potential electrodes on the
cell constant value. For the real sensor, with the potential
electrodes 0.5 cm wide, the observed variations of the cell
constant can be even 10%. Only the sensor with
dimensionless potential electrodes has its cell constant
value invariant and proportional to the interelectrode
distance. That confirms theoretical expectations that it is
advantageous to minimize the size of the potential
electrodes.

Fig. 8. Polarization impedance for one platinum electrode surface

(0.07 cm2) in 2–77 mmol/l NaCl solutions [9]

Influences of the electrode polarization manifest

themselves also in the four-electrode method of
measurement. At least three harmful effects can be
observed: increase in the impedance of the current
electrodes, increase in the impedance of the potential
electrodes and variations of the cell constant resulting from
the finite size of the potential electrodes. The first effect Fig. 9. Variations of the cell constant of the four-electrode conductivity
means the necessity of applying a higher voltage for sensor versus the ratio Rp/Re and the width of the potential electrodes
supplying the current electrodes, the second one – necessity – comparison of the modeled and measured values [1]
of applying the differential input voltmeter of high input
impedance and large CMMR (Common Mode Rejection 3. PROBLEMS ASSOCIATED
Ratio) value. These problems are commonly well known [2] WITH MINIATURIZATION OF SENSORS
unlike the effect of the finite size of the potential electrodes
– which therefore needs more detailed explanation here [1]. 3.1. Polarization error in miniature sensors
Two identical potential electrodes, made up of the same Miniaturization of the conductivity sensors requires a
material, should generate identical distortions of the cautious approach to their design. Because the measured
potential distribution and hence compensate each other. conductivity is equal to
However, in practice, it is difficult (if at all possible) to
ensure the identity of the electrodes. Potential electrodes of K (3)
κ=
large size and large electrode impedances can produce the R
the error of its determination, resulting from polarization,
can be calculated as
K K − 2R p (4)
∆κ p = − =K
R e + 2R p Re R e (R e + 2R p )

The relative value of this error is

∆κ p 2Rp 2Vp (5)
δκ p = =− =−
κ Re Ve

One electrode of surface area S and unit polarization Fig. 10. Top view and side view of planar two-electrode (a)
and four-electrode (b) conductivity sensor [2]
resistance rp has its polarization resistance
rp At present there are known conductivity sensors, the
Rp = (6) electrodes of which are provided with a thin insulating film,
S
e.g. from Ta2O5. The specific nature of the Ta2O5–solution
therefore the measured sensor resistance is interface provides the electrode with a stable and relatively
rp rp
low impedance [10]. Insulation coating also protects
K L (7)
Rs = Re + 2Rp = +2 = +2 electrodes against dirt, is easy for washing and drying, etc. –
κ S S⋅κ S
however it requires application of higher measuring
and then rp (8) frequencies
δκ = − 2 ⋅κ
L
3.2. Instrumental errors
An important conclusion arises from formula (8) –
Most of instrumental errors occurring in the conductivity
miniature two-electrode conductivity sensors of the
measurements performed using classic “large” conductivity
“classic” design (i.e. with the electric field distribution close
sensors (dimensions in centimetres) will occur when using
to uniform) cannot be used for measurement of highly
miniature ones (dimensions of order 10-100 µm), usually
conducting solutions. Assuming electrodes of a given
more intensively.
polarization resistance rp per unit of the area, too small a
length L of the current paths makes impossible the reduction Very small electrode surfaces significantly increase
of the error caused by electrode polarization to a reasonable electrode impedances. At lower measuring frequencies the
value – this is a problem of impossibility of designing such current applied to the supplying (current) electrodes
sensors of sufficiently high value of the cell constant and, produces a much higher voltage drop at the sensor, hence
simultaneously, of sufficiently large electrodes. much higher power is dissipated in it and a noticeable effect
Example: considering 0.01 mol/l NaCl aqueous solution at of self-heating can occur. Simultaneously, the maximum
25oC, measured at frequency 4 kHz using two polished acceptable power dissipation is limited then by the very
platinum electrodes of unite polarization resistance 4 Ωcm2, small dimensions of the sensor. This problem applies to the
from formula (8) we obtain δκ = - 0.8% for L = 1 cm and two- and four-electrode sensors. Hence the measuring
δκ = - 80% for L = 0.01 cm (100µm). For better conducting current has to be adequately limited. Currents of the order of
solutions the situation will be worse. Therefore, when using 10 µA or even lower are applied (higher currents are also
the miniature sensor under these conditions, application of undesirable because they could produce excessive current
the four-electrode method is required. Of course, increasing density at the supplying electrodes). Consequently, the
the frequency will reduce the polarization error. measured voltage drops are very low and very good quality,
A possible and feasible solution is applying sensors of high-gain and low-noise amplifiers are required.
such a design that their cell constant value depends on the
Very high impedances of the current and potential
non-uniform distribution of the electric field rather than on
electrodes involve also a necessity of increasing
their geometric dimensions. Examples of such designs, in
considerably the input impedance and CMMR of the input
the form of two- and four-electrode planar interdigitated
amplifier used for measurement of the voltage occurring
sensors, feasible to manufacture as thin or thick film
between the potential electrodes. Simultaneously, at higher
devices, are discussed in detail in [2] and are shown
frequencies a stronger influence of the parasitic stray
schematically in fig. 10. The cell constant of such sensors
capacitances can be observed (smaller distances between the
can be estimated theoretically or modelled with an accuracy
leads, thinner insulation, etc. can considerably increase these
of the order of 20-30 % [10,2]. The planar conductivity
capacitances). That gets worse given the high frequency
sensors mentioned operate with the “open” (stray) electric
characteristics of such measurements.
field and therefore they have to be used in a sufficiently
large volume of the medium if the absolute value of the
4. SUMMARY
conductivity has to be determined. When this volume is
limited, the application of suitable electrical shields A demand for accurate microscale conductivity sensors,
concentrating the lines of the field (current paths) is robust and low-cost (disposable) can be observed in
advisable [1]. contemporary applications of electrolytic conductivity
measurements. It requires a new approach in designing
conductometric instrumentation, mainly with regard to the
miniature conductivity sensors but also concerning the co- [6] IEC, “Expression of performance of electrochemical
operating electronics. Such sensors, in the form of two- and analyzers, Part 3: Electrolytic conductivity”, IEC
four-electrode planar interdigitated ones, feasible to Publication 607046, July 2002
manufacture as thin or thick film devices and integrated with
[7] Z. Moron, “Investigations of van der Pauw method
temperature sensors have been developed.
applied for measuring of electrolyte solutions”,
Application of the four-electrode method of Measurement, Vol. 33, Issue 3, pp. 281-290, April 2003
measurement is a recognized remedy for avoiding the
[8] J. Braunstein, G. D. Robbins, “Electrolytic conductance
influence of the electrode polarization in the microscale
measurements and capacitive balance”, J. Chem.
measurements but it must be applied carefully to avoid
Education 1971, Vol. 48, No 1, pp. 52-59
(minimize) residual and parasitic effects, especially when
the measurements are carried out across a wide range of [9] P.Mirtaheri, S. Grimnes, O. G. Martinsen: „Electrode
frequencies. polarization impedance in weak NaCl aqueous
solutions”, IEEE Trans. Biomed. Eng., Vol. 52, No. 12,
The problem of the absolute measurement of local
Dec. 2005, pp. 2093-2099
conductivities is still unsolved – miniature, well calculable
conductivity sensors are still unavailable. [10] W. Olthuis, W. Streekstra, P. Bergveld, Theoretical and
experimental determination of cell constants of planar-
Miniaturization of the electrodeless conductivity sensors
interdigitated electrolyte conductivity sensors,
is at present not feasible to the same extent as for the
Sensors&Actuators B 24-25 (1995), pp. 252-256
electrode ones, however it would be very advantageous for
many reasons, e.g. nonoccurrence, in principle, of
polarization influences, the possibility of non-invasive local
measurements at different depths, etc.
The subject as stated in the title of this paper is too wide
for presentation in full in the space available here and
therefore this paper should be regarded as an introduction to
the problem – further details will be provided in subsequent
publications.

ACKNOWLEDGMENTS
The author would like to thank Professor Ewart R.
Carson from City University for his help in the correction of
the manuscript.

REFERENCES
[1] Z. Moron, “Measurements of electrical conductivity of
liquids at low frequencies” (in Polish), Oficyna
Wydawnicza Politechniki Wrocławskiej, Wrocław 2003
[2] D. He, M.A. Shannon, N.R. Miller, “Micromachined
silicon electrolytic conductivity probes with integrated
temperature sensor”, IEEE Sensors Journal 2005, Vol.
5, No. 6, pp. 1185-1196
[3] A. Ivorra, R. Gomez, N. Noguera, R. Villa, A. Sola, L.
Palacios, G. Hotter, J. Aguilo, „Minimally invasive
silicon probe for electrical impedance measurements in
small animals”, Biosensors and Bioelectronics 391
(2003) pp. 391-399
[4] Z. Moron, “Differential three-electrode measurement of
electrolytic conductivity”, J. Phys. E: Sci. Instr. 1981,
Vol. 14, pp. 686-688
[5] S. L. Schiefelbein, N. A. Fried, K. G. Rhoads, D. R.
Sadowayc, “A high-accuracy, calibration-free technique
for measuring the electrical conductivity of liquids”,
Rev. Sci. Instr., Vol. 69, No 9, pp. 3308-3313, Sept.
1998