IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL.

70, 2021 1008222

Review on Electrolytic Conductivity Sensors

Carsten Thirstrup and Lisa Deleebeeck

Abstract— Electrolytic conductivity (EC) sensors are widely (i.e., its permittivity [10], and viscosity [7], [8]). Additionally,
used in diverse industries, including: pharmaceuticals, food, conductivity is strongly dependent on temperature [11], and
power, manufacturing, and environmental monitoring. This arti- to a lesser extent, pressure [9], [10].
cle presents an overview and analysis of EC sensors used in these
industries from the lowest to the highest range of conductivities Conductivity can be calculated from resistance (direct cur-
and in applications, where there are strong requirements for rent, dc) or impedance (alternating current, ac) of electrolyte
accurate and traceable conductivity measurements of water solutions. Diverse phenomena are distinguished based on
and electrolyte solutions. Included are reported sensor designs the pulse duration or frequency of an applied electric field.
based on two- and four-electrodes with parallel plates, coaxial Conductivity (κ) is seen at lower frequencies, while dielectric
cylinders, van der Pauw configurations, planar and miniature
configurations, and contactless inductive or capacitive coupling. permittivity (ε ) and loss (ε ) are distinguished at high fre-
A comprehensive list of exact analytical formulas for com- quencies (>MHz) [4], [5], [12], [13]. Dielectric spectroscopy
putation of cell constants for the configurations is provided. focuses on dipole relaxations of the solvent, and its ability to
The advantages and disadvantages of using frequency domain absorb electromagnetic radiation [6].
or time domain measurement schemes in these sensors are EC has units of S · m−1 (−1 · m−1 ), with the units
discussed, and the reader is guided to select the suitable method
of sensor calibration among the methods available, computing μS · cm−1 or mS · cm−1 commonly used in industrial applica-
uncertainty budgets, ensuring traceability to the International tions [14], [15] (see Table I). Conductivity sensors are low-cost
System of Units (SI) and compliance with standards, such items, their measurement is reliable, and usage is mandated by
as Pharmacopeias. Finally, trends in the development of EC international regulations (e.g., drinking water standards [16]).
sensors and the related challenges in relation to calibration Related parameters can be calculated from conductivity, with
and compliance with standards are presented. This article also
introduces the reader to the EC cell of a sensor, which can the addition of solution species information [10], [17]. These
be modeled by solving the Poisson–Nernst–Planck equations or include: conductance ratios (for determination of practical
using appropriate equivalent circuit diagrams. It is stepwise salinity (Sp) [3]), the historical parameter equivalent conduc-
described how the models can be extended to include effects tivity (S · cm−1 · L−1 ), molar (S · mol · cm−1 · L−1 ) and molal
such as electrode polarization and steric effects of ions, and it is (S · mol · cm−1 · kg−1 ) conductivity () of a salt (e.g., KCl,
illustrated by means of an ideal EC cell, where guards eliminate
fringe fields, that the simplest equivalent circuit with a single KCl ), limiting molar (or molal) conductivity at infinite dilu-
resistor and a single capacitor models electrolytic sensor data tion (0 ) [7], ionic molar (or molal) conductivity (λ) of a
very well in the lower range of electrolytic conductivities. single ion (e.g., Cl− , λCl ), and limiting ionic molar conductiv-
Index Terms— Chemical sensors, conductivity measure- ity (λ◦ ). Tabulated λ◦ values are calculated based on values of
ment, environmental monitoring, measurement uncertainty, λ0Cl [9], [17], while the conductivity of ultrapure water (UPW)
standardization. at 25 ◦ C (0.055 0 μS · cm−1 ± 0.000 1 μS · cm−1 [11]) is
calculated from λ0H and λ0OH derived from the conductivities
I. I NTRODUCTION of HCl and NaOH with concentration [11], [18]. Additionally,
ionic transport (or transference, t) number [9], ion mobility
E LECTROLYTIC conductivity (EC) sensors measure a
property of electrolyte solutions, which has been termed
electrical conductivity [1] or ionic conductivity [2], depending
(μ) [8], [10], and the degree of dissociation of a salt [8], [17],
[19], [20] can be calculated from EC.
Conductivity data, and these calculated quantities, have been
on the application. It is a nonspecific measure of a solution’s
important inputs to the development of various theoretical
ability to conduct charge through the movement of dissolved
models and databases [7], [17], including: theory of conduc-
charge carriers [3]. All charged species contribute to the
tance in water and other solvents [5], [21], theory of solution
overall EC, dependent on their charge (z), mobility (μ), and
at infinite dilution [7], determining the solubility of sparingly
interactions with other ions and the solvent [4]. EC depends
soluble salts [19], and means of ion interaction (e.g., hydra-
on: 1) ion species present in solution [5], [6]; 2) their con-
tion [9]) [4]. EC sensors enable detection of small traces of
centrations [1], [7]; and 3) their interactions (e.g., extent of
impurities or contamination from undesirable compounds as
dissociation [8], ion-pair formation [4]), and the solvent [9]
they are extremely sensitive, simple, and robust, which means
Manuscript received December 3, 2020; revised April 19, 2021; accepted that practical applications are numerous and diverse. Table I
May 7, 2021. Date of publication May 28, 2021; date of current version is a nonexhaustive list of such applications.
June 22, 2021. This work was supported by the Danish Agency for Institutions
and Educational Grants, part of the Ministry of Higher Education and Science. EC measurements are most often an indication value for
The Associate Editor coordinating the review process was Güvenir Kaan Esen. the presence of contaminants (Table I), such as the presence
(Corresponding author: Carsten Thirstrup.) of ionic impurities in UPW applications. These particular
The authors are with DFM A/S, DK-2970 Hørsholm, Denmark (e-mail:
[email protected]). measurements are extremely sensitive to concentration when
Digital Object Identifier 10.1109/TIM.2021.3083562 the ionic species is known [14]. However, the value of using
1557-9662 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

TABLE I
P RACTICAL A PPLICATIONS E MPLOYING EC S ENSORS

conductivity to infer concentration is limited by the availability This article serves as an introduction and overview to
of experimental data, a priori knowledge of the contents of the EC sensors and covers the topics of models, development,
solution, and the quality of generated model correlations [17]. calibration, traceability, and usage of the sensors. Section II
The uncertainties assigned to these inferred quantities (e.g., is a brief introduction to the basic operating principles of an
ionic species concentration) are beyond the scope of this arti- EC cell and an overview of mathematical models describing
cle, which limits itself to a discussion of the EC measurement its behavior. Section III describes the various measurements
and its uncertainties. schemes reported in frequency and time domains, and gives

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

With the application of an electrical field, ions will move

through diffusion and migration as ions are attracted to and
accumulated in the vicinity of the electrode/solution interfaces,
resulting in the formation of an ionic double layer. This is
schematically illustrated in Fig. 1 for two plane-parallel plates
separated by a distance (2L). When a dc voltage is applied,
the current will eventually be blocked and the voltage will drop
within a Debye length [50] from the electrodes (see Fig. 1(b)).
The voltage applied is sufficiently low (<0.8 V) to not activate
any chemical reactions at the electrodes (e.g., electrolysis).
Studies of this electrical double layer (EDL) and charge
build-up at the electrodes (electrode polarization) have been
reported [12], [47], [51]–[55]. These phenomena are important
for many biological and electrochemical systems. In Fig. 1(a),
negatively and positively charged hydrated ions are illustrated
schematically. A hydrated ion consists of a central ion sur-
rounded by a spherical shell of water molecules, referred to
as an “ion” in the following. At each electrode, there is a
compact layer of adsorbed ions of opposite charge compared
to the electrode polarity with an effective thickness of the
order of half a diameter of a (hydrated) ion (Stern layer) [56],
and there is an equal number of (hydrated) ions of opposite
charge in the solution. The charges of the adsorbed ions are
Fig. 1. (a) Schematics of EC cell with an applied voltage (V ) and a screened by ions of the opposite charge. Adsorbed ions will
length (2L) between two plane parallel electrodes. (b) Potential (ϕ) plotted diffuse into the liquid to a distance of the order of a Debye
as function of position (x) along the cell.
length, forming the EDL.
The ion-concentration and potential variation in an EC cell
an overview of equivalent circuit diagrams used in modeling can be modeled by solving the Poisson–Nernst–Planck (PNP)
experimental data. At low and at high conductivities, different equations. In 1-D, these equations can describe the overall
calibration methods may be employed ensuring the traceability behavior of the cell [50], [57], [58]
of EC measurements. This is the topic of Section IV, where
∂ci (x, t) ∂
examples of uncertainty budgets are presented. In Section V, =− ji (x, t) (2a)
the sensor configurations are reviewed including reference to ∂t  ∂x 
available sensors, and the current trends in their development. ∂ci e ∂ϕ
ji = −D i + z i ci (2b)
This article is finally summarized in Section VI. ∂x kB T ∂x
   N
∂ ∂ϕ
II. EC AND D OUBLE L AYER T HEORY ε = −e z k ck . (3)
∂x ∂x k=1
At low ion concentrations, the EC (κ) can be expressed
as [49]
In (2) and (3), ji is the mass flux, and ci is the normalized

N
ion concentration of ion (i ), which is normalized to the bulk
κ= ci0 e|z i |μi (1) ion number concentration ci0 [m−3 ]. The index (i = 1 to N)
i=1
includes the number of ion species N, t [s] is time, x [m] is
where ci0 is the number concentration [m−3 ], z i is the valence, the position in the direction perpendicular to the electrodes,
and μi [m2 · V−1 · s−1 ] is the mobility of ion (i ), and the Di [m2 · s−1 ] is the diffusion coefficient for ion (i ), which
index (i ) is summed over the ionic species (N) in solution. is related to the ion mobility through the Einstein relation
In (1), ion–ion interactions are neglected, which is valid at as Di = (kB T /e)μi , z i is the charge of ion (i ), ϕ [V] is
low conductivities for strong electrolytes (i.e., salt which the electrostatic potential, ε [F · m−1 ] is the permittivity of
dissociates completely). the solution, kB [J · K−1 ] is the Boltzmann’s constant, e [C]
Here we distinguish between the terms “conductivity sen- is the elementary charge, and T [K] is the temperature. The
sor” (which includes both contact and noncontact types), Nernst–Planck equations (see (2)) comprise time derivatives
and “conductivity cell.” A simple conductivity cell is that and describe the dynamics of the electrolytes, while the Pois-
shown in Fig. 1: a volume of electrolytic solution bounded son equation (see (3)) describes the distribution of electrolytes
by two electrodes. A voltage or current is applied to the cell, in the electric field.
the corresponding current/voltage is measured and analyzed by Equation (3) can be solved analytically for symmetric
the sensor electronics and software. The conductivity value binary electrolytes (z 1 = −z 2 ) (Gouy–Chapman solu-
is displayed by a sensor readout unit and/or the data are tion [57]). This is the standard model employed to describe the
transferred to an external computer, mobile device, or network. capacitor-like diffuse layer near the electrode surfaces, with a

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

Debye length (λD ) Equations (2) and (3), optionally with substitutions in (7)
 and (8), must be solved with proper boundary conditions for
εk B T the potential (ϕ) and the ionic mass flux ( ji ) at x = ±L
λD = . (4)
2z i2 e2 ci0 (see Fig. 1) [55]–[58], [60], [61], [68], [69].
The corresponding diffuse layer capacitance (CD ) is
III. E QUIVALENT C IRCUIT M ODELS AND
ε M EASUREMENT S CHEMES
CD = . (5)
λD
A variety of measurement techniques are available in the
Equation (2) combined with (3) have been solved for various time and frequency domains, the selection of which is depen-
configurations and applications, with specific boundary condi- dent on the solvent, conductivity level, and desire level of
tions accounting for surface effects [56]–[68]. The compact uncertainty. For high precision (lowest uncertainty) apply-
layer of adsorbed ions on the surface is known as the Stern ing frequency domain techniques, such as electrochemical
layer [56], its capacitance is impedance spectroscopy (EIS), are recommended: normally ac
ε (Section III-A3), but dc measurements at highest conductivities
CS = (6) (Section III-A1). The majority of commercial EC sensors mea-
xS
sure in the time domain, which is the subject of Section III-B.
where x S is the thickness of the Stern layer.
The descriptions in Sections III-A–III-B include differences
For KCl (aq), λD ∼ 0.3 nm at 1 mol · L−1 and λD ∼ 1 μm
in the published models covering the low conductivity range
at 10−7 mol · L−1 (see (4)). Since typically x S < 1 nm, the
(UPW ∼ 5.5 μS · m−1 ) versus higher conductivity levels
total surface capacitance Csurf = ε/(λD + x S ) is dominated
(several S · m−1 ), and the different frequency regimes
by CD at low-to-moderate ion concentrations. x S and λD are
employed. This section will also discuss challenges encoun-
indicated in Fig. 1. For most EC sensors, the separation of
tered in conductivity measurements.
electrodes is much larger than (λD + x S ). For an EC cell, the
potential variation (ϕ) involves an approximate linear drop
within the Stern layer and an exponential drop through the A. Frequency Domain Measurements
diffuse layer, approaching zero in the bulk solution (Fig. 1). 1) DC Measurements: When a voltage is applied to an
In a practical EC cell, the electrode surfaces are not EC cell, the electrode polarization eventually prevents a dc
smooth and uniform, containing surface defects, charge inho- current from flowing (Fig. 1)—making dc measurements
mogeneities, adsorbed species and variations in composition unsuitable for low and intermediate conductivity measure-
and stoichiometry. These can be modeled by an anomalous ments. Electrode polarization is minimized through use of
PNP equation (PNPA model) based on a fractional derivative ac measurements. However, at very large ion concentrations
in time [59]–[61]. The effects become more important at (conductivities ≥10 S · m−1 ), dc measurements can be made
higher ion concentrations and higher applied voltages, where accurately with the proper choice of electrodes [71]. A four-
the time derivative (∂ci /∂t) in (2) is replaced by a fractional electrode configuration based on Ag/AgCl electrodes has been
diffusion term of distributed order [61] used, because the configuration is reversible with respect to the
 1 applied voltage/current and with minimum electrode polariza-
∂ci ∂ γ ci
→ dγ p(γ ) γ (7) tion due to the separation of applied current and voltage sens-
∂t 0 ∂t ing electrodes. Such a system comprises an EC cell made of
where p(γ ) is a distribution function of γ with precision bored glass, a constant temperature bath, a precision
1
0 dγ p(γ ) = 1.
digital voltmeter, an accurate standard resistor, and a constant
Equations (2) and (3) allow for arbitrarily large ion concen- current source [71], [72]. A constant current is applied to an
trations, because the ions are assumed to be point charges. For outer pair of electrodes, it passes through the standard resistor
higher ion concentrations and voltages, the finite size of the in order to obtain a traceable value for the applied current. The
ions and their mutual interactions limit the ion concentrations voltage is measured by an inner pair of electrodes. Very low
in the vicinity of the electrodes. The PNP equations are voltages must be measured with low uncertainty, increasing
modified (modified PNP equations (MPNP)) to include steric the cost of the required equipment significantly. In order
effects taking account of the finite size of ions. These have to eliminate the effects of different charge build up at the
been analyzed theoretically [57], [62]–[67] to include ion–ion voltage electrode pair, the reverse current measurement is also
interaction in (2), by adding a term to the derivative of the made. Using a high precision voltmeter enabled conductivity
normalized ion concentration (∂ci /∂ x) of the form measurements of a 1 mol · kg−1 KCl (aq) solution with an
expanded uncertainty better than 0.04% [72]. Advantages of
∂ci ∂ci 1 
N
∂ck dc versus ac measurements are the avoidance of the reactive
→ + ak3 (8)
∂x ∂x 1− N 3 ∂x effects of ac circuits (seen at higher electrolyte concentrations)
k=1 ck ak k=1
and avoidance of contribution of lead resistance to the overall
where the second term in (8) includes the interaction between resistance [73].
ion (i ) and the other ions (k = i ), and ak is the effective 2) Equivalent Circuit Models: Equivalent circuit models are
diameter of ion (k) normalized to the average of the initial useful, because they are often simple and provide good fits to
ion concentration [m−3 ]. measurement data. However, it is difficult to interpret the role

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

porosity [54]. In Fig. 2(c), a CPE element has been added to

R0 (Fig. 2(a)) in order to provide a simple phenomenological
model [18].
Fig. 2(d) takes account of the lead resistor (R∞ ) and
inductor (L) [75]. Using this circuit, Sanabria and Miller [75]
were able to fit data of admittance (inverse impedance) from
1 Hz to 1 MHz of electrolyte solutions of KCl, NaCl, MgCl2 ,
and CaCl2 of 1 μmol · L−1 to 10 mmol · L−1 .
Apart from R0 representing the solution resistance, the
circuit in Fig. 2(e) comprises a CPE coupled in parallel with a
charge transfer resistance Rct in series with a Warburg element
(W ), which is a CPE with α = 1/2. When the CPE in Fig. 2(e)
is replaced by a capacitor, the diagram is reduced to a Randles
circuit [18], [52]. In [54], W represents a convective diffusion
process of Fe3+ (aq): 1 mol · L−1 NaOH + 0.1 mol · L−1
K3 [Fe(CN)6 ]. Fig. 2(f) has been used at the Danish National
Fig. 2. Equivalent circuit diagrams. (a) Simplest diagram. (b) [60]. (c) [18]. Metrology Institute (DFM), Hørsholm, Denmark [82] to fit
(d) [75]. (e) [54]. (f) [76]. impedance spectra of UPW and KCl solutions up to 140 mS ·
m−1 . The diagram represents two branches in a Nyquist plot,
where R0 and CPE1 dominate at low conductivities and R0
of each circuit element, and different equivalent circuits may and CPE2 dominate at higher conductivities. R∞ and L only
describe the same electrical response [50], [55]. The challenge have an effect at the higher conductivities.
is to model the polarization effects at the electrodes, which 3) Electrochemical Impedance Spectroscopy: For accurate
becomes increasingly difficult at higher conductivity levels. conductivity measurements, EIS is normally used. Data of real
Selection of the appropriate equivalent circuit to fit EIS data and imaginary parts of the impedance are recorded as function
depends on conductivity level and frequency range, which will of frequency, usually using an inductance–capacitance–
be discussed further below. Additionally, quality of fit can be resistance (LCR) meter. In order to derive the conductivity
evaluated through the calculation of Kramers–Kronig residuals from the spectra recorded, the data may be fitted by PNP mod-
(see [74]), which should be small. Additionally, the quality of els (Section II) [55], [59]–[61], [70], [83] or fitted by equiv-
fit must be taken into account when calculating uncertainty alent circuit models (Section III-A2) [13], [18], [43], [51],
(see Section IV-G). [52], [54], [75], [76], [78]–[82], [84], [85].
Different equivalent circuits modeling the electrode– Fig. 3 illustrates plots of impedance data with the negative
electrolyte interface have been reported (Fig. 2) [13], [18], of the imaginary part of the impedance (−Im{Z }) plotted as
[43], [51], [52], [54], [60], [75]–[81]. Fig. 2(a) models the function of real part of the impedance (Re{Z }) for: 1) UPW
bulk of an electrolyte solution; it comprises a resistor (R0 ) and for KCl solutions corresponding to nominal conductivities
and a capacitor (C) [18] of 2) 130 μS · m−1 ; 3) 1500 μS · m−1 ; and 4) 10 mS · m−1 .
K cell In each plot, there is an indication of the frequency corre-
R0 = (9a)
κ sponding to a data point close to the top of the semicircle.
ε In each plot, the arrow indicates the direction of increasing
C = (9b)
K cell frequency. The data were recorded using a coaxial cell with
where K cell is the cell constant of the EC cell, which is geometrically computed K cell [82], [85] (see also Section V-C)
dependent on the dimensions (see Section IV-A1). The time and an Agilent E4980A LCR meter having an available fre-
constant of the cell (R0 C = ε/κ) does not depend on these quency range from 20 Hz to 2 MHz. The voltage amplitude
dimensions. was 0.5 V. With K cell = 1.190 56 m−1 , the approximate
Electrode polarization effects are taken into account resistances in Fig. 3 vary from 218 k to 120 . The
in Fig. 2(b) by serial addition of an entity comprising a bulk resistance (R0 ) is derived from the intersection of the
capacitor (C1 ) and a constant phase element (CPE) in par- impedance data with the Re{z}-axis at approximate frequencies
allel [60]. These circuit elements model unusual diffusive listed in Table II, which are sufficiently low to be measured
processes equivalently to solving the PNPA (see (7)). The with high accuracy by an LCR meter. With increasing κ,
impedance of a CPE (Z Q ) can be expressed [54], [77] i.e., from Fig. 3(a)–(d), and equivalently according to (9a),
when employing cells of decreasing K cell , the impedance
1
ZQ = (10) data move from the high-frequency arc represented by a
Q(i ω)α semicircle toward the low-frequency arc, observable as an
where Q is an amplitude, i is the imaginary unit, ω is the almost straight line to the right in Fig. 3(d). In order to derive
angular frequency, and α is an exponent. CPEs have also the bulk resistance (R0 ), a suitable frequency range should be
been interpreted to be related to the fractal nature of surface chosen according to K cell and κ (Fig. 3) [18], [85]. Optimal
roughness of the electrodes [18], [75], current, and potential frequency range further depends on cell geometry, solvent, and
distributions from interface inhomogeneities and electrode electrolyte species. The solid semicircles seen in Fig. 3(a)–(d)

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

Fig. 3. Negative of the imaginary part of the impedance (−I m{Z}) plotted as function of real part of the impedance (Re{Z}) for (a) UPW (20 Hz–10 kHz),
and KCl solutions with nominal conductivities of (b) 130 μS ·m −1 (20 Hz–140 kHz), (c) 1500 μS ·m −1 (20 Hz–2 MHz), and (d) 10 mS ·m −1 (20 Hz–2 MHz).
Inset: the corresponding plots of parallel real part of the impedance (R p ) as function of inverse frequency are shown.

TABLE II
Q UANTITIES E MPLOYED IN E VALUATION OF D ATA IN F IG . 3

are computed data using Fig. 2(a). In this case, the impedance by the CPE1 (approaching the behavior of a capacitor) and
is given by R0 (Z (CPE2) → 0). The dashed curves can only be observed
ω R02 C R0 in Fig. 3(c) and (d), because they coincide with the solid curves
−Im{Z } = , Re{Z } = (11) in Fig. 3(a) and (b).
1 + (ω R0 C) 2
1 + (ω R0 C)2
Fig. 2(a) models the bulk EC with only one adjustable
where R0 and C are given by (9), the permittivity of water parameter (κ), provides good fits for κ < 1 500 μS · m−1 .
ε = ε0 εr , where εr = 78.5 and ε0 is the vacuum permittivity. At 10 mS · m−1 , polarization effects are significant and the
This simple model only includes the high-frequency arc, and simplest model does not fit well. However, the dashed curve
as ω → 0: −Im{Z } → 0 and Re{Z } → R0 , which is the point based on Fig. 2(f) gives a good fit.
where the semicircle intersects the Re{z}-axis. The inserts are plots of the real part of the parallel
When (9) and (11) are combined, there is a single adjust- impedance (Rp ) as function of inverse frequency, where Rp
ment parameter (κ); using these the semicircle fits to the data can be expressed as
in Fig. 3, conductivities are listed in Table II. The dashed
 
curves are plots based on fitting the data using parameters −Im{Z } 2
from Fig. 2(f) [82]. The model in Fig. 2(f) covers two arcs Rp = Re{Z } 1 + . (12)
Re{Z }
in the impedance spectra, including a low frequency arc,
which is dominated by CPE2, R0 and L (Z (CPE1) → ∞) The R0 is then determined as the interception with the ordinate
and a high frequency arc to the left, which is dominated axis in the “RP versus 1/ f ” plot [49], [86]. A similar procedure

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

conductivity measurements have been reported including use

of triangular signals and analyzing the time derivative [89],
applying a step current and measuring the voltage across the
electrodes [90], applying current pulses of opposite signs [91],
and using a single square wave and measuring the voltage at
three different times [53]. A double pulse injection technique
has also been proposed, where application of a voltage pulse
was followed by application of a current pulse, where the
amplitude of the second pulse was adjusted based on the
first pulse in order to obtain optimal amplitude for the data
acquisition system [92]. Time domain measurements include
dielectric spectroscopy [93] and monitoring of electrolysis
reactions [87].
Commercial conductivity measurement systems are often
Fig. 4. Current step pulse and voltage response: EC measurement method
using a current-step-pulse applied to a conductivity cell and measuring the based on current-pulse measurements and monitoring of the
resultant voltage across the electrodes. voltage at various times. At high conductivity levels, electrode
polarization has an appreciable effect. In order to get reliable
is widely used at metrology institutes [18], [73], where Re{Z } results, the algorithm used to derive the conductivity from
is plotted as function of inverse frequency (1/ f ). These current–voltage measurements needs to be optimized to the
approaches become equivalent as Im{Z} approaches 0. cell constant, electrode materials, and conductivity level [76].
In the ideal case of Fig. 2(a) with Z given by (11), Rp is Time domain pulse measurements can be transformed into
reduced to Rp = R0 according to (12), i.e., Rp is a constant the frequency domain and vice versa through the use of
and independent of frequency. The relative derivatives of the Laplace and inverse Laplace transform equations [76]. Analyt-
slopes of Rp versus 1/ f in the insets of Fig. 3 are given in ical and numerical schemes can be used to convert expressions
Table II. The data for UPW is closest to the ideal case with based on equivalent circuit models (Section III-A2) into time
no electrode polarization, where d Rp /d(1/ f ) = 0. domain step-responses [76]. Regardless of pulse shape (square
wave (Fig. 4) or sine wave (ac)), the frequency or the time
duration, depending on the domain, must be chosen such
B. Time Domain Measurement that information is obtained in the region of the local min-
Most commercial conductivity sensors are based on time imum between high frequency (semicircle) and low frequency
domain measurements; because circuitry can be implemented (straight line) regions of the impedance (see Fig. 3(d)). This
in a simpler and more compact manner, with faster response allows minimization of measurement uncertainties over a wide
times compared to recording frequency spectra by LCR conductivity range. The measurement signal (as produced by
meters. The disadvantage with impedance spectroscopy is that the EC meter) must be adjusted as conductivity increases.
the system being measured must be at steady-state throughout The value of indication of an EC sensor system can only be
the time required to measure the spectrum, which may rise to verified through the use of EC standards, i.e., solutions for
the order of minutes, depending on the frequency range [87]. which the EC is known. The need for calibration of sensor
According to Fig. 2(a), applying a square current pulse systems using electrolyte solutions measurements, as opposed
with an amplitude of I0 to the circuit results in charging up to purely electrical calibration of the meter, has been shown
the voltage to R0 I0 with a time constant τ = R0 C, which at low conductivity values (e.g., 5 μS · m−1 [76]).
is τ = ε/κ (see (9)). For Fig. 3, τ are given in Table II.
Fig. 4 illustrates schematically a train of square current pulses IV. S ENSOR C ALIBRATION
applied to the electrodes of an EC cell with a corresponding
voltage response; the resistance is derived as R0 = V0 /I0 and A. Measurement Method
conductivity from (9a). The voltage across the cell increases EC (κ), bulk resistance (R0 ), conductance (G), dielectric
to V0 within time τ (phase (1) in Fig. 4). Subsequently, permittivity (ε), and loss (ε ) values are calculated from
the electrode polarization represented by a CPE in series with measurements of current and potential (voltage) as described
the bulk resistor (see Fig. 2(b)–(f)) causes an approximately in Section III. The conductance of an electrolyte solution is
linear increase in the voltage (phase (2) in Fig. 4) [53], [76]. the inverse of solution resistance (1/R0 ) and it can then be
The pulse width (tw ) should be sufficiently short to minimize expressed as the inverse of (9a).
effects of polarization, but τ . For small polarization effects, Routine EC measurements require determination of K cell
the voltage almost reaches a plateau in phase 2, and the by means of calibration, as described in published standards,
conductivity can easily be derived from a well-defined ratio of such as ISO 7888 [94], IEC 60746-3:2002 [95], American
V0 /I0 . For large polarization effects, more involved algorithms Society for Testing and Materials (ASTM) D1125-14 [96],
and/or current/voltage measurement schemes are required. ASTM D5391-14 [97], and other national standards (see [28]).
Electrode polarization is also a challenge in time-domain As it is relatively common for various measurands: com-
measurements, and it has been observed to depend on cur- mon practice within industrial EC measurements is dic-
rent density [52], [53], [88]. Variations on time-domain tated by these types of standards, rather than following the

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

eliminates any fringe effects, and K cell can be computed

from (13), and the conductivity is given by (14) [72]
d
K cell = (13)
A
d
κ= (14)
A· R
where R is the difference in the resistance measurement
with and without the center piece. Conductivity measurements
between 25 mS · m−1 and 6 S · m−1 have been made with this
primary cell [73]. For details on the use of removable center
piece primary cells, on which the International Union of Pure
Fig. 5. (a) Primary EC at DFM and (b) schematic of cylindrical center piece
which can be inserted into the primary cell; the center piece is measured
and Applied Chemistry (IUPAC) recommended conductivity
geometrically providing traceability for cell constant. values of molal (mol· kg−1 ) KCl solutions are based, see [86].
Primary cells are most commonly employed to assign K cell
mis en practique of the metrological unit, such as described values to other EC cells, so-called “secondary” cells, through
by Section IV-B [72], [86]. Determination of K cell , as well as the use of transfer fluids (also referred to as transfer standards).
routine measurements, is not without their challenges, which This ensures traceability to secondary EC cells, which are used
are discussed in more detail in Section IV-F. To comply to produce CRMs.
with important industry standards (e.g., US and EU Pharma- K cell values can be assigned to secondary cells using a
copeia [32], [38], [98], [99]), appropriate documentation are transfer fluid (a stable electrolyte solution, most commonly
required, including means of metrological traceability. KCl (aq)) in one of two ways against a primary cell, or other
1) Determination of K cell : The parameter K cell given in cell of known K cell . In the latter case, the cell under test
units of m−1 (or cm−1 ) is designated as property of an may become a tertiary or quaternary cell depending on the
individual EC cell. While termed a “constant,” this property level in the calibration chain. In one method, the conductance
value changes with time [38], requiring that it be redetermined ratio between the transfer fluid measured in the secondary
periodically. Changes in K cell can be the result of physical or and primary (or reference) cells are determined under near
chemical changes in the electrodes (e.g., surface area) and/or identical conditions (temperature control, minimal time and
their alignment [1], [32], [38], [44], [100]. space variability, etc.) [3]. When calculating the uncertainty of
Determination of K cell is performed by: 1) geometric mea- this method, potential influences of transfer solution handling,
surements [73]; 2) using a transfer fluid [29]; 3) using a and temperature differences must be included [108], [109].
certified reference material (CRM) with an assigned κref [28]; Alternatively, the transfer fluid may be circulated through
or 4) making a capacitive measurement of a fluid of known a closed-loop, which includes both primary (or reference)
dielectric constant [101]. The latter has been employed in and secondary cells, thus avoiding potential introduction of
extremely low conductance (pS) fluids, such as aviation contaminants (e.g., impurities, CO2 ) [108]. The use of a
fuels [39] and is not discussed further in this work. circulated transfer fluid is required for K cell employed in pure
water applications (κ < 5 mS · m−1 ) due to the dominating
B. Geometrically Measured Conductivity Cells effect of atmospheric CO2 [82].
EC cells with K cell determined by measure of geometric
measurement are termed “primary” cells, as measurements C. Certified Reference Materials
do not derive their traceability to a higher order standard The most common means of K cell calibration in routine
of the same quality [85]. Primary EC (κ) is traceable EC applications is through the use of CRMs (i.e., fluids
to: 1) length [m] and 2) resistance []. Primary cells assigned a reference conductivity (κref ) value) [15], [31].
include extremely demanding construction and handling, and Historically, solutions of aqueous KCl have been selected
are limited to several well characterized geometries [73], as EC CRMs [20], [100] as high purity KCl and water are
[85], [102]. Primary cell geometries include two-electrode easily accessible, and the long-term stability of KCl (aq)
(removable center pieces [20], [28], [72], [73], [86]; piston- solutions [105]. CRMs may be purchased as solutions; these
type [103]–[105]; co-axial [85]) and four-electrode (reversible typically have larger uncertainties than transfer fluids of sim-
electrodes [20], [86]; nonreversible electrodes [102]; van der ilar composition due to the inclusion of homogeneity and
Pauw [106], [107]) arrangements. In Fig. 5(a), the primary stability uncertainties [28], [38], [105], [109]. Alternatively,
differential EC cell at DFM is depicted, and in Fig. 5(b), a calibration solutions may be prepared according to recipes
schematic of the cylindrical centerpiece, which can be inserted published in international standards: IUPAC 2001 recommen-
into the primary cell, is illustrated. The design is based on an dations [72] and International Organization of Legal Metrol-
original proposal by the National Institute of Standards and ogy (OIML) standards [20], [100]. Both set of standards
Technology (NIST), Gaithersburg, MD, USA [72]. (molal · (mol · kg−1 ) [72] and demal (“equivalents” · dm−3 )
In Fig. 5, the length (d) and the area ( A) of the removable [100] solutions) require the use of: 1) KCl of known
cylindrical center piece are measured geometrically. Making purity (from argentometry [28], [101]; 2) water of suffi-
conductivity measurements with and without the center piece cient purity (low conductivity) equilibrated with atmospheric

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

laboratories might be kept sufficiently low to enable UPW to

be used as a reference material.
It is desirable to have other conductivity values than the
value at 25.00 ◦ C. Light et al. [11] have measured the con-
ductivity of pure water from 0 ◦ C to 100 ◦ C. They found
that the value of resistivity changes from 86.19 M · cm
at 0 ◦ C to 1.274 M · cm at 100 ◦ C, a conductivity range of
1.16 μS · m−1 to 78.5 μS · m−1 . The accuracy was ∼0.25%
below 60 ◦ C and ∼0.5% between 60 ◦ C and 100 ◦ C, although
the precise definition of “accuracy” is not included [11].
It should be noted that K cell of an EC cell changes with
temperature, but this is normally a negligible effect in routine
applications (see Section IV-F1).

E. Metrological Traceability
Fig. 6. Plot of EC of UPW as function of time as measured at DFM at
25.00 ◦ C. The data are plotted with 0.3% error bars illustrating the expanded Increasingly the means of traceability to the International
uncertainty. The standard deviation of the data plotted is 0.042%.
System of Units (SI), or recognized international reference(s)
(e.g., traceability of Sp measurements to International Asso-
ciation for the Physical Sciences of the Oceans (IAPSO)
CO2 [86]; and 3) calibrated balance(s). These standards
Standard Seawater (SSW) [23]), is required to meet national
rely on preparation according to recipe, with suitable cau-
and international requirements, such as the US or EU Pharma-
tion [86] (e.g., suitable drying of KCl [110]), with traceability
copeia [32], [98], [99]. Traceability requires: 1) documentation
achieved through gravimetry (weighing), and conductivity
of the calibration chain and 2) assignment of uncertainty
values assigned based on prior measurements—i.e., historical
to reported values [3], [16], [112]. Since the pioneering
measurements (see [100]). The correct use of the molal
research by Kohlrausch [162] in 1869 on “electrolytic con-
conductivity solutions (IUPAC 2001 Recommendation [72])
ductance” [19], [20], key parameters on which the current
must include the assigned uncertainties of the conductivity
measurement and application of EC rely have changed as a
values. The equivalence between molal conductivity solution
result of more accurate measurement and/or changing inter-
assigned and measured (using primary cells) values has been
national consensus. As an illustrative example, the IUPAC
demonstrated [3], [28]. By contrast, the demal conductivity
recommended molal conductivity solutions rely on the atomic
solutions (OIML standards [111]) do not include uncertainty
weights of potassium and chloride [86], as well as a consensus
contributions, and are not recommended for the calibration of
understanding of temperature. Since these conductivity values
K cell [20], [86].
were verified to a high level of precision (see [86]) between
1989 and 1991, the common usage scale of temperature
D. UPW as a Reference Material has undergone a change (International Practical Temperature
Brinkmann et al. [73] have pointed out that the charac- Scale (IPTS-68) to International Temperature Scale (ITS-90)),
terization of UPW through traceable EC measurement would requiring the recalculation of conductivity values (see [72]).
open the way for its application as an important standard for Traceability to the SI allows intercomparison of measurements
the analysis of chemical and physical properties in aqueous across time and space. This is in contrast to Sp values
solutions. Bevilacqua [31] also suggested that UPW be used (a conductance-derived measurement), which are traceable
as a reference. This is complicated as the conductivity of UPW only to an international reference, where intercomparison on a
is extremely sensitive to the presence of impurities, which decadal timescale become complicated by the required inclu-
may come into solution through leaching of UPW in contact sion of uncertainty as a result of changes to the international
with most materials [2], or through exposure to atmospheric reference [22], [23].
CO2 [86]. The proper assignment of uncertainty values is a key aspect
At DFM, the EC of UPW has been measured at 25.00 ◦ C of metrological traceability [28], [46], [85], [108], [110],
using a closed-flow-loop measurement system with a Milli-Q [112]. Influences which contribute to the uncertainty value of
system (MilliPore, Merck, Darmstadt, Germany) over several measured EC are often those which complicate its measure-
years [82]. Fig. 6 includes a laboratory relocation (12 km) in ment and/or the calibration of K cell . The influences should be
November 2017. A reproducible value for the conductivity identified, quantified, and uncertainty contributions propagated
of UPW can be achieved. At 25.00 ◦ C, UPW could be in the appropriate manner [109], [113].
used as a good reference. This does require validation as to An EC sensor comprises two components: a conductivity
whether the same value and purity can be achieved at various cell (with a calibrated K cell ) and an analyzer/measurement
laboratories. The standard deviation of the data is 0.042%, device. These two components should be calibrated together
which is significantly less than the uncertainty of the data as a system calibration to account for influences from the
in Fig. 6. This suggests that the variation in purity of UPW lead wires, stray capacitance, the resolution, and data analy-
as function of time and arising from different locations of sis algorithms, etc. [40], [86], [101]. The accuracy of the

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

analyzer component, which may also be called a readout leach into solution [29], [108]. It has also been suggested
unit, reader, meter, or transmitter, is assured through the use that the presence of dissolved gases (e.g., N2 ) influence the
of calibrated resistors [3], [114]. The analyzer measurement conductivity of water-based solutions [21]. The presence of
range, and K cell , should be matched to the expected solution dissolved atmospheric CO2 has a well-documented influence
resistance, to ensure that the analyzer remains within its linear on lower conductivity electrolytes and UPW [86]. Atmospheric
range [32], [78]. A more in-depth discussion of the potential CO2 dissolves in water, forming carbonic acid. Its dissociation
influence of the analyzer component in high precision appli- into H+ and HCO− 3 results in measurable changes to the
cations can be found in [114]. For routine EC measurements, conductivity of UPW [101], [114]. This effect makes the
calibration of K cell , or the contribution of a calibrant CRM measurement of lowest electrolytic conductivities difficult.
assigned uncertainty, is typically the dominant component to It requires the use of close-loop calibration system for pure
uncertainty. water applications [38], [82], and necessitates the use of mixed
The measurement challenges and the corresponding solvent CRMs at low conductivities (<8 mS · m−1 ). These
contributions to EC uncertainty budgets are the topics of CRMs are typically based on either dilute HCl–water solutions
Sections IV-F and IV-G. (minimizing the dissociation of carbonic acid) or inclusion of
organic solvents in order to increase viscosity [16], [105]. The
F. Measurement Challenges effect of CO2 on the conductivity of water does, however,
Measurement challenges commence with the choice of permit the development of CO2 sensor, which is based on
the appropriate measurand (e.g., EC, conductance, etc.). changes in measured conductivity [46]. For the preparation of
This depends on the application, taking into consideration: CRMs and calibrant KCl (aq), the water is first equilibrated
the solvent being measured, the conductivity range, and tem- with atmospheric CO2 [115], and a component added to
perature. Further, based on the desired measurand, and desired uncertainty [72].
level of precision, the appropriate conductivity cell geometry The interactions between the electrolyte solution and the
should be selected (see Section V). measurement electrodes lead to polarization effects which
Calibration of K cell should be made using a CRM that must be accounted for when extracting solution bulk resis-
matches the expected EC measurement range [3]. While K cell tance. Electrode polarization effects are dependent on sol-
is often assumed to be invariant with conductivity level, there vent [18], electrolyte species and concentration [12], and
is a consistently reported [18], [28], [38], [86] influence of the electrodes [8]. Influences associated with the electrodes
CRM conductivity level on the evaluated K cell . Further, some include the electrode surface area [1], [2], [100], [106], [110]
authors have suggested matrix matching, where possible, the and any changes to area during measurement. Area is changed
CRM employed for calibration of K cell and routine mea- by the formation of bubbles on the electrode surface [20], [32],
surement electrolyte solutions. This has been suggested in as well as electrolyte “creep” around the edges of the elec-
such applications as seawater and equally conductive KCl trodes [45]. Sharp electrode edges should be avoided [45].
(aq) [3]. Matrix matching has also been identified as a Polarization is additionally a function of the distance between
potentially problematic issue in the low conductivity range the electrodes [2], [8], [12], [43], [110], [116] and the electrode
(<8 mS · m−1 ), where mixed solvent CRMs are typically material [2], [14], [43], [45], [117], with platinum being
employed [32]. CRMs should be used appropriately [16], the best choice for high precision applications [29]. The
according to the recommendations of the manufacturer. Several magnitude of applied voltage should be lower than the reaction
influences on CRM usage have been investigated, including potentials of all possible electrochemical reactions at the
storage conditions [86], [103], [109], and multiple uses of a electrode/electrolyte interface. Any current density [52] should
CRM bottle [105], resulting in the introduction of a variable be minimal in order to prevent any reaction from occurring,
amount of air “headspace” above the liquid [32]. Matching such as electrolysis [29], [36], and any possibilities of self-
conductivity levels in application and calibration is compli- heating (high resistance applications) [38].
cated at both high and lowest conductivity (κ < 5 mS · m−1 ) At high conductivities, where polarization effects are pre-
due to the lack of CRMs. At high conductivities, use of various dominant and typically R < 100 , calibrated resistors are
high concentration HCl (aq) solutions has been suggested [7]. used in the calibration of an analyzer/readout unit, which is
At lower conductivities, UPW has been suggested as a cal- based on time domain techniques. This does not necessarily
ibrant (see Section IV-D). However, use of these calibrants reflect the ability of the sensor system to accurately measure
is not currently directly traceable to the SI [18], [101], impedance of electrolyte solutions, and a system calibration
and so should not be used in combination with SI-traceable should be performed based on electrolyte solutions [76].
conductivity results. When applying ac techniques (e.g., EIS), the frequency range
Contaminants, in the form of ions or gases, influence mea- is carefully chosen to match the conductivity range of the
sured conductivity. The cleanliness of the conductivity cell, electrolyte solutions [85], [114], the desired measurand (e.g.,
especially when changing between electrolyte solutions, will conductivity or dielectric permittivity [5], [12]), and the sensor
influence measurement [86]. Electrolyte solutions in contact type [106]. It is chosen to minimize the contributions of
with the materials of the conductivity cells and the storage electrode polarization, the lead wires, etc. The contribution
vessel, including CRM storage bottle, may results in the to uncertainty of these influences is generally included in
inclusion of ion contaminants. When in contact with glass the “inference” [85] or “extrapolation” [112], [113] term of
for extended periods of time, components from glass may solution resistance from the EIS measurement.

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

1) Temperature: EC is strongly influenced by changes method used in metrology. A sensor may be calibrated as a
in temperature [11], [111]. Electrolyte solutions in the system, or K cell and the sensor analyzer/readout unit may be
mid-conductivity range are typically assumed to have a tem- calibrated separately.
perature coefficient of 2%·◦ C−1 [8], and this parameter may be In Table III, two examples of uncertainty budgets are listed:
employed for automatic temperature compensation (i.e., com- (A) a sensor cell calibration based on a closed-flow-loop
pensating EC readings to a reference temperature, typically calibration system (using a circulated transfer fluid) and (B)
25 ◦ C [14]) in “smart” conductivity sensors [110]. However, a sensor system calibration based on CRM solutions certified
the temperature dependence of EC is dependent on solvent, using secondary cells (see Section V-B). The traceability for
electrolyte species, and conductivity range. Low conductivity method (A) originates from geometrical measurements on a
mixed solvent CRMs have much higher temperature coeffi- coaxial primary cell (see Section V-C), and the traceability
cients [32], as does UPW [29]. The temperature coefficient of for method (B) originates from geometrical measurements on
UPW, and at conductivities near UPW, is strongly dependent a differential primary cell as described in Section IV-B. The
on the presence of ionic contaminants, such that measuring data in Table III(A) have been taken from [82, Table I]. There
the temperature coefficient serves as an indicator of water within, the device-under-test (DUT) sensor was mounted in a
purity [11], [31]. For example, 1-ppm change in salt con- closed-flow-loop together with the coaxial primary cell. The
centration changes the temperature compensation slope from uncertainty contributions of K cell are: 1) geometric measure-
approximate 5%·◦ C−1 for UPW to 2.5% ·◦ C−1 for 1 ppm ments of the primary cell (K cell ); 2) resistance measurements
NaCl [14]. (R) including calibration using standard resistors, impedance
Conductivity sensors are constructed with a minimal influ- measurements by LCR meters of solutions and curve fitting to
ence of temperature on the K cell value. The calibration of derive R0 ; 3) temperature gradient; 4) conductivity gradients
K cell and the routine measurement of EC values should be between primary cell and DUT; and 5) reproducibility [82].
performed under the same temperature conditions. K cell is The uncertainty arising from the (2e) curve fitting was calcu-
dependent on temperature, and primary conductivity cells lated from the comparison of the semicircle fitting (Fig. 3(a))
are made from materials with the lowest possible thermal relative to the Rp versus 1/ f , combined with statistical vari-
expansion coefficients as to have K cell values which are tem- ation of the fit parameters (simulated by random number
perature insensitive [86]. The temperature is best monitored in generator of frequency range) and including a variation in the
close proximity to the conductivity cell [15], [38]. For in-lab input real part of the impedance with the assigned uncertainty
measurements, the use of an appropriate means of temperature of each measurement point. The relative differences in the
stabilization (thermostating) is advisable. Thermostating may derived values of κ between the fits (using Fig. 2(f)) to
be performed through control of air temperature [28] or the impedance spectra (dashed curves in Fig. 3) and the
immersing the cell in a thermostating fluid. This fluid should derived values from plots of “Rp versus 1/ f ” in the inserts
be nonconductive in order to minimize the influence of stray (Fig. 3) are −0.0018%, −0.010%, −0.018%, and −0.014%
currents [101], [111]. For 0 ◦ C measurements, the conductivity in Fig. 3(a)–(d).
cell may be packed in ice [100], or placed in an ethanol For conductivity between 5.5 μS · m−1 and 140 mS · m−1 ,
bath [11]. Higher temperature applications, where measure- the estimated maximum standard uncertainty is 0.17%–0.13%,
ments are relevant for boiler water [10] and hydrothermal largest for UPW because of its larger temperature coefficient
vents [9], more robust temperature control solutions such as (Section IV-F1). For a system calibration of a sensor based
oils are required. on calibration against a primary cell, point 2 “Resistance
When employing temperature compensation, the correct measurements” (Table III(A)) is replaced by the standard
temperature coefficient must be used [43]. It is preferable to deviation and resolution of the DUT sensor readout unit.
measure the temperature coefficient of individual electrolyte For a system calibration (Table III(B)), the data read from
solutions. If the electrolyte solution composition and concen- the readout device is compared with the certified EC of the
tration range are known, a temperature compensation function CRM, which contributes: 1) to the uncertainty budget with a
(see [17], [72]) may be suitable. If temperature compensation value in the range from 0.13% to 0.05% for the conductivity
is performed, the fit of the compensation function must be range from 10 mS · m−1 to 25 S · m−1 . The other contributions
included as a component of the uncertainty [16]. This is in include: 1) the readout from the DUT sensor system with
addition to other temperature-related uncertainty contributions standard deviation of the readings and 2) digital resolution.
(temperature stability and homogeneity (time and space), Contributions from temperature include: 1) offset from the
and temperature probe calibration) [28], [72], [110]. When correct temperature and thermometer calibration uncertain;
reporting EC values, temperature must be included; it should 2) temperature gradient in the temperature-controlled bath; and
be stated when temperature compensation is employed, and 3) reproducibility of the measurements.
actual measurement temperature preferably specified. Table III(A) and (B) cover the calibration range from
5.5·10−6 S·m−1 to 25 S·m−1 (6.7 decades). Total uncertainties
have very similar values for case (A) covering the low con-
G. Uncertainty Budgets ductivity range and case (B) covering the higher conductivity
Calibration of an EC sensor involves computation of the range. From 10 mS · m−1 to 140 mS · m−1 is covered by both
uncertainty of the calibration; the “Guide to the Expression measurement methods. A conductivity sensor can be calibrated
of Uncertainty in Measurement” (GUM) [118] is the standard over a wide range of EC with similar uncertainties. However,

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

TABLE III
U NCERTAINTY B UDGETS W ITH R ELATIVE S TANDARD U NCERTAINTIES FOR C ALIBRATION OF A DUT EC S ENSOR
BASED ON (A) G EOMETRICALLY M EASURED P RIMARY C ELL AND (B) CRM

a sensor normally cannot operate with high accuracy over more 3) For highest conductivity (≥10 S · m−1 ): the
than ∼four decades. It is necessary to use at least two sensors four-electrode (Ag/AgCl) geometry combined with
to cover the full range in Table III. dc (Section III-A1).
In a two-electrode configuration, the same pair of electrodes
V. S ENSOR C ONFIGURATIONS is used for exciting the current flow and measuring the voltage
A. Two Electrode and Four Electrode Configurations across the electrodes. The limitation of the two-electrode
configuration is the polarization effects at high concentrations
Sensor configuration is chosen based on the applications,
(Section III-B), and the bulk resistance of the solution cannot
for example, strongly dimensionally constrained application
simply be derived as the ratio between V0 and I0 (Fig. 4). In a
will benefit from the use of planar electrodes (Section V-E),
four-electrode configuration, I is supplied by a set of two-
while system constraints such as limiting contact between
electrodes and V is measured by another set of electrodes,
analyte and sensor will favor the use of noncontact sensors
which experience essentially zero current. Bulk resistance can
(Section V-F). Choice of appropriate sensor configuration
be derived accurately from the ratio (V0 /I0 ). A schematic
can be based on available K cell for a given geometry (e.g.,
drawing of the two configurations can be seen in Fig. 7.
coaxial cylindrical, Section V-C) and acceptable level of
A four-electrode sensor is more resistant to fouling of the
uncertainty for a given application. For high precision applica-
electrodes [14] and exhibits a larger measurement range [119].
tions, the following configurations are recommended based on
Minimal current flowing at the inner electrodes eliminates
EC range.
polarization effects, such that four-electrodes sensor are used
1) For the lowest conductivity range (e.g., UPW applica-
at higher ion concentrations. However, the elimination of
tions and κ ≤ 0.1 mS · m−1 ): The use of a flow-through
polarization is not complete [120]–[122] and the effective K cell
coaxial cylindrical electrode (Section V-C) in order
of a four-electrode cell does vary with the ratio between the
to eliminate fringe effects and to isolate the calibrant
polarization and bulk resistance [43].
solution from contaminants, including air (see also
Section IV-D).
B. Plane-Parallel Electrodes
2) For low conductivity and high range (0.1 mS · m−1 to
∼10 S · m−1 ): the plane-parallel electrodes geometry The design of a plane-parallel two-electrode EC cell, termed
combined with ac (Section V-B). Jones-type cell [123], is used at metrology institutes as

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

Fig. 7. Schematic of EC sensor configurations with (a) two-electrodes and

(b) four-electrodes.

Fig. 9. Coaxial cylindrical EC cell. (a) Optical micrograph. (b) Drawing

with cross-sectional view.

in the range from 100  to 100 k (100 μS·m−1 to 60 S·m−1

for K cell = 1 m−1 , see (9a)). Below 100 , polarization effects
dominate; above 100 k, leakage currents and stray capacitor
contributions become important.
Parallel electrode configurations are used in dielectric
spectroscopy of conductive liquids either in two-electrode or
four-electrode configurations [120], but for practical working
conductivity sensors parallel electrodes are not commonly
used.

C. Coaxial Cylindrical Electrodes

At low conductivities (≤100 μS · m−1 ), a plane parallel
configuration is not practical due to small electrode separation
arising from measurement constraints of the LCR meter,
which becomes less accurate at high resistance (>100 k).
A coaxial cylindrical configuration is more appropriate
(Fig. 9) [85], [101]. It comprises an inner cylindrical electrode
with an outer diameter (din ) and a concentric outer cylindrical
electrode with an inner diameter (dout ) and a height (h). K cell is
Fig. 8. Jones-type cell comprising two-electrodes each with an area ( A) and given by [85]
separated by the distance (d).
ln(dout /din )
K cell = (15)
2πh
secondary cells for manufacture of CRMs (Fig. 8). K cell can be where ln is the natural logarithm. The two guards above and
estimated by (13); but this does not take into account fringe below the outer cylindrical electrode and thin electrical isola-
fields at the electrode circumference, nor field lines in the tion layers of Mylar at the interfaces between each electrode
environment of glass/air. Jones-type cells are calibrated against help ensure that the electric field lines are perpendicular to
a primary cell (Section IV-F). the cell walls over the length. For an isolation layer with a
During design, electrode area ( A) and separation (d) are thickness g, h should be replaced by h + g in (15). With
matched to the conductivity of the solutions under test. In prac- this cell, calibration of EC sensors has been performed in the
tice, the values of K cell are between 10 m−1 and 6000 m−1 . range from 5.5 μS · m−1 to 140 mS · m−1 , with an expanded
For high accuracy, it is desirable to measure resistance values uncertainty of 0.3% [82]. Agreement of 0.1% to conductivity

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

Fig. 11. Planar electrodes for conductivity sensors. (a) Two-electrodes

interdigitated. (b) Four-electrodes interdigitated. (c) Pair of disk electrodes.
(d) Pair of annular electrodes.

near the wall of a cell body, thin strips entered into grooves or
Fig. 10. Schematics of measurement configuration using van der Pauw
method. placing the electrodes in individual chambers, separated from
the measurement vessel by gaps of small widths [106].
Zhang et al. [107] made a configuration with adjustment
measurements made by Jones-types cells in the range from screws in order to enable a symmetric configuration, which
10 mS · m−1 to 140 mS · m−1 was achieved [82]. could be verified by applying current to stainless steel
Golnabi [124] has reported construction of a cylindrical electrodes 1, 4 and 1, 2 and measuring the voltage at the elec-
cell probe system and analyses of electrical parameters for trodes 2, 3 and 3, 4, respectively. For the symmetric configura-
a number of water samples. Most common configurations for tion, the voltage/current ratio should be the same in both cases.
a commercial conductivity sensor use a central rod as one Conductivity measurements were made on KCl solutions
electrode and a cylinder as the second electrode (Fig. 7(a)) (1 mS · m−1 –200 mS · m−1 ), with a maximum relative error
fabricated with a specification of K cell within 2%. The design of 1% and a repeatability error of 0.1%. Gurusamy et al. [48]
does not allow a simple mathematical expression for K cell , developed a van der Pauw-based system for measuring varia-
and these must be calibrated with a liquid calibrant to ensure tion in urea concentration in diesel exhaust fluid. Urea is used
traceable measurements [82]. Usually, a temperature sensor is to catalytically reduce NOx emissions from diesel engines,
integrated, allowing compensation for temperature variations, and can be measured to concentration changes of 0.25% in
when the temperature coefficient of the liquid is known urea–water mixtures.
(Section IV-F1).

D. Van der Pauw Type E. Planar Electrodes

An alternative EC cell configuration, which takes advantage Planar electrodes are suitable for miniaturization, enabling
of the separation of current and potential sensing elec- smaller sample volumes, important for applications in pharma-
trode (four electrode configuration), is a van der Pauw cell. cology, blood and urea analysis, and tissue perfusion measure-
K cell depends ideally only on one parameter, the height ments [125]. Both thick film (screen-printed) electrodes [126]
(h) [106] (Fig. 10). and thin film (vacuum deposited) electrodes are reported,
AC current with constant amplitude (frequency on various substrates: Ta2 O5 -insulating film [127], Pyrex sub-
10 Hz–10 kHz) is applied to the electrodes 1 and 4, strates [128], borosilicate glass [129], printed-circuitboards
and voltage is measured at the electrodes 2 and 3. The width (PCBs) [130], [131], silicon micro-electromechanical system
of each electrode is much smaller than the cell circumference. (MEMS)-based sensors [125], and complementary-metal–
For a perfectly symmetrical configuration, K cell can be oxide–semiconductor (CMOS)-based wireless integrated sen-
computed according to [106] sors with electrodes formed in polysilicon [132]. Mostly
platinum is chosen as the electrode material, as it is chem-
ln 2
.
K cell = (16) ically inert with no oxide formation, and exhibits a linear
πh temperature coefficient [129]. Fig. 11 illustrates four different
The van der Pauw technique appears simple and elegant, configurations of planar electrodes, (a) two-electrodes inter-
but there are some considerations: influence of the finite digitated [119], [125]–[128], [132], (b) four-electrodes inter-
size of electrodes, nonideal electrode location, and electrode digitated [119], [129], [132], (c) pair of disk electrodes [133],
impedance. A practical realization includes thin rods placed and (d) pair of annular electrodes [134].

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

For low conductivity application, where CRMs are lacking K cell and the bulk solution resistance are also scaled by F, but
(κ < 0.1 mS · m−1 ), it can be advantageous to calculate the electrode impedance increases with F 2 [125].
K cell . Olthuis et al. [127] derived an expression for K cell of Measurements with miniaturized planar electrodes are chal-
Fig. 11(a) lenging at low conductivities, and a design with minimal
1 2K (k) electrode impedance is important. Platinum-black increases
K cell = (17) the effective electrodes surface area, and reduces the
(N − 1)L K [(1 − k 2 )(1/2) ]
electrode impedance by two orders of magnitude [125].
where N is the number of electrode fingers, L is the length Platinum-black electrodes in a four-probe measurement con-
of each electrode, and K (k) is a mathematical function, figuration, 2% linearity and repeatability for 100 mS · m−1
the complete elliptic integral of the first kind. The parameter to 20 S · m−1 NaCl solutions were reported [125]. Ag/AgCl
(k) can be approximated by k = (s/(s + w)) when N = 2, electrodes are relatively nonpolarizable electrodes and have
and k = cos((π/2)(w/(s + w))), when N > 2, where w been used in microfluidic-based sensors [133], although,
is the width of the electrodes, and s is the distance between reduction-oxidation (redox) processes tend to occur at the elec-
the electrodes. In Fig. 11(a) N is six electrode fingers. Equa- trodes and ions may diffuse into the solution being measured.
tion (17) agrees better with experimental data for larger N. By optimizing the design of interdigitated electrodes,
Experimentally determined K cell were systematically 0%–30% Timmer et al. [128] developed a conductivity sensor with
lower than computed values, which could be explained by the K cell as low as 7.9 m−1 , enabling measurements of UPW.
presence of fringing effects at the Ta2 O5 /SiO2 interface [127]. Recently, Brom-Verheijden et al. [119] fabricated microm-
Using N = 2 electrode fingers, where the electrode fingers eter sized interdigitated planar Pt electrodes in two- and
comprise semi-infinite cylindrical electrodes of radius (a), four-electrode configurations. Good agreement was achieved
K cell is given by [135] between derived conductivities using (KCl (aq) solutions)
2 s between 1 mS · m−1 and 50 mS · m−1 with two-electrodes and
K cell = arccosh . (18)
πL 2a 0.3 mS · m−1 to 1.2 S · m−1 with four-electrodes.
For the configuration illustrated in Fig. 11(c), K cell is given Small electrode surfaces increase the electrode impedances,
by [133] such that higher power is dissipated, causing a self-heating
 effect. It is important to keep current low, implying the need
π − 2arctan(1/ (d/b)2 − 1) for accurate measurements of small voltage amplitudes [43].
K cell = (19)
2πb With most planar structures, it is extremely difficult to ana-
where b is the radius and d is the distance between the centers lytically evaluate K cell exactly: numerical methods normally
of the two circular disks (Fig. 11(c)). need to be used in the computation. Stray electric fields,
Assuming an average potential along the inner electrode effects of substrate material and the finite volume of the
disk in Fig. 11(d), Cheng et al. [134] gives an approximate solutions must be taken into account for high accuracy.
expression for the capacitance of the pair of annular electrodes. To date, for planar configurations, corrections to computed
The agreement with finite element computations was relatively K cell have been made in order to match model to experimental
good. Using (9b), K cell can be written data [119], [125]–[127], [129], [131], [133].
⎡ 2

2 ∞ J1 (ξrc ) rao J1 (ξrao ) − rai J1 (ξrai )
K cell = ⎣ −   F. Noncontact Conductivity Sensors
π 0 ξrc ξ rao
2 − r2
ai
⎤ There are two types of sensors, in which there is no galvanic
contact between the electrodes and the solution, as illustrated
/tanh(ξ h)⎦dξ (20) schematically in Fig. 12: with (a) capacitive coupling and
(b) inductive coupling.
where J1 is the first-order Bessel function of the first kind, 1) Capacitively Coupled Sensors: A capacitively coupled
rc is the radius of the inner electrode, rai is the inner radius sensor (CCS) is very similar to the configurations described
and rao is the outer radius of the outer electrode, and h is the in Sections V-B, V-C, and V-E. K cell and capacitance (C)
height of the dielectrics above the electrode, i.e., the electrolyte originate from the same definition apart from the dielectric
solution. Equation (20) agrees better with numerical finite function (ε) (see (9)). However, in a CCS, an isolation layer
element computations for small values of rc and relatively is introduced between electrode and solution as illustrated
large values of rai [134]. schematically in Fig. 12(a). This layer corresponds to coupling
Polarization effects, stray impedance, and temperature the EC cell in series with a capacitor (Cg ). In a simplest model,
effects are more pronounced as the dimensions scale down. Cg is coupled in series with R0 in Fig. 2(a), and for Cg C,
The electrode polarization impedance increases with decreas- Cg can be coupled in series with R0 and C. A guard should
ing dimensions, and lead capacitance can become the dom- preferably be integrated with the sensor in order to reduce
inant component of stray impedance contributions. K cell of stray electric fields (Fig. 12(a)).
sensors based on interdigitated electrodes requires correction CCS are used extensively for automotive applications,
depending on conductivity; the measurement range of con- such as for measuring the height of liquid levels [136],
ductivities must be narrow in order to achieve an acceptable and in capillary electrophoresis [137], [138]. Measurements
accuracy [126]. If the dimensions are scaled by a factor F, then of conductivity (coaxial cylindrical [139]) and permittivity

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

reports other types of ICS, including a sensor comprising two

solenoid inductors, where the sensing mechanism is based on
a variation in the permeability of water as function of salt
concentration.

G. Trends
EC sensors are being used in a large number of fields
including water quality and environmental monitoring, health
care, and multisensor applications (Table I). In these fields,
there is a trend toward cost reduction, miniaturization, easy
operation, low power consumption, and development of smart,
integrated sensors.
Miniaturization and integration mean sensors with
Fig. 12. Schematics of two noncontact conductivity sensors with (a) capac- planar configurations (Section V-E). Recent reports
itive coupling and (b) inductive coupling. include: a four-electrode planar configuration based on
laser-induced-graphene fabricated on flexible polyimide [24]
(plane-parallel electrodes [140]) of water samples have been for monitoring conductivity and temperature; a MEMS-based
reported. Frequency ranges are from 500 Hz to 2000 Hz, integrated sensor chip for monitoring pH, conductivity,
similar to contact conductivities sensors [139]. temperature, oxidation-reduction potential (ORP), and Cu2+
The advantages of using CCS include: no fouling and corro- concentration [26]; a high precision, low power consumption,
sion of electrodes, and no or limited polarization effects [140]. low-cost and integrated miniaturized temperature, and
A disadvantage is increased interference with external electro- conductivity sensor based on a two-fingers four-electrode cell
magnetic fields, because the impedance level is increased by MEMS [25]. MEMS technology is advantageous, because it
the effect of Cg . consumes little power, takes relatively smaller space, shows
2) Inductively Coupled Sensors: Inductive conductivity sen- high performance, and can be mass produced [150].
sors (ICSs) have several applications in industry and for For smaller production series, PCB sensors have a lower
saline measurements [14], [35], [141]–[145]. The conductivity cost than manufacturing of MEMS devices [130]. Werner
measurement is based on the electric field coupling to the fluid and Dean [130] reported two planar electrodes integrated
by electromagnetic induction, with no electrode polarization. onto a PCB for measuring conductivity. For low-cost,
In Fig. 12(b), a schematic drawing illustrates the principles single use conductivity sensing in health care applications,
of a sensor based on two mutually coupled toroid inductors. Majumder et al. [150] reported a low volume (15 μL),
The primary coil is fed with an input ac current, while an self-powered paper-based battery for sensing ionic conductiv-
output current is induced in the secondary coil via flow of ity by using the solution under test as battery electrolyte.
charged species (ions) in the solution in which the toroids The development of wireless technologies, Internet of
are immersed. K cell depends on the cross sections of volumes Things (IoT), and Cloud computing has paved the way
inductively excited by electromagnetic coupling. It is not pos- for the development of wireless sensors measuring con-
sible to make designs for a geometrically measurable K cell for ductivity among other parameters in particular for applica-
accurate measurements; K cell of ICS needs to be determined tions within monitoring water quality and agriculture [26],
experimentally. [27], [30], [152]–[154]. IoT refers to low-cost devices interact-
Agranovich et al. [142] developed an analytical expression ing over the Internet in an intelligent way to process informa-
for the impedance of ICSs comprising three parameters, which tion. Smart and distributed sensing systems can be considered
must be determined by a measurement procedure on three as a major technological corner stone of IoT [155]. A number
known solutions. Using frequencies from 1 kHz to 1 MHz gave of communication protocols for wireless sensors have been
fairly good agreement of conductivity measurements on KCl reported including Zigbee, WIFI, and Bluetooth [26], [155]
solution standards (0.4 S · m−1 –4 S · m−1 ) between a toroidal for actively powered sensors, and near-field communica-
sensor and a commercial conductivity meter. Conductivity tion (NFC) and radio frequency-identification (RFID) for
measurements on alcohols (<50 μS · m−1 ) were made, using passive sensors (transponders) [30].
frequencies above ∼105 Hz in order to get sufficiently good Cloud computing is also playing a role in the development
signal-to-noise ratios. of wireless sensors. The Cloud collects and stores sensor data
ICS have been used for more than 60 years in oceanographic corresponding to the measured values. The Cloud includes data
applications [146]–[148]. Commercially, these types of sensors processing, and may perform predictive analysis, search for
are available from a large number of companies for various trends, and data correctness assessment [155]. For multisens-
industrial applications [149]. Conductive ranges of ICS are ing including EC sensing, the data recorded in the Cloud may
from 5 mS·m−1 to 200 S·m−1 , in the high conductivity range. be processed using multivariate techniques, artificial neural
At low conductivities, where electrode fouling is less prob- networks, and machines learning, often employed for analysis
lematic, contact-electrode sensors are preferred. The designs of water quality data [152], [154]. Challenges include privacy
of ICS are mostly based on toroid configurations. Within (not all data can be accessed freely), security (improper
applications for ground water monitoring, Parra et al. [143] access, use, or malfunctioning), and integrity (incorrect or

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

manipulated data) [153], [155]. In addition, there is a lack approach of measuring the electrical properties from two or
of standardization, which is a major challenge for sensor more electrodes and fitting the data by computer models seems
calibration. to have reached its limits. In order to obtain deeper insight into
Commercially, wireless conductivity sensors run on the physical and chemical processes, and reactions involved in
batteries are available with a typical Bluetooth range between EC, there is a need for the development of new measurement
sensor and readout unit of 10 m [156]. Integration decreases techniques. To date, there are only few complementary data
wiring/cables, which means a reduction of stray capacitances from other types of measurements, which could support further
and inductances. Presently, calibration of such sensors is chal- development of EC theory; these include spatial distribution of
lenging according to standards. In ASTM D1125-14 [96] and charges and electrical fields, adsorption/desorption at electrode
D5391-14 [97] (see Section IV-A), a conductivity instrument surfaces, dynamics of ion migration etc. As the level of system
should be calibrated using standard resistors in place of the EC integration increases and EC sensors become smaller, the
cell and temperature compensator. However, with integrated solution volumes also become smaller, the surface to volume
transducer electronics, there is no separation between the EC ratio shrinks, and solution surfaces and interfaces become
cell and the readout unit. K cell cannot readily be derived: more important.
ideally, K cell is related to the EC cell and it should be The thickness of the EDL is usually too small to be
independent of the readout electronics. By introducing wireless measured. Methods to clarify such local electric fields in
readout from a monolithic chip, this separation is no longer solutions have not yet been fully established [47], [158].
possible, and calibration can only be made at a system level, Doi et al. [47] have recently demonstrated measurements of
making it more difficult to identify any variation and drift electric fields in solutions locally in a channel using glass
in the measured conductivity. Integrated sensors are simply microelectrodes with 1-μm-diameter tips and Ag/AgCl wire
difficult to calibrate using both solutions for the sensor and electrodes. The authors achieved good agreement between
resistors for the readout/analyzer—two separate processes that experimental electrical potential distribution data and a numer-
are required by the current ASTM standards [96], [97]. ical model based on the PNP equations (see Section II). Other
As conductivity sensors are miniaturized and other designs, ways of gaining deeper insight into EC may include optical
such as planar devices, become more appropriate, simple methods with nanometer scale spatial resolution for imaging
analytical solutions for the determination of K cell can no electrochemical reactions and processes as recently reviewed
longer be obtained, and finite-element methods [134], [135] by Wang et al. [159]. Another possibility could be the use of
or others [130] are needed. Values of K cell can be obtained surface plasmons excited at one electrode of an EC cell and
through calibration with appropriate transfer fluids or CRMs monitoring the surface plasmon resonance as a function of
(where available), but with higher uncertainties. EC sensors time, when applying a voltage or current pulse to the electrodes
have the potential of being fabricated with high precision and as we recently proposed [67].
in large numbers using MEMS-based technologies. Rather For an EC cell with dimensions in the nanometer and
than employing laborious solution calibrations, it is more micrometer range, the EDL thickness then becomes compa-
attractive to use traceable geometric measurements such as rable to the geometric dimensions of the cell. In addition,
scatterometry [157], which is a fast, precise, and low-cost polarization effects, electric field fringe effects, and effects
method to determine geometrical parameters with lateral res- of contaminants become more important. Novel characteriza-
olution of a few nanometers of periodic structures includ- tion techniques of the sensors and more advanced computer
ing multiple-electrode interdigitated conductivity sensors models are needed. In the case of small electrolyte volumes
(see Fig. 11(a) and (b)). utilized in microfluidic-based sensors, even a small amount
When calibrating an EC sensor, it is preferable that the of contamination causes a proportionate larger effect on the
metrological traceability can be achieved from geometrically conductivity. If the size becomes smaller than the width of the
measured reference cells with K cell based on a simple ana- EDL, the EDLs at each electrode will overlap, and the apparent
lytical expression as inferred by (13), (15), (16), and (18). EDL will start to disappear [158]. In this case, the EC cell will
Miniature sensors with calculable K cell have not yet been be dominated by the nanointerfacial space, and it will enter
realized. The accuracy of geometric measurements at the nm into a different regime of energy and mass transport [160].
and μm scales, usually performed by atomic force or confocal This would require a different formulation of the metrolog-
interference microscopy, are of the order of 0.1%, which is ical definition of EC, ideally based on quantum metrology,
almost an order of magnitude less accurate than a coordinate expected to improve the quality of sensor measurements [161].
measurement machine (CMM). CMMs are normally used for
measuring cm scale macroscopic geometrically measured con-
VI. C ONCLUSION
ductivity cells. However, for some applications, lower accuracy
of miniaturized cells may be sufficient. In this article, we have summarized the various models
Conductivity is a macro parameter, and acquiring data solely reported in the literature for EC cells, which are used to under-
from EC measurements makes it difficult to develop unam- stand and predict the behavior of the majority of EC sensors.
biguous models, whether they are based on PNP equations Computer models include solving the PNP equations with
or equivalent circuit diagrams. A lot of work has been done modifications related to the finite size of electrolyte ions and
toward developing such models; there is also a large amount electrode properties. Alternative models are based on equiv-
of data available from impedance spectroscopy. However, the alent circuit diagrams. There are a large number of different

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

TABLE IV
S UMMARY OF S ENSORS C ONFIGURATIONS , I NCLUDING THE N UMBER OF E LECTRODES , T YPICAL M EASUREMENT S CHEMES ( AC , DC , AND T IME
V ERSUS F REQUENCY D OMAINS ), M EANS OF T RACEABILITY, E XPECTED L EVEL OF A CCURACY, A PPLICABLE C ONDUCTIVITY R ANGE , AND S OME
E XAMPLES OF A PPLICATIONS

models reported in the literature, and good agreement can be effects contribute to the uncertainty, the most prominent being
obtained between solutions to the PNP equations, equivalent temperature. The advent of wireless conductivity sensors has
diagrams, and experimental data. However, since most data added a new challenge to calibration procedures based on
are only relationship between macroscale voltage–current data, existing standards.
different models may fit experimental data equally well. This There are a large number of different configurations for EC
applies to the experimental data in the literature as reported in sensors with two- or four electrode-based measurements and
the frequency domain and in the time domain. There is a need with electrode designs ranging from parallel plates, coaxial
for complementary and new measurement method in order to cylindrical, van der Pauw, and planar designs. Contactless
deepen the insight into the physical and chemical processes sensors either based on inductively coupling or capacitive
involved in EC. coupling are also available, and these have the advantage
There are a number of standard methods used in calibration of no fouling of the electrodes. Sensor geometries, typically
of EC sensors comprising an EC cell and a readout unit. These employed measurement schemes, and some examples of appli-
methods are based on either calibration against a conductivity cations in which these sensors are employed are summarized
sensor with known cell constant, typically in a closed-flow- in Table IV. Recent trends in EC sensors are toward miniature,
loop, or using CRMs with an assigned conductivity value. wireless sensors integrated with other sensors and the IoT,
At low conductivities, the first option is normally used, and with data collection, storage and processing in the Cloud.
because the high solubility of CO2 in aqueous solutions makes This puts new challenges to metrological traceability and to
it difficult to produce stable CRMs at low conductivities. the formulation of standards for calibration methods tailored
However, UPW in a closed-flow loop has been suggested for to the new types of EC sensors.
potential use as a standard at ∼5.5 μS·m−1 at 25 ◦ C, although Of interest to manufacturers, recent trends within research
this is not traceable to the SI. and development of EC systems have included miniaturization
In order to be able to trust the readings of an EC sensor, of sensors, primarily of planar geometry, and smart (versus
as well as comply with international regulations, it is necessary analog) sensors. Small, planar electrodes allow for use in
to know the route of metrological traceability, including the numerous applications, including measurement in applica-
use of an uncertainty budget. A number of challenges and tion where space and volume are at a premium. Difficulties

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

associated with this geometry of sensors include the non- [17] R. B. McCleskey, D. K. Nordstrom, J. N. Ryan, and J. W. Ball,
straightforward determination of cell constant from geometric “A new method of calculating electrical conductivity with applications to
natural waters,” Geochimica Cosmochimica Acta, vol. 77, pp. 369–382,
parameters, requiring the use of CRMs for calibration. For Jan. 2012.
EC systems intended for low conductivity applications (e.g., [18] S. Seitz, A. Manzin, H. D. Jensen, P. T. Jakobsen, and P. Spitzer, “Trace-
<100 μS · m−1 ), calibration must be performed in a flow-loop ability of electrolytic conductivity measurements to the international
system of units in the sub mS m−1 region and review of models of
configuration, as CRMs are not available. While smart sensors, electrolytic conductivity cells,” Electrochimica Acta, vol. 55, no. 22,
which incorporate the EC meter and EC electrode into one pp. 6323–6331, Sep. 2010.
unit, increase the ease of use for industrial applications, this [19] K.-H. Lubert and K. Kalcher, “History of electroanalytical methods,”
Electroanalysis, vol. 22, nos. 17–18, pp. 1937–1946, Sep. 2010.
combination does complicate calibration. Metrologically and
[20] Y. C. Wu, W. F. Koch, W. J. Hamer, and R. L. Kay, “Review of
to comply with ASTM standards [96], [97], the EC electrode electrolytic conductance standards,” J. Solution Chem., vol. 16, no. 12,
(or EC system) should be calibrated with liquid standards [96] pp. 985–997, Dec. 1987.
or in a flow-loop [97], while the EC meter should be separately [21] M. J. Francis, “Effect of degassing on the electrical conductivity of pure
water and potassium chloride solutions,” J. Phys. Chem. C, vol. 112,
calibrated against reference resistor(s). Inclusion of access no. 37, pp. 14563–14569, Sep. 2008.
and functionality allow calibration laboratories or industrial [22] R. Pawlowicz, R. Feistel, T. J. McDougall, P. Ridout, S. Seitz, and
metrology departments to perform these two calibration tasks H. Wolf, “Metrological challenges for measurements of key climatolog-
ical observables Part 2: Oceanic salinity,” Metrologia, vol. 53, no. 1,
separately on smart EC sensors would be an advantageous pp. R12–R25, Feb. 2016.
development. [23] S. Seitz, R. Feistel, D. G. Wright, S. Weinreben, P. Spitzer, and
P. De Bièvre, “Metrological traceability of oceanographic salinity mea-
surement results,” Ocean Sci., vol. 7, no. 1, pp. 45–62, Jan. 2011.
R EFERENCES [24] A. Kaidarova, M. Marengo, N. R. Geraldi, C. M. Duarte, and
J. Kosel, “Flexible conductivity, temperature and depth sensor for marine
[1] G. Reade, G. Ottewill, and F. Walsh, “Understanding electrical and elec- environmental monitoring,” in Proc. IEEE SENSORS, Montreal, QC,
trolytic conductivity,” Trans. IMF, vol. 78, no. 2, pp. 89–92, Jan. 2000. Canada, Oct. 2009, pp. 1–4.
[2] R. Ahmed and K. Reifsnider, “Study of influence of electrode geom- [25] C. Wu, W. Gao, J. Zou, Q. Jin, and J. Jian, “Design and batch microfabri-
etry on impedance spectroscopy,” Int. J. Electrochem. Sci., vol. 6, cation of a high precision conductivity and temperature sensor for marine
pp. 1159–1174, Apr. 2011. measurement,” IEEE Sensors J., vol. 20, no. 17, pp. 10179–10186,
[3] S. Seitz, P. T. Jakobsen, and M. Mariassy, “Metrological advances Sep. 2020.
in reference measurement procedures for electrolytic conductivity,” [26] Z. Wu, J. Wang, C. Bian, J. Tong, and S. Xia, “A MEMS-based multi-
Metrologia, vol. 56, no. 3, Jun. 2019, Art. no. 034003. parameter integrated chip and its portable system for water quality
[4] R. Buchner and G. Hefter, “Interactions and dynamics in electrolyte detection,” Micromachines, vol. 11, no. 1, p. 63, Jan. 2020.
solutions by dielectric spectroscopy,” Phys. Chem. Chem. Phys., vol. 11, [27] S. Vignesh, D. Surya, R. PraveenRaj, and M. Somu, “Water distrib-
no. 40, pp. 8984–8999, 2009. ution System,” Int. J. Res. Adv. Develop., vol. 4, no. 1, pp. 57–61,
[5] V. G. Artemov, A. A. Volkov, N. N. Sysoev, and A. A. Volkov, 2020.
“Conductivity of aqueous HCl, NaOH and NaCl solutions: Is water just [28] T. Asakai, I. Maksimov, S. Onuma, T. Suzuki, T. Miura, and A. Hioki,
a substrate?” Europhys. Lett., vol. 109, no. 2, p. 26002, Jan. 2015. “New japanese certified reference materials for electrolytic conduc-
[6] R. Gulich, M. Köhler, P. Lunkenheimer, and A. Loidl, “Dielectric spec- tivity measurements,” Accreditation Qual. Assurance, vol. 22, no. 2,
troscopy on aqueous electrolytic solutions,” Radiat. Environ. Biophys., pp. 73–81, Apr. 2017.
vol. 48, no. 1, pp. 107–114, Feb. 2009. [29] E. Orrù, F. Durbiano, and M. Ortolano, “Reference measurement system
[7] A. de Diego, “Application of the electrical conductivity of concen- for low electrolytic conductivity values with a flowing solution,” Meas.
trated electrolyte solutions to industrial process control and design: Sci. Technol., vol. 24, no. 3, Mar. 2013, Art. no. 035903.
From experimental measurement towards prediction through modelling,”
[30] M. D. Steinberg, B. Tkalčec, and I. M. Steinberg, “Towards a passive
Trends Anal. Chem., vol. 20, no. 2, pp. 65–78, Feb. 2001.
contactless sensor for monitoring resistivity in porous materials,” Sens.
[8] L. Coury, “Conductance measurements Part 1: Theory,” Current Sepa- Actuators B, Chem., vol. 234, pp. 294–299, Oct. 2016.
rat., vol. 18, no. 3, pp. 91–96, 1999.
[31] A. C. Bevilacqua, “Ultrapure water—The standard for resistivity mea-
[9] J. Plumridge, H. Arcis, and P. R. Tremaine, “Limiting conductivities of
surements of ultrapure water,” in Proc. 17th Annu. Semiconductor Pure
univalent cations and the chloride ion in H2 O and D2 O under hydrother-
Water Chem. Conf., Santa Clara, CA, USA, 1998, pp. 1–25.
mal conditions,” J. Solution Chem., vol. 44, no. 5, pp. 1062–1089,
May 2015. [32] J. Barron and C. Ashton, “Stable low level conductivity standards,” Intl.
J. Metrol., vol. 12, no. 1, pp. 1–8, 2005.
[10] S. Hipple, “High temperature conductivity probe for monitoring contam-
ination levels in power plant boiler water,” M.S. thesis, Graduate School [33] D. Spencer and H. Morgan, “High-speed single-cell dielectric spec-
Energy Geo-Environ. Eng., Pennsylvania State Univ., Philadelphia, PA, troscopy,” ACS Sensors, vol. 5, no. 2, pp. 423–430, Feb. 2020.
USA, 2008. [34] T. Petitclerc, “Recent developments in conductivity monitoring of
[11] T. S. Light, S. Licht, A. C. Bevilacqua, and K. R. Morash, “The funda- haemodialysis session,” Nephrol. Dialysis Transplantation, vol. 14,
mental conductivity and resistivity of water,” Electrochem. Solid-State no. 11, pp. 2607–2613, Nov. 1999.
Lett., vol. 8, no. 1, p. E16, 2005. [35] M. Berger, F. Sellering, H. Röhrich, H. Mansour,
[12] P. B. Ishai, M. S. Talary, A. Caduff, E. Levy, and Y. Feldman, “Electrode T. Perl, and S. Zimmermann, “A differential transformer for noninvasive
polarization in dielectric measurements: A review,” Meas. Sci. Technol., continuous sodium monitoring during dialysis treatment,” in Proc. IEEE
vol. 24, no. 10, Oct. 2013, Art. no. 102001. SENSORS, Montreal, QC, Canada, Oct. 2019, pp. 1–4.
[13] L. F. Lima, A. L. Vieira, H. Mukai, C. M. G. Andrade, and [36] G. Liu et al., “A wearable conductivity sensor for wireless real-time
P. R. G. Fernandes, “Electric impedance of aqueous KCl and NaCl solu- sweat monitoring,” Sens. Actuators B, Chem, vol. B227, pp. 34–42,
tions: Salt concentration dependence on components of the equivalent May 2016.
electric circuit,” J. Mol. Liquids, vol. 241, pp. 530–539, Sep. 2017. [37] M. F. Mabrook and M. C. Petty, “Effect of composition on the electrical
[14] K. M. Queeney, “Principles and hardware of electrolytic conductivity conductance of milk,” J. Food Eng., vol. 60, no. 3, pp. 321–325,
measurements,” in Proc. ISA 48th Anal. Division Symp., Calgary, AB, Dec. 2003.
Canada, vol. 1, May 2003, pp. 1–10. [38] A. C. Bevilacqua, “The effect of temperature, temperature error, and
[15] V. M. Braga, “Conductivity sensor calibrations to meet water industry impurities on compensated resistivity measurements,” in Proc. 16th
requirements,” in Ultrapure Water. Singapore: Tall Oaks Publishing Inc., Annu. Semiconductor Pure Water Chem. Conf. (SWCC), Santa Clara,
Mar. 2004. CA, USA, 1997, pp. 131–160.
[16] P. Arquint, H. Bühler, and S. Worbs, “Traceable standards for electrolytic [39] Standard Test Methods for Electrical Conductivity of Avia-
conductivity measurements required for pharmaceutical production,” tion and Distillate Fuels, Standard D2525-95, ASTM Int.,
CHIMIA Int. J. Chem., vol. 63, no. 10, pp. 647–649, Oct. 2009. West Conshohocken, 2015.

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

[40] P. Spitzer, P. Fisicaro, S. Seitz, and R. Champion, “PH and electrolytic [63] T.-L. Horng, T.-C. Lin, C. Liu, and B. Eisenberg, “PNP equations with
conductivity as parameters to characterize bioethanol,” Accreditation steric effects: A model of ion flow through channels,” J. Phys. Chem.
Qual. Assurance, vol. 14, no. 12, pp. 671–676, Dec. 2009. B, vol. 116, no. 37, pp. 11422–11441, Sep. 2012.
[41] White Paper on Interntionally Compatible Biofuel Standards, Tripartite [64] H. Wang and L. Pilon, “Intrinsic limitations of impedance measurements
Task Force Comprising Brazil, B. T. T. Force, EU, Brussels, Belgium, in determinining electric double layer capacitances,” Electrochimica
2007. Acta, vol. 63, pp. 55–63, Feb. 2012.
[42] M. Park, X. Zhang, M. Chung, G. Less, and A. Sastry, “A review of [65] B. Roling and M. Drüschler, “Intrinsic limitations of impedance mea-
condiction phenomena in Li-ion batteries,” J. Power Sources, vol. 195, surements in determining electric double layer capacitances,” Elec-
no. 24, pp. 7904–7929, 2020. trochimica Acta, vol. 76, pp. 526–528, Aug. 2012.
[43] Z. Moroń, “Considerations on the accuracy of measurements of elec- [66] H. Wang and L. Pilon, “Instrinsic limitations of impedance measure-
trical conductivity of liquids,” in Proc. 18th IMEKO World Congr., ments in determining electric double layer capacitance,” Electrochimica
Rio de Janeiro, Brazil, vol. 22, Sep. 2006, pp. 1–6. Acta, vol. 76, pp. 529–531, Feb. 2012.
[44] T. Asakai, Y. Yamauchi, and T. Suzuki, “High-throughput and precise [67] C. Thirstrup, L. C. Deleebeeck, and A. M. N. Lima, “Ion-specific quan-
measurement method for electrolytic conductivity in a higher conduc- titative measurement scheme using transit-time surface plasmon reso-
tivity range (10 S m−1 to 0.1 S m−1 ) by chromatography system with nance,” Meas. Sci. Technol., vol. 30, no. 10, Oct. 2019, Art. no. 105102.
conductivity detector,” Microchem. J., vol. 143, pp. 312–318, Dec. 2018. [68] I. Lelidis, J. R. Macdonald, and G. Barbero, “Poisson–Nernst–Planck
[45] P. Mirtaheri, S. Grimnes, and O. G. Martinsen, “Electrode polarization model with Chang-Jaffe, diffusion, and ohmic boundary conditions,”
impedance in weak NaCl aqueous solutions,” IEEE Trans. Biomed. Eng., J. Phys. D, Appl. Phys., vol. 49, no. 2, Jan. 2016, Art. no. 025503.
vol. 52, no. 12, pp. 2093–2099, Dec. 2005. [69] G. Barbero, M. Becchi, and F. C. M. Freire, “Contribution of the
[46] T. Sonsa-Ard, N. Chantipmanee, N. Fukana, P. Hauser, P. Wilairat, and electrode-electrolyte interface to the impedance of an electrolytic cell,”
D. Nacapricha, “Contactless conductivity sensor employing moist paper J. Appl. Phys., vol. 104, no. 11, Dec. 2008, Art. no. 114111.
as absorbent for in-situ detection of generated carbon dioxide gas,” Anal. [70] A. R. Duarte, F. Batalioto, G. Barbero, and A. M. F. Neto, “Measurement
Chimica Acta, vol. 1118, pp. 45–51, Jun. 2020. of the impedance of aqueous solutions of KCl: An analysis using
[47] K. Doi, N. Asano, and S. Kawano, “Development of glass micro- an extension of the Poisson-Nernst-Planck model,” Appl. Phys. Lett.,
electrodes for local electric field, electrical conductivity, and pH mea- vol. 105, no. 2, Jul. 2014, Art. no. 022901.
surements,” Sci. Rep., vol. 10, no. 1, p. 4110, Dec. 2020. [71] Y. C. Wu, W. F. Koch, D. Feng, and L. A. Holland, “A DC method
[48] P. Gurusamy, U. Gandhi, U. Mangalanathan, and K. M. Lr, “Electrical for the absolute determination of conductivities of the primary standard
conductivity based quality measurement of diesel exhaust fluid,” J. Anal. KCl solutions from 0 ◦ C to 50 ◦ C,” J. Res. Nat. Inst. Stand. Technol.,
Chem., vol. 73, no. 11, pp. 1128–1135, Nov. 2018. vol. 99, pp. 241–246, May 1994.
[49] R. Robinson and R. Stokes, Electrolyte Solutions, 2nd ed. New York, [72] K. Pratt, W. Koch, Y. Wu, and P. Berezansky, “Molality-based primary
NY, USA: Dover, 2002. standards of electrolytic conductivity,” Pure Appl. Chem., vol. 73,
[50] J. Macdonald and W. Johnson, “Fundamentals of impedance spec- pp. 1783–1793, Nov. 2001.
troscopy,” in Impedance Spectroscopy, E. Barsoukov, J. R. Macdonald, [73] F. Brinkmann et al., “Primary methods for the measurement of elec-
Eds. New York, NY, USA: Wiley, 2005, pp. 1–26. trolytic conductivity,” Accreditation Qual. Assurance, vol. 8, nos. 7–8,
[51] F. Grasso, F. Musumeci, and A. Triglia, “Impedance spectroscopy of pp. 346–353, Jul. 2003.
pure water in the 0.01 Hz to 100 kHz range,” Il Nuovo Cimento D, [74] L. Deleebeeck and S. Veltzé, “Electrochemical impedance spec-
vol. 12, no. 8, pp. 1117–1130, Aug. 1990. troscopy study of commercial Li-ion phosphate batteries: A metrol-
[52] L. A. Geddes, “Historical evolution of circuit models for the electrode- ogy perspective,” Int. J. Energy Res., vol. 44, no. 9, pp. 7158–7182,
electrolyte interface,” Ann. Biomed. Eng., vol. 25, no. 1, pp. 1–14, Jul. 2020.
Jan. 1997. [75] H. Sanabria and J. H. Miller, Jr., “Relaxation processes due to the
[53] J. Lario-García and R. Pallàs-Areny, “Constant-phase element identifica- electrode-electrolyte interface in ionic solutions,” Phys. Rev. E, Stat.
tion in conductivity sensors using a single square wave,” Sens. Actuators Phys. Plasmas Fluids Relat. Interdiscip. Top., vol. 74, no. 5, Nov. 2006,
A, Phys., vol. 132, no. 1, pp. 122–128, Nov. 2006. Art. no. 051505.
[54] P. Córdoba-Torres, T. J. Mesquita, O. Devos, B. Tribollet, V. Roche, and [76] C. Thirstrup, A. Snedden, and L. C. Deleebeeck, “Addressing the
R. P. Nogueira, “On the intrinsic coupling between constant-phase ele- challenges of traceable electrolytic conductivity measurements in water,”
ment parameters α and Q in electrochemical impedance spectroscopy,” Meas. Sci. Technol., vol. 28, Nov. 2017, Art. no. 124001.
Electrochimica Acta, vol. 72, pp. 172–178, Jun. 2012. [77] I. D. Raistrick, D. R. Franceschetti, and J. R. Macdonald, “Theory,” in
[55] A. Sawada, “Introduction of effective dielectric constant to the Poisson- Impedance Spectroscopy, Hoboken, NJ, USA: Wiley, 2005, pp. 27–128.
Nernst-Planck model,” Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat. [78] O. Bottauscio, P. P. Capra, F. Durbiano, and A. Manzin, “Modeling of
Interdiscip. Top., vol. 93, no. 5, May 2016, Art. no. 052608. cells for electrolytic conductivity measurements,” IEEE Trans. Magn.,
[56] H. Wang, A. Thiele, and L. Pilon, “Simulations of cyclic voltammetry for vol. 42, no. 4, pp. 1423–1426, Apr. 2006.
electric double layers in asymmetric electrolytes: A generalized modified [79] G. H. Zimmerman and H. Arcis, “Extrapolation methods for AC
Poisson–Nernst–Planck model,” J. Phys. Chem. C, vol. 117, no. 36, impedance measurements made with a concentric cylinder cell on
pp. 18286–18297, Sep. 2013. solutions of high ionic strength,” J. Solution Chem., vol. 44, no. 5,
[57] M. S. Kilic, M. Z. Bazant, and A. Ajdari, “Steric effects in the pp. 912–933, May 2015.
dynamics of electrolytes at large applied voltages. II. Modified [80] J. A. Giacometti, N. Alves, and M. Y. Teruya, “Impedance of aqueous
Poisson-Nernst-Planck equations,” Phys. Rev. E, Stat. Phys. Plas- solutions of KCl at the ultra-low frequency range: Use of cole-cole
mas Fluids Relat. Interdiscip. Top., vol. 75, no. 2, Feb. 2007, impedance element to account for the frequency dispersion peak at
Art. no. 021503. 20 mHz,” Brazilian J. Phys., vol. 46, no. 1, pp. 50–55, Feb. 2016.
[58] A. Flavell, M. Machen, B. Eisenberg, and X. Li, “A conservative finite [81] S. Xiaoping, P. Spitzer, and U. Sudmeier, “Novel method for bulk
difference scheme for Poisson–Nernst–Planck equations,” J. Comput. resistance evaluation in conductivity measurement for high-purity
Electron., vol. 15, pp. 1–15, Mar. 2013. water,” Accreditation Qual. Assurance, vol. 12, no. 7, pp. 351–355,
[59] J. R. Macdonald, “Three to six ambiguities in immittance spectroscopy Jul. 2007.
data fitting,” J. Phys., Condens. Matter, vol. 24, no. 17, May 2012, [82] C. Thirstrup, A. Snedden, and H. D. Jensen, “Bridging the gap between
Art. no. 175004. ultrapure water and reference materials in electrolytic conductivity mea-
[60] E. K. Lenzi, J. L. de Paula, F. R. G. B. Silva, and L. R. Evangelista, surements,” IEEE Trans. Instrum. Meas., vol. 66, no. 10, pp. 2745–2754,
“A connection between anomalous Poisson–Nernst–Planck model and Jul. 2017.
equivalent circuits with constant phase elements,” J. Phys. Chem. C, [83] J. R. Macdonald, L. R. Evangelista, E. K. Lenzi, and G. Barbero,
vol. 117, no. 45, pp. 23685–23690, Nov. 2013. “Comparison of impedance spectroscopy expressions and responses
[61] E. K. Lenzi, H. V. Ribeiro, R. S. Zola, and L. R. Evangelista, of alternate anomalous Poisson–Nernst–Planck diffusion equations
“Fractional calculus in electrical impedance spectroscopy: Poisson– for finite-length situations,” J. Phys. Chem. C, vol. 115, no. 15,
Nernst–Planck model and extensions,” Int. J. Electrochem. Sci., vol. 12, pp. 7648–7655, Apr. 2011.
pp. 11677–11691, Dec. 2017. [84] M. Becchi, L. Callegaro, F. Durbiano, V. D’Elia, and A. Strigazzi,
[62] B. Eisenberg, Y. Hyon, and C. Liu, “A mathematical model for the hard “Novel impedance cell for low conductive liquids: Determination of
sphere repulsion in ionic solutions,” Commun. Math. Sci., vol. 9, no. 2, bulk and interface contributions,” Rev. Sci. Instrum., vol. 78, no. 11,
pp. 459–475, 2011. Nov. 2007, Art. no. 113902.

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
THIRSTRUP AND DELEEBEECK: REVIEW ON EC SENSORS 1008222

[85] H. D. Jensen and C. Thirstrup, “Direct traceability for ultra-pure water [109] L. Deleebeeck, J. Avnskjold, and A. Snedden, “Short- and long-
conductivity,” NCSLI Measure, vol. 9, no. 1, pp. 68–72, Mar. 2014. term stability of electrolytic conductivity certified reference mate-
[86] R. H. Shreiner and K. W. Pratt, “Standard reference materials: Primary rials,” Accreditation Qual. Assurance, vol. 25, no. 2, pp. 127–138,
standards and standard reference materials for electrolytic conductivity,” Apr. 2020.
NIST Special Publication, vols. 142–260, pp. 1–26, May 2004. [110] G. Lazar, C. Campureanu, I. Cirneanu, and D. I. Vaireanu, “Good
[87] G. A. Snook, K. McGregor, A. J. Urban, and M. A. Cooksey, “Fast practices in evaluating the uncertainty of measurements for the con-
Fourier transform current pulse method for dynamic measurements of ductivity of the electrolyte solutions,” Revista Chimie, vol. 68, no. 11,
cell ohmic resistance during electrolysis,” Electrochimica Acta, vol. 54, pp. 2482–2487, Dec. 2017.
no. 21, pp. 4925–4932, Aug. 2009. [111] Y. C. Wu and W. F. Koch, “Absolute determination of electrolytic
[88] S. Mayer, L. A. Geddes, J. D. Bourland, and L. Ogborn, “Faradic resis- conductivity for primary standard KCl solutions from 0 to 50◦ C,”
tance of the electrode/electrolyte interface,” Med. Biol. Eng. Comput., J. Solution Chem., vol. 20, no. 4, pp. 391–401, Apr. 1991.
vol. 30, no. 5, pp. 538–542, Sep. 1992. [112] U. Breuel, B. Werner, and D. Jehnert, “Metrology in chemistry for pH
[89] J. Wu and J. P. W. Stark, “A low-cost approach for measuring electrical and electrolytic conductivity traceability dissemination,” CHIMIA Int.
conductivity and relative permittivity of liquids by triangular wave- J. Chem., vol. 63, no. 10, pp. 643–646, Oct. 2009.
form voltage at low frequencies,” Meas. Sci. Technol., vol. 16, no. 5, [113] N. Hamim, F. S. H. Krimastuti, A. Hindayani, and Y. Aristiawan,
pp. 1234–1240, May 2005. “Certified reference materials for calibration of conductivity meter at
[90] B. Carkhuff and R. Cain, “Corrosion sensors for concrete the measuring of electrolytic conductivity in water: Preparation and its
bridges,” IEEE Instrum. Meas. Mag., vol. 6, no. 2, pp. 19–24, measurement,” AIP Conf. Proc., vol. 2175, Nov. 2019, Art. no. 020065.
Jun. 2003. [114] L. Callegaro, F. Durbiano, E. Orru, and B. Trinchera, “An impedance
[91] P. H. Daum and D. F. Nelson, “Bipolar current method for determination spectrometer for the metrology of electrolytic conductivity,” IEEE Trans.
of solution resistance,” Anal. Chem., vol. 45, no. 3, pp. 463–470, Instrum. Meas., vol. 62, no. 6, pp. 1766–1770, Jun. 2013.
Mar. 1973. [115] R. H. Shreiner, “Preparation and uncertainty calculations for
[92] A. Meijía-Aguilar and R. Pallàs-Areny, “Electrical impedance mea- the molality-based primary standards for electrolytic conductivity,”
surement using voltage/current pulse excitation,” in Proc. 19th IMEKO American Lab., Nat. Inst. Standards Technol., Gaithersburg, MD, USA,
World Congr., Fundam. Appl. Metrol., Lisbon, Portugal, Sep. 2009, Tech. Rep, Feb. 2004, vol. 36, no. 3, pp. 28–32.
pp. 662–667. [116] F. Batalioto, O. G. Martins, A. R. Duarte, and A. M. F. Neto,
[93] Y. Feldman, E. Polygalov, I. Ermolina, Y. Polevaya, and B. Tsentsiper, “Erratum to: The influence of adsorption phenomena on the impedance
“Electrode polarization correction in time domain dielectric spec- spectroscopy of an electrolytic cell,” Eur. Phys. J. E, vol. 34, no. 2,
troscopy,” Meas. Sci. Technol., vol. 12, no. 8, pp. 1355–1364, Aug. 2001. p. 10, Feb. 2011.
[94] Water Quality—Determination of Electrical Conductivity, [117] J. S. Maràz, F. Fernández, and L. Magaña, “Electrolytic resistivity
Standard 7888, ISO, Geneva, Switzerland, 1985. measurement using alternating current with high frequency,” Revista
[95] Expression of Performance of Electrochemcal Analyzers—Part 3: Mexicana Física, vol. 55, no. 2, pp. 149–152, 2009.
Electrolytic Conductivity, Standard IEC 60746-3, 2nd ed. Geneva, [118] Working Group 1 of the Joint Committee for Guides in Metrology,
Switzerland, International Electrotechnical Commission, 2002. “Evaluation of measurement data—Guide to the expression of uncer-
[96] Standard Test Methods for Electrical Conductivity and Resitivity of tainty in measurement,” Joint Commitee Guides Metrol., Bur. Int. des
Water, Standard D1125-14, ASTM International, West Conshohocken, Poids et Mesures, Sèvres, France, Tech. Rep. 100:2008, 1995, vol. 100.
PA, USA, 2014. [119] G. J. A. M. Brom-Verheijden, M. H. Goedbloed, and
[97] Standard Test Method for Electrical Conductivity and Resistivity of M. A. G. Zevenbergen, “A microfabricated 4-electrode conductivity
a Flowing High Purity Water Sample, Standard D5391-14, ASTM sensor with enhanced range,” Multidisciplinary Digit. Publishing Inst.
International, West Conshohocken, PA, USA, 2014. Proc., vol. 2, no. 13, p. 797, Dec. 2018.
[98] USP <644> Conductivity of Solutions & USP <645> Water Conduc- [120] B. A. Mazzeo and A. J. Flewitt, “Two- and four-electrode, wide-
tivity, United States Pharmacopeial Conv., Rockwille, MA, USA, 2019. bandwidth, dielectric spectrometer for conductive liquids: Theory, lim-
[99] Water for Injections, European Pharmacopeia 9.1, Eur. Directorate Qual. itations, and experiment,” J. Appl. Phys., vol. 102, no. 10, Nov. 2007,
Medicines Healthcare, Strasbourg, France, 2017. Art. no. 104106.
[100] H. C. Parker and E. W. Parker, “The calibration of cells for conductance [121] S. Grimnes and Ø. G. Martinsen, “Sources of error in tetrapolar
measurements. III. Absolute measurements on specific conductance of impedance measurements on biomaterials and other ionic conductors,”
certain potassium chloride solutions,” J. Amer. Chem. Soc., vol. 46, no. 2, J. Phys. D, Appl. Phys., vol. 40, no. 1, pp. 9–14, Jan. 2007.
pp. 312–335, Feb. 1924. [122] H. Ma, Y. Su, and A. Nathan, “Cell constant studies of bipolar
[101] M. Máriássy, K. W. Pratt, and P. Spitzer, “Major applications of and tetrapolar electrode systems for impedance measurement,” Sens.
electrochemical techniques at national metrology institutes,” Metrologia, Actuators B, Chem., vol. 221, pp. 1264–1270, Dec. 2015.
vol. 46, no. 3, pp. 199–213, Jun. 2009. [123] G. Jones and S. M. Christian, “The measurement of the conductance of
[102] V. Gavrylkin, A. Glukhenkly, and A. Mikhal, “An analysis of the error electrolytes. VI. Galvanic polarization by alternating current,” J. Amer.
when determining the constant of the primary standard conductometric Chem. Soc., vol. 57, no. 2, pp. 272–280, Feb. 1935.
cell,” Meas. Techn., vol. 56, pp. 935–940, Nov. 2013. [124] H. Golnabi, “Measurements of the electrical parameters for differ-
[103] I. C. S. Fraga, J. C. Lopes, L. R. Cordeiro, L. F. da Silva, and ent water samples,” Measurement, vol. 44, no. 10, pp. 2175–2184,
P. P. Borges, “Evaluation of the stability of solutions of low electrolytic Dec. 2011.
conductivity by primary measurements,” J. Solution Chem., vol. 44, [125] D. He, M. A. Shannon, and N. R. Miller, “Micromachined silicon
no. 9, pp. 1920–1936, Sep. 2015. electrolytic conductivity probes with integrated temperature sensor,”
[104] P. Spitzer and S. Seitz, “Electrolytic conductivity,” in Handbook of IEEE Sensors J., vol. 5, no. 6, pp. 1185–1196, Dec. 2005.
Metrology and Testing, H. Czichos, T. Saito, and L. Smith, Eds. [126] J. Hubálek, “Iterative precise conductivity measurement with IDEs,”
Heidelberg, Germany: Springer, 2011, pp. 498–507. Sensors, vol. 15, no. 5, pp. 12080–12091, May 2015.
[105] K. D. C. Cunha, L. D. S. Pardellas, and F. B. Gonzaga, “Stability moni- [127] W. Olthuis, W. Streekstra, and P. Bergveld, “Theoretical and experi-
toring of electrolytic conductivity reference materials under repeated use mental determination of cell constants of planar-interdigitated electrolyte
conditions by the primary measurement method,” J. Solution Chem., conductivity sensors,” Sens. Actuators B, Chem., vol. 24, nos. 1–3,
vol. 49, no. 3, pp. 306–315, Mar. 2020. pp. 252–256, Mar. 1995.
[106] Z. Moroń, “Investigations of van der Pauw method applied for mea- [128] B. Timmer, W. Sparreboom, W. Olthuis, P. Bergveld, and
suring electrical conductivity of electrolyte solutions,” Measurement, A. van den Berg, “Optimization of an electrolyte conductivity detec-
vol. 33, no. 3, pp. 281–290, Apr. 2003. tor for measuring low ion concentrations,” Lab Chip, vol. 2, no. 2,
[107] B. Zhang, Z. Lin, X. Zhang, X. Yu, J. Wei, and X. Wang, “System for pp. 121–124, May 2002.
absolute measurement of electrolytic conductivity in aqueous solutions [129] C. S. Rustomji et al., “Thin-film electrochemical sensor electrode
based on van der Pauw’s theory,” Meas. Sci. Technol., vol. 25, no. 5, for rapid evaluation of electrolytic conductivity, cyclic voltammetry,
May 2014, Art. no. 055005. and temperature measurements,” J. Appl. Electrochem., vol. 46, no. 1,
[108] E. Orrù, F. Durbiano, and M. Ortolano, “Development of an in-line pp. 59–67, Jan. 2016.
calibration system for flow-through cells for low electrolytic conductivity [130] F. T. Werner and R. N. Dean, “Characterising a PCB electrical conduc-
values,” Accreditation Qual. Assurance, vol. 19, no. 1, pp. 11–16, tivity sensor using electromagnetic simulation and a genetic algorithm,”
Feb. 2014. IET Sci., Meas. Technol., vol. 11, no. 6, pp. 761–765, Sep. 2017.

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.
1008222 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 70, 2021

[131] B. Zhang, Q. Cheng, Z. Bai, Q. Zhao, and Z. Yang, “An integrated [152] B. K. Jha, “Cloud-based smart water quality monitoring system using
miniature conductivity and temperature sensor chip with open-cell IoT sensors and machine learning,” Int. J. Adv. Trends Comput. Sci.
annular electrode,” in Proc. 20th Int. Conf. Solid-State Sensors, Actuat. Eng., vol. 9, no. 3, pp. 3403–3409, Jun. 2020.
Microsyst. Eurosensors (TRANSDUCERS EUROSENSORS XXXIII), [153] Y.-P. Lin et al., “Real-time identification of irrigation water pollution
Berlin, Germany, Jun. 2019, pp. 1341–1344. sources and pathways with a wireless sensor network and blockchain
[132] F. Segura-Quijano, J. Sacristán-Riquelme, J. García-Cantón, framework,” Sensors, vol. 20, no. 13, p. 3634, Jun. 2020.
M. T. Osés, and A. Baldi, “Towards fully integrated wireless [154] J. Park, K. T. Kim, and W. H. Lee, “Recent advances in information and
impedimetric sensors,” Sensors, vol. 10, no. 4, pp. 4071–4082, communications technology (ICT) and sensor technology for monitoring
Apr. 2010. water quality,” Water, vol. 12, no. 2, p. 510, Feb. 2020.
[133] Y.-L. Tan and C.-E. Chong, “Resistivity measurement of a small- [155] J. V. Capella, A. Bonastre, J. C. Campelo, R. Ors, and M. Peris, “IoT
volume sample using two planar disc electrodes and a new geometric & environmental analytical chemistry: Towards a profitable symbiosis,”
factor,” IEEE Sensors J., vol. 8, no. 5, pp. 516–521, May 2008. Trends Environ. Anal. Chem., vol. 27, Sep. 2020, Art. no. e00095.
[134] H. Cheng, Y. Zhang, X. Huang, J. A. Rogers, and Y. Huang, “Analysis [156] Free and Connected—IDS Goes Wireless (Product Brochure), WTW
of a concentric coplanar capacitor for epidermal hydration sensing,” Wissenschaftlich-Technische Werkstatten GmbH, Xylem, New York,
Sens. Actuators A, Phys., vol. 203, pp. 149–153, Dec. 2013. NY, USA, Mar. 2016.
[135] A. Hyldgård, D. Mortensen, K. Birkelund, O. Hansen, and [157] M. H. Madsen and P.-E. Hansen, “Scatterometry—Fast and robust mea-
E. V. Thomsen, “Autonomous multi-sensor micro-system for measure- surements of nano-textured surfaces,” Surf. Topogr., Metrol. Properties,
ment of ocean water salinity,” Sens. Actuators A, Phys., vol. 147, no. 2, vol. 4, Apr. 2016, Art. no. 023003.
pp. 474–484, Oct. 2008. [158] R. Hatsuki, F. Yujiro, and T. Yamamoto, “Direct measurement of elec-
[136] B. Kumar, G. Rajita, and N. Mandal, “A review on capacitive-type tric double layer in a nanochannel by electrical impedance spectroscopy,”
sensor for measurement of height of liquid level,” Meas. Control, vol. 47, Microfluidics Nanofluidics, vol. 14, no. 6, pp. 983–988, Jun. 2013.
no. 7, pp. 219–224, Sep. 2014. [159] Y. Wang, Z. Cao, Q. Yang, W. Guo, and B. Su, “Optical methods for
[137] P. Kubáň and P. C. Hauser, “20th anniversary of axial capacitively studying local electrochemical reactions with spatial resolution: A crit-
coupled contactless conductivity detection in capillary electrophoresis,” ical review,” Analytica Chimica Acta, vol. 1074, pp. 1–15, Oct. 2019.
Trends Anal. Chem., vol. 102, pp. 311–321, May 2018. [160] C. Sun, R. Zhou, Z. Zhao, and B. Bai, “Nanoconfined fluids: What
[138] X.-Y. Tang, J. Huang, H. Ji, B. Wang, and Z. Huang, “New contactless can we expect from them?” J. Phys. Chem. Lett., vol. 11, no. 12,
conductivity detection (CCD) sensor for fluid conductivity measure- pp. 4678–4692, Jun. 2020.
ment,” IEEE Sensors J., vol. 20, no. 19, pp. 11256–11264, Oct. 2020. [161] E. O. Göbel and U. Siegner, The new International System of Units
[139] H. Golnabi and M. Sharifian, “Investigation of water electrical parame- (SI): Quantum Metrology and Quantum Standards, Weinheim, Germany:
ters as a function of measurement frequency using cylindrical capacitive Wiley, 2019.
sensors,” Measurement, vol. 46, no. 1, pp. 305–314, Jan. 2013. [162] F. Kohlrausch, “Ueber die Gültigkeit der Ohm’schen Gesetze für Elek-
[140] J. Wahid and Q. Ahsan, “Detection of impurities in water by measuring trolyte,” Nachrichten von der Königl. Gesellschaft der Wissenschaften
capacitance,” in Proc. 8th Int. Conf. Electr. Comput. Eng., Dhaka, und der Georg-Augusts-Universität zu Göttingen, 1869, pp. 14–16.
Bangladesh, Dec. 2014, pp. 20–22.
[141] A. L. Ribeiro, H. M. G. Ramos, P. M. Ramos, and J. D. Pereira,
“Inductive conductivity cell for water salinity monitoring,” in Proc. 18th
IMEKO World Congr., Metrol. Sustain. Develop., Rio de Janeiro, Brazil, Carsten Thirstrup received the Ph.D. degree from
Sep. 2006, pp. 1–4. the Technical University of Denmark, Kongens
[142] D. Agranovich, E. Polygalov, I. Popov, P. B. Ishai, and Y. Feldman, Lyngby, Denmark, in 1991.
“Inductive dielectric analyzer,” Meas. Sci. Technol., vol. 28, no. 3, From 1990 to 1995, he was with the Micro-
Mar. 2017, Art. no. 035103. electronics Centre, Kongens Lyngby, where he was
[143] L. Parra, S. Sendra, J. Lloret, and I. Bosch, “Development of a involved in the field of photonic integrated circuits.
conductivity sensor for monitoring groundwater resources to optimize From 1995 to 1998, he was a Research Fellow
water management in smart city environments,” Sensors, vol. 15, no. 9, with Riken, Wako, Japan, within nanolithography.
pp. 20990–21015, Aug. 2015. In 1998, he co-founded Vir Biosensor, Taastrup,
[144] S. Wu et al., “Investigation of the performance of an inductive Denmark. From 2004 to 2012, he was with the
seawater conductivity sensor,” Sensors Transducers, vol. 186, pp. 43–48, Medical Device Company Coloplast, Humlebæk,
Mar. 2015. Denmark. He is currently with the Danish National Metrology Institute
[145] S. K. Hui, H. Jang, C. K. Gum, C. Y. Song, and H. K. Yong, (DFM A/S), Hørsholm, Denmark, where he is involved in the field of
“A new design of inductive conductivity sensor for measuring electrolyte electrochemistry. His research focuses on developing measurements systems
concentration in industrial field,” Sens. Actuators A, Phys., vol. 301, within electrolytic conductivity of chemical and biological systems.
Jan. 2020, Art. no. 111761.
[146] H. Hinkelmann, “Ein Verfahren zur elektrodenlosen Messung der
elektrischen Leitfähigkeit von Elektrolyten,” Z. Angew. Physik, vol. 10,
no. 11, pp. 500–503, 1958. Lisa Deleebeeck received the M.Sc. degree from
[147] K. Park and W. V. Burt, “Electrolytic conductance of sea water and the University of Calgary, Calgary, AB, Canada,
the salinometer: Part 2,” J. Oceanograph. Soc. Jpn., vol. 21, no. 3, in 2010, and the Ph.D. degree from the Technical
pp. 124–132, 1965. University of Denmark, Roskilde (Risø Campus),
[148] K. Striggow and R. Dankert, “The exact theory of inductive con- Denmark, in 2015.
ductivity sensors for oceanographic application,” IEEE J. Ocean. Eng., She has been employed as the Team Leader for
vol. OE-10, no. 2, pp. 175–179, Apr. 1985. Electrochemistry and Contact Thermometry with the
[149] Direct Industry. Inductive Conductivity Sensors. Virtual Danish National Metrology Institute (DFM A/S),
Expo Group. Accessed: Jul. 16, 2020. [Online]. Available: Hørsholm, Denmark, since 2016. She is the Danish
https://www.directindustry.com/industrial-manufacturer/inductive- Representative to EURAMET TC-MC (Chemistry &
conductivity-sensor-131915.html Biology) within the European Regional Metrology
[150] B. D. Majumder, J. K. Roy, and S. Padhee, “Recent advances in Organization (EURAMET e.V.), Braunschweig, Germany, and the Consulta-
multifunctional sensing technology on a perspective of multi-sensor tive Committee for Amount Substance: Metrology in Chemistry and Biology
system: A review,” IEEE Sensors J., vol. 19, no. 4, pp. 1204–1214, [CCQM, Bureau International des Poids et Mesures (BIPM)], Sèvres, France.
Feb. 2019. Her research focuses on the development of methods and reference materials
[151] L. Ortega, A. Llorella, J. P. Esquivel, and N. Sabaté, “A paper-based within electrochemical metrology areas of pH and electrolytic conductivity.
battery as conductivity sensor for single use applications,” ACS Sensor, Her work further includes calibrations within coulometry (amount of sub-
vol. 5, pp. 1743–1749, May 2020. stance) and contact thermometry instruments.

Authorized licensed use limited to: Univ of Calif Santa Barbara. Downloaded on July 01,2021 at 07:49:17 UTC from IEEE Xplore. Restrictions apply.