Subscriber access provided by University of Texas Libraries

Article
Electrochemical Impedance Spectroscopy of Ion-Selective Membranes:
Artifacts in Two-, Three-, and Four-Electrode Measurements
Evan L. Anderson, and Philippe Buhlmann
Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.6b02641 • Publication Date (Web): 09 Sep 2016
Downloaded from http://pubs.acs.org on September 9, 2016

Just Accepted

“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth

Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.
Page 1 of 9 Analytical Chemistry

1
2
3
4
5
6
7 Electrochemical Impedance Spectroscopy of Ion-Selective
8
9
Membranes: Artifacts in Two-, Three-, and Four-Electrode
10 Measurements
11
12 Evan L. Anderson and Philippe Bühlmann*
13
14 Department of Chemistry, University of Minnesota, 207 Pleasant Street South East, Minneapolis, Minnesota
15 55455, United States
16 +
electrochemical impedance spectroscopy (EIS), K -selective membranes, membrane impedance, ion-selective
17 electrodes (ISEs)
18
19 ABSTRACT: Electrochemical impedance spectroscopy is frequently used to characterize, optimize, and monitor ion-
20 selective membranes. However, because of the relatively high resistance of ion-selective membranes, their impedance
21 spectra often contain artifacts that can cause misinterpretation. While in the high-frequency range artifacts are often
22 readily identifiable by the occurrence of inductive features or negative resistances, artifacts are easy to overlook in the
23 low-frequency range, where telltale characteristics are typically missing. Some artifacts can be avoided by use of two-
24 electrode cells, but this experimental design makes it hard to distinguish the impedance of the ion-selective membrane
25 from that of the measuring electrodes. This work shows that experimental data can be analyzed accurately with the use
26 of models that account for the capacitive leakage present in the reference channels of the impedance spectrometer. To
27 test these models, valinomycin-doped K+-selective membranes were studied by electrochemical impedance spectros-
28 copy with two-, three-, and four-electrode cells, using several measuring electrodes with low to high impedances. The
29 models were found to correctly predict experimental data and provide an intuitive understanding of the cause of the
30 impedance artifacts. This understanding can be applied to design electrochemical impedance spectroscopy experi-
31 ments of ion-selective membranes with three- and four-electrode cells that minimize artifacts.
32
33
34 INTRODUCTION measurements.14-19 The expectation is that the applied
35 AC current flows from one working electrode through the
Electrochemical impedance spectroscopy (EIS) has
36 been used to study ion-selective electrodes since the ion-selective membrane to another working electrode,
37 1980s.1-4 The frequency dependence of the impedance while the resulting AC voltage across the ion-selective
38 of ion-selective electrodes (ISEs)5-9 makes it possible to membrane is observed with two separate reference elec-
39 determine a number of membrane and interfacial proper- trodes. However, as discussed in this contribution, four-
40 ties, including bulk resistances, double layer capacitanc- electrode EIS may also suffer from artifacts (i.e., features
41 es, charge transfer resistances, and Warburg impedanc- in the EIS spectra that are not associated with the sens-
42 es. Often, impedance spectroscopy is used in combina- ing membrane but result from the measuring electrodes
43 tion with other experimental techniques to study complex and/or instrumentation), which may or may not be ap-
44 phenomena, such as in the case of solid-contact ion- parent to the user.14,20,21 Moreover, even though four-
selective membranes.10-13 Early studies used cells with electrode cells are generally expected to provide data of
45
two AgCl-coated Ag plates of large surface areas as higher quality, they are often avoided by users due to the
46
electrodes, immersed into two electrolyte solutions sepa- difficulty of introducing a fourth electrode into an existing
47 device, as for example in the case of solid-contact
48 rated by the ion-selective membrane of interest.1 Be-
cause of their large surface area, these electrodes have ISEs.11-13 This has led to numerous reports of EIS with
49 three-electrode cells, which exhibit their own artifacts.11-
low charge transfer resistances and low bulk resistanc-
50 13
In addition, many reports that include EIS characteri-
es. They provide stable surface potentials, making them
51 ideal electrodes for EIS. However, impedance spectra of zation of ISEs lack experimental details, often not even
52 ion-selective membranes acquired using two-electrode mentioning the number of electrodes used. Curiously,
53 cells may exhibit artifacts related to these electrodes,14 the first publication on EIS of ISE membranes shows a
54 which can lead to inaccurate data interpretation. These four-electrode cell but data from two-electrode meas-
55 artifacts are particularly relevant in the frequency range urements are presented.1
56 where Warburg impedances are observed. Impedance artifacts in measurements of thin films, solid
57 It is now more common for EIS measurements to be electrolytes, and fuel cells have been studied.22-28 Also,
58 acquired using four-electrode cells, which allows for bet- the effects of reference electrodes29 and cell geome-
59 ter control of the applied potential and reduction of un- tries30 on artifacts in three-electrode cells filled with elec-
60 desired impedance artifacts that arise in two-electrode trolyte solutions were reported. However, impedance
artifacts in measurements of ion-selective membranes
ACS Paragon Plus Environment
 Analytical Chemistry
using four-electrode cells have not been reported in the
1 literature. Moreover, a comprehensive comparison of
2 artifacts in EIS spectra of ion-selective membranes ob-
3 tained with two-, three-, and four-electrode cells has
4 been missing.
5 The goal of this contribution is, therefore, to provide an
6 intuitive understanding of the cause of low and high fre-
7 quency artifacts when measuring impedance spectra of
8 ion-selective membranes in cells with two, three or four
9 electrodes. In addition, straightforward ways to reduce
10 and/or avoid artifacts in EIS measurements are dis-
11 cussed.
12 EXPERIMENTAL
13 Materials. Sample solutions were prepared using deion-
14 ized and charcoal-treated water (>18.2 MΩ cm specific
15 resistance) obtained with a Milli-Q PLUS reagent grade
16 water system (Millipore, Bedford, MA). High molecular
17 weight poly(vinyl chloride), 2-nitrophenyl octyl ether, and
18 potassium tetrakis(4-chlorophenyl)borate were pur-
19 chased from Fluka (Buchs, Switzerland), valinomycin
20 was purchased from Sigma-Aldrich (St. Louis, MO), and
21 KCl (99.5 % purity) was purchased from Fisher Scientific
(Waltham, MA).
22
Preparation of Ion-Selective Membranes. A tetrahy-
23
drofuran solution (1 mL) of the membrane components
24
25 (200 mg) was poured into a glass dish with an inner di-
ameter of 30 mm, and the tetrahydrofuran was allowed
26
to evaporate overnight. The membranes were composed
27
of 33% w/w poly(vinyl chloride), 65.8% w/w 2-nitrophenyl
28 octyl ether, and 50 mol % potassium tetrakis(4-
29 chlorophenyl)borate relative to valinomycin. Membrane
30 discs (1.1 cm) were cut and glued onto Tygon tubing (9.5
31 mm i.d.) using tetrahydrofuran. The membranes were
32 equilibrated overnight with 1.0 M KCl.
33 Electrochemical Impedance Spectroscopy. All im-
34
pedance measurements were performed on a Solartron
35 1255B frequency response analyzer with an SI 1287
36 electrochemical interface (Farnborough, Hampshire,
37 U.K.) controlled by ZPlot software (Scribner Associates,
38 Southern Pines, NC). The SI 1287 electrochemical inter-
39 face lists a reference channel input capacitance of 50 pF
40 and a reference channel input resistance greater that 10
41 GΩ, and the 1255B frequency response analyzer lists a
42 input capacitance of less than 10 nF. In addition to the
43 listed input capacitances, the SI 1287 is connected to
44 the 1255B with coaxial cables, which further increase the
input capacitance. Measurements were performed in the
45
frequency range from 1 MHz to 0.1 Hz at frequencies
46
separated by 0.1 units on the logarithmic scale. Meas-
47 urements were performed at the open circuit potential
48 with an AC amplitude of 100 mV (unless otherwise stat-
49 ed). Membranes were kept in a 1.0 M KCl solution. Data
50 was collected using a variety of working, counter, and
51 reference electrodes, i.e., a 0.25 mm coiled Pt wire
52 (99.998%, Alfa Aesar, Ward Hill, MA; electrode A), Pt
53 gauze (2” by 0.5”; electrode B), AgCl-coated 0.5 mm Ag
54 wire (Ag wire, 99.9%, Alfa Aesar, Ward Hill, MA; elec-
55 trode C), electrode B inserted into a glass tube with a
56 Vycor glass frit31,32 (filled with 3 M KCl; electrode D),
57 electrode B inserted into a flame-pulled Pasteur pipette
58 (filled with 1.0 M KCl; electrode E), and a Mettler Toledo
59 DX200 free-flowing double-junction33 reference electrode
60
ACS Paragon Plus Environment
Page 2 of 9
(with a 1.0 M LiOAc bridge electrolyte and AgCl-
saturated 3.0 M KCl inner reference electrolyte; elec-
trode F). All simulations and data fits were performed
using complex number notation with the mathematics
software Mathematica 10.1 (Champaign, IL), which al-
lows for very complex models that take into account not
only the electrochemical cell but also the measuring cir-
cuitry.
RESULTS AND DISCUSSION
Ion-Selective Membrane Impedance Spectra. Electro-
chemical impedance spectra are acquired by applying an
AC voltage of controllable and variable frequency across
a sample while simultaneously measuring the resulting
current and the phase shift between the applied voltage
and current.16 From the observed current and phase
shift, the complex impedance of the sample can be cal-
culated. Multiple texts are available15-19 for a detailed
introduction to EIS.
An equivalent circuit can be used to approximate the
electrochemical impedance, ZLoad, of an ion-selective
membrane (Figure 1).16,18,19 For the valinomycin-doped
K+-selective membrane that was used as a representa-
tive example in this study, four circuit elements can be
used to describe its impedance spectrum: Rs (bulk re-
sistance), Cdl (double layer capacitance), Rct (charge-
transfer resistance), and Zw (Warburg impedance).1 Be-
cause these equivalent circuit elements are descriptive
of the electrochemical properties of the membrane bulk,
membrane interface, and electrolyte solution, an accu-
rate measurement of each element is necessary to
quantitatively understand current transport through the
electrochemical cell.16
Using complex numbers, the total impedance represent-
ed by the equivalent circuit presented in Figure 1 is given
by the following function:

 
 
  √1  1
where ω is the angular frequency of the applied voltage.
The values of each coefficient can be fitted to experi-
mental data and used to draw useful conclusions about
the properties of the membrane. Unfortunately, Equation
1 is insufficient for fitting experimental impedance spec-
tra that are affected by artifacts, which is true all too of-
ten. Therefore, a more detailed analysis that takes into
account the instrumentation used to acquire EIS spectra
is required.
Figure 1. Ideal impedance spectrum of a valinomycin-
+
doped K -selective membrane. Data is plotted from 1 MHz
Page 3 of 9 Analytical Chemistry

to 1 Hz for Rs = 5 kΩ, Cdl = 1 pF, Rct = 100 kΩ, and a the reference electrodes (represented in Figure 2 by
1 Warburg coefficient of 0.5. their impedances ZCR and ZWR) to affect the measured
2 impedance spectra.20-22,34
Experimental Parameters of Impedance Measure-
3 The capacitive leakage inherent to impedance spec-
ments. Four instrument channels are used in commer-
4 cial impedance spectrometers to measure the current trometers is intensified by the cabling used to connect
5 and voltage across a sample. These are typically denot- the measurement electrodes to the impedance spec-
6 ed the working electrode (WE), working electrode refer- trometer.22,35 Due to the capacitance of the cable shield-
7 ence (WR), counter electrode (CE), and counter elec- ing, an additional path to virtual ground is present. These
8 trode reference (CR) channels. The WR and CR chan- two capacitances can add up to an apparent capacitive
9 nels are used to measure the potential. The WE and CE short to virtual ground of hundreds of picofarads.22 In
10 channels are used to provide and measure the current addition to these capacitive leakages changing the
11 needed to apply a desired voltage across the sample measured reference potential, they provide ground loops
12 load. In the case of the Solartron 1255B frequency re- that can lead to inductive features in the resulting im-
13 sponse analyzer used in this study, the current is meas- pedance spectra that are not related to the sample.
14 ured in the WE rather than the CE channel, as shown in The leakage currents (ICRleak and IWRleak) in Figure 2 allow
15 Figure 2. (Note that the distinction between WE and CE for part of the supplied current to bypass the current
16 is arbitrary and that an instrument that measures the measurement circuitry (Imeasured), which is located in a
current on the CE channel is still represented adequately different location of the spectrometer than the voltage
17
by Figure 2 if the labels CE and WE are swapped. measurement circuitry (Vmeasured). Consequently, when
18 Therefore, the discussion given here is—with the excep- the complex impedance is calculated as Vmeasured/Imeasured,
19 tion of the channel labeling—relevant to instruments a result is obtained that does not only reflect ZLoad.
20 from a wide range of suppliers.)
21 Using complex number notation, the entire equivalent
circuit shown in Figure 2 can be analyzed in a straight-
22
forward manner using Kirchhoff’s voltage and current
23
laws. This analysis will be used in the following to de-
24 scribe impedance artifacts encountered when using two-,
25 three-, and four-electrode cells.
26
Two-Electrode Measurements. The simplest and his-
27
torically first used method to measure the impedance of
28 ion-selective membranes was based on a two-electrode
29 cell.1,36 The circuit presented in Figure 2 can be modified
30 Figure 2. Simplified schematic consistent with a Solartron
to that of a two-electrode cell (Figure 3).
31 1255B frequency response analyzer: WE, WR, CR, and CE
32 represent the working electrode, working electrode
reference, counter electrode reference, and counter
33 electrode channels, respectively. ZWE, ZWR, ZCR, and ZCE
34 represent the complex impedance of electrodes connected
35 to each channel. ZLoad is the impedance of the measured
36 sample, and Vmeasured is the voltage measured by the
37 spectrometer. ZCRleak and ZWRleak are the impedances to
38 ground at the reference channels. Imeasured is the current
39 measured by the insturment, and ICRleak and IWRleak are the
leakage currents that cause impedance artifacts. The dotted Figure 3. Schematic of an impedance spectrometer with the
40
line encloses components internal to the spectrometer. WE and WR channels shorted and the CE and CR channels
41
In an ideal system, the input impedance of the reference shorted for a two-electrode measurement.
42
43 channels is so large that no current can pass through By removing the reference electrodes, shorting the WE
44 them. This allows the impedance of the sample inserted to the WR channel, and shorting the CE to the CR chan-
45 in between the WE and CE channels to be measured. nel, ground loops no longer contribute to a substantial
46 However, each impedance spectrometer has reference capacitive leakage. This allows for measurements of
47 channels with a finite input impedance (represented in samples with a wide range of load impedances, ZLoad, at
Figure 2 by ZCRleak and ZWRleak). Due to the large input high frequencies without the need to optimize the exper-
48
resistance, current through the reference electrodes is imental setup. However, when using a two-electrode cell,
49
negligible at low frequencies. However, at a frequency of the impedance characteristics of the working and coun-
50 1 MHz, the reference channel input capacitance provides ter electrode contribute to the acquired spectrum; the
51 a path to a virtual ground that has an impedance of ap- observed cell impedance is not just ZLoad but the sum ZCE
52 proximately 3 kΩ. + ZLoad + ZWE. This is illustrated by the impedance spec-
53   tra of K+-selective membranes (see Figure 4) as meas-
54 |
|     3.2 kΩ (2)
√ ∗ ! "#$ ∗ %&' ( ured using a AgCl-coated Ag wire (electrode C) as the
55 WE electrode and a range of different CE electrodes
56 This high frequency path to virtual ground allows for cur- (i.e., electrode types C-F, which were chosen because
57 rent to flow through the WR and CR channels (referred the values of their impedances cover a wide range from
58 to as capacitive leakage, leakage current, or stray ca- 0.1 to 15 kΩ and which included a AgCl-coated wire, an
59 pacitance) and causes the impedance characteristics of
60
ACS Paragon Plus Environment
 Analytical Chemistry Page 4 of 9

electrode with a Vycor glass frit31,32 as liquid junction, a channels are not symmetrical in terms of the impedance
1 narrow-tip pipette, and a free-flowing double-junction33 artifacts generated, a fact that experimentalists must be
2 electrode, as described in detail in the Experimental Sec- aware of.
3 tion). The equivalent circuit of Figure 2 can be modified in two
4 ways, either shorting the WE to the WR channel or the
5 CE to the CR channel. First, let us consider shorting WE
6 to the WR channel (Figure 6).
7
8
9
10
11
12
13
14
+
15 Figure 4. Impedance spectra of a K -selective membrane
16 measured using a two-electrode cell. The WE was a AgCl-
Figure 6. Schematic of an impedance spectrometer with the
17 coated Ag wire and the CE was of type C (
), D (), E (
),
WE and WR channels shorted to produce a three-electrode
F (). Inset shows data expanded from the boxed region.
18 measurement.
(For graphs with the individual impedance spectra of
19 electrodes D–F, see the Supplemental Information, Figure Shorting the WE to the WR channel eliminates any volt-
20 S1.) age drop across ZWRleak; as IWRleak vanishes, it no longer
21 causes artifacts in measured spectra. However, in this
22 Ignoring the impedances of the CE and WE electrodes
can lead to poor fits of Rs, Cdl, Rct, and Zw if the experi- configuration, the WE electrode will influence the meas-
23 ured impedance in the same way as for the two-
mental data (as shown in Figure 4) are fitted with Equa-
24 electrode spectra. Also, the CR electrode causes arti-
tion 1 (which ignores properties of the spectrometer), as
25 facts that depend on its impedance, ZCR, as illustrated by
shown in Figure 5 for a particularly poor choice of refer-
26 ence electrodes. Note that the high frequency artifact at Figure 7.
27 the bottom left of the spectrum could easily be misinter-
28 preted as a feature associated with the membrane under
29 study, such as charge transfer resistance at the sam-
30 ple/membrane interface.
31
32
33
34
35
36
37
+
38 Figure 7. Impedance spectra of a K -selective membrane
39 as measured using a three-electrode cell, with the WE and
40 WR channels shorted. WE/WR was a AgCl-coated Ag wire,
CE was a Pt gauze, and the WR electrode was of type C
41 +
Figure 5. Impedance spectra of a K -selective membrane (
), D (), E (
), or F (). Inset shows data expanded
42 measured using a two-electrode cell. The WE was a AgCl- from the boxed region.
43 coated Ag wire and the CE a free-flowing double-junction
44 reference electrode (). Solid line is the data fitted to The spectra in Figure 7 show that an increasing re-
45 Equation 1. sistance of the CR electrode leads, for high frequencies,
46 to greater excursions into negative real impedance. The-
Although two-electrode measurements can introduce se artifacts not only affect the high frequency data, but
47 unwanted impedance features into measured spectra, also cause easily recognized deviations in frequencies
48 careful selection of the WE and CE electrodes leads to as low as 5 kHz. This can be visualized by fitting the
49 meaningful data. Ideal WE and CE electrodes have low spectrum from Figure 7 with the largest excursion into
50 charge transfer resistances, low bulk resistances, and negative real impedance with Equation 1, as shown in
51 stable surface potentials. However, ideal electrodes with Figure 8. The fitted semicircle shows on its left side an
52 large surface areas and a low charge transfer resistance excursion into negative real impedance (corresponding
53 (such as AgCl-coated Ag plates in contact with a chloride to an unreasonable negative value for the resistance Rs),
54 solution) are not always applicable when studying non- and it also fits the experimental data on its right half quite
55 aqueous samples or non-chloride containing solutions. poorly, resulting in avoidable errors in the determination
56 Therefore, the use of four-electrode cells is often prefer- of Cdl and Rct.
57 able.
One approach to reduce these artifacts is to use a refer-
58 Three-Electrode Measurements. In this section data ence electrode with a much lower impedance than the
59 will be presented to demonstrate that the WR and CR
60
ACS Paragon Plus Environment
Page 5 of 9 Analytical Chemistry

one of the free-flowing double-junction reference elec- Importantly, contrary to the impedance spectra of Figure
1 trode with a ZCR of about 15 kΩ, as used for the data 7, spectra collected with the CE and CR channels short-
2 shown in Figure 8. This approach successfully minimizes ed did not exhibit impedance artifacts with any of the
3 the impedance artifacts, as shown by the data collected reference electrodes studied. This can be explained by
4 with a AgCl-coated Ag wire (type C) in Figure 7. the properties of simplified test circuits and predictions
5 from mathematical models, as discussed in the following.
6 Test Circuit Analysis. To further clarify the origin of the
7 impedance artifacts arising in the three-electrode meas-
8 urements of the K+-selective membrane, a test circuit
9 was assembled using a resistor in series with a parallel
10 combination of a resistor and capacitor to represent the
11 load impedance of an ion-selective membrane in an
12 electrochemical cell. Impedance spectra collected using
13 this test circuit are presented in Figure 11.
14
15 +
Figure 8. Impedance spectrum of a K -selective membrane
16 measured using a three-electrode cell with the WE and WR
17 channels shorted. WE/WR was a AgCl-coated Ag wire, CE
18 was Pt gauze, and CR was a free-flowing double-junction
19 reference electrode (). The solid line is a fit of all
20 experimental data to Equation 1.
21 The alternative to shorting the WE to the WR channel is
22 shorting the CE to the CR channel. The corresponding
23 schematic is presented in Figure 9.
24 Figure 11. Impedance spectra of a 2 kΩ resistor in series
with a parallel combination of a 150 kΩ resistor and 300 pF
25 capacitor. Two-electrode measurement (
), three-electrode
26 measurement with the CE and CR channels shorted and an
27 in-series 15 kΩ resistor added into the WR channel (
),
28 three-electrode measurement with the WE and WR shorted
29 and an in-series 15 kΩ resistor added to the CR channel
30 (). Inset shows data expanded from the boxed region.
31 Impedance spectra collected with the test circuit exhibit-
32 Figure 9. Schematic of an impedance spectrometer with the ed the same artifacts as measurements with the electro-
33 CE and CR channels shorted to produce a three-electrode chemical cell comprising the K+-selective membrane.
34 measurement. Adding an in-series resistor into the CR channel (as
35 shown in Figure 12, but with a resistance for ZCR) caused
By shorting the CE to the CR channel, it is this time ICR-
36 excursions into negative real impedance (Figure 11, ),
leakthat no longer causes artifacts in the measured spec-
37 while adding the same in-series resistor into the WR
tra. However, the impedance characteristics of the CE
38 channel caused no change (

). The data shown in Fig-
electrode now contribute to the observed impedance
39 ure 10 was further analyzed in Mathematica 10.1 using
spectra, and the WR electrode has the potential to cause
40 models comprising also equivalent circuit elements rep-
artifacts in the EIS spectrum. Impedance spectra collect-
41 resenting the spectrometer. The first model was devel-
ed with this three-electrode cell are shown in Figure 10.
42 oped using the schematic shown in Figure 6 to deter-
mine how artifacts are affected by changing the imped-
43
ance of the CR electrode (for the Mathematica notebook,
44 see the Supplemental Information). As it became appar-
45 ent, the same conclusions as from this model could also
46 be obtained from a simplified model in which the imped-
47 ance contributions of the CE electrode and WE electrode
48 were removed. A schematic of this simplified circuit is
49 presented in Figure 12.
50
51
52
+
53 Figure 10. Impedance spectra of a K -selective membrane
54 measured using a three-electrode cell, with the CE and CR
55 channels shorted. CE/CR was a AgCl-coated Ag wire, WE
56 was a Pt gauze, and WR was of type C (
), D (), or E
(
). Inset shows data expanded from the boxed region.
57
58
59
60
ACS Paragon Plus Environment
 Analytical Chemistry Page 6 of 9

Figure 12. Schematic used to model a three-electrode Figure 13 shows that at high frequencies the impedance
1 impedance spectrum with the WR and WE channels ZCR can cause Vmeasured to approach zero as the applied
2 shorted. AC voltage across ZCR cancels the voltage across the
3 As the figure shows, ZLoad and ZCR + ZCRleak form parallel load. Figure 13 also shows that a reference electrode
4 paths from CE to ground. Therefore, the capacitive leak- with a low real impedance (i.e., resistance) will cause a
5 age (ICRleak) has the largest effect on the observed im- reduction in impedance artifacts (see also Supplemental
6 pedance spectrum when the load impedance (ZLoad) is Information, Figure S2). This shows that reference elec-
7 large relative to ZCR + ZCRleak, diverting a substantial frac- trodes connected to the CR channel should have a low
8 tion of the current Itotal coming from CE towards ZCR ra- resistance.
9 ther than ZLoad. This is exactly the situation that is en- In the opposite situation (i.e., shorting the CE to the CR
10 countered when ion-selective membranes with their typi- channel; see Figure 14), the current coming from the
11 cally high bulk resistance are analyzed.14 Importantly, load impedance is split among the parallel paths ZWR +
12 the leakage current (ICRleak) that flows at high frequencies ZWRleak and ZWE. (For our instrument, ZWRleak was deter-
13 through ZCR and ZCRleak (the voltmeter has a very high mined to be 500 ± 40 pF by fitting the data in Figure S3,
14 input impedance and does not draw appreciable current) , with the equivalent circuit developed from the circuit
does not cause substantial changes in the measured in Figure 14 and 350 ± 10 pF by fitting data collected
15
potential (Vmeasured) as long as ZCR is small relative to ZCR- with shortened leads; see Figure S8.) If the impedance
16
leak. (For our instrument, ZCRleak was determined to be of ZWE is much larger than ZWR + ZWRleak, then Vmeasured
17
1000 ± 100 pF by fitting data of Figure 11, , with the will differ substantially from the potential present across
18 equivalent circuit shown in Figure 12; shortening the ZLoad.
19 leads reduced this value to 380 ± 60 pF, see Figure S7.) Having a WE electrode that has a much greater imped-
20 This can be seen readily from the voltage divider equa- ance than the WR electrode is rarely encountered in
21 tion, which shows that the voltage across ZCR equals the practice, but an illustration of impedance artifacts for
22 voltage across the load multiplied with ≈ZCR/(ZCR+ZCRleak), such a situation is included in the Supplemental Infor-
23 which is insignificant as long as ZCR << ZCRleak. However, mation (see Figures S3–S5 and S8).
24 by introducing an additional impedance with the CR elec-
25 trode that is on the order of or larger than ZCRleak, an ad-
26 ditional potential difference is present across that elec-
27 trode (ICRleak ZCR) that prevents the actual potential of the
28 load impedance (Imeasured ZLoad) from being measured
29 accurately, resulting in artifacts in the EIS spectra. The
30 total potential that is measured can be calculated as –
31 ICRleak ZCR + Imeasured ZLoad. As the frequency of the applied
32 AC current is decreased, the ratio ZCR/(ZCR+ZCRleak) be-
gins to approach zero because the frequency-dependent
33
impedance of the capacitance component of ZCRleak in- Figure 14. Simplified schematic used to model three-
34 creases, and the measured voltage used to determine electrode impedance spectra with the WR and WE channels
35 the impedance of interest approaches the ideal value of shorted.
36 Imeasured ZLoad. Any stray impedance associated with the
37 An important conclusion for the practitioner of EIS who
CE channel that may exist does not contribute to imped-
38 cannot avoid using a three-electrode setup is that the
ance artifacts. This process can be graphically repre-
two measurement options represented by Figures 6 and
39 sented by plotting the absolute potential across the CR
9 are not equivalent. If a large reference electrode im-
40 channel (i.e., ICRleak ZCR) relative to the voltage across the
pedance cannot be avoided, the setup represented by
41 load (i.e., Imeasured ZLoad; see Figure 13).
Figure 9 is preferable, except for the unlikely case that
42 the impedance of WE is very large. It is noteworthy that,
43 in the setup represented by Figure 9 (and, therefore, the
44 often preferable setup), the reference electrode WR is on
45 the same side of the load as WE, opposing CE on the
46 other side of the load. This setup may appear counterin-
47 tuitive to someone very familiar with three-electrode volt-
48 ammetry, but one should keep in mind that this is the
49 result of the arbitrary labeling of the two current carrying
50 channels as CE and WE.
51 Four-Electrode Measurements. Four-electrode meas-
52 urements exhibit the same type of artifacts as those col-
53 Figure 13. Calculated ratio of the absolute potential across lected using the two types of three-electrode measure-
54 ZCR and across ZLoad (as modeled with the circuit shown in ments. Impedance spectra collected with a four-
Figure 12). ZLoad is a 2 kΩ resistor in series with a parallel electrode cell, where the WE was a Pt wire, WR was a
55
combination of a 150 kΩ resistor and a 300 pF capacitor, AgCl-coated Ag wire, CE was a Pt gauze, and the CR
56 ZCRleak is 1000 pF, and ZCR is a 1 kΩ (dotted line), 10 kΩ
57 electrodes were of type C to F are presented in Figure
(dashed line), or 150 kΩ (solid line) in-series resistor added 15. As the figure shows, these spectra exhibit the same
58 into the CR channel.
59
60
ACS Paragon Plus Environment
Page 7 of 9 Analytical Chemistry

artifacts that are present in three-electrode measure-

1 ments with the WE and WR channels shorted (Figure 7).
2 Impedance artifacts are minimized by reducing the im-
3 pedance of the CR electrode, as expected from the
4 mathematical simulations (see Figure 13). In addition,
5 these spectra display a larger Warburg impedance than
6 those measured with three electrodes. This can only be
7 due to the removal of the impedance of the WE elec-
8 trode. Therefore, for a complete analysis of an ion-
9 selective membrane, particularly the Warburg imped-
10 ance, a four-electrode cell is the most appropriate.
11 Impedance spectra were also measured with a four-
12 electrode cell where, as above, the WE was a Pt wire
13 and the CE was a Pt gauze, but the types of electrodes +
Figure 16. Subset of impedance spectra of a K -selective
14 for the CR and WR channels were switched, i.e., the CR
membrane measured using a four-electrode cell. WE was a
15 was a AgCl-coated Ag wire, and the WR electrode was Pt wire, WR was a AgCl-coated Ag wire, CE was a Pt
16 of type C–F. These spectra did not display any artifacts gauze, and CR was a AgCl-coated Ag wire. Data was
17 as the impedance of the WE channel with a Pt wire is collected for applied voltages of 10 mV (
), 100 mV (),
much smaller than that of the reference electrodes con- and 1 V (
). Adjacent data points are separated by 0.1
18
nected to the WR channel. This result is expected from units on a logarithmic scale. The dotted line indicates a 45
o
19
the mathematical models presented in Figure 12 and 14. angle.
20
21 A suitable voltage amplitude must be determined exper-
22 imentally. By applying voltages that are too large, chemi-
23 cal artifacts will be seen, while applying voltages that are
24 too low will cause increased noise. The K+-selective
25 membranes studied here did not show deviations from
26 ideal behavior until voltage amplitudes larger than 100
27 mV were applied.
28 CONCLUSIONS
29 EIS measurements of ion-selective electrode mem-
30 branes have been carried out using two-, three-, and
31 four-electrode cell for decades. However, to date, there
32 has not been a comprehensive discussion of instrumen-
33 tal artifacts that arise in the EIS characterization of such
34 membranes, which typically exhibit large bulk resistanc-
35 es.
36 With this study, we showed that two-electrode meas-
37 +
urements include varying degrees of impedance artifacts
Figure 15. Impedance spectra of a K -selective membrane caused by the measuring electrodes. The occurrence of
38 measured using a four-electrode cell. WE was a Pt wire,
39 such artifacts can be detected by measuring spectra with
WR was a AgCl-coated Ag wire, CE was a Pt gauze, and CR different electrodes. If the selection of the electrode af-
40 electrode was of type C (
), D (), E (
), or F (). Inset
fects the EIS spectra, a three- or four-electrode cell
41 shows data expanded from boxed region.
should be used.
42
Chemical Artifacts. Users must select an amplitude for Three-electrode EIS was carried out on valinomycin-
43 the AC modulation of the voltage used to measure im- doped K+-selective membranes to illustrate the causes of
44 pedance spectra. Common voltage amplitudes reported impedance artifacts, i.e., the leakage current inherent to
45 in the literature vary from 0.01 to 1 V.2-4,37,38 Importantly, impedance spectrometers.14-16 For a spectrometer that
46 the voltage amplitude applied to an ion-selective mem- measures the current on the WE channel, the leakage
47 brane can have a direct effect on experimental data if it current has a large effect on impedance spectra if resis-
48 is too large. By applying a large voltage across an ion- tive reference electrodes are connected to the CR chan-
49 selective membrane, large numbers of ions are forced nel, causing inaccurate measurements. In addition, WR
50 out of or into the membrane. This was found to introduce electrodes were found to introduce impedance artifacts if
51 systematic errors in the impedance spectra of K+- their impedance is much larger than that of the WE elec-
52 selective membranes at low frequencies. Most im- trode.
53 portantly, these artifacts cause the Warburg impedance
It was also shown that impedance artifacts in four-
54 element to deviate from the ideal 45º angle. The experi-
electrode measurements can be analyzed with the un-
55 mental observation of this characteristic is illustrated in
derstanding of artifacts present in three-electrode meas-
56 Figure 16, which shows that the Warburg impedance at
urements. By acquiring impedance spectra with multiple
an applied voltage amplitude of 1 V deviates from the
57 reference electrodes, a user can determine the cause of
theoretical 45º angle.
58 the artifact. Combining this with an understanding of the
59 circuitry inherent to the impedance spectrometer allows
60
ACS Paragon Plus Environment
 Analytical Chemistry Page 8 of 9

for accurate measurements of equivalent circuits of ion- Assembly and Supramolecular Devices, John Wiley & Sons:
1 selective membranes. This can be achieved by using (i) Chichester, UK, 2012; pp 2539−80.
2 low impedance CR electrodes and (ii) WR electrodes (10) Fierke, M. A.; Lai, C. Z.; Bühlmann, P.; Stein, A. Anal.
Chem. 2010, 82, 680-688.
3 with a larger impedance than the WE electrode, or the
(11) Veder, J. P.; Patel, K.; Sohail, M.; Jiang, S. P.; James,
4 opposite if current is measured with the CE channel (see M.; De Marco, R. Electroanalysis 2012, 24, 140-145.
5 Figure 9). Users who do not know which channel is used (12) De Marco, R.; Ng, A.; Panduwinata, D. Electroanalysis
6 for the current measurement in their instrument may test 2008, 20, 313-317.
7 in a preliminary experiment both setups to empiricially (13) De Marco, R.; Jiang, Z. T.; Martizano, J.; Lowe, A.; Pe-
8 identify the preferred setup. In those cases where ex- jcic, B.; van Riessen, A. Electrochim. Acta 2006, 51, 5920-
changing the electrodes connected to the WE, WR, CE, 5925.
9 (14) Lvovich, V., AC Impedance Spectroscopy Measurement
or CR channels is not feasible, EIS spectra affected by
10 - Understanding the Effects of 2-, 3-, and 4-Electrode Cells and
artifacts can be fitted with an equivalent circuit that takes
11 Cables on Experimental Data. Knovel, 2014, Chapter 4.
into account the impedances of the reference channels, (15) Macdonald, J. R., Impedance Spectroscopy Emphasiz-
12 permitting the determination of accurate membrane im-
13 ing Solid Materials and Systems. John Wiley and Sons: New
pedances. York, New York, 1987.
14 (16) Orazem, M.; Tribollet, B., Electrochemical Impedance
15 ASSOCIATED CONTENT Spectroscopy. John Wiley and Sons: Hoboken, New Jersey,
16 Supplemental Information is listed in the text. This material 2008.
17 is available free of charge via the Internet at (17) Yuan, X.-Z.; Song, C.; Wang, H.; Zhang, J., Electro-
18 http://pubs.acs.org chemical Impedance Spectroscopy in PEM Fuel Cells. Spring-
er: Heidelberg, New York, 2010.
19 (18) Lvovich, V., Impedance Spectroscopy: Applications to
20 AUTHOR INFORMATION Electrochemical and Dielectric Phenomena. John Wiley and
21 Corresponding Author Sons: Hoboken, New Jersey, 2012.
22 (19) Lasia, A., Electrochemical Impedance Spectroscopy and
* Philippe Bühlmann (buhlmann@umn.edu) its Applications. Springer Science: New York, New York, 2014.
23
Author Contributions (20) Vanýsek, P. ECS Trans. 2008, 13, 101-113.
24 (21) Vanysek, P.; Tavassol, H. ECS Trans. 2009, 19, 43-54.
25 The manuscript was written through contributions of all au- (22) Veal, B. W.; Baldo, P. M.; Paulikas, A. P. J. Electro-
26 thors. chem. Soc. 2015, 162, H47-H57.
27 (23) Hsieh, G.; Ford, S. J.; Mason, T. O.; Pederson, L. R.
Funding Sources Solid State Ionics 1996, 91, 191-201.
28
This work was supported by the University of Minnesota (24) Hsieh, G.; Mason, T. O.; Pederson, L. R. Solid State
29 Initiative for Renewable Energy and the Environment. Ionics 1996, 91, 203-212.
30 (25) Hsieh, G. Solid State Ionics 1997, 96, 153-172.
31 ACKNOWLEDGMENT (26) Xie, Z.; Song, C.; Andreaus, B.; Navessin, T.; Shi, Z.;
32 The authors thank Yu Ishige, Central Research Laboratory,
Zhang, J.; Holdcroft, S. J. Electrochem. Soc. 2006, 153, E173.
33 (27) Cimenti, M.; Co, A. C.; Birss, V. I.; Hill, J. M. Fuel Cells
Hitachi Ltd., for stimulating discussions. 2007, 7, 364-376.
34
(28) Horvat-Radošević, V.; Kvastek, K. J. Electroanal. Chem.
35 ABBREVIATIONS 2008, 613, 139-150.
36 EIS, electrochemical impedance spectroscopy; WE, working (29) Tran, A. T.; Huet, F.; Ngo, K.; Rousseau, P. Electro-
37 electrode; WR, working electrode reference; CE, counter chim. Acta 2011, 56, 8024-8039.
38 electrode; CR, counter electrode reference (30) Battistel, A.; Fan, M.; Stojadinović, J.; La Mantia, F.
39 Electrochim. Acta 2014, 135, 133-138.
REFERENCES (31) Mousavi, M. P. S.; Bühlmann, P. Anal. Chem. 2013, 85,
40 8895-8901.
41 (1) Armstrong, R. D.; Covington, A. K.; Evans, G. P. J. Elec- (32) Mousavi, M. P. S.; Anderson, E. L.; Saba, S. A.; Hillmy-
42 troanal. Chem. Interfacial Electrochem. 1983, 159, 33-40. er, M.; Bühlmann, P., Avoiding Errors in Electochemical Meas-
(2) Crawley, C. D.; Rechnitz, G. A. J. Membr. Sci. 1985, 24, urements: Effect of Frit Material on the Performance of Refer-
43 201-219.
44 ence Electrodes with Porous Frits. Presented at SEAC Poster
(3) Horvai, G.; Graf, E.; Toth, K.; Pungor, E.; Buck, R. P. Session, Chicago, Illinois, March 9th, 2016
45 Anal. Chem. 1986, 58, 2735-2740. (33) Dohner, R. E.; Wegmann, D.; Morf, W. E.; Simon, W.
46 (4) Toth, K.; Graf, E.; Horvai, G.; Pungor, E.; Buck, R. P. Anal. Chem. 1986, 58, 2585-2589.
47 Anal. Chem. 1986, 58, 2741-2744. (34) Mansfeld, F. J. Electrochem. Soc. 1988, 135, 906.
(5) Bakker, E.; Bühlmann, P.; Pretsch, E. Chem. Rev. 1997, (35) Balabajew, M.; Roling, B. Electrochim. Acta 2015, 176,
48 97, 3083-3132.
49 907-918.
(6) Bühlmann, P.; Pretsch, E.; Bakker, E. Chem. Rev. 1998, (36) Figaszewski, Z. J. Electroanal. Chem. Interfacial Elec-
50 98, 1593-1687. trochem. 1982, 139, 309-315.
51 (7) Amemiya, S., Potentiometric Ion-Selective Electrodes, (37) Fortunato, R.; Branco, L. C.; Afonso, C. A. M.; Benaven-
52 In Handbook of Electrochemistry, Zoski, C., Ed.; Elsevier: Bos- te, J.; Crespo, J. G. J. Membr. Sci. 2006, 270, 42-49.
ton, Massachusetts, 2007, pp 261–294. (38) Radu, A.; Anastasova-Ivanova, S.; Paczosa-Bator, B.;
53
(8) Bobacka, J.; Ivaska, A.; Lewenstam, A. Chem. Rev. Danielewski, M.; Bobacka, J.; Lewenstam, A.; Diamond, D.
54 2008, 108, 329-351. Anal. Methods 2010, 2, 1490.
55 (9) Bühlmann, P.; Chen, L. D. Ion-Selective Electrodes with
56 Ionophore-Doped Sensing Membranes, In Supramolecular
57 Chemistry: From Molecules to Nanomaterials; Vol. 5: Self-
58
59
60
ACS Paragon Plus Environment
Page 9 of 9 Analytical Chemistry

For TOC only

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment