PRIMER

Electrochemical impedance
spectroscopy
Shangshang Wang   1, Jianbo Zhang   1, Oumaïma Gharbi   2, Vincent Vivier   2,
Ming Gao   3 and Mark E. Orazem   3 ✉
Abstract | Electrochemical impedance spectroscopy (EIS) is a powerful tool to investigate
properties of materials and electrode reactions. This Primer provides a guide to the use of
EIS and a comparison with other electrochemical techniques. The analysis of impedance data
for reduction of ferricyanide in a KCl supporting electrolyte is used to demonstrate the error
structure for impedance measurements, the use of measurement and process models, and the
sensitivity of impedance to the evolution of electrode properties. This Primer provides guidelines
for experimental design, discusses the relevance of accuracy contour plots to wiring and
instrumentation selection, and emphasizes the importance of the Kramers–Kronig relations
to data validation and analysis. Applications of EIS to battery performance, metal and alloy
corrosion, and electrochemical biosensors are highlighted. Electrochemical impedance
measurements depend on both the mechanism under investigation and extrinsic parameters,
such as the electrode geometry. Experimental complications are discussed, including the
influence of non-stationary behaviour at low frequencies and the need for reference electrodes.
Finally, emerging trends in experimental and interpretation approaches are also described.

Cyclic voltammetry
Electrochemistry is a discipline encompassing all hetero­ required to facilitate evaluation of electrochemical systems.
An electrochemical technique geneous chemical reactions involving electron transfer, The EIS technique is broadly applicable because it can
in which the current response homogeneous reactions that influence reactions at elec­ provide an understanding of the electrochemical mecha­
of an electrochemical system trode surfaces, properties of the interfaces (the double nisms occurring at an electrified interface in a single meas­
is measured as a function
layer) and bulk properties of electrolytes. Electrochemical urement. These underlying mechanisms include those
of the potential, which is
swept in positive and negative reactions are at the root of several research areas includ­ involved in operating commercial batteries, the corrosion
directions at a given rate. ing energy conversion and storage, corrosion, sensors of metals and alloys, and electrochemical biosensors.
and biosensing applications, as well as the production Textbooks, monographs2–5 and review articles have
Chronoamperometry of commodity materials such as aluminium and chlo­ been dedicated to EIS that are specific to the applica­
An electrochemical technique
in which a potential step is
rine. The development of several electrochemical tech­ tions6–12. The history of EIS has also been retraced13,14,
applied to a working electrode niques such as cyclic voltammetry, chronoamperometry and highlighting the evolution of the technique from the first
and the current response is chronopotentiometry, scanning electrochemical microscopy measurement of electrolyte conductivity using alternat­
recorded as a function of time. and electrochemical impedance spectroscopy (EIS) was ing current in the mid-nineteenth century15,16 to the
required to enable studying these reactions across this measurement of interfacial capacitance17 or the measure­
broad range of application settings. Each of these tech­ ment of diffusivity18. EIS has benefited from the devel­
1
School of Vehicle and
Mobility, Tsinghua University,
niques has expanded in recent decades, as a result of opment of instruments, such as the first potentiostats and
Beijing, China. advances in instrumentation and modelling1,2. the rotating disk electrode. Potentiostats enable control
2
Sorbonne Université, CNRS, From a mechanistic point of view, an electrochem­ of the working electrode potential referenced to an elec­
Laboratoire Interfaces et ical reaction at an electrode–electrolyte interface can be trode at its equilibrium redox potential19 and the rotat­
Systèmes Electrochimiques decomposed into a series of multistep processes (mass ing disk electrode affords the well-defined transport of
(LISE), Paris, France.
transport, charge-transfer processes, adsorption), each reactants and products to the electrode surface20. The
3
Department of Chemical occurring at distinct rates. The individual steps are development of mathematical models, such as those for
Engineering, University of
Florida, Gainesville, FL, USA.
time-dependent and may occur at different timescales. porous electrodes that describe the impedance response
✉e-mail: morazem@ Therefore, the use of transient techniques such as EIS, as a function of the pore geometry in the presence or
che.ufl.edu which enables the analysis of time-dependent mecha­ absence of electrochemical reactions, have enabled a
https://doi.org/10.1038/ nisms based on the response (current or potential) of the better understanding of complex electrochemical sys­
s43586-021-00039-w electrochemical system collected at selected frequencies, is tems21,22. EIS has also allowed for better comprehension

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 1

0123456789();:
Primer

Chronopotentiometry
of intermediate adsorption on electrode surfaces23,24, coefficient was also observed by cyclic voltammetry, but
An electrochemical technique particularly for multistep mechanisms, including elec­ this requires the use of convolution sweep voltammetry41
in which a current step is trochemical reactions coupled by adsorbed reaction or the comparison of cyclic voltammetry data (the shape of
applied to a working electrode intermediates, electrochemical reactions coupled with the voltammogram, the peak separation, the peak current)
and the potential response is
recorded as a function of time.
homogeneous or heterogeneous chemical reactions and with digital simulations42. Cyclic voltammetry and EIS
photoelectrochemical reactions25,26. Examples include are therefore complementary techniques, and the choice
Scanning electrochemical iron dissolution in an acidic solution27,28 and corrosion of which technique is best to use depends on the process
microscopy of magnesium29,30. EIS is a technique of choice for the that each user is seeking to characterize.
A local electrochemical
study of porous electrodes21,22,31 because the output is EIS is a type of transfer-function measurement,
technique that allows sensing
of the surface topography
directly affected by the porosity of the electrode, which commonly used in the analysis of linear time-invariant
and reactivity. may allow the pore size to be evaluated32–34. Together, systems2. In the case of an electrochemical system, the
advances in instrumentation and models provide the main difficulty is that the system must remain in a sta­
Interfacial capacitance capability for detecting multistep mechanisms and tionary state throughout the measurement. EIS uses a
The capacity of the electrical
double layer or the double
define modern EIS2,35. small-amplitude potential or current periodic perturba­
layer in series with a thin Cyclic voltammetry is often preferred to EIS for the tion to excite the electrochemical system at different fre­
film on the electrode surface study of electrochemical steps coupled with chemical quencies, as illustrated in Fig. 1. By measuring the response
(measured in farads per reactions, whose kinetics can be measured by varying (current or potential) of the system to this perturbation,
square metre).
the potential scan rate36–39. In contrast to cyclic voltam­ a transfer function is calculated that is the electrochemical
Potentiostat
metry where the entire potential domain is scanned impedance of the system in the case of an electrochemical
Electronic hardware for at a given scan rate, EIS offers the unique advantage of cell. The impedance, Z, can be expressed as2:
electrochemical experiments being able to perform measurements at different poten­ ~
maintaining a constant tials. Interestingly, the potential dependence of the V (ω)
potential difference between
Z (ω) = ~
charge-transfer coefficient was inferred from EIS measure­ I (ω)
the working electrode and the
ments on aromatic compounds with fast reaction kinet­ ~
reference electrode. V (ω) (1)
ics, revealing good agreement with the predicted values = ~ (cos ϕ(ω) + j sin ϕ(ω))
I (ω)
Rotating disk electrode using the Marcus theory for outer-sphere electron trans­
A set-up allowing control
fers40. The potential dependence of the charge-transfer = Z r + jZ j
of the rotation rate of a disk
electrode embedded in an
insulating circular plane in
order to enhance and tune a Electrochemical system b Perturbation signal c Electrochemical response
mass transport by generating Potentiostat
a thin diffusion layer with I(t) = I0+ ΔIsin(ωt –ϕ)
uniform thickness. RE WE
CE
Linear domain
selection
I (mA cm–2)

E0
E(t)

ΔI(t)

E(t)
E(t) =
Time E0+ ΔEsin(ωt)
High Low ΔE(t)
frequencies frequencies E (V(ref))

d Graphical representation e Equivalent circuit and model

CPEdl
|Z| (Ω cm2)

10–1
–Zj (Ω cm2)

WE Re
f (Hz)
φ (degree)

ZF

104
ZF
Zr (Ω cm2) 10–1 104 Ztot (ω) = Re +
f (Hz) 1 + ZF (jω)αQ

Fig. 1 | Required steps to acquire an EIS measurement. a | An electrochemical system with a working electrode (WE),
a counter electrode (CE) and a reference electrode (RE). The potential E(t) is applied between the working and reference
electrode, and the resulting current is measured at location (A). b,c | In this example, a periodic perturbation signal with
amplitude (ΔE) is applied between the WE and RE from high to low frequencies (part b), and the electrochemical response
to this perturbation is measured in the linear domain (part c). d,e | Impedance data are often presented in Nyquist or
Bode representations (part d) and may be represented by an equivalent circuit in which a mechanistic interpretation
of the system under study is used to extract a meaning for the faradaic impedance, ZF (part e). α, parameter associated
with the CPE; ω, angular frequency; CPEdl, double-layer constant-phase element; EIS, electrochemical impedance
spectroscopy; f, frequency; ϕ, phase angle; Q, parameter associated with the CPE; Re, ohmic resistance; Zj, imaginary
part of the impedance; Zr, real part of the impedance.

2 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

where ω is the angular frequency, related to the fre­ Box 1 | Kramers–Kronig transformations
quency f (in Hertz) by ω = 2πf, ϕ is the phase angle
between the input and output signals, and j = −1 is the Kramers–Kronig transformations were derived
∼ ∼ under the assumptions that a system is stable, such
imaginary number. The variables V and I are phasors,
which are complex time-invariant numbers that account that perturbations to the system do not grow, responds
linearly to a perturbation and is causal where a response
for the amplitude and phase of a sinusoidal function.
to a perturbation cannot precede the perturbation.
The electrochemical impedance, as defined by Eq. (1), These relations are widely used as part of a validation
is a frequency-dependent complex number, whose real procedure to identify the frequency ranges that may
part, Z r, is a frequency-dependent resistance and imag­ have been corrupted by instrument artefacts or by
inary part, Z j, is a frequency-dependent reactance . non-stationary behaviour.
Although the International Union of Pure and Applied The imaginary part of the impedance, Zj, can be
Chemistry (IUPAC) conventions43 hold that the real calculated from the real part through:
part should be represented by Z′ and the imaginary part 2ω ∞ Z r ( x ) − Z r ( ω)
denoted by Z″, the use of primes is too easily confused Zj(ω) = −
π
∫0 x2 − ω2
dx
with the notation for spatial derivatives in the engineer­
ing literature. Thus, the notation used in the present where ω is the angular frequency, Zj the imaginary part
work is that the real part is denoted by subscript ‘r’ and of the impedance and Zr the real part of the impedance.
the imaginary part is denoted by subscript ‘j’. If the high-frequency limit of the impedance is known
(Zr(∞)), the real part can be obtained from the imaginary
EIS measurements should be designed to conform to
part through:
the Kramers–Kronig relations (Box 1), which are derived
under the assumptions that the system under investiga­ 2 ∞ xZj(x) − ωZj(ω)
tion is linear, stable and causal (Box 2). The perturbation
Zr(ω) = Zr(∞) +
π
∫0 x2 − ω2
dx

amplitude required to achieve a linear response with an

appropriate signal-to-noise ratio is generally determined Direct integration of the Kramers–Kronig relations
experimentally for each system under study. Although for electrochemical impedance spectroscopy (EIS)
measurements has been replaced, in modern use,
a typical amplitude for potentiostatic modulation may
by regression of Kramers–Kronig-consistent circuit
be in the order of 10 mV, amplitudes as large as 1 V may be
models that provide extrapolation to zero and infinite
used for the impedance response of high-resistivity frequency limits that are consistent with the behaviour
lubricating fluids. The Kramers–Kronig relations may of electrochemical systems (see the figure).
be applied as a tool to determine which part of the meas­
40
ured frequency range is uncorrupted by instrument or Measurement
non-stationary artefacts. Kramers–Kronig
30
This Primer provides a description of the EIS tech­
–Zj (Ω)

nique and its principle, including the main aspects of 20

the measurement procedure, data representation and
analysis, but is not intended to cover the fundamentals 10
of electrochemistry that can be found elsewhere1,44,45. The
tools that will help design and perform a valid EIS meas­ 0
30
10 20 40 50 60 70 80
urement are described with specific attention drawn to Zr (Ω)
the choice of cell configuration, measurement settings and
data validation. Several examples and applications are pre­
sented in order to introduce the different methods of data between the working electrode and the counter electrode,
representation and interpretation. Finally, the challenges which complicates the interpretation of the impedance
of this technique and an outlook of how EIS will likely response. The three-electrode arrangement (Fig. 2b) is
evolve in the near future are presented and discussed. the classic device used for all analytical electro­chemical
measurements48,49 that employs a working electrode,
Experimentation which is the sample of interest, a counter electrode and
The use of EIS requires numerous steps, from the meas­ a reference electrode. The use of a reference electrode
urement procedure to data analysis. Despite the technique enables studying the working electrode response inde­
being incorporated into most modern potentiostat soft­ pendent of the processes taking place at the coun­
Complex number
An ordered pair of real
ware, the first step is to make a correct measurement, which ter electrode. A four-electrode set-up (Fig. 2c) is used
and imaginary numbers. is the focus of this section. The EIS measurement system for the characterization of electrolyte conductivity50,
includes the samples and the cells, the instruments and free-standing films51, an embedded rebar in concrete52
Resistance the connection of cells with the instruments46. and the interface between two immiscible electrolyte
The real part of the
solutions53. Two reference electrodes are positioned on
impedance of an electrical
circuit (measured in ohm Test samples and cells either side of the interface to measure the potential and
square metres). A cell with two, three or four electrodes can be used the two other electrodes allow for passage of the cur­
depending on the measurement to be carried out, as rent. The investigation of electrolytes or films without
Reactance shown in Fig.  2 (refs1,47). The two-electrode system the confounding influence of working electrode and
The imaginary part of the
(Fig. 2a) is used in cases where it is difficult to insert a counter electrode impedances can be done using the
impedance of an electrical
circuit (measured in ohm reference electrode, such as batteries and fuel cells. four-electrode system. The chosen cell configuration
square metres). The measured response is the sum of all contributions depends on the system being evaluated.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 3

0123456789();:
Primer

Ohmic resistance
The area of the counter electrode is normally much conductivity, the ohmic resistance and the electrode
An ionic resistance of the larger than that of the working electrode to minimize its dimension56:
electrolyte and electronic contribution to the cell impedance. This size difference is π r0
resistance of the electrode, especially important for two-electrode measurements in Re = (3)
wire and connection 4 κ
(measured in ohm square
which the measured impedance includes contributions
metres). from both the working and counter electrodes. Although where r0 is the disk radius and κ is the electrolyte con­
a uniform current and potential distribution on the ductivity. The high-frequency ohmic resistance can be
Capacitance working electrode are desirable, practical considerations estimated using Eq. (3) because the electrolyte conduc­
The ability of the
often force the use of electrode configurations in which tivity is often known, thus providing a check on the
electrochemical system,
specifically the electrode–
the current and potential distributions are not uniform. measurement quality.
electrolyte interface, to hold Frequency dispersion caused by non-uniform current High-frequency artefacts may appear when using
electrical charge (measured and potential distributions distort the high-frequency an ordinary reference electrode for measurements
in farads per square metre). impedance response, which complicates the interpre­ taken using conventional three-electrode arrangements
tation of the impedance data. The geometry-induced (Fig. 2b). The impedance of the reference electrode may
frequency dispersion can be avoided by using small decrease in the high-frequency region (above ~100 kHz),
working electrodes or by eliminating frequencies larger increasing the current flowing through it, which would
than the characteristic frequency, fc: result in a fluctuation of the reference potential and
afford abnormal impedance results. However, placing a
1 capacitor (the value of which depends on the electrolyte
fc = (2)
2πReC conductivity) in series with a platinum wire that is con­
nected in parallel to the reference electrode to bypass
where R e is the high-frequency ohmic resistance the high-frequency component of current or noise can
between the working and reference electrodes and relieve this abnormality57.
C is the capacitance of the working electrode–electrolyte
interface54. Instrumentation
The counter electrode is sometimes placed in a Along the course of developing the EIS measure­
separate compartment to prevent its reaction pro­ ment technique, instruments with different meas­
ducts from interfering with the working electrode. This urement principles have been proposed, including
compartment is ionically connected with the working alternating-current bridges 14,58, Lissajous curves 2,
electrode compartment, usually with a glass/ceramic phase-sensitive detection and lock-in amplifiers 59,
frit or asbestos threads. The reference electrode should frequency-response analysers (FRAs) 4,60, Laplace
provide a stable potential measurement against which transforms61 and wavelet transforms62. The dominant
to control the potential of the working electrode. There approach used by modern EIS instruments, provided by
are two approaches for the placement of the reference vendors such as Solartron, Biologic, Gamry, Metrohm
electrode. Placing the reference electrode close to the and Zahner, falls in the category of FRAs that employ
working electrode, such as using a Luggin tube in a the orthogonality of sines and cosines to determine the
glass cell55 (Fig. 2b), results in a smaller contribution real and imaginary parts of the complex impedance at a
of the ohmic resistance. By contrast, placing the ref­ specified frequency. The discussion in the present work
erence electrode far from the working electrode has accordingly emphasizes FRAs.
the advantage that, for some geometries such as a disk
electrode, a relationship exists between the electrolyte Accuracy contour plots. EIS accuracy depends on the
measurement frequency and the impedance of the sam­
ple. Each combination of device and associated cables
Box 2 | Linearity, stability and causality can be characterized by an accuracy contour plot, which
Linearity depicts the levels of accuracy for the magnitude and the
The constraint of linearity means that the response to an input perturbation is a linear phase angle of impedance based on the frequency and
function of the single-frequency or multi-frequency input. Because the current is a the magnitude of impedance. Such accuracy contour
non-linear function of potential for electrochemical systems, linearity is achieved using plots should be evaluated to see whether the test sample
a small input perturbation such that the current can be expressed by the first term of its of interest falls into the area with adequate accuracy63.
Taylor series expansion. The choice of an input perturbation magnitude is a compromise A generic accuracy contour plot for potentiostatic mod­
between the need to achieve linearity and the need for a sufficient signal-to-noise ratio. ulation is shown in Fig. 2d. Line A is based on the mini­
Stability mum current resolution and can be moved by changing
The constraint of stability means that the response to a perturbation cannot grow the potential perturbation amplitude. The absolute
with time. The current response to a step in potential must eventually decay to a new high-impedance limits of a potentiostat and associated
steady value. cables can be obtained by measuring the impedance in
Causality which the wires are not connected, simulating an infinite
The constraint of causality means that the response to a perturbation cannot cell impedance. The result yields the effective capaci­
precede the perturbation. This constraint ensures that the measured alternating- tance of the leads and instrumentation, shown as line B.
current response of the system to an external modulation must be correlated only to Line C corresponds to the maximum frequency capabil­
the applied alternating-current perturbation. Causality is especially important for ity of the instrument. The absolute low-impedance limits
electrochemical systems because measurements can take minutes to hours, depending can be obtained by measuring the impedance in which
on the frequency range used, and non-stationary behaviour leads to failure of causality.
the wires are shorted, simulating zero cell impedance.

4 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

a b c
WE/WSE CE/RE WE/WSE CE WE CE

RE WSE RE

Electrochemical cell WSE RE WE RE

f
g
d
A Twisted
cables

B √
log (|Z|(Ω cm2))

<1%,<2º >1%,
>2º C Parallel
cables

x
D
E
log (f (Hz))

V1 V2 A1 A2
e Cwire

Electrochemical workstation with impedance module

Rwire Lwire

Zcell Faraday cage

Fig. 2 | Guidelines for EIS measurements. a–c | Schematic of the two-electrode set-up (part a), three-electrode set-up
(part b) and four-electrode set-up (part c). d | Accuracy contour plot: line A, maximum measurable impedance; line B,
capacitive limit; line C, maximum measurable frequency; line D, inductive limit; line E, lowest measurable impedance.
e | Circuit representation showing the influence of wire capacitance, resistance and inductance on the measured
impedance. For most measurements, the contribution of wire properties can be neglected, and measured impedance
is the same as the cell impedance. Acceptable measurements may be made in the region with <1% error in magnitude and
<2° error in phase. Measurements in the region marked >1%, >2° will show anomalous features caused by the instrument
and cabling. f | Schematic of a wire connection to the electrochemical cell. g | Twisted wires suppress mutual induction.
A1, A2, current connectors; Cwire, capacitance of wire; CE, counter electrode; EIS, electrochemical impedance spectroscopy;
f, frequency; Lwire, inductance of wire; RE, reference electrode; Rwire, resistance of the wire; V1, V2, voltage connectors;
WE, working electrode; WSE, working sense electrode; Zcell, cell impedance.

The result yields the effective inductance (line D) and where ZC,wire, ZL,wire and Rwire are the impedances associated
resistance (line E) of the leads and instrumentation. with the capacitance, inductance and resistance of the
A zone is indicated in which impedance measurements wires, respectively. Instrument vendors usually use short
may be made with error less than 1% in magnitude and shielded wires to reduce the contribution of the wire
2° in phase. A second zone corresponds to errors larger capacitance and inductance, and often provide accuracy
than 1% and 2° for magnitude and phase, respectively. contour plots with wires coming with the instrument.
Measurements outside the boundary formed by lines However, the positions of the limits in accuracy contour
A–E are not possible. plots are strongly affected by the wires used to connect
The measured impedance can be represented by the the instrument to the cell. To this end, users should
electrical circuit (Fig. 2e), and the relationship between always generate a new accuracy and check whether they
measured impedance, Zmeas, and cell impedance, Zcell, need to use longer wires for special purposes, such as
is given by: connecting the cell far away from the instrument.
The wires connecting the equipment and cells
in EIS measurements form two loops, one includes
(R wire + Z L,wire + Zcell ) × Z C,wire
Z meas = (4) the two wires bearing the current and the other bears the
R wire + Z L,wire + Zcell + Z C,wire two wires sensing the voltage (Fig. 2f). The contribution

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 5

0123456789();:
Primer

of wire properties can be neglected and the meas­ changes associated with galvanostatic modulation
ured impedance is the same as the cell impedance for depend on the frequency-dependent impedance of the
most measurements. However, for systems exhibiting system. As the impedance of the systems may change
small impedance, the use of long wires may give rise by orders of magnitude with frequency, the potential
to high-frequency artefacts because the magnetic field swing at low frequencies from a fixed galvanostatic
generated by the alternating current-carrying leads will modulation may cause non-linear behaviour. In this
drive an alternating potential in the sense leads (Fig. 2f). case, the amplitude of the current perturbation should
This phenomenon is called mutual inductance, which be adjusted as a function of frequency. Algorithms for
can be reduced by minimizing the loop area between frequency-dependent galvanostatic modulation have
the two current wires and the two voltage wires via a been developed65,66. Distortions of the expected ellipse
tight-braided twist64 (Fig. 2g). Another solution includes in low-frequency Lissajous plots (current as a function
maximizing the distance between the current-carrying of potential) provide a good indicator for non-linear
and the voltage-sensing wires63. Finally, orientating the response67,68. If distortions from a perfect ellipse are
clamp (bayonet) of the current and voltage wires in a observed, the perturbation amplitude should be reduced.
cross-configuration — that is, perpendicular to each Monitoring of the spectral content at higher harmonics
other — may help to further reduce this interaction. of the response signal may also be used to determine the
suitable range of perturbation amplitude. These exper­
Settings. The quality of the impedance measurement imental approaches are suitable for both potentiostatic
is controlled by a series of parameters, including the and galvanostatic modulation.
selection of potentiostatic or galvanostatic modula­ The power of EIS lies in its ability to probe processes
tion, the perturbation amplitude, the frequency range in systems across a wide frequency range. Ideally, the fre­
and the number of cycles used to perform the measure­ quency range should be set to match the dynamic range
ment at each frequency. In principle, the EIS meas­ of the system under study. Modern EIS instruments can
urement can be obtained either by applying a current easily cover a broader range from megahertz to micro­
perturbation and measuring the potential response, in hertz, spanning 12 orders of magnitude. In practice,
the galvanostatic mode, or by applying a potential per­ the accuracy of the equipment may drop significantly
turbation and measuring the current response, in the at high frequencies, as shown in Fig. 2d, and the meas­
potentiostatic mode. The choice of potential or current urement duration may become prohibitively long or the
regulation depends on the electrochemical systems, system may no longer be stationary in the low-frequency
and a first criterion for this selection is the shape of the limit (Box 1). As a result, the typical frequency range
steady-state current potential curves, which should be for electrochemical measurements, including energy
generated before the EIS measurement as a best prac­ conversion as well as storage device and corrosion,
tice. When the slope of the current–potential curve is 10 kHz–10 mHz. Usually, seven to ten points per
is very sharp (for example, the dissolution of zinc), frequency decade (for example, from 100 Hz to 10 Hz),
the galvanostatic mode is usually preferred because the equally spaced logarithmically, are required for meas­
potentiostatic mode may lead to overloading of the cell uring an impedance spectrum with sufficient accuracy
and the current measurement circuits of the potentio­ for a detailed data analysis. It is preferable to start the
stat. For steep slopes of the current–potential curve, measurement from the high-frequency limit sweeping
a small change in potential might lead to a large change towards the low-frequency limit, because the operator
in current. In general, galvanostatic impedance is more has the opportunity to grasp the main behaviour of the
suitable for non-invasive probing of metal corrosion system in a short period of time. If anything goes wrong,
at the open-circuit potential and for measuring most such as current overloads or excessively noisy signals,
high-energy electrochemical devices, where imped­ the operator can abort the measurement and make any
ance is low and current levels are high. However, for necessary change or adjustment to the settings or the cell.
electrochemical systems with larger impedance, the The stochastic error structure or noise level of the
potentiostatic mode is generally employed. measurement is influenced by the number of cycles used
The amplitude of the perturbing current or poten­ to measure the impedance at each frequency. The preci­
tial must be determined experimentally. The pertur­ sion of the measurement increases with the square root
bation amplitude should be small enough to make the of the integration cycles. This effect from the number of
system linear, but as large as possible to get the best cycles used is particularly worthwhile at high frequen­
signal-to-noise ratio. Selection of the amplitude depends cies because the noise content is high and the time over­
both on the polarization curve of the system and the head for repetition is low. At very low frequencies, this
noise level of the measurement system. The optimal effect is less significant owing to the increased time over­
perturbation amplitude is not necessarily the same over head, on the one hand, and the change in the nature of
the entire frequency range, and it may be compulsory to the predominant noise from external electromagnetic
adapt it to make a good measurement because electro­ interference to the system non-stationarity on the other.
Frequency decade chemical systems involve different processes. For the For some instruments, an auto-integration mode is used
A unit for measuring frequency potentiostatic EIS mode, an alternating-current poten­ to select the number of cycles, which is usually three or
ratios on a logarithmic scale, tial signal with a peak to peak amplitude of 5–15 mV larger. By contrast, the user is allowed to choose between
with one decade corresponding
to a ratio of ten between
is commonly used. However, for a system with large fast, normal and low-noise options that influence the
two frequencies (an order impedance or significant noise, the amplitude may be number of cycles used for the measurement for other
of magnitude difference). increased so long as linearity is assured. The potential instruments.

6 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

Introducing a delay of one or two cycles before the stochastic error structure and the frequency range that
impedance measurement will help reduce the errors is consistent with the Kramers–Kronig relations71–73.
because a transient response may occur in the cell during
the change from one frequency to another. Similarly, Results
there is usually a start-up transient before the entire The first step towards data analysis is the use of graph­
sequence of the measurement, which tends to corrupt ical methods to visualize and interpret the impedance
the first frequency measured. A simple way of obviating data. To emphasize a specific feature or behaviour,
this problem is to ignore the first frequency in subse­ impedance data need to be presented in different for­
quent regression analyses. In addition, the measured mats that include the Nyquist representation for mass
results within ±3 Hz of the line frequency and its first transfer and kinetic behaviour; the Bode representation
harmonic (for example, 50 Hz and 100 Hz in Europe for frequency-dependent behaviour; the admittance
or 60 Hz and 120 Hz in North America) are prone to format for capacitive behaviour at high frequency; and
stochastic noise2. Data at these frequencies should be the complex-capacitance format for capacitive behaviour
discarded in regression analyses, or, preferably, not of a dielectric system. The parameters associated with
measured. The user can check the EIS plot to spot any resistance and capacitance can be extrapolated by graph­
dispersed points near the line frequencies. If there is ical methods. Regression and statistical analysis provide
dispersion, just remove these data before regression. more advanced quantitative analysis of impedance data.
The analysis includes checking for consistency with the
Environment control Kramers–Kronig relations, finding the error structure
Many electrochemical parameters have a sensitive of impedance measurements and extrapolating param­
dependence on temperature. To ensure reliable and eters by regression of process models. The methods and
repeatable measurements, tight control of the cell temper­ application to sample data are illustrated in this section.
ature is required. Temperature control may be achieved
using a thermal bath for glass cells, a heater and fan feed­ Graphical representation
back control for fuel cells and an environmental cham­ Interpretation relies partly on data representation, which
ber for batteries. This precaution is especially necessary is a critical aspect in EIS analysis. The commonly used
for measurements at low frequencies. For low-current representations are shown and discussed here.
measurement, cells must be placed in a Faraday cage
to shield the electrical noise from the environment. Potential–current versus time and Lissajous curves.
The Faraday cage should be connected to the ground The relation of the sinusoidal potential and current as
to eliminate the voltage difference between the interior a function of time at a frequency of 20 Hz is shown in
of the cage and the ground reference. Otherwise, this Fig. 3a. The amplitude and phase shift contain impor­
voltage difference tends to couple capacitively to the tant information of impedance data. The phase shift
electrodes, leading to only partial shielding of the exter­ between the input and output signals can be seen more
nal electromagnetic interference69. Mechanical vibra­ clearly in the Lissajous curve (Fig. 3b), obtained by plot­
tion may become an issue for systems with extremely ting the output signal as a function of the input signal at
small current and high sensitivity to electrical contact, a specific frequency. In the early years of EIS, Lissajous
such as when probing EIS of a single particle. In these plots were used to perform EIS measurement at low and
and similar situations, an anti-vibration table should high frequencies, respectively. Moreover, the Lissajous
be used70. plot (Fig. 3b) has several advantages when monitored
during the course of the experiment. If distortions of
Debugging and correction the Lissajous figure are seen, the perturbation ampli­
When doubt arises about the quality of a measurement, tude applied to the system should be reduced, at least in
debugging tests could be performed on an electrical the frequency domain where the degeneration of the
circuit that comes with the potentiostat or an electri­ ellipse is observed. If the Lissajous plot can be made to
cal circuit with similar electrical characteristics to the have an elliptical shape by adjusting the perturbation
electrochemical system under study. The use of an amplitude, the measurement can be said to be linear.
appropriate electrical circuit with known impedance In addition, a shift of the electrical signal along the x
response allows for identifying instrument and wir­ or y axis, usually occurring at the low-frequency range,
ing artefacts. An oscilloscope, preferably analogue, is implies a shift of the system to a non-stationary state,
extremely helpful in identifying the presence, magni­ resulting in an erroneous measurement in this frequency
tude, frequency range and nature of the electromagnetic domain. For example, this can be the case for a battery, if
noise. In addition, the non-linearity can be captured if the amount of charge changes during the measurement,
a significant component is detected at high multiples or for a corroding system evolving naturally, even slowly,
of the perturbation frequency. Some EIS instruments over time. Accordingly, even if the current measuring
have a digital oscilloscope in the menu that should be devices are becoming more and more efficient, their
exploited in the initial phase to determine the appro­ mode of operation is not generally known to the user.
Oscilloscope priate range of the amplitude of the perturbing signal. The recorded current signal, for example, can be an
Electronic hardware monitoring Distortions of the expected ellipse in low-frequency average over a few cycles or a period of time. Therefore,
the time-domain signals
that are processed in the
Lissajous plots provide a good indicator for non-linear connecting an oscilloscope to trace the live Lissajous
impedance measurement response67,68. Impedance data can also be evaluated by curve is a way to guarantee the validity of the measure­
(for example, Lissajous plot). regression of a measurement model to determine the ments. If the traces of the live Lissajous curves overlap,

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 7

0123456789();:
Primer

a b E0
(a) f = 20 Hz f = 20 Hz
0.42
E0 11

E (V(SCE))
0.40

I (μA cm–2)
0.38 ∆T I0
10
T=1/f
11
I0
I (μA cm–2)

10
9
9

0.0 0.1 0.2 0.38 0.40 0.42

Time (s) E (V(SCE))

c 30
(a)
d –50
Raw data
Corrected for Re
20
|Z| (Ω cm2)

–40
10 0.1 Hz
–30

Zj (Ω cm2)
0
0
–20
φ (degree)

–30

–60 –10
3.3 kHz
–90
0
10–2 10–1 100 101 102 103 104 105 0 10 20 30 40 50
f (Hz) Zr (Ω cm2)

e f g –100
–100
20 Faradaic
–75 time domain
e– –75

Double-layer
10 –50
Oxel Oxsol
charging
I (mA cm–2)

I (mA cm )
–2

Mass –50 –25

transfer 0
Redel Redsol 0 –6
10 10–4 10–2 100
–25
–10 Scan rate (Vs–1)
1
4
10 0
–20
–0.6 –0.4 –0.2 0 0.2 0.4 0 5 10 15
E (V(ref)) Time (s)

Fig. 3 | EIS data of an ideally polarizable electrode obtained for a redox couple in solution. a | Temporal measurements
of potential and current as a function of time at a frequency, f, of 20 Hz (red line, current density; blue line, potential).
b | Lissajous curve (current as a function of potential) plotted from the results presented in part a. c | Bode representation:
modulus and phase of impedance as a function of frequency. d | Nyquist representation: imaginary part of impedance, Zj,
as a function of the real part, Zr, in a complex plane with frequency as a parameter. e | Representation of a mass transfer
controlled electrochemical reaction involving an oxidized species Ox and a reduced species Red. f | Cyclic voltammetry
with scan rate as a parameter. g | Chronoamperometry with time on a linear and logarithmic (inset plot) scale. All results
shown in parts f and g were obtained for the same set of parameters for a planar-disk electrode (A = 0.2 cm2) under
assumption of a Butler–Volmer relationship for a quasi-reversible process with semi-infinite diffusion: k0 = 10−2 cm s−1;
α = 0.5; E0 = –0.1 V/NHE; Dox = Dred = 10–5 cm2 s−1; Cox* = 10 mM; Cred* = 0 mM; Re = 5 Ω; Cdl = 10 µF. A, working electrode area; α,
parameter associated with the CPE; Cdl, double-layer capacitance; Cox and Cred, concentration of the oxidized and reduced
species, respectively; CPE, constant-phase element; Dox and Dred, diffusion coefficient of the oxidized and reduced species,
respectively; E0, standard reaction potential; el, subscript indicating electrode; EIS, electrochemical impedance
spectroscopy; k0, reaction rate constant; ϕ, phase angle; NHE, normal hydrogen electrode; Re, ohmic resistance;
SCE, saturated calomel electrode; sol, subscript indicating solution; Z, impedance; Zj, imaginary part of the impedance;
Zr, real part of the impedance.

8 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

the measurement at that frequency is not experiencing used to obtain the characteristic time constant. The
non-stationary behaviour, which means the system does low-frequency straight line is the Warburg impedance
not change with time. for a mass transfer process. The time constant, τ, is
determined by:
Bode and Nyquist representations. The Bode (Fig. 3c) and
Nyquist (Fig. 3d) representations are obtained over a wide 1
τ= (5)
frequency range (typically between 100 kHz and 10 mHz) 2πfc
and are therefore 3D representations of the experimental
results, namely frequency (f), modulus (|Z|) and phase where the time constant τ = RC dl (where C dl is the
angle (ϕ) for the Bode diagram or frequency, real part double-layer capacitance) and fc is the characteris­tic
(Zr) and imaginary part (Zj) for the Nyquist diagram. relaxation frequency. Several processes exhibit similar
The Bode and Nyquist representations are the most relaxation frequencies, which results in overlapping
commonly used in the literature, and they generally time constants. However, it is still possible to distinguish
allow a preliminary analysis of the system, leading to these different time constants using strategies such as
the identification of elementary processes involved in the regression analysis, different graphical representa­
mechanism. A typical Bode plot, where the phase angle tions2,74 or a method based on the distribution function
and modulus are plotted as a function of frequency, is of relaxation times75–77.
shown in Fig. 3c. The phase angle and modulus are cal­
culated from the real and imaginary parts of the meas­ Other representations. Other methods for representing
urements. The Bode plot is an intuitive illustration of impedance, such as the complex-admittance format and
the impedance change with frequency. The phase angle the complex-capacity format, are less used compared
usually tends towards 0° at high frequency owing to the with the representation methods discussed above. The
ohmic resistance of the electrolyte, shown as blue dots choice of representation depends on the specific needs
in Fig. 3c. The ohmic resistance-adjusted phase angle has for data visualization and interpretation. The admit­
an asymptotic value of −90° at high frequency for the tance is also a transfer function with real and imaginary
ideally polarizable electrode. If there is constant-phase parts, which is inversely related to the impedance. The
element (CPE) behaviour, the asymptotic value at high admittance format emphasizes the capacitive behaviour
frequency would be lower than 90°. Thus, plots of the at high frequencies. It is commonly used in solid-state
ohmic resistance-adjusted phase angle provide a direct systems. Complex capacitance, C(ω), is defined as:
representation format for CPE behaviour or frequency
dispersion behaviour (Box 3). A typical Nyquist plot, in 1
C (ω) = Cr + jC j = (6)
which the imaginary part is plotted against the real part jω(Z − Re )
of impedance data, is shown in Fig. 3d. The Nyquist rep­
resentation emphasizes larger values of impedance that for an impedance corrected by the ohmic resistance.
show the influence of mass transfer and reaction kinet­ Like the Nyquist plot, the complex-capacitance format
ics. Each symbol is the data point measured at a specific obscures the frequency dependence of the data. The
frequency and solid dots are used to designate some typi­ high-frequency asymptotic value can be used to deter­
cal frequencies. In the case for these sample data, 1.6 kHz mine the capacitance. Therefore, the complex-capacitance
is the characteristic frequency for the charge-transfer format is usually employed to represent systems for
process of the electrochemical reaction, which can be which the capacity is of interest.

EIS and electrical circuits. Electrochemical impedance

Box 3 | Constant-phase element
spectra can be represented as a combination of electrical
The constant-phase element (CPE) is often used to improve the fit of models to impedance (resistance, capacitance or inductance) and electrochem­
data. CPE parameters α and Q cannot be used directly to extract a capacitance; however, ical (faradaic) impedances. In particular, thin films on
a capacitance may be extracted under assumption of a distribution of time constants an electrode surface behave like dielectrics, as does the
along the electrode surface or through a film (see the figure). Not all time-constant
double layer that forms spontaneously on an electrified
distributions give rise to a CPE. The ohmic resistance-corrected phase angle, ϕ, provides
interface. Therefore, electrical circuits are often used
a convenient way to determine whether a time-constant distribution is represented by a
CPE. ω, angular frequency; Z, impedance; Zj, imaginary part of the impedance; Zr, real part to analyse experimental results, but such an approach
of the impedance. masks the physical and chemical properties of the inter­
face2,4,13. All Kramers–Kronig-consistent impedance
α=1 α<1 1 R spectra may be fit by an equivalent circuit, which is the
ZCPE = Z=
(jω)αQ 1 + (jω)αQR point of using a measurement model (discussed further
in Reproducibility and data deposition). The issue is
that such models are not unique, and fitting by use of
α=1 an electrical circuit does not necessarily provide useful
–Zj

–Zj

information concerning the system under investigation.

α<1
For example, batteries are most often analysed in terms of
equivalent electrical circuits because of the complexity
of the system78; however, mechanistic models make it
possible to take into account all physical chemistry in
Zr Zr an increasingly precise manner79–82.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 9

0123456789();:
Primer

Users often prefer one technique over others, but it is also available, such as a recent Python-based program
important to note that EIS and other widely used electro­ for regression that allows both a measurement model
chemical techniques are complementary. For illustration, analysis of the data and the ability to regress arbitrary
a system involving reduced species dissolved in solution functions73.
(Fig. 3e) is presented using cyclic voltammetry (Fig. 3f), The quality of the fit may be determined using a
chronoamperometry (Fig. 3g) and EIS (Fig. 3c,d). graphical comparison of the model result and the imped­
The sweep-rate-dependent increase and decrease ance data or, if the experimental error structure is known,
in value of the current with increasing potential in the by the weighted χ2 statistic. Both simplex regression and
cyclic voltammetry curves (Fig. 3f) reveals the presence Levenberg–Marquardt regression strategies are used87,88.
of a diffusion-limited system. The difference between Despite being very sensitive to selection of initial guesses,
the peak potentials at a given scan rate allows the deter­ the Levenberg–Marquardt regression method provides
mination of electron-transfer kinetics. The variations confidence intervals for regressed parameters, thus alert­
in peak intensities as a function of the square root of ing the user to situations where regressed parameters do
the potential scan rate allow calculation of the diffusion not have statistical significance. The simplex routine is
coefficient. less sensitive to initial guesses and, as a result, is easily
When the same electrochemical system is studied used, but the absence of confidence intervals represents
by applying a potential step, little information is read­ a severe shortcoming. A hybrid approach is preferred in
ily apparent from the resulting current as a function of which simplex is used to provide good initial guesses for
time, shown in Fig. 3g. Inflections in the current response the Levenberg–Marquardt regression.
can be seen when time is plotted on a logarithmic scale,
shown in the inset of Fig. 3g, and these inflections high­ Application to sample data
light the presence of two time constants The current To demonstrate a systematic procedure for interpretation
response is dominated by the cell time constant for the of impedance data, the data analysis methods discussed
shorter times, that is, the electrolyte resistance, Re, in above are applied to sample impedance measurements
series with the double-layer capacitance, Cdl. By con­ in the following sections. The error analysis associated
trast, the diffusion of electroactive species is responsible with these sample data is presented in Reproducibility
for the current decay for longer times owing to their con­ and data deposition.
sumption at the electrode surface. The inset in Fig. 3g
shows the current response to the ReCdl time constant, Sample data. The sample data were collected for the
characteristic of the double-layer charging. This value impedance response of a platinum disk electrode 5 mm
defines the lower limit timescale (two to three times in diameter rotating at 120 rpm in an electrolyte con­
the ReC dl value) at which kinetic information can be taining 0.01 M K3Fe(CN)6, 0.01 M K4Fe(CN)6 and 1 M
retrieved, and can be tuned using an electrode of appro­ KCl as the supporting electrolyte. Overall, this system
priate size. The use of microelectrodes reduces the ohmic is widely used as a reference for electrochemical meas­
resistance and, thus, the ReCdl time constant (as shown urements and further experimental details are avail­
by Eq. (3)), making it possible to study systems with fast able elsewhere89. The first frequency measurement
charge transfer, down to the 100 µs domain83. was collected after 14 s, and the subsequent measure­
Similarly, detailed analysis of the EIS diagram ments were collected after 8.9 h of immersion. Each
(Fig.  3d) allows these same parameters to be deter­ measurement required about 19.5 min for acquisition.
Warburg semi-infinite mined. The width of the high-frequency capacitive As shown in Fig. 4a, a substantial difference is seen
diffusion impedance loop yields the charge-transfer resistance associated between the first measurement made and the measure­
An impedance element
describing the diffusion
with the kinetics of the electrochemical reaction, and ments made after 8.9 h. A film is hypothesized to form
behaviour of the electrolyte the frequency at the peak of the high-frequency loop on the platinum disk, and this film is seen to influence
in the absence of convection yields the double-layer capacitance (Eq. (5 )). The the impedance results89. The regression program used
with a diffusion layer that can tail forming a 45° slope in the low-frequency range to analyse the data in the present work is available for
spread to infinity.
is attributed to the diffusion of electroactive species free non-commercial use73. The installation file includes
Simplex regression (Warburg semi-infinite diffusion impedance). Interestingly, all of the sample data discussed in this subsection, sam­
A robust mathematical the current response to a small-amplitude potential step ple process models including the one used to analyse
algorithm used to solve results in a curve (Fig. 3g) that contains the same informa­ the data and the Python code used to calculate the
non-linear least-squares tion as the EIS response, which is enabled by transform­ convective diffusion impedance response of a rotating
curve-fitting problems that
is less sensitive to initial
ing data from the time domain to the frequency domain disk electrode.
values but does not provide using the Fourier transform84,85.
confidence intervals for the Process model. The depressed high-frequency semicir­
resulting parameter estimates. Regression cular loop seen in Fig. 4a can be attributed to the faradaic
Complex non-linear least-squares regression is used to reaction, and the shape of the larger low-frequency loop,
Levenberg–Marquardt
regression fit mathematical models to impedance data. The com­ with a 45° line at higher frequencies, can be associated with
A mathematical algorithm plex non-linear least-squares approach has the advan­ mass transfer. For the ferri/ferrocyanide system, mass
used to solve non-linear tage that a common set of parameters are obtained by transfer is expected to influence the faradaic reactions.
least-squares curve-fitting simultaneous regression of the model to both real and The depressed character of the high-frequency loop
problems that is sensitive
to initial values and provides
imaginary data86. Commercial programs such as ZPlot is attributed to frequency dispersion that is approx­
confidence intervals for the or ZSimpWin are available to fit electrical circuit mod­ imated by a CPE (Box 3). The origin of the observed
resulting parameter estimates. els to impedance data. Non-commercial programs are CPE is uncertain and the usual explanations in terms

10 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

a –20 b α, Q
Start time
14 s Re
32,135 s (8.93 h)
–15 33,305 s (9.25 h) 0.146 Hz
34,472 s (9.57 h)
Zj (Ω cm2)

Zd
–10
Rt

–5 0.99 Hz
0.022 Hz

0
0 5 10 15 20 25 30 35 40
Zr (Ω cm2)

Fig. 4 | Sample analysis of impedance data. a | Impedance data in a Nyquist format for reduction of ferricyanide on a
platinum disk 5 mm in diameter rotating at 120 rpm after 14 s, 32,135 s, 33,305 s and 34,472 s. Lines represent the weighted
fit of Eq. (7) to the data. The difference between the first measurement (14 s) and measurements taken after 8.93 h can be
attributed to slow formation of a film on the electrode surface. b | Electrical circuit corresponding to convective diffusion
to a rotating disk electrode with high-frequency constant-phase element (CPE) behaviour. α, parameter associated with
the CPE; Re, ohmic resistance; Rt, charge-transfer resistance associated with electrode kinetics; Q, parameter associated
with the CPE; Zd, diffusion impedance associated with transport of reactive species to the electrode surface; Zj, imaginary
part of the impedance; Zr, real part of the impedance. Graph generated in panel a uses data reported in ref.89.

of distributions along the surface of the electrode90 or subscript i2. The dimensionless diffusion impedance for
through a film91 are not likely to apply. However, the cou­ a stationary film in terms of a diffusion time constant
pling between faradaic and charging currents may be a τ = δN,i2 / Di, where δN,i is the Nernst stagnant diffusion
possible reason92, but more work is needed to explore layer thickness:
this hypothesis.
The circuit representation of the process model is −1 tanh jωτ
= (10)
shown in Fig. 4b, and the corresponding mathematical ′
θ (0) jωτ
expression is given as:
is often used to account for the impedance of a rotating
R t + Zd disk. The advantage of Eq. (9) over Eq. (10) is that it pro­
Z = Re + (7)
1 + (jω)α Q(R t + Z d ) vides a more accurate model for the diffusion impedance
response of a rotating disk electrode with no additional
where Re is the ohmic resistance, Rt is the charge-transfer parameters.
resistance, Zd is the diffusion impedance, and α and Q
are parameters for a CPE, that is, Z CPE = ((jω)α Q )−1. The Regression. The regression was weighted by the error
diffusion impedance is expressed in terms of a diffusion structure deduced from the three measurements collected
resistance, Rd, and a dimensionless diffusion impedance, after 8.9 h of immersion, such that:
–1/θ′(0), as: 2
σr = σj = β Z r| + γ| Z (11)
 −1 
Z d = R d  ′  (8)
 θ (0)  where σ is the standard deviation with subscripts r and
j representing the real and imaginary parts, respec­
where θ is the dimensionless concentration phasor tively, β = 0.000291 and γ  = 3.76 × 10−5 Ω−1 cm−2. Five
scaled to its value at the electrode surface, and θ′(0) is the low-frequency data points were found to be incon­
derivative with respect to position evaluated at the elec­ sistent with the Kramers–Kronig relations and were
trode surface. The dimensionless diffusion impedance removed from the first data set (measured at 14 s). No
is given by: low-frequency data were removed from the data col­
lected after 8.9 h. Data with frequencies above 4.3 kHz
−1 Z (1) Z (2) were removed to eliminate the contribution of ohmic
′
= Z (0) + 1/3 + 2/3 + … (9)
θ (0) Sc Sc impedance associated with non-uniform current and
potential distributions. The resulting regression results
where Z(0), Z(1) and Z(2) are the three complex solutions are illustrated by lines in Fig. 4a and the values of the
to the convective diffusion impedance corresponding to fitted parameters are summarized in Table 1, which
three terms in the series expansion for the axial veloc­ were all statistically significant. The Schmidt number
ity near the disk electrode, and the Schmidt number is for the measurement at 14 s is 1076 ± 14, and is on the
expressed as Sc = μ/ρDi, where μ is the electrolyte vis­ order of the value expected for a bare electrode under
cosity, ρ is the electrolyte density, and Di is the diffu­ the measurement conditions. There was an observed
sion coefficient for the reacting species denoted by the increase in the Schmidt number after 8.9 h of immersion.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 11

0123456789();:
Primer

Table 1 | Results from weighted regression of Eq. (7) to the sample data presented in Fig. 7
Parameter (units) 14 s 32,135 s 33,305 s 34,472 s
Sc 1,076 ± 14 1,235.7 ± 7.8 1,244.8 ± 8.0 1,256.5 ± 8.3
Re (Ω cm2) 1.252 ± 0.072 1.381 ± 0.011 1.383 ± 0.011 1.382 ± 0.011
Q (µF sα–1 cm−2) 460 ± 68 59.8 ± 1.2 59.4 ± 1.2 59.6 ± 1.3
Rd (Ω cm2) 31.719 ± 0.066 33.199 ± 0.031 33.289 ± 0.032 33.390 ± 0.033
Rt (Ω cm2) 1.273 ± 0.078 3.137 ± 0.013 3.161 ± 0.013 3.188 ± 0.013
α 0.702 ± 0.026 0.9048 ± 0.0.0028 0.9057 ± 0.0.0028 0.9052 ± 0.0.0029
χ2/ν 139 53 56 59
C (µF cm–2) 14.4 21.4 21.5 21.4
fc (kHz) 8.8 5.4 5.4 5.4
Current (mA) 90.3 91.3 91.3 90.2
Potential (mV) 215.6 215.7 215.7 215.7
Confidence intervals given are ±1σ. The χ2/ν statistic should be in the order of unity for perfect regression. These results show
that the model could be further improved. The effective capacitance and the characteristic frequency were calculated from
the regressed parameters. The current and potential were measured after each electrochemical impedance spectroscopy (EIS)
measurement. α, parameter associated with the CPE; C, capacitance; CPE, constant-phase element; fc, characteristic relaxation
frequency; Q, parameter associated with the CPE; Rd, diffusion resistance; Re, ohmic resistance; Rt, charge-transfer resistance; σ,
standard deviation; Sc, Schmidt number; ν, degree of freedom for the regression.

Similarly, the charge-transfer resistance and the diffu­ diagnostics in situ and in operando, to identify the pres­
sion resistance increased with elapsed time. The effective ence and evaluate the relative importance of different
capacitance, Ceff, was evaluated based on the values from processes or to extract detailed physical and chemical
regression results, following the Brug formula90: information about a process of interest.

α α Process decoupling and diagnosing

 ReR t  (1− )/
1/α 
Ceff = Q   (12) In the initial stage or pretest, EIS may be used to probe the

 Re + R t  general behaviour of the system, similar to the use of
the cyclic voltammetry technique in the time domain.
where Re is the ohmic resistance, Rt is the charge-transfer In cyclic voltammetry, voltage is swept across a wide
resistance, and α and Q are the parameters associated range (1~2 V) to expose current peaks corresponding
with the CPE. Equation (12) relates the CPE parameters to different electrochemical reactions, in the anodic
to capacitance for a surface distribution of time con­ and cathodic scans. In EIS, frequency is swept across
stants. The values presented in Table 1 are in the range several orders of magnitude to expose impedance fea­
expected for double-layer capacitance. The characteristic tures and signatures at different timescales. Major
frequencies, calculated using Eq. (2), are larger than the features can be easily identified, whereas abnormal
maximum frequency used in the regression, confirming behaviours can be easily spotted.
that the influence of ohmic impedance was avoided. EIS provides a powerful means for decoupling pro­
The analysis presented here demonstrates the dis­ cesses occurring at different timescales as compared
criminating power of impedance spectroscopy. The with time-domain methods. The typical EIS spectra
method allows an in situ interrogation of an electro­ and the corresponding physical processes of iron cor­
chemical system. The method is non-invasive, such rosion in sulfuric acid solution93, solid oxide fuel cells94,
that the potential excursion is very small compared lithium-ion batteries80 and polymer electrolyte fuel cells95
with cyclic voltammetry, and can be used to evaluate are shown in Fig. 5. Generally, electrolyte resistance
the ohmic resistance, the electrode capacitance, the dominates at the high-frequency range, diffusive imped­
charge-transfer resistance, which is inversely related to ance dominates at the low-frequency range and interfa­
the reaction rate constant, and the Schmidt number. The cial processes dominate at the intermediate-frequency
EIS measurement is very sensitive to the evolution of range. Inductive loops are frequently observed, as
electrode properties, in this case showing the influence shown here for a corrosion process and for a polymer
of growth of a film on the electrode. The presence of a electrolyte fuel cell. The inductive loops for the corro­
film increases the charge-transfer resistance, increases sion of iron in sulfuric acid were attributed to the cou­
the electrode capacitance and influences mass trans­ pling of electrochemical reactions by three intermediate
fer. The systematic changes revealed by the impedance species28,96. Depending on the operating conditions, the
response are not visible in the steady-state current at the inductive loops for the polymer electrolyte fuel cell have
applied potential given in Table 1. been attributed to water transport97, reactions coupled
by oxidation of the platinum catalyst98 or adsorbed
Applications intermediates (discussed further in Limitations and
EIS is widely applied across diverse electrochemical optimizations).
cells, devices and systems, including energy storage and These physical processes can be further decoupled.
conversion, corrosion, sensors and biomedical systems. For example, a physics-based model for lithium-ion bat­
This method can be used to provide simple system teries was able to resolve the controversial attribution

12 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

of EIS features at low frequency to diffusion in either the electrolyte) may be differentiated even without the
electrolyte or active particles79,80,99. The contributions use of a reference electrode100,101. The diagnostic results
from different parts of the cell (the anode, the cathode, from EIS can help identify the rate-determining process,

a Typical EIS of iron corrosion in sulfuric acid solution b Typical EIS of solid oxide fuel cells
Solid electrolyte

Anode
Cathode
~10 nm

H2

Current density
2O2– +
3 3
O2 1 2 O2 H2O

~100 nm

–Zj
quency
sing fre
–Zj

Activation Passivation Decrea

domain domain 3
Potential / V(SSE) 2
1

Zr

1 Ion conduction within grains

2 Ion conduction across or along grain boundaries
3 Charge transfer at the electrode–electrolyte interface
Zr

c Typical EIS of lithium-ion batteries d Typical EIS of polymer electrolyte fuel cells

y quency
nc sing fre
qu
e Decrea
fre
ng
asi
cre
De
–Zj

3
5 2
–Zj

4
1
Insertion capacitance

3
2 45º
1 Zr

4
45º
Zr

1 Lithium-ion conduction in electrolyte of porous 1 Proton conduction in electrolyte membrane

electrode 2 Charge transfer at the electrode–electrolyte
2 Lithium-ion conduction through SEI film interface
3 Intercalation/deintercalation at the electrode– 3 Gas diffusion through gas diffusion layer
electrolyte interface 4 Water transport across membrane or
4 Lithium-ion diffusion in electrode phase platinum oxide formation or relaxation of
5 Lithium-ion diffusion in electrolyte phase adsorbed intermediates for ORR

Fig. 5 | Typical EIS spectra and the corresponding physical processes. a | Iron corrosion in sulfuric acid solution92.
b | Solid oxide fuel cells93. c | Lithium-ion batteries79. d | Polymer electrolyte fuel cells. Circled numbers (parts b–d) refer to
transport and reaction mechanisms and their influence on the respective Nyquist plots. EIS, electrochemical impedance
spectroscopy; ORR, oxygen reduction reaction; SEI, solid electrolyte interphase; SSE, saturated sulfate electrode; Zj,
imaginary part of the impedance; Zr, real part of the impedance. Panel a is adapted with permission from ref.28, IOP.
Panel b is adapted with permission from ref.94, RSC. Panel c is adapted with permission from ref.80, IOP.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 13

0123456789();:
Primer

and are hence valuable for the design and evaluation of frequencies or even a single frequency can be used to
electrochemical devices. monitor the status of the system and, subsequently,
to control the operation condition.
Mechanism identification In a proton-exchange membrane fuel cell, imped­
EIS can be employed to identify reaction or failure ance at a single frequency (around 1 kHz) is used to
mechanisms from EIS signatures or the evolution of EIS monitor the water content of the membrane, based on
features. Once the mechanism is identified, EIS can be which the operation conditions are adjusted to prevent
employed to estimate the reaction parameters, such as both flooding and drying out111. In lithium-ion batter­
charge-transfer resistance and the diffusion coefficient. ies, EIS is used to estimate the internal temperature of
A systematic way to identify the reaction mechanism the cell112, which constitutes a more reliable and earlier
composed of different elementary steps has been estab­ indicator for possible thermal runaway than the sur­
lished, showing that the development of impedance face temperature. For the online monitoring of EIS,
models from assumed reaction kinetics is superior to a simplified equivalent electric circuit model, using the
the interpretation of impedance data in terms of an intu­ general purpose CPE, usually proves to be adequate.
itively assumed electrical circuit102. One way to validate EIS is also widely used for biosensing applications, as
the proposed mechanism is to check whether a single it allows real-time monitoring of tiny changes at the
set of parameters can describe the current–voltage electrode–electrolyte interface113–115. For example, EIS
curve and the electrochemical impedance measured at methods enabled detection of the breast cancer cell con­
different potentials28,96,103–105. centration via covalent grafting of the specific antibody
The underlying mechanism for the deposition of directed against the receptor ERBB2 on electrogenerated
hydrogen under potential was explored using EIS106. polymers115 (Fig. 7). The resistance determined from the
Dynamic EIS and dynamic cyclic voltammetry were zero-frequency limit of the impedance is correlated to
used to resolve the controversy over the failure mech­ the concentration of cancer cells, as shown in Fig. 7. The
anism during sub-zero start up of a proton-exchange impedimetric biosensor can measure the EIS continu­
membrane fuel cell107. The exchange current density, ously and can monitor the concentration of cancer cells
related to the charge-transfer resistance, can be esti­ with high sensitivity and efficiency.
mated from EIS at the open circuit voltage108. The reac­
tion order of different species can be estimated from the Material and process characterization
exchange current density at different reactant concen­ Once the frequency response of the system being evalu­
trations (activities)109. Finally, the activation energy, Ea, ated is well understood, EIS can be used to characterize
can be estimated from the exchange current density at the materials, structures and processes of interest, and
several reaction temperatures108,110, as shown in Fig. 6. extract the associated structural, kinetic and mass trans­
fer parameters. To improve the accuracy, it is advisable to
Monitoring devise simplified systems for such purposes.
The features in EIS can be attributed to specific com­ For material characterization, it is common prac­
ponents or processes by examining the trend of the tice in the laboratory to use EIS with a four-electrode
impedance data while systematically varying the param­ set-up to measure the electrolyte conductivity, be it
eters. The adjusted parameter can be external, such liquid or solid. The four-electrode configuration was
as operation conditions, or internal, such as the cell used to explore the solvation/desolvation behaviour of
design parameters. After the EIS features of a system are the lithium ion, identifying desolvation of the Li+ ion
thoroughly understood, impedance results at several as the major contributor to the interfacial resistance in

a I/Il b 1
Il
0.8 Rct ≈ RT
ῆ nFI0
0.6
log(I0)

EISatOCV
0.4
η
I ≈ –I0η nF 0.2
Ĩ RT

0
–Il 3 3.1 3.2 3.3 3.4 3.5 3.6
1,000/T (1/K)

Fig. 6 | Use of EIS measurements at different temperatures to extract activation energy. a | Current–overpotential
curve for electrochemical systems showing the relationship between the exchange current and the slope of the current–
potential plot. b | Activation energy, Ea, is obtained from the slope of the exchange current density as a function of inverse
absolute temperature. The charge-transfer resistance is obtained from the high-frequency loop of the inset impedance
diagram108. EIS, electrochemical impedance spectroscopy; F, Faraday’s constant 96,487 Coulombs/equivalent;
OCV, open-circuit potential; Rct, charge-transfer resistance. Adapted with permission from ref.108, Elsevier.

14 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

a 60
b 0.15
R1 R2

5
40 CPE1 CPE2 4 0.10
3

∆R2 (MΩ)
–Zj (kΩ) 2
20
1 0.05

y = –0.0467 + 0.039 log(x)

R2 = 0.99956
0 0
0 20 40 60 101 102 103 104 105 106
Zr (kΩ) MCF-7 (cells ml–1)

Fig. 7 | Use of EIS to detect the concentration of breast cancer cells. Covalent grafting of the specific antibody directed
against the receptor ERBB2 on electrogenerated polymers. a | Electrochemical impedance spectroscopy (EIS) of system
with MCF-7 cancer cells114: (1) 0 cells ml–1, (2) 100 cells ml–1, (3) 1,000 cells ml–1, (4) 10,000 cells ml–1, (5) 100,000 cells ml–1.
b | Evolution of the resistance change as compared with case i, ΔR2, as a function of the concentration of MCF-7 cancer
cells. Error bars show the relative standard deviation (RSD = 5%) of three identical experiments. CPE, constant-phase
element; Zj, imaginary part of the impedance; Zr, real part of the impedance. Adapted with permission from ref.115, Elsevier.

lithium-ion batteries116. For structure characterization, where Z represents the measured impedance, Ẑ is the
an electrochemical porosimeter enabled estimating model value of the impedance, εfit is the contribution
the pore size distribution in a porous electrode using to the residual error that can be attributed to the use
EIS117,118. Different pore-size distributions correspond of an inadequate model, εstoch is the contribution due to
to different CPE behaviour, as shown in Fig. 8. Also, for stochastic errors and εbias is the contribution due to bias
process characterization, EIS has been used to estimate errors. The stochastic errors have a mean value equal
the charge and mass transfer properties of a lithium-ion to zero and can be characterized by the standard devi­
battery component119. To improve the power of process ation. The bias errors may be considered those errors
differentiation, a generalized transmission-line model that cause the impedance data to be inconsistent with
for porous electrodes80 was developed. the Kramers–Kronig relations. Error structure and sam­
ple variability should always be evaluated for a scien­
Reproducibility and data deposition tific analysis of impedance data, although this is not yet
As EIS measurements are very sensitive to sample prepa­ commonplace.
ration and the conditions to which the electrode has The step by step procedure for assessing the error
been exposed, the use of data repositories as a means to structure of replicated impedance spectra involves fitting
enhance reproducibility has not been widely adapted. the Voigt measurement model shown in Fig. 9a to each
Although round-robin tests have been attempted120, impedance measurement using modulus weighting71.
these have shown scattered results, partly because the The impedance is given by:
protocols used emphasized the speed of measurement
over data quality. To this end, impedance measure­
K Rk
Z = R0 + ∑ (14)
ment and interpretation are considered specific to each k 1 + jωτk
individual application.
Two types of error structure may be considered for where R0 is the leading resistance, similar to the ohmic
impedance measurements. Sample to sample variability resistance used in Eq. (7), τk = RkCk, k = 1 … K, and K is
is important for use of impedance data for quality control, the number of Voigt elements. The estimate of the stand­
such as evaluating the durability of coatings121. A mini­ ard deviation of the stochastic part of the error structure
mum of eight samples has been suggested for the com­ is obtained from the standard deviation of the residual
mercial evaluation of the relative performance of different errors for the regression. It is important that the fit to
coatings, and 12–18 samples for assessing properties of each spectrum includes the same number of Voigt ele­
an individual coating. The impedance of these coatings ments. The resulting standard deviation for the data pre­
would be evaluated as a function of time under accelerated sented in Fig. 4a is presented as 100σ|Z|−1 in Fig. 9b, where
degradation protocols to assess film durability121. the standard deviation for the real part of the imped­
The second type of error structure is that associated ance is shown as circles and the standard deviation for
with each individual measurement. The error struc­ the imaginary part of the impedance as triangles. The
ture of impedance data may be considered in terms standard deviation is seen to be a small percentage of
of the residual errors for the regression of a model to the modulus, ranging from about 0.2% at low frequency
impedance data. The residual errors are expressed as2: to about 0.02% at high frequency. This standard deviation
σ is the error structure used to weight regressions and
Z − Zˆ = ε fit + εstoch + ε bias (13) must be determined for each type of experiment.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 15

0123456789();:
Primer

a 1 b
σ→0 σ = σ0 σ→∞

Probability
–Zj / Rpore
0.5

σ = σ0

0
0 0.5 1 r
–Zr / Rpore Pore size

Fig. 8 | Influence of pore size distribution on impedance behaviour. a | Impedance of porous materials with various
pore size distributions117,118. b | Corresponding pore size distribution of porous materials. σ, standard deviation of the radius
distribution; r , average radius; Rpore, pore resistance; Zj, imaginary part of the impedance; Zr, real part of the impedance.
Adapted with permission from ref.118, Elsevier.

Application of the Kramers–Kronig integrals to an oxide or polymeric film. In principle, the interfacial
impedance data was considered controversial in the late capacitance obtained by EIS and cyclic voltammetry for
twentieth century because the integrand included an the same system should be the same, but that equality
interpolated function of frequency ranging from zero has not yet been demonstrated experimentally. Both
to infinity, requiring extrapolation into the unmeasured techniques are in situ and measure a total current and
frequency domain (Box 1). This objection was resolved an averaged potential. However, cyclic voltammetry
by the regression of a measurement model to the data. measures the global effective observation of the CPE
This measurement model (Fig. 9a) satisfied the Kramers– behaviour over a broad range of potential, whereas EIS
Kronig relations and was shown to provide a satisfac­ provides a measurement at a specified potential.
tory fit to the impedance data, which supports that data Analytical electrochemistry experiments usually
fit by the measurement model could be understood employ disk electrodes that are studied by means of a
to be consistent with the Kramers–Kronig relations72. classical three-electrode potentiostat and a frequency
Most vendors of impedance instrumentation provide response analyser for measuring the impedance response.
a regression-based assessment of the consistency with A three-electrode device enables monitoring the electro­
the Kramers–Kronig relations that is derived from a chemical processes occurring only at the working elec­
linear measurement model analysis122. The non-linear trode, dissociating them from the processes taking place
Voigt measurement model approach described here has at the counter electrode. However, even with these sim­
greater sensitivity to inconsistency with the Kramers– ple systems, the impedance response is not trivial. The
Kronig relations, making it easier to identify frequen­ observed EIS response depends on the electrochemical
cies for which a given data set does not satisfy the reactions at the working electrode, the geometry of the
Kramers–Kronig relations123. cell as well as the electrode and the electrolyte conduc­
tivity, which often results in a non-ideal response of
Limitations and optimizations the interfacial capacitance54,128–131. Independent of this
The set of measured signals in cyclic voltammetry and frequency dispersion that is ascribed to the cell geom­
EIS contain the same information, which should be etry and that gives rise to a high-frequency ohmic
easy to deconvolute depending on the physicochemical impedance54, the capacitive behaviour of the interface is
process and whether cyclic voltammetry or EIS is used. rarely ideal and requires the use of a CPE for its descrip­
Nevertheless, it is sometimes difficult to compare results tion91,132,133. Also, commercial reference electrodes usu­
obtained from both techniques. Currently, the evalua­ ally have high impedance that leads to high-frequency
tion of interfacial capacitance from cyclic voltamme­ artefacts, but this drawback can be overcome by using a
try and EIS for the case of a non-ideal behaviour124–127, dual reference electrode57. Overall, the measured electro­
which is commonly depicted as a CPE, remains an chemical impedance depends on the mechanism under
unresolved issue. The CPE is considered to result from investigation in addition to the extrinsic parameters
a position-dependent distribution of a time constant, that are not associated with the mechanism, such as the
represented as τ (x ) = R(x )C (x )128, and is often used in geometry of the cell and the electrode.
an electrical circuit model to improve the fitting to the Some systems are inherently more complex. For
impedance data (Box 3). The distribution of the time con­ example, the study of corrosion134,135 or corrosion pro­
stant along the surface of the electrode may be caused by tection136,137 systems using coatings have small current
the variation in surface distribution of current and poten­ densities, and thus require measurements using large
tial due to the electrode geometry. The time-constant electrodes138,139. Measurements on batteries are most often
distribution may also be in the direction normal to the performed on two-electrode systems. The contributions
electrode as a result of a distribution of properties for of both electrodes are then measured simultaneously,

16 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

which makes the analysis of the response even more to instru­mentation, where the most common is the FRA4.
complex because it includes the sum of the positive and However, other approaches such as multi-sine69,144,145,
negative electrode contributions. This experimental white noise146,147 or wavelet148,149 measurements are also
difficulty can be circumvented by inserting a reference used, but their description is outside the scope of this
electrode in the device46,140,141 or using a symmetrical Primer. These techniques have also been used to eval­
cell142,143. Other artefacts may exist, such as an induc­ uate other transfer functions involving electrochemical
tive contribution caused by the connecting wires in the systems such as thermal impedance spectroscopy150,
high-frequency domain when low impedances are meas­ electrohydrodynamic impedance151–153, modulation of
ured (for battery applications) or a significant noise in the the interfacial capacitance154–157 and local EIS158–162.
low-frequency range when measuring high impedance
(for anti-corrosion coatings). In the latter case, a potenti­ Field-specific limitations
ostat that allows the measurement of small currents must A general purpose EIS module may not have adequate
be used and the measurement must be carried out in a accuracy for the test sample one wants to measure, and
Faraday cage. A simple way to eliminate measurement depending on the EIS application, different issues are
errors related to the equipment is to perform the same likely to arise.
measurement on an equivalent electrical circuit exhibit­
ing similar characteristics (time constants and impedance Low-frequency range. For measurement at low fre­
magnitude) to the electrochemical system under study. quency, the major problem is the long measurement
Different techniques have been developed to time and the drifting or non-stationarity of the system.
perform impedance measurements with respect The lower the frequency, the more pronounced this
problem. The multi-sine method has been proposed
a to alleviate the problem by reducing the measurement
τ1 τ2 τk τK
C1 =
R1
C2 =
R2
Ck =
Rk
CK =
RK time163, but the result represents an averaged behaviour
that is always consistent with the Kramers–Kronig rela­
tions123. Thus, the Kramers–Kronig relations cannot be
R0
used to detect the influence of non-stationary behaviour
1 2 ... k ... K on multi-sine measurements.
In lithium-ion batteries, potentiostatic EIS at low
frequency will involve the charging and discharging of
R1 R2 Rk RK
the cell, revealing a capacitive behaviour70. For galva­
nostatic EIS, the cell is under charging or discharging,
b 0.6 hence non-stationary, conditions, violating one of the
Real three prerequisites for impedance. However, continu­
Imaginary ous efforts to measure the EIS of a non-stationary sys­
tem, frequently under the rubric of dynamic EIS, have
been reported over the years. Following the first report
of studying a non-stationary system in 1985 (ref.164),
a proposed time correction method was used to study
0.1 rechargeable lithium-ion batteries165 and the criterion
for the restriction on the amplitude was explored in a
100σ/|Z|

discharging/charging process166.
For proton-exchange membrane fuel cells, additional
features appear at low frequencies with controversial
explanations. For example, inductive loops attributed to
different mechanisms — such as humidity-dependent ion­
omer properties167, poisoning of platinum oxide growth168
or relaxation of intermediate adsorbed species in a multi­
0.01 step oxygen reduction reaction reaction169 — are observed
in the low-frequency range. However, the dominant
mechanism for these inductive loops may vary with the
0.005
10–2 10–1 100 101 102 103 104 105 working condition. The lower the humidity, the more sig­
f (Hz) nificant the effect of water transport across the membrane;
the higher the potential, the greater the effect of platinum
Fig. 9 | Measurement model analysis to identify the error structure of impedance oxide formation and intermediate species adsorption. In a
data. a | Schematic of a Voigt circuit used as a measurement model177. To improve stability distributed EIS measurement, local EIS downstream even
of the regressions, the model, as shown in Eq. (14), is written in terms of resistors and time ventured into the second quadrant (negative real and neg­
constants. The Voigt measurement model may be used to assess the error structure of
ative imaginary), a peculiar behaviour that is associated
impedance measurements and to provide preliminary physical properties such as ohmic
resistance and capacitance. b | Standard deviation of the stochastic part of the error with the consumption and interaction of O2 along the
structure for the data presented in Fig. 7. The measurement model approach was used channel170. Dedicated complementary observations are
to filter lack of replication in repeated measurements. σ, standard deviation of the radius needed to prove or disprove the proposed mechanisms
distribution; τ, time constant; C, capacitance; f, frequency; K, number of Voigt elements; and the conjecture that the dominant mechanism may
R0, leading resistance; Z, impedance. Panel a adapted with permission from ref.178, Elsevier. vary with the working condition.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 ­ 17

0123456789();:
Primer

Reference electrodes. Reversible hydrogen electrodes of their state of charge, such as battery materials. The
and dynamic hydrogen electrodes47 are employed as use of multiple transfer functions combining the tradi­
reference electrodes in proton-exchange membrane tional electrochemical measurement with that of one or
fuel cells. The tip of the reference electrode is usually more other parameters such as temperature, pressure
located outside the active area171, owing to the narrow or hydrodynamic conditions of the solution is another
space (~10 µm) between the anode and the cathode. area of development.
In coin cells or pouch-type lithium-ion batteries, refer­ The analysis and interpretation of the large quanti­
ence electrodes are either placed between the electrodes ties of impedance data that can be generated in a lab­
or outside the active area of the electrodes172. However, oratory or industrial setting is another area of interest.
artefacts such as current leaking and short circuits have Automated circuit modelling of impedance data has
been associated with placing the lithium metal reference been proposed, but circuit models obscure the under­
electrode between the two electrodes173. For durable use lying physics and chemistry. In particular, impedance
of reference electrodes in lithium-ion batteries, in situ models are not unique and EIS is not a stand-alone tech­
electrochemical alloying of a thin insulated gold wire nique. Hypothesis-driven models can guide selection of
with lithium174 and plating lithium in nickel mesh175 are complementary experiments that can be used to allow
a couple of potential routes. Such reference electrodes the user to gain confidence in their model or disprove it.
can last reliably for months, which is extremely valuable Interpretation of impedance measurements requires
for diagnosing cells under cycling. both a model that accounts for the chemistry and physics
of the system under study and the error structure of the
Outlook data. The quality of the measurement may be assessed
The development of EIS over the past few decades has using various approaches. The stochastic and bias con­
enabled a better understanding of many electrochemical tributions to the error structure may be identified by
systems involved in our modern technologies. The ver­ use of recently published software73. Manufacturers of
satility of EIS and its ability to retrieve in situ quantitative impedance instrumentation are urged to include esti­
information from complex electrochemical systems have mates of the standard deviation of impedance measure­
made it an essential and complementary tool for electro­ ment as part of their standard output, and researchers
chemists. However, careful considerations on the choice are encouraged to consider sample to sample variability
of experimental conditions and appropriate mathemat­ when appropriate. Overall, verifying the consistency of
ical models that would accurately describe the physics analysed data with the Kramers–Kronig relations should
and reaction mechanisms involved in the system consid­ become a standard for publication of impedance data.
ered are necessary. EIS is used as a routine tool for electro­ EIS data interpretation will move in two major
chemical characterization, but the analysis is often directions. The first is the rapid automated interpre­
limited to regression of equivalent electrical circuits176. tation of data, which can be quantitative if driven by a
Since the 1980s, the EIS technique has benefited solid understanding of the system under study, but is
from substantial advances in electronic instrumentation otherwise a qualitative assessment. The other direc­
that have facilitated impedance measurements. These tion includes a more systematic analysis that sees EIS
advances will continue, often with specific objectives as one of a set of complementary tools that can be used
such as enabling measurement of local electrochemical to explore the physics and chemistry of electrochemical
impedance or coupling EIS with other transfer-function systems95. The automated interpretation will facilitate
measurements. Different routes can be considered to use of impedance for diagnostics such as understand­
measure the local impedance. The use of a bi-electrode ing the health of batteries in electric vehicles, enhancing
allows these measurements, but the spatial resolution is signal processing for subcutaneous glucose sensors or
currently limited to about 1 µm. Optical measurements rapid processing of data collected from multichannel test
provide an interesting way to improve this resolution, stations. The system-specific modelling will enhance the
but they are limited to an interface whose topography ability of research laboratories to understand the broad
does not change with time, which is thus not suitable for range of systems for which EIS may be used.
the study of many systems such as corroding materials
Published online xx xx xxxx
or those that experience volume changes as a function

1. Bard, A. J. & Faulkner, L. R. Electrochemical 7. Bredar, A. R. C., Chown, A. L., Burton, A. R. 12. Margarit-Mattos, I. C. P. EIS and organic
Methods: Fundamentals and Applications 2nd edn & Farnum, B. H. Electrochemical impedance coatings performance: revisiting some key points.
(Wiley, 2001). spectroscopy of metal oxide electrodes for energy Electrochim. Acta 354, 136725 (2020).
2. Orazem, M. E. & Tribollet, B. Electrochemical applications. ACS Appl. Energy Mater. 3, 66–98 13. Macdonald, D. D. Reflections on the history
Impedance Spectroscopy 2nd edn (Wiley, 2017). (2020). of electrochemical impedance spectroscopy.
This book is a good introduction to hypothesis- 8. Xu, Y. et al. A review of impedance measurements Electrochim. Acta 51, 1376–1388 (2006).
driven modelling of impedance spectroscopy. of whole cells. Biosens. Bioelectron. 77, 824–836 14. Gabrielli, C. Once upon a time there was EIS.
3. Macdonald, D. D. Transient Techniques in (2016). Electrochim. Acta 331, 135324 (2020).
Electrochemistry (Plenum, 1977). 9. Bahadir, E. B. & Sezgintürk, M. K. A review on This article presents a detailed description
4. Gabrielli, C. Identification of Electrochemical Process impedimetric biosensors. Artif. Cells Nanomed. of EIS history, especially the evolution of EIS
by Frequency Response Analysis (Solartron Electronic Biotechnol. 44, 248–262 (2016). measurement equipment.
Group, 1980). 10. Bandarenka, A. S. Exploring the interfaces between 15. Kohlrausch, F. & Nippoldt, W. A. Ueber die Gültigkeit
5. Barsoukov, E. & Macdonald, J. R. Impedance metal electrodes and aqueous electrolytes with der Ohm’schen Gesetze für Elektrolyte und eine
Spectroscopy:  Theory, Experiment, and Applications electrochemical impedance spectroscopy. Analyst numerische Bestimmung des Leitungswiderstandes
2nd edn (Wiley, 2005). 138, 5540–5554 (2013). der verdünnten Schwefelsäure durch alternirende
6. Deleebeeck, L. & Veltze, S. Electrochemical impedance 11. Sacco, A. Electrochemical impedance spectroscopy: Ströme [German]. Ann. Phys. 214, 280–298 (1869).
spectroscopy study of commercial Li-ion phosphate fundamentals and application in dye-sensitized solar 16. Kohlrausch, F. Ueber die elektromotorische Kraft sehr
batteries: a metrology perspective. Int. J. Energy Res. cells. Renew. Sustain. Energy Rev. 79, 814–829 dünner Gasschichten auf Metallplatten [German].
44, 7158–7182 (2020). (2017). Ann. Phys. 224, 143–154 (1873).

18 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

17. Wien, M. Ueber die polarisation bei Wechselstrom 40. Garreau, D., Saveant, J. M. & Tessier, D. 62. Itagaki, M., Ueno, M., Hoshi, Y. & Shitanda, I.
[German]. Ann. Phys. 294, 37–72 (1896). Potential dependence of the electrochemical Simultaneous determination of electrochemical
18. Warburg, E. Ueber die polarisationscapacität des transfer-coefficient — impedance study of the impedance of lithium-ion rechargeable batteries with
platins [German]. Ann. Phys. 311, 125–135 (1901). reduction of aromatic-compounds. J. Phys. Chem. measurement of charge–discharge curves by wavelet
19. Hickling, A. Studies in electrode polarisation. 83, 3003–3007 (1979). transformation. Electrochim. Acta 235, 384–389
Part IV.—The automatic control of the potential of 41. Saveant, J. M. & Tessier, D. Variation of the (2017).
a working electrode. Trans. Faraday Soc. 38, 27–33 electrochemical transfer-coefficient with potential. 63. Gamry Instruments. Accuracy contour plots —
(1942). Faraday Discuss. 74, 57–72 (1982). measurement and discussion. Gamry https://
20. Koutecky, J. & Levich, V. G. The use of a rotating disk 42. Corrigan, D. A. & Evans, D. H. Cyclic voltammetric www.gamry.com/application-notes/EIS/accuracy-
electrode in the study of electrochemical kinetics and study of tert-nitrobutane reduction in acetonitrile contour-plots-measurement-and-discussion/ (2020).
electrolytic processes. Zh. Fiz. Khim. 32, 1565–1575 at mercury and platinum electrodes: observation 64. Zhong, G., Koh, C. K., Roy, K. in IEEE/ACM Int. Conf.
(1958). of a potential dependent electrochemical transfer Computer Aided Design. ICCAD-2000. IEEE/ACM
21. de Levie, R. On porous electrodes in electrolyte coefficient and the influence of the electrolyte cation Digest of Technical Papers (Cat. No. 00CH37140)
solutions: I. Capacitance effects. Electrochim. Acta on the rate constant. J. Electroanal. Chem. 106, 406–411 (IEEE, 2000).
8, 751–780 (1963). 287–304 (1980). 65. Wojcik, P. T., Agarwal, P. & Orazem, M. E. A method
22. de Levie, R. On porous electrodes in electrolyte 43. Sluyters-Rehbach, M. Impedances of electrochemical for maintaining a constant potential variation during
solutions — IV. Electrochim. Acta 9, 1231–1245 systems: terminology, nomenclature and galvanostatic regulation of electrochemical impedance
(1964). representation — part I: cells with metal electrodes measurements. Electrochim. Acta 41, 977–983
23. Gerischer, H. & Mehl, W. Zum mechanismus der and liquid solutions (IUPAC Recommendations 1994). (1996).
kathodischen wasserstoffabscheidung an quecksilber, Pure Appl. Chem. 66, 1831–1891 (1994). 66. Wojcik, P. T. & Orazem, M. E. Variable-amplitude
silber und kupfer [German]. Z. Elektrochem. Berichte 44. Newman, J. S. & Thomas-Alyea, K. E. Electrochemical galvanostatically modulated impedance spectroscopy
Bunsenges. Physikalische Chem. 59, 1049–1059 Systems 3rd edn 647 (Wiley, 2004). as a tool for assessing reactivity at the corrosion
(1955). 45. Bockris, J. O. M., Reddy, A. K. N. & Gamboa-Aldeco, M. potential without distorting temporal evolution
24. Gerischer, H. Intermediates in electrochemical Modern Electrochemistry: Electrodics 2nd edn Vol. 2A of the system. Corrosion 54, 289–298 (1998).
reactions. Faraday Discuss. Chem. Soc. 56, 7–15 764 (Kluver Academic, 2000). 67. Hirschorn, B., Tribollet, B. & Orazem, M. E. On
(1973). 46. Talaie, E. et al. Methods and protocols for selection of the perturbation amplitude required to
25. Klahr, B., Gimenez, S., Fabregat-Santiago, F., electrochemical energy storage materials research. avoid nonlinear effects in impedance measurements.
Hamann, T. & Bisquert, J. Water oxidation at Chem. Mater. 29, 90–105 (2017). Isr. J. Chem. 48, 133–142 (2008).
hematite photoelectrodes: the role of surface states. 47. Bard, A. J., Inzelt, G. & Scholz, F. Electrochemical 68. Hirschorn, B. & Orazem, M. E. On the sensitivity
J. Am. Chem. Soc. 134, 4294–4302 (2012). Dictionary (Springer Science & Business Media, of the Kramers–Kronig relations to nonlinear effects
26. Wang, H., Guerrero, A., Bou, A., Al-Mayouf, A. M. 2008). in impedance measurements. J. Electrochem. Soc.
& Bisquert, J. Kinetic and material properties 48. Song, J. Y., Lee, H. H., Wang, Y. Y. & Wan, C. C. 156, C345–C351 (2009).
of interfaces governing slow response and long Two- and three-electrode impedance spectroscopy 69. Van Ingelgem, Y., Tourwé, E., Blajiev, O., Pintelon, R.
timescale phenomena in perovskite solar cells. of lithium-ion batteries. J. Power Sources 111, & Hubin, A. Advantages of odd random phase
Energy Environ. Sci. 12, 2054–2079 (2019). 255–267 (2002). multisine electrochemical impedance measurements.
This article presents a comprehensive review of 49. Engebretsen, E. et al. Localised electrochemical Electroanalysis 21, 730–739 (2009).
perovskite solar cells, with emphasis on the use impedance measurements of a polymer electrolyte 70. Dokko, K. et al. Kinetic characterization of single
of impedance spectroscopy for studying kinetics of fuel cell using a reference electrode array to give particles of LiCoO2 by AC impedance and potential
bulk and interface processes. cathode-specific measurements and examine step methods. J. Electrochem. Soc. 148, A422
27. Epelboin, I., Gabrielli, C., Keddam, M. & Takenouti, H. membrane hydration dynamics. J. Power Sources (2001).
Model of the anodic behavior of iron in sulfuric acid 382, 38–44 (2018). 71. Agarwal, P., Crisalle, O. D., Orazem, M. E. &
medium. Electrochim. Acta 20, 913–916 (1975). 50. Sone, Y., Ekdunge, P. & Simonsson, D. Garcia-Rubio, L. H. Application of measurement
28. Keddam, M., Mattos, O. R. & Takenouti, H. Reaction Proton conductivity of Nafion 117 as measured models to impedance spectroscopy. II. Determination
model for iron dissolution studied by electrode by a four-electrode AC impedance method. of the stochastic contribution to the error structure.
impedance. I. Experimental results and reaction J. Electrochem. Soc. 143, 1254–1259 (1996). J. Electrochem. Soc. 142, 4149–4158 (1995).
model. J. Electrochem. Soc. 128, 257–266 (1981). 51. Deslouis, C., Musiani, M. M. & Tribollet, B. 72. Agarwal, P., Orazem, M. E. & Garcia-Rubio, L. H.
This article presents a detailed description Free-standing membranes for the study of Application of measurement models to impedance
of the procedure to follow for EIS interpretation electrochemical reactions occurring at conducting spectroscopy. III. Evaluation of consistency with the
in the case of complex mechanisms, using iron polymer/electrolyte interfaces. J. Phys. Chem. 100, Kramers–Kronig relations. J. Electrochem. Soc. 142,
dissolution as an example. 8994–8999 (1996). 4159–4168 (1995).
29. Baril, G. et al. An impedance investigation of 52. Keddam, M., Novoa, X. R. & Vivier, V. The concept This article introduces the use of Voigt measurement
the mechanism of pure magnesium corrosion in of floating electrode for contact-less electrochemical models to assess the consistency of impedance data
sodium sulfate solutions. J. Electrochem. Soc. 154, measurements: application to reinforcing steel-bar to the Kramers–Kronig relations.
C108–C113 (2007). corrosion in concrete. Corros. Sci. 51, 1795–1801 73. Watson, W. & Orazem, M. E. Measurement model
30. Gomes, M. P. et al. On the corrosion mechanism (2009). program. ECSArXiv https://doi.org/10.1149/osf.io/
of Mg investigated by electrochemical impedance 53. Samec, Z. Electrochemistry at the interface kze9x (2020).
spectroscopy. Electrochim. Acta 306, 61–70 (2019). between two immiscible electolyte solutions (IUPAC 74. Orazem, M. E., Pebere, N. & Tribollet, B. Enhanced
31. de Levie, R. in Advances in Electrochemistry and technical report). Pure Appl. Chem. 76, 2147–2180 graphical representation of electrochemical impedance
Electrochemical Engineering Vol. 6 (eds Delahay, P. (2004). data. J. Electrochem. Soc. 153, B129–B136 (2006).
& Tobias, C.W.) 329–397 (Wiley, 1967). 54. Huang, V. M.-W., Vivier, V., Orazem, M. E., Pebere, N. 75. Ivers-Tiffée, E. & Weber, A. Evaluation of
This pioneering work develops an EIS model for & Tribollet, B. The apparent constant-phase-element electrochemical impedance spectra by the distribution
a porous electrode. behavior of an ideally polarized blocking electrode a of relaxation times. J. Ceram. Soc. Jpn. 125, 193–201
32. Lasia, A. in Electrochemical Impedance Spectroscopy global and local impedance analysis. J. Electrochem. (2017).
and its Applications 1–367 (Springer, 2014). Soc. 154, C81–C88 (2007). This article is a significant contribution showing
33. Lasia, A. Impedance of porous electrodes. 55. Shchukin, E. D., Vidensky, I. V. & Petrova, I. V. Luggin’s the interest in using the distribution function of
J. Electroanal. Chem. 397, 27–33 (1995). capillary in studying the effect of electrochemical relaxation times to identify the different time
34. Barcia, O. E. et al. Further to the paper “Application reaction on mechanical properties of solid surfaces. constants of EIS spectra.
of the impedance model of de Levie for the J. Mater. Sci. 30, 3111–3114 (1995). 76. Wan, T. H., Saccoccio, M., Chen, C. & Ciucci, F.
characterization of porous electrodes” by Barcia et al. 56. Newman, J. S. Resistance for flow of current to a disk. Influence of the discretization methods on the
[Electrochim. Acta 47 (2002) 2109]. Electrochim. Acta J. Electrochem. Soc. 113, 501–502 (1966). distribution of relaxation times deconvolution:
51, 2096–2097 (2006). 57. Tran, A. T., Huet, F., Ngo, K. & Rousseau, P. implementing radial basis functions with DRTtools.
35. Orazem, M. E. & Tribollet, B. A tutorial on Artefacts in electrochemical impedance measurement Electrochim. Acta 184, 483–499 (2015).
electrochemical impedance spectroscopy. ChemTexts in electrolytic solutions due to the reference electrode. 77. Dierickx, S., Weber, A. & Ivers-Tiffée, E. How the
6, 12 (2020). Electrochim. Acta 56, 8034–8039 (2011). distribution of relaxation times enhances complex
36. Savéant, J. M. Elements of Molecular and This article is a key paper on the way to equivalent circuit models for fuel cells. Electrochim.
Biomolecular Electrochemistry: An Electrochemical circumventing the artefact linked to the Acta 355, 136764 (2020).
Approach to Electron Transfer Chemistry (Wiley-VCH, reference electrode in the high-frequency mode. 78. Nara, H., Yokoshima, T. & Osaka, T. Technology
2006). 58. Hills, G. & Payne, R. Improved method for measuring of electrochemical impedance spectroscopy for an
37. Costentin, C., Drouet, S., Robert, M. & Saveant, J. M. the double layer capacity at a dropping mercury energy-sustainable society. Curr. Opin. Electrochem.
A local proton source enhances CO2 electroreduction electrode. Application to measurements at high 20, 66–77 (2020).
to CO by a molecular Fe catalyst. Science 338, 90–94 pressure. Trans. Faraday Soc. 61, 316–325 (1965). 79. Huang, J., Li, Z., Ge, H. & Zhang, J. Analytical solution
(2012). 59. EG&G Princeton Applied Research. A lock-in to the impedance of electrode/electrolyte interface
38. Oleinick, A., Svir, I. & Amatore, C. A few key primer. EG&G Princeton Applied Research in lithium-ion batteries. J. Electrochem. Soc. 162,
theoretical issues of importance in modern molecular https://www.physik.uni-wuerzburg.de/fileadmin/ A7037–A7048 (2015).
electrochemistry. Curr. Opin. Electrochem. 13, 33–39 physik-fpraktikum/_imported/fileadmin/11999999/ 80. Huang, J. & Zhang, J. Theory of impedance response
(2019). Lock-In_Primer.pdf (1986). of porous electrodes: simplifications, inhomogeneities,
39. Manbeck, G. F., Fujita, E. & Concepcion, J. J. 60. Wellstead, P. Frequency Response Analysis non-stationarities and applications. J. Electrochem.
Proton-coupled electron transfer in a strongly (Solartron Instruments, 1983). Soc. 163, A1983–A2000 (2016).
coupled photosystem II-inspired chromophore– 61. Pilla, A. A. A transient impedance technique for This article presents a general framework of
imidazole–phenol complex: stepwise oxidation the study of electrode kinetics: application to EIS models that can be applicable to the porous
and concerted reduction. J. Am. Chem. Soc. 138, potentiostatic methods. J. Electrochem. Soc. 117, electrodes in lithium-ion batteries, fuel cells and
11536–11549 (2016). 467 (1970). capacitors.

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 19

0123456789();:
Primer

81. Fuller, T. F., Doyle, M. & Newman, J. Simulation impedance, and stability. J. Electroanal. Chem. 45, 127. Sadkowski, A. On the ideal polarisability of
and optimization of the dual lithium ion insertion cell. 267–277 (1973). electrodes displaying CPE-type capacitance
J. Electrochem. Soc. 141, 1–10 (1994). 105. Bertocci, U. & Ricker, R. E. in ASTM Special Technical dispersion. J. Electroanal. Chem. 481, 222–226
82. Gruet, D., Delobel, B., Sicsic, D., Lucas, I. T. & Publication 1154 143–161 (ASTM International, (2000).
Vivier, V. On the electrochemical impedance response 1992). This article discusses the physical meaning
of composite insertion electrodes — toward a better 106. Losiewicz, B., Jurczakowski, R. & Lasia, A. Kinetics of of a CPE.
understanding of porous electrodes. Electrochim. Acta hydrogen underpotential deposition at polycrystalline 128. Orazem, M. E. et al. Dielectric properties of materials
295, 787–800 (2019). platinum in acidic solutions. Electrochim. Acta 80, showing constant-phase-element (CPE) impedance
83. Amatore, C. & Maisonhaute, E. When voltammetry 292–301 (2012). response. J. Electrochem. Soc. 160, C215–C225
reaches nanoseconds. Anal. Chem. 77, 303A–311A 107. Wang, S. S., Sun, Y., Huang, F. S. & Zhang, J. B. (2013).
(2005). Freezing site of super-cooled water and failure This article presents an overview of methods
84. Popkirov, G. S. & Schindler, R. N. A new impedance mechanism of cold start of PEFC. J. Electrochem. Soc. used to extract capacitance from CPE parameters,
spectrometer for the investigation of electrochemical 166, F860–F864 (2019). with emphasis on the use of a power-law model to
systems. Rev. Sci. Instrum. 63, 5366–5372 (1992). 108. Takahashi, M., Tobishima, S., Takei, K. & Sakurai, Y. extract the thickness of oxide films.
85. Wiegand, G., Neumaier, K. R. & Sackmann, E. Reaction behavior of LiFePO4 as a cathode material 129. Newman, J. Frequency dispersion in capacity
Fast impedance spectroscopy: general aspects for rechargeable lithium batteries. Solid State Ion. measurements at a disk electrode. J. Electrochem. Soc.
and performance study for single ion channel 148, 283–289 (2002). 117, 198–203 (1970).
measurements. Rev. Sci. Instrum. 71, 2309–2320 109. Vieil, E., Maurey-Mey, M. & Cauquis, G. Current 130. Huang, V. M.-W., Vivier, V., Frateur, I., Orazem, M. E.
(2000). log-transform vs reactant concentration as a tool & Tribollet, B. The global and local impedance
86. Macdonald, J. R., Schoonman, J. & Lehnen, A. P. for chemical reaction order and electrochemical response of a blocking disk electrode with local
Applicability and power of complex nonlinear least mechanisms determination — I. General theory constant-phase-element behavior. J. Electrochem. Soc.
squares for the analysis of impedance and admittance in steady-state voltammetry and application to 154, C89–C98 (2007).
data. J. Electroanal. Chem. 131, 77–95 (1982). electro-oxidation of thianthrene in presence of 131. Huang, V. M.-W., Vivier, V., Orazem, M. E., Pebere, N.
87. Box, G. E. P. & Draper, N. R. Empirical Model-Building water. Electrochim. Acta 27, 1565–1583 (1982). & Tribollet, B. The apparent constant-phase-element
and Response Surfaces (Wiley, 1987). 110. Laidler, K. J. The development of the Arrhenius behavior of a disk electrode with Faradaic reactions.
88. Draper, N. R. & Smith, H., S. Applied Regression equation. J. Chem. Educ. 61, 494 (1984). A global and local impedance analysis.
Analysis (Wiley, 1981). 111. Hong, P., Xu, L. F., Jiang, H. L., Li, J. Q. & J. Electrochem. Soc. 154, C99–C107 (2007).
89. Orazem, M. E., Durbha, M., Deslouis, C., Takenouti, H. Ouyang, M. G. A new approach to online AC 132. Hirschorn, B. et al. Determination of effective
& Tribollet, B. Influence of surface phenomena on impedance measurement at high frequency of capacitance and film thickness from constant-
the impedance response of a rotating disk electrode. PEM fuel cell stack. Int. J. Hydrog. Energy 42, phase-element parameters. Electrochim. Acta 55,
Electrochim. Acta 44, 4403–4412 (1999). 19156–19169 (2017). 6218–6227 (2010).
90. Brug, G. J., van den Eeden, A. L. G., 112. Zhu, J. G., Sun, Z. C., Wei, X. Z. & Dai, H. F. 133. Hirschorn, B. et al. Constant-phase-element
Sluyters-Rehbach, M. & Sluyters, J. H. The analysis of A new lithium-ion battery internal temperature behavior caused by resistivity distributions in
electrode impedances complicated by the presence on-line estimate method based on electrochemical films: II. Applications. J. Electrochem. Soc. 157,
of a constant phase element. J. Electroanal. Chem. impedance spectroscopy measurement. J. Power C458–C463 (2010).
176, 275–295 (1984). Sources 274, 990–1004 (2015). 134. Díaz, B., Guitián, B., Nóvoa, X. R. & Pérez, M. C.
91. Hirschorn, B. et al. Constant-phase-element behavior 113. Khan, M. Z. H., Hasan, M. R., Hossain, S. I., The effect of chlorides on the corrosion behaviour
caused by resistivity distributions in films: I. Theory. Ahommed, M. S. & Daizy, M. Ultrasensitive detection of weathered reinforcing bars. Electrochim. Acta 336,
J. Electrochem. Soc. 157, C452–C457 (2010). of pathogenic viruses with electrochemical biosensor: 135737 (2020).
92. Wu, S. L., Orazem, M. E., Tribollet, B. & Vivier, V. state of the art. Biosens. Bioelectron. 166, 112431 135. Hernández, R. D. P. B., Aoki, I. V., Tribollet, B. &
The influence of coupled faradaic and charging (2020). De Melo, H. G. Electrochemical impedance
currents on impedance spectroscopy. Electrochim. 114. Prodromidis, M. I. Impedimetric immunosensors — spectroscopy investigation of the electrochemical
Acta 131, 2–12 (2014). a review. Electrochim. Acta 55, 4227–4233 (2010). behaviour of copper coated with artificial patina layers
93. Gabrielli, C. & Group, S. I. Identification of 115. Seven, B. et al. Impedimetric biosensor for cancer and submitted to wet and dry cycles. Electrochim. Acta
Electrochemical Processes by Frequency Response cell detection. Electrochem. Commun. 37, 36–39 56, 2801–2814 (2011).
Analysis (Solartron Electronic Group, 1980). (2013). 136. Capelossi, V. R. et al. Corrosion protection of clad
94. Tuller, H. L., Litzelman, S. J. & Jung, W. Micro-ionics: 116. Abe, T., Sagane, F., Ohtsuka, M., Iriyama, Y. & 2024 aluminum alloy anodized in tartaric-sulfuric
next generation power sources. Phys. Chem. Chem. Ogumi, Z. Lithium-ion transfer at the interface acid bath and protected with hybrid sol-gel coating.
Phys. 11, 3023–3034 (2009). between lithium-ion conductive ceramic electrolyte Electrochim. Acta 124, 69–79 (2014).
95. Huang, J. et al. Impedance response of porous and liquid electrolyte — a key to enhancing the rate 137. Bouali, A. C. et al. Layered double hydroxide based
electrodes: theoretical framework, physical models capability of lithium-ion batteries. J. Electrochem. Soc. active corrosion protective sealing of plasma
and applications. J. Electrochem. Soc. 167, 166503 152, A2151–A2154 (2005). electrolytic oxidation/sol-gel composite coating on
(2020). 117. Song, H.-K., Jung, Y.-H., Lee, K.-H. & Dao, L. H. AA2024. Appl. Surf. Sci. 494, 829–840 (2019).
This article is a systematic review of physics- Electrochemical impedance spectroscopy of porous 138. Calado, L. M., Taryba, M. G., Carmezim, M. J. &
based EIS models for porous electrodes with electrodes: the effect of pore size distribution. Montemor, M. F. Self-healing ceria-modified coating
substantial mathematic details that may help Electrochim. Acta 44, 3513–3519 (1999). for corrosion protection of AZ31 magnesium alloy.
model development. 118. Song, H.-K., Hwang, H.-Y., Lee, K.-H. & Dao, L. H. Corros. Sci. 142, 12–21 (2018).
96. Keddam, M., Mattos, O. R. & Takenouti, H. Reaction The effect of pore size distribution on the frequency 139. Roggero, A. et al. Thermal activation of impedance
model for iron dissolution studied by electrode dispersion of porous electrodes. Electrochim. Acta 45, measurements on an epoxy coating for the
impedance. II. Determination of the reaction model. 2241–2257 (2000). corrosion protection: 2. electrochemical impedance
J. Electrochem. Soc. 128, 266–274 (1981). 119. Suthar, B., Landesfeind, J., Eldiven, A. & spectroscopy study. Electrochim. Acta 305, 116–124
97. Pivac, I. & Barbir, F. Inductive phenomena at low Gasteiger, H. A. Method to determine the in-plane (2019).
frequencies in impedance spectra of proton exchange tortuosity of porous electrodes. J. Electrochem. Soc. 140. Karden, E., Buller, S. & De Doncker, R. W. A method
membrane fuel cells — a review. J. Power Sources 165, A2008–A2018 (2018). for measurement and interpretation of impedance
326, 112–119 (2016). 120. Ritter, S. et al. Results of an international round-robin spectra for industrial batteries. J. Power Sources 85,
98. Roy, S. K., Orazem, M. E. & Tribollet, B. Interpretation exercise on electrochemical impedance spectroscopy. 72–78 (2000).
of low-frequency inductive loops in PEM fuel cells. Corros. Eng. Sci. Technol. 56, 254–268 (2020). 141. Costard, J., Ender, M., Weiss, M. & Ivers-Tiffée, E.
J. Electrochem. Soc. 154, B1378 (2007). 121. Tait, W. S. Using electrochemical measurements Three-electrode setups for lithium-ion batteries: II.
99. Huang, J. et al. An analytical three-scale impedance to estimate coating and polymer film durability. Experimental study of different reference electrode
model for porous electrode with agglomerates in J. Coat. Technol. 75, 45–50 (2003). designs and their implications for half-cell impedance
lithium-ion batteries. J. Electrochem. Soc. 162, 122. Boukamp, B. A. A linear Kronig–Kramers spectra. J. Electrochem. Soc. 164, A80–A87 (2017).
A585–A595 (2015). transform test for immittance data validation. 142. Osaka, T., Mukoyama, D. & Nara, H. Review —
100. Sabet, P. S. & Sauer, D. U. Separation of predominant J. Electrochem. Soc. 142, 1885–1894 (1995). development of diagnostic process for commercially
processes in electrochemical impedance spectra of 123. You, C., Zabara, M. A., Orazem, M. E. & Ulgut, B. available batteries, especially lithium ion battery,
lithium-ion batteries with nickel–manganese–cobalt Application of the Kramers–Kronig relations to by electrochemical impedance spectroscopy.
cathodes. J. Power Sources 425, 121–129 (2019). multi-sine electrochemical impedance measurements. J. Electrochem. Soc. 162, A2529–A2537 (2015).
101. Boillot, M. et al. Effect of gas dilution on PEM fuel J. Electrochem. Soc. 167, 020515 (2020). 143. Itou, Y., Ogihara, N. & Kawauchi, S. Role of conductive
cell performance and impedance response. Fuel Cell 6, 124. Allagui, A., Freeborn, T. J., Elwakil, A. S. & carbon in porous Li-ion battery electrodes revealed
31–37 (2006). Maundy, B. J. Reevaluation of performance of electric by electrochemical impedance spectroscopy using a
102. Harrington, D. A. & Conway, B. E. AC impedance double-layer capacitors from constant-current charge/ symmetric cell. J. Phys. Chem. C 124, 5559–5564
of faradaic reactions involving electrosorbed discharge and cyclic voltammetry. Sci. Rep. 6, 38568 (2020).
intermediates. 1. Kinetic theory. Electrochim. Acta (2016). 144. Ji, G., Macia, L. F., Allaert, B., Hubin, A. & Terryn, H.
32, 1703–1712 (1987). This article presents a detailed description Odd random phase electrochemical impedance
This paper presents a systematic methodology of the use of a CPE in cyclic voltammetry. spectroscopy to study the corrosion behavior of
to construct EIS models from a multistep reaction 125. Gharbi, O., Tran, M. T. T., Tribollet, B., Turmine, M. hot dip Zn and Zn-alloy coated steel wires in
mechanism. & Vivier, V. Revisiting cyclic voltammetry and sodium chloride solution. J. Electrochem. Soc.
103. Epelboin, I., Keddam, M. & Lestrade, J. C. Faradaic electrochemical impedance spectroscopy analysis 165, C246–C257 (2018).
impedances and intermediates in electrochemical for capacitance measurements. Electrochim. Acta 145. Sanchez, B., Fernandez, X., Reig, S. & Bragos, R.
reactions. Faraday Discuss. Chem. Soc. 56, 264–275 343, 136109 (2020). An FPGA-based frequency response analyzer
(1973). 126. Sadkowski, A. Time domain responses of constant for multisine and stepped sine measurements
104. Gabrielli, C. & Keddam, M. Heterogeneous reactions phase electrodes. Electrochim. Acta 38, 2051–2054 on stationary and time-varying impedance.
coupled by diffusion. Multiple stationary states, (1993). Meas. Sci. Technol. 25, 015501 (2014).

20 | Article citation ID: (2021) 1:41 www.nature.com/nrmp

0123456789();:
 Primer

146. Gabrielli, C., Huet, F. & Keddam, M. Comparison of 159. Ferrari, J. V. et al. Influence of normal and radial 172. Raccichini, R., Amores, M. & Hinds, G. Critical review
sine wave and white noise analysis for electrochemical contributions of local current density on local of the use of reference electrodes in Li-ion batteries:
impedance measurements. J. Electroanal. Chem. 335, electrochemical impedance spectroscopy. a diagnostic perspective. Batteries 5, 24 (2019).
33–53 (1992). Electrochim. Acta 60, 244–252 (2012). 173. Hoshi, Y. et al. Optimization of reference electrode
147. Chang, B. Y. & Park, S. M. Electrochemical impedance 160. de Abreu, C. P. et al. Multiscale electrochemical position in a three-electrode cell for impedance
spectroscopy. Annu. Rev. Anal. Chem. 3, 207–229 study of welded Al alloys joined by friction stir measurements in lithium-ion rechargeable battery
(2010). welding. J. Electrochem. Soc. 164, C735–C746 by finite element method. J. Power Sources 288,
148. Smulko, J. M., Darowicki, K. & Zielinski, A. (2017). 168–175 (2015).
On electrochemical noise analysis for monitoring 161. Shkirskiy, V., Volovitch, P. & Vivier, V. Development 174. Solchenbach, S., Pritzl, D., Kong, E. J. Y., Landesfeind, J.
of uniform corrosion rate. IEEE Trans. Instrum. Meas. of quantitative local electrochemical impedance & Gasteiger, H. A. A gold micro-reference electrode for
56, 2018–2023 (2007). mapping: an efficient tool for the evaluation of impedance and potential measurements in lithium ion
149. Cottis, R. A. Interpretation of electrochemical noise delamination kinetics. Electrochim. Acta 235, batteries. J. Electrochem. Soc. 163, A2265 (2016).
data. Corrosion 57, 265–285 (2001). 442–452 (2017). 175. Chu, Z. Y. et al. Testing lithium-ion battery with the
150. Abada, S. et al. Safety focused modeling of lithium-ion 162. Gharbi, O., Ngo, K., Turmine, M. & Vivier, V. internal reference electrode: an insight into
batteries: a review. J. Power Sources 306, 178–192 Local electrochemical impedance spectroscopy: the blocking effect. J. Electrochem. Soc. 165,
(2016). a window into heterogeneous interfaces. Curr. Opin. A3240–A3248 (2018).
151. Deslouis, C., Tribollet, B., Mengoli, G. & Musiani, M. M. Electrochem. 20, 1–7 (2020). 176. Orazem, M. E. Electrochemical impedance
Electrochemical behavior of copper in neutral aerated 163. Creason, S. C. & Smith, D. E. Fourier transform spectroscopy: the journey to physical understanding.
chloride solution. II. Impedance investigation. J. Appl. Faradaic admittance measurements: I. Demonstration J. Solid State Electrochem. 24, 2151–2153 (2020).
Electrochem. 18, 384–393 (1988). of the applicability of random and pseudo-random 177. Agarwal, P., Orazem, M. E. & Garci-Rubio, L. H.
This article presents a comprehensive analysis of noise as applied potential signals. J. Electroanal. Chem. Measurement models for electrochemical impedance
EIS and transfer functions applied to the corrosion 36, A1–A7 (1972). spectroscopy. I. Demonstration of applicability.
of copper in chloride solution. 164. Stoynov, Z. B. & Savova-Stoynov, B. S. Impedance J. Electrochem. Soc. 139, 1917–1927 (1992).
152. Robertson, B., Tribollet, B. & Deslouis, C. Measurement study of non-stationary systems: four-dimensional 178. Liao, H. et al. Physical properties obtained
of diffusion coefficients by DC and EHD electrochemical analysis. J. Electroanal. Chem. 183, 133–144 from measurement model analysis of impedance
methods. J. Electrochem. Soc. 135, 2279–2284 (1985). measurements. Electrochim. Acta 354, 136747 (2020).
(1988). 165. Itagaki, M. et al. In situ electrochemical impedance
153. Tribollet, B., Newman, J. & Smyrl, W. H. Determination spectroscopy to investigate negative electrode of Author contributions
of the diffusion coefficient from impedance data in lithium-ion rechargeable batteries. J. Power Sources Introduction (S.W., J.Z., O.G., V.V., M.G. and M.E.O);
the low frequency range. J. Electrochem. Soc. 135, 135, 255–261 (2004). Experimentation (S.W., J.Z., O.G., V.V., M.G. and M.E.O);
134–138 (1988). 166. Huang, J., Li, Z. & Zhang, J. B. Dynamic electro­ Results (S.W., J.Z., O.G., V.V., M.G. and M.E.O); Applications
154. Antano-Lopez, R., Keddam, M. & Takenouti, H. chemical impedance spectroscopy reconstructed from (S.W., J.Z., O.G., V.V., M.G. and M.E.O); Reproducibility and
A new experimental approach to the time-constants continuous impedance measurement of single frequency data deposition (S.W., J.Z., O.G., V.V., M.G. and M.E.O);
of electrochemical impedance: frequency response during charging/discharging. J. Power Sources 273, Limitations and optimizations (S.W., J.Z., O.G., V.V., M.G. and
of the double layer capacitance. Electrochim. Acta 1098–1102 (2015). M.E.O); Outlook (S.W., J.Z., O.G., V.V., M.G. and M.E.O);
46, 3611–3617 (2001). 167. Wiezell, K., Holmstrom, N. & Lindbergh, G. Studying Overview of the Primer (M.E.O.).
155. Cachet, H. et al. Capacitance probe of the electron low-humidity effects in PEFCs using EIS II. modeling.
displacement in a dye sensitized solar cell by J. Electrochem. Soc. 159, F379–F392 (2012). Competing interests
an intermodulation technique: a quantitative 168. Setzler, B. P. & Fuller, T. F. A physics-based The authors declare no competing interests.
model. Electrochim. Acta 49, 2541–2549 impedance model of proton exchange membrane
Peer review information
(2004). fuel cells exhibiting low-frequency inductive loops.
Nature Reviews Methods Primers thanks S. Ciampi, A. Lasia
156. Larios-Duran, E. R. et al. Dynamics of double-layer J. Electrochem. Soc. 162, F519–F530 (2015).
and I. Shitanda for their contribution to the peer review of
by AC modulation of the interfacial capacitance and 169. Kulikovsky, A. A. Analysis of Damjanovic kinetics of
this work.
associated transfer functions. Electrochim. Acta 55, the oxygen reduction reaction: stability, polarization
6292–6298 (2010). curve and impedance spectra. J. Electroanal. Chem. Publisher’s note
157. Antaño-López, R., Keddam, M., Turmine, M. & 738, 130–137 (2015). Springer Nature remains neutral with regard to jurisdictional
Vivier, V. The impedance response of a passive 170. Schneider, I. A., Kramer, D., Wokaun, A. & Scherer, G. G. claims in published maps and institutional affiliations.
film revisited by a double modulation technique. Oscillations in gas channels — II. Unraveling the
ChemElectroChem 6, 202–210 (2019). characteristics of the low frequency loop in air-fed
158. Zou, F., Thierry, D. & Isaacs, H. S. A high-resolution PEFC impedance spectra. J. Electrochem. Soc. 154, Related links
probe for localized electrochemical impedance B770–B782 (2007). ZPlot: http://www.scribner.com/software/146-zplot-and-
spectroscopy measurements. J. Electrochem. Soc. 171. Kuhn, H., Andreaus, B., Wokaun, A. & Scherer, G. G. zview-for-windows
144, 1957–1965 (1997). Electrochemical impedance spectroscopy applied ZSimpWin: https://www.ameteksi.com/products/software/
This study reports the design of dual to polymer electrolyte fuel cells with a pseudo zsimpwin
microelectrodes for performing local EIS to reference electrode arrangement. Electrochim. Acta
enhance the spatial resolution of the technique. 51, 1622–1628 (2006). © Springer Nature Limited 2021

NATurE REvIEwS | METhodS PRImERS | Article citation I­D:­ (2021) 1:41 21

0123456789();: