Application of Electrochemical Impedance Spectroscopy To Degradation and Aging Research of Lithium-Ion Batteries

pubs.acs.
org/JPCC Article
Application of Electrochemical Impedance Spectroscopy to

Degradation and Aging Research of Lithium-Ion Batteries
Published as part of The Journal of Physical Chemistry virtual special issue “Zhao-Wu Tian Festschrift”.
Wenxuan Hu, Yufan Peng, Yimin Wei, and Yong Yang*
Cite This: J. Phys. Chem. C 2023, 127, 4465−4495 Read Online

See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
ACCESS Metrics & More Article Recommendations

Downloaded via NATL UNIV OF SINGAPORE on August 23, 2023 at 05:55:54 (UTC).
ABSTRACT: An in-depth understanding of battery degradation and aging in-Operando not

only plays a vital role in the design of battery managing systems but also helps to ensure safe
use and manufacturing optimization of lithium-ion batteries (LIBs) in large-scale
applications. Electrochemical impedance spectroscopy (EIS) is a nondestructive method
which unravels electrode kinetic processes inside the batteries in different time domains,
including charge-transfer reactions, interfacial evolutions, and mass diffusions. It has become
a powerful diagnosis and pre/prognosis tool in battery aging research, as it provides
important insight into the changes of internal electrochemical processes by correlating the
impedance evolution to degradation mechanisms. This review gives a critical overview on
rapidly developing impedance techniques for degradation and aging investigation of Li-ion
batteries. The EIS variations of LIBs at different aging conditions of calendar aging and
accelerated aging are systematically summarized. In addition, the working principles, data
validation, and modeling methods, including equivalent circuit model (ECM), distribution
of relaxation times (DRT), and transmission line model (TLM), of classical EIS and dynamic EIS are elaborately concluded. Finally,
the challenges and perspectives of further application of EIS in the aging research of LIBs are presented.
1. INTRODUCTION understanding of the degradation mechanisms of LIBs, which

Nowadays, lithium-ion batteries (LIB) play a critical role in highly depend on utilization and development of advanced
most aspects of human society due to their excellent energy characterization methods. X-ray-based techniques,2 such as X-
storage performance. Since the LIB industry has flourished, the ray absorption spectroscopy (XAS), X-ray diffraction (XRD),
concerns of academia and industry have long been focused on and X-ray photoelectron spectroscopy (XPS), could determine
how to develop and produce LIBs with better electrochemical the change of chemical composition and elements’ valence and
performance on different scales. With the potential of electrode crystal structure. Electron and scanning probe microscopy
materials having been gradually exploited to the limit, an techniques,3 such as scanning electron microscopy (SEM),
equally, if not more, important issue is brought to light, i.e., transmission electron microscopy (TEM), and cryo-TEM,
how to use these batteries efficiently and safely. To answer this could observe the morphology change of the electrode and the
question properly, an in-depth understanding of degradation thickness of solid electrolyte interface (SEI). Titration-based
mechanisms of LIBs in different operation conditions should techniques, such as titration gas chromatography (TGC)4 and
be established.1 Furthermore, monitoring the state of charge mass spectrometry titration (MST),5,6 could quantify the dead
(SOC), state of health (SOH), and internal physicochemical Li metal and SEI components. In addition, solid-state nuclear
parameters such as internal temperature/pressure plays a vital magnetic resonance (NMR)7 and electron paramagnetic
role in the diagnosis of the batteries. To achieve these targets, resonance (EPR)8 are able to achieve a time-resolved and
in situ or operando electrochemical and nonelectrochemical quantitative detection of Li plating. However, most of these
techniques are indispensable to detect the working state and methods are invasive technologies and need additional
degradation tendency of the battery accurately, which are still
great challenges for battery researchers.
LIBs are a kind of complicated electrochemical system with Received: January 3, 2023
several components, including cathode, anode, electrolyte, Revised: February 12, 2023
separator, and current collector. The complexity of the systems Published: March 1, 2023
causes difficulties in decoupling and identifying the degrada-
tion mechanisms during battery aging. Recent rapid develop-
ments in large-scale applications of LIBs call for the systematic
© 2023 American Chemical Society https://doi.org/10.1021/acs.jpcc.3c00033

4465 J. Phys. Chem. C 2023, 127, 4465−4495
The Journal of Physical Chemistry C pubs.acs.org/JPCC Article
Figure 1. Schematic graph shows capability and custom procedures of electrochemical impedance techniques for the battery aging study.
expensive as well as precise instruments with special cell evolution and relating them to degradation mechanisms.
designs for tests, which are not feasible in practical application. Therefore, the EIS technique has been widely used in different
Besides these physical characterizations, electrochemical applications for battery research, including internal temper-
methods, such as differential voltage analysis (DVA),9,10 ature monitoring,19 degradation mechanism diagnosis,20
differential charging voltage (DCV),11 incremental capacity battery states estimation,21 and remaining useful life (RUL)
analysis (ICA),12 and Coulombic efficiency (CE),13,14 also play prediction.22
instrumental roles in battery aging research. Nevertheless, they There are several good reviews published which have
are quite complex determination methods that need a long concluded the principles, modeling, interpretation, and
monitoring time. application of EIS in LIBs;23−31 e.g., Wang et al. provided a
Electrochemical impedance spectroscopy (EIS) is a powerful detailed method primer of EIS, 23 and Vivier et al.
tool in the analysis of electrochemical systems. By employing a comprehensively elucidated the definition, graphical represen-
sinusoidal current or voltage with a small amplitude as input tation, error analysis, and application of electrochemical
signal, the electrochemical systems would generate an impedance.24 When it comes to the interpretation of EIS
approximately linear response so that the impedance of cells data, there are mainly three types of methods, i.e., equivalent
over the time domain and frequency domain is obtained. Since circuit model (ECM), DRT, and physics-based impedance
the internal electrochemical processes in LIBs possess different model. Ciucci summarized different modeling approaches and
relaxation times to respond to the perturbation, the impedance their advantages and disadvantages.25 Lu et al. introduced
spectroscopy could detect convolutedly and be decoupled with DRT methods in the views of the basic time scale property in
different components by further modeling. Compared with batteries for battery diagnosis.26 Gaddam et al. covered the
physical characterization and other electrochemical character- physics-based modeling method and its application in different
ization methods, EIS has several unique advantages. First, it battery systems.27 Other reviews in this area focused on the
relates voltage with current in the time domain and frequency application of EIS to LIB research.28−31 McCarthy et al.
domain so as to obtain the kinetic parameters of internal provided a review on the use of impedance for assessing
processes in LIBs. Second, it is a nondestructive technique, battery parameters of SOC, SOH, and internal temperature.28
which is available in in situ and operando research. It can play Wang et al. reviewed the onboard impedance acquisition,
an important role in monitoring the states of the battery during modeling, and applications under the premise of electric
operation. Moreover, EIS could effectively deconvolute the vehicle implementation.29 While most of the present reviews
complex electrochemical processes into a series of basic focused on the EIS acquisition, modeling, and interpretation in
processes based on different relaxation times.15 The resolution different electrochemical systems, few of them have thoroughly
of EIS is relatively high, especially with the assistance of new summarized the application of EIS on degradation mechanism
modeling methods, such as distribution of relaxation times research of LIBs. Iurilli et al. concluded the EIS variations
(DRT),16 distribution function of differential capacity before and after cycling aging and calendar aging from the
(DDC),17 and distribution of diffusion times (DDT).18 All perspective of the battery chemistry based on ECM.20 Though
of those characteristics make it possible to monitor the change the review gives a critical assessment of EIS interpretation to
in internal kinetic processes in LIBs by tracking the impedance model aging phenomena, the conclusion is mostly qualitative
4466 https://doi.org/10.1021/acs.jpcc.3c00033
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 2. Excitation signals, output responses, advantages, and disadvantages of different impedance techniques: (a) classical EIS (reprinted with
permission from ref 23. Copyright 2021 Springer Nature); (b) linear DEIS (reprinted with permission from ref 150. Copyright 2021 Elsevier.
Reprinted with permission from ref 170. Copyright 2021 The Electrochemical Society); (c) single-frequency DEIS (reprinted with permission from
ref 43. Copyright 2015 Elsevier); (d) nonlinear frequency response analysis (NFRA) (reprinted with permission from ref 50. Copyright 2017
Elsevier); (e) odd random-phase EIS (ORP-EIS) (reprinted with permission from ref 62. Copyright 2017 Elsevier).
due to the difficulty of deconvoluting EIS by ECM and the vacancy for comprehensively summarizing the latest develop-
discrepancy in EIS results for the same battery chemistry ment of EIS applications for LIBs’ degradation and aging
between different references. To conclude, there is still a research.
J. Phys. Chem. C 2023, 127, 4465−4495
This review aims at providing a guideline for choosing and (1)

employing the appropriate impedance methods to achieve an
in-depth study of the aging process of LIBs. In section 2, the (2)
working principles, measurements and data validation of
classical EIS and dynamic EIS are elaborately overviewed.
Section 3 focuses on the data modeling, interpretation, and (3)
general battery aging diagnosis procedure by ECM, DRT, and
TLM. In section 4, the applications of EIS in the LIB aging (4)
study are comprehensively summarized in three aspects,
including EIS variation at different battery aging conditions (5)
(calendar aging and accelerated aging), new impedance
techniques for studying specific degradation mechanisms
such as Li plating and SEI growth, and impedance-based
(6)
machine learning model for battery states estimation. Section 5
gives a summary of this review and the perspectives of EIS 2.1.2. Data Validation. Since the quality of EIS data highly
application to battery degradation and aging research. The influences the modeling and interpretation process, data
capabilities and custom procedures to employ different validation plays an important role in improving the credibility
electrochemical impedance techniques for battery degradation and reliability of EIS. To acquire reliable EIS data, three
and aging research are shown in Figure 1. conditions should be sanctified, i.e., linearity, causality, and
stability. Linearity means that the relation between the input
2. EIS MEASUREMENT signal and output response is linear. The impedance spectrum
should be independent of the amplitude. However, the kinetic
It is known that the classical EIS test is operated on the characteristic of the electrochemical process, described by the
electrochemical system at a stable or quasi-potentiostatic state Butler−Volmer equation, is intrinsically nonlinear. Thus, a
(e.g., open-circuit state), so as to obtain a time-invariant small amplitude of the perturbation signal should be employed
impedance spectrum. Moreover, a small perturbation ampli- to approach the linear situation. There are several experimental
tude should be applied to keep the formation of harmonics. methods to conduct data validation, including varying the
Different from classical EIS, dynamic EIS (DEIS) that operates signal’s amplitude to observe the consistency of impedance
at nonstationary conditions is considered as a good alternative spectrum, comparing the value of total harmonic distortion
or auxiliary tool for unveiling the kinetics of complex (THD) with fundamental frequency, or recording the temporal
electrochemical systems. In this section, the two major response of the potential and current to make sure that the
categories, classical EIS and DEIS, are presented in detail. Lissajous curve is not distorted.32 Causality means the
2.1. Classical EIS. 2.1.1. Principle. For the classical EIS, a response of the system completely results from the inputs.
sinusoidal potential or current signal with a small amplitude is Faraday cages can prevent the measured system from noises
used as the input signal as shown in Figure 2a. Based on the derived from an external electric field and electromagnetic
input signals, the classical EIS could be further divided into waves. Stability means that the response is independent of
potentiostatic EIS (PEIS) and galvanostatic EIS (GEIS). While time, which can be easily checked by repeated tests. Besides
PEIS fulfills most of the requirements, GEIS is developed for the above three measurement prerequisites, finiteness means
the systems with low impedance, like commercial batteries, to the values of the imaginary and real parts of the impedance are
avoid overflow. Although the input signals between GEIS and finite over the frequency range. When the four conditions are
PEIS are different, the mechanism is the same, as shown in eqs met, the imaginary and real part of the complex functions
1 and 6. In a linear system, a sinusoidal potential signal at a satisfy the Kramers−Kronig (K−K) relations.33 The expres-
specific frequency could lead to a sinusoidal current output at sions of the K−K relations are shown as follows:
the same frequency with a phase angle (φ). Based on Ohm’s
law, the impedance of the electrochemical system at different
frequencies could be obtained. By putting the result into a (7)
complex plane and employing Euler’s formula, the impedance
function could be transformed to the sum of real part (Z′) and
imaginary part (Z″).
(8)
Nyquist and Bode plots are now the most widely used
methods to clearly display the features of impedance. By
applying the real part of impedance to the x-axis and the
(9)
negative imaginary part to the y-axis, a Nyquist plot, i.e.,
electrochemical impedance spectroscopy, is obtained to Considering the complexity of the calculation processes,
illustrate the variation of impedance over the frequency several commercial and open-access software programs have
range. Although the Nyquist plot displays the features of real been developed to help researchers achieve this work.
and imaginary parts of the impedance, it can not directly reflect Schönleber et al. developed the open-assess “Lin_KK” tool
the information on frequency unless marks are added in the to conduct the K−K validity test.34 By conducting the K−K
figure. The Bode plot, as a supplementary Nyquist plot, takes transforms, the residuals between real and calculated values of
the modulus (|Z(ω)|, eq 5) and phase angle (φ(ω), eq 6) of every single point could be obtained based on eqs 10 and 11.
the impedance as ordinate and logarithmic frequency as Only if all of the residuals (residuals of the real part ΔRe(ω)
abscissa, which directly displays the information on frequency and the imaginary part ΔIm(ω)) are smaller than 1% can the
in graphs. EIS data be considered as credible.35 Not until the raw data are
J. Phys. Chem. C 2023, 127, 4465−4495
validated by K−K transforms can the researcher go through be an indicator of electrode materials with fast charge−
further data analysis and modeling. discharge capability.
In addition to linear DEIS, single-frequency DEIS is another
kind of technique for LTV conditions. As shown in Figure 2c,
(10) it uses an intermodulation signal with a single-sine current
signal with a single frequency and a large direct current to
observe the impedance variation of the specific frequency over
a large SOC range during charging or discharging. Since SOC
(11) varies significantly during the measurement, the multi-
frequency test is meaningless and inaccurate. Different from
2.2. Dynamic EIS. Since batteries are inherently nonlinear linear DEIS in which the SOC during measurement is nearly
and unstable under operation, classical EIS measurement has unchanged in a short test time, the single-frequency DEIS
been limited in LIBs’ research. In practical tests, the cell would could be conducted over the whole SOC and closer to working
be charged or discharged to the specific SOC and then a conditions. Huang et al. first employed the single-frequency
classical EIS measurement would be conducted after DEIS method to measure a continuous impedance change of
relaxation, which is relatively time-consuming, inconvenient, LIBs with charge and discharge over a full SOC range.43 A
and deflected from operation conditions. Therefore, a dynamic small ac current amplitude (0.01 A, 1% dc current) was used to
EIS (DEIS) technique has received more and more attention meet the linear condition. However, the disadvantage of this
for its promising potential in operando techniques. method is that a single test could only obtain the impedance of
Based on the differences in the input signal, DEIS could be a single frequency, which makes it inconvenient to get a full-
divided into two categories: linear time-varying (LTV) systems frequency-range impedance spectrum. Therefore, its applica-
and nonlinear time-varying (NLTV) systems. In the LTV tion scenarios should be delicately selected to maximize its
systems, the test conditions are often compromised so as to advantages. Actually, the single-frequency DEIS has a unique
approximately fulfill the linearity of the EIS frequency advantage when investigating a specific process with a stable
response. The data analysis is still based on the linear time constant, in which case the measured impedance variation
framework of classical EIS, while in the NLTV systems the could be ascribed to the process. Lyu et al. used single-
input signals are controlled to deliberately expose the nonlinear frequency DEIS to find the indicator of predicting thermal
information on the processes in batteries. Since the linear runaway caused by overcharge.19 The impedance at 70 Hz is
framework of classical EIS is no longer suitable for these tests, found to be correlated with overcharging as the slope of
different mathematical treatments, models, and frameworks are impedance evolution turns from negative to positive when the
necessary to validate and interpret the results. cell starts to overcharge. By online monitoring of the dynamic
2.2.1. Linear Time-Varying Conditions. For the LTV impedance at 70 Hz, the thermal runaway could be prevented
conditions, the input current signal is usually the intermodu- by cutting off the charging once the slope turns positive. It
lation of a direct current signal with an alternating current provides a protocol for single-frequency DEIS application. It is
signal as shown in Figure 2b.36,150 In this way, the impedance better to use single-frequency DEIS with a selective frequency
under operation conditions rather than stationary conditions as an indication method for detecting specific degradation
could be detected, which is helpful to unravel hidden kinetic mechanisms.
mechanisms and parameters. It is worth noting that the 2.2.2. Nonlinear Time-Varying Conditions. For the NLTV
frequency range used in linear DEIS usually discards the low conditions, the output response deviates from the ideal linear
frequency range to shorten the test time because the short test framework owing to the nonlinear distortion. As a result, it is
time could decrease the change of SOC during measurement necessary to detect and quantify this deviation to evaluate the
to guarantee the data quality. The data of linear DEIS could validity of impedance measurement and achieve data
also be validated by the K−K relations. Itagaki et al. first interpretation.44 Nonlinear EIS can provide additional
employed DEIS in LIBs to analyze the mechanism of Li+ information to distinguish the physicochemical processes in
intercalation into graphite and the formation of SEI.37,38 certain cases where linear EIS is incapable, since the faradaic
Huang et al. measured DEIS of a LiMn2O4/Li half-cell in the and nonfaradaic reactions show considerably different
galvanostatic mode (0.1 mA alternating current) while the cell contributions to higher harmonics.45 Nonlinear EIS has already
was charging or discharging at a series of direct currents (0.2− been used in solid oxide fuel cells,46 electrochemical reaction
1.6 mA).39 It is found that the charge-transfer resistance mechanisms,47,48 and electrocatalytic reactions.49 However,
decreases with the rise of direct current, which can be only a few groups have been devoted to the research of
explained by the Butler−Volmer equation. Moreover, the Rct employing the techniques for LIBs due to the complexity of
during the charge process is smaller than that during the battery chemistry and the difficulty of nonlinear EIS measure-
discharge process at the same SOC. The underlying reason, ment and data analysis. At present, two kinds of methods have
proposed based on a single particle model, is attributed to a been proven to be feasible in LIBs, i.e., nonlinear frequency
significant surface lithium-ion concentration variation caused response analysis (NFRA) based on higher harmonic response
by direct current. The phenomenon, called Rct hysteresis, has and odd random phase electrochemical impedance spectros-
also been reported in other works.40−42 In another work, the copy (ORP-EIS) based on multisine excitations.
phenomenon is ascribed to the change in the Li+ concentration NFRA is an effective method to obtain dynamic information
in the electrolyte, which increases during deintercalation and about nonlinear processes by investigating higher harmonic
decreases during intercalation.41 In addition, Watanabe et al. responses. It is known that a small sinusoidal perturbation
also found that both the Rct and the magnitude of the Rct could lead to a sinusoidal response of the same frequency (f)
hysteresis increased with the particle size of cathode with a phase angle. But if a large amplitude perturbation is
materials.42 They considered that the hysteresis of Rct could employed, the response would be the sum of sinusoids at f, 2f,
J. Phys. Chem. C 2023, 127, 4465−4495
Table 1. Symbols, Circuits, Figures in Complex Plane, and Physical Meanings or Applications of the Electrical Elements and
Simple Equivalent Circuits
J. Phys. Chem. C 2023, 127, 4465−4495
3f, etc., while the sinusoids of higher frequencies nf (n ≥ 2) are an operando ORP-EIS test is too massive to store in a
called higher harmonics. For NFRA, a sinusoidal current with a commercial electrochemical station.
large amplitude is used to excite the system, as shown in Figure Apparently, varieties of DEIS techniques are very attractive,
2d. Thus, the amplitude of individual higher harmonic Yn and since they can disclose dynamic impedance variations and
the sum of them ∑i n= 2Yn can be obtained on the basis of the hidden kinetic mechanisms during operation. Nevertheless,
Fast Fourier Transform (FFT). The value of Yn reflects the researchers should be very careful when using these
strength of the nonlinear responses of different processes. techniques, as the error derived from nonlinear and nonsta-
Generally, physical processes like ionic transport show a tionary states may lead to a misleading conclusion. The validity
relatively weak nonlinear response, while the redox reactions, of the DEIS data needs to be assessed based on the basis of a
charge-transfer processes at heterogeneous interfaces, deliver robust theoretical framework.
strong nonlinear responses. Harting et al. investigated the
amplitude and temperature dependency of higher harmonic 3. EIS DATA ANALYSIS FOR GENERAL BATTERY
responses as well as the impact of aging of LIBs with NFRA AGING DIAGNOSIS
and EIS.50 Although it is difficult for EIS to deconvolute the The interpretation of EIS data is the most important but also
impedance of electrochemical reactions and the ionic transport the hardest part of its application. When using a symmetric cell
process in the SEI, NFRA could separate them and distinguish or three-electrode system to measure the EIS of a single
their typical frequency range by a local minimum in Y3. electrode in LIB, the elementary processes that could be
Admittedly, NFRA is a powerful dynamic measurement detected involve (1) transfer of electrons from current
method that allows in-depth kinetic studies and diagnosis in collector to active materials; (2) transfer of electrons across
LIBs. Nevertheless, there are still some limitations of this the composite electrode layer; (3) diffusion of ion through the
method. First, it is more of a qualitative method than a electrolyte; (4) diffusion and migration of ions through the
quantitative method, as it usually gives a relatively broad separator; (5) diffusion of ions in the porous electrode; (6)
frequency range convoluting a class of processes, such as double-layer charging in the solid/electrolyte interface; (7)
diffusion or electrochemical reactions. In addition, the clear electrochemical insertion process; and (8) coupled diffusion of
interpretation of spectra requires physical models of the ion and electron inside the active material particles.15,64
electrochemical systems. A robust correlation between Yn and Although EIS could distinguish these individual processes in a
certain physical meaning or physicochemical processes should single spectrum in the ideal case, many of them would
be constructed if they are used as indicators for certain convolute and overlap with each other in the realistic
degradation mechanisms.52−57 measurement. Therefore, it is a critical issue to deconvolute
In addition to NFRA, ORP-EIS is another state-of-the-art the inside information and achieve an unambiguous
operando impedance technique. It was developed by the interpretation of the EIS data by modeling. The battery
research group from the Vrije Universiteit Brussel in 2004.58 aging could be diagnosed as different degradation modes or
Not until recent years has it been successfully used for LIBs. mechanisms by monitoring the impedance evolution during
The method is able to obtain the full-frequency impedance aging. Edge et al. identified five principal degradation
change of LIBs over a wide SOC range during normal mechanisms, i.e., SEI layer growth, lithium plating, structural
operation. A specially designed broadband periodic odd decomposition of the positive electrode, and particle fracture at
random phase multisine excitation is imposed on a direct both electrodes.65,66 Furthermore, these degradation mecha-
current to detect the impedance and the nonlinear distortions. nisms couple with each other and lead to observable
The time series for odd random phase multisine currents are degradation modes at the cell level, including loss of lithium
generated in MATLAB, then written to the text file, and finally inventory (LLI), loss of active materials (LAM), conductivity
applied through a potentiostat. Halleman et al. extended the loss (CL), and stoichiometric drift. In this section, the most
theoretical framework of ORP-EIS to NLTV systems by commonly used modeling methods, ECM, DRT, and TLM,
developing a nonlinear distortion quantification method, which and their application for general battery aging diagnosis are
makes it possible for ORP-EIS to be applied in LIBs.59−61 Zhu introduced.
et al. used ORP-EIS to study the continuous evolution of the 3.1. Equivalent Circuit Model. 3.1.1. Data Modeling.
charge-transfer resistance over SOC for a NMC/graphite coin ECM is the most widely used method to analyze the EIS data
cell during charging at different temperatures, C-rates, and at present. It is an a priori method. The model should be
SOH.62 Since the method delivers a continuous function of constructed on the basis of the priori knowledge about the
resistance and SOC, the differential function of dRct/dSOC vs battery system and characteristics of the Nyquist plot. The
SOC is obtained in Figure 2e, where the aged cell showed a electrical elements are supposed to relate to the individual
larger upward slope than the new cell at SOC range of 70− physicochemical processes in the battery. By constructing
100%. Halleman et al. conducted ORP-EIS in a commercial ECM and using complex nonlinear least-squares (CNLS) to fit
potentiostat to monitor the impedance evolution of LIBs EIS data, the impedance of different electrical elements could
during charging and discharging.63 It is found that the be obtained. There are several fundamental electrical elements,
nonlinear contribution dominates at low SOC. The physical including in resistor (R), capacitor (C), inductor (L), constant
origins of the differences between classical EIS and operando phase element (CPE, Q) and Warburg (W). Table 1 shows the
EIS are also discussed. The operando ORP-EIS is a promising symbols, figures in complex plane, physical meanings of the
way to monitor the impedance in real-time during LIBs electrical elements, and simple equivalent circuits.
operation. Nevertheless, several challenges need to be Resistance, capacitance, and inductance are three types of
overcome. First, both the testing hardware and data processing fundamental electric behaviors that can be simulated by
software of the system are specially designed and self- resistor, capacitor, and inductor. CPE is developed to rectify
assembled. Second, the high threshold makes it hard to the deviation of realistic capacitance, which is also called
popularize at present, even in the lab. Last, the data obtained in capacitance dispersion. Its origin is still uncertain. Several
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 3. Typical Nyquist plot of LIBs and the dominant processes in different frequency ranges.
reasons have been discussed to cause the phenomenon, Despite high-frequency inductance, midfrequency and low-
including nonuniform potential and current distribution, frequency inductance is relatively rare to see but also exists in
surface heterogeneity and roughness, and electrode porosity.67 LIBs.68 Similarly, the low-frequency inductance arc could also
The figure of a capacitor in the Nyquist plot is a vertical ray be modeled by an RL-element coupled in a Randles circuit. At
with a 90° phase angle from the origin, while that of a CPE is a the border of superhigh- and high-frequency regions lies the
ray with a phase angle ranging from 0 to 90° from the origin. intersection of the Nyquist plot and real axis, which represents
Warburg impedance is used to simulate the faradaic impedance the ohmic resistance of electrolyte and electrode. At high- and
resulting from the diffusion process. Due to the complexity of midfrequency regions, the capacitive reactance section is
diffusion process, various types of Warburg impedance have composed of several semicircles or semiarcs. Each semiarc
been developed, such as semi-infinite diffusion, finite diffusion, represents a single electrochemical process, which can be
spherical diffusion Warburg impedance, and so forth. Based on described by an RC-element or a ZARC element. In general,
these electric elements, different ECMs could be constructed the semiarcs in the high- and midfrequency region are assigned
to fit practical electrochemical systems. to SEI and charge-transfer reactions of the cathode and anode,
A typical Nyquist plot of commercial LIBs as shown in separately. At the low-frequency region, the diffusion
Figure 3 could be divided into three sections, the inductive impedance represents the diffusion process of ions in the
reactance in the superhigh-frequency region, the capacitive solid phase, which is usually simulated by the Warburg
reactance in the high- and midfrequency regions, and diffusion element.
impedance in the low-frequency region. The second-order 3.1.2. ECM for Degradation Mode Diagnosis. ECM exactly
Randles circuit is usually used to fit this Nyquist plot. The divides the Nyquist plot into several regions that are bound up
corresponding electrochemical processes involved in each with different degradation modes. By relating the impedance
region are discussed in detail below. variations to the degradation modes, we can identify and
At the superhigh-frequency region, the inductive signal quantify the batteries’ internal evolution during aging can be
appears at the fourth quadrant with a phase angle nearly −90°. identified and quantified. Specifically, the relationships in each
It is commonly observed in those systems with small region are shown as below:
impedance, especially in commercial LIBs. The phenomenon In the high-frequency region, the increase of ohmic
derives from the porous structure of electrodes, an outer lead, resistance (Rohm) is generally ascribed to CL. To be specific,
and spirally wound electrodes. Simply deleting the data of the CL is mainly caused by electrolyte decomposition and
inductive section is not a good way, although this way is now consumption. Thus, the change of Rohm is often used to reflect
widely used in EIS data processing. The presence of inductance the degradation of the electrolyte. In addition, the corrosion of
could pull the first semiarc near the intercept of real axis down current collector and stripping of conductive carbon can also
to the fourth quadrant. Therefore, deleting the range could cause a loss of conductivity, though their proportions are small.
lead to fitting error of the model parameter of the first semiarc. At the midfrequency range, the first semicircle modeled by an
To maintain the accuracy of fitting results, a series inductor or RC element relates to SEI, with RSEI representing Li+ diffusion
an RL-element should be added in ECM to fit the inductance. ionic resistance and CSEI representing the capacitance of the
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 4. Coupling relationships between degradation mechanisms and modes and corresponding impedance evolution with the resolution
capability of the data analysis methods (DRT and ECM). Reproduced with permission from ref 65. Copyright 2021 Royal Society of Chemistry.
SEI layer. The second semicircle modeled by an RC element the increase of Rct and Rw to LAM for two reasons. First, it has
relates to the charge-transfer reaction of the anode and been reported that Li plating could lead to the decrease of Rct
cathode, with Rct and Cdl denoting charge-transfer resistance of the anode.145 Thus, assigning the increase in Rct to LLI is
and double-layer capacitance, respectively. However, the not reasonable. Moreover, the plated Li would form SEI and
causes of the evolution of these elements are so complicated dead Li, while Xu et al. employed DSC to quantify the two
that they cannot be simply assigned to LLI and LAM. The processes and found that the formation of SEI accounted for a
coupling relationship between degradation mechanisms and larger proportion.72 Brown et al. also used the increased point
degradation modes is shown in Figure 4. As the growth of SEI of RSEI as the onset of Li plating.73 Therefore, the increase in
leads to LLI, the first semicircle is assigned to LLI. However, RSEI is able to reflect the degradation mechanisms of LLI, i.e.,
the second semicircle is influenced by several different SEI growth and Li plating. Based on this proposal, the metric,
degradation mechanisms, such as Li plating, structural GEIS, is defined as the impedance growth in percentage to
decomposition of cathode, and particle cracking. It contains quantify the effects of degradation modes, as shown in eq 12.
degradation mechanisms that lead to both LLI and LAM. In
the low-frequency region, the increase of Warburg resistance
(Rw) is caused by structural decomposition and particle
fracture, which should be assigned to the LAM of both
electrodes.
Based on the above analysis, it could be found that the point
of controversy is the ascription of Rct. In some studies, the sum
of RSEI and Rct is assigned to LLI, while Rw is assigned to
LAM.69,70 In the others, RSEI is assigned to LLI, while the sum
of Rct and Rw is assigned to LAM.71 Recently, Iurilli et al.
reviewed dozens of articles and concluded that Rct was assigned
to LAM in 61% of the cases.20 The controversy is derived from (12)
the low resolution of ECM method since the second semicircle Despite being most commonly used in EIS data analysis and
contains extensive information. The intrinsic limitation of the battery aging diagnosis, the ECM method has some inevitable
ECM makes this quantification method more of a rough limitations. First, there is no one-to-one correspondence
evaluation than a precise calculation. In this review, the authors between the ECM and kinetic behavior of the electrochemical
propound a proposal to assign the increase of RSEI to LLI and system, which may lead to serious misinformation about
J. Phys. Chem. C 2023, 127, 4465−4495
battery aging. Second, there is no one-to-one correspondence

between ECM and EIS data as well. The ECM may vary with (14)
the state of battery aging (SOC, temperature, pressure, etc.),
and a single EIS plot could be described by different ECMs.
The option on different models is artificial to some extent. (15)
Furthermore, it is relatively hard for ECM to deconvolute the
information between different kinetic processes with a similar (16)
time constant. The adjacent semiarch arcs usually overlap with
each other in a practical test, preventing researchers from
quantitative analysis and in-depth understanding about the (17)
battery systems. To overcome the difficulty and circumvent the
shortcomings of ECM, several strategies are developed and To accomplish DRT calculations, several kinds of algorithms
proved to be effective,90−94 including the three-electrode have been developed, including the Monte Carlo method,75
system, symmetric cell, and distribution of relaxation times. Fourier transform,76 genetic programming,77,78 and the most
3.2. Distribution of Relaxation Times. DRT is a time- widely used Tikhonov regularization.79,80 Recently, new
scale characterization that transforms the impedance of an strategies such as deep neural network81,82 and the Gaussian
electrochemical system from frequency domain to time process83,84 have also been developed and proved to be
domain. In the electrochemical system, the relaxation time of efficient and effective to realize DRT. Herein, the mechanism
a single electrochemical process corresponds to its time and effects of parameters in Tikhonov regularization are
constant. Time constant is an important concept in EIS, which introduced.
represents the time that response signals attenuate to 1/e of its There are two main processes in the Tikhonov regulariza-
maximum. The physical meaning is the time that the system tion. First, the Gaussian function is used as the radial basic
transits from the transient state to stationary state, reflecting function to discretize G(τ). The reconstructed G(τ) could be
the kinetic rate of the process. A larger time constant means a written as eq 18, where xm represents the weight coefficient
longer relaxation time. The relationship of the time constant, and Φm represents the basic function in discretization. In other
resistance, capacitance, and characteristic frequency (fc, Hz) in words, G(τ) is transformed to the weighted sum of numerous
a RC-element satisfies the following equation: Gaussian functions centered on τm. xm could be estimated by
minimizing optimization function S(x). As shown in eq 19,85
the first term is the fitting residue term exhibiting the
difference between experimental and modeled impedance,
(13) while the second term is the regularization term to obtain the
optimal solution of x, the vector of weight coefficient xm (x =
For the same process in a certain electrochemical system, its [x1, x2, x3, ..., xm]). A is a matrix consisting of An,m as defined in
time constant is determinate and nearly invariable, unless the eq 20.16 The product of An,m and x is the modeled impedance
component and structure of the interface have changed after discretization. In the regularization term, λ and P(x)
significantly. Each time constant corresponds to a single represent the regularization parameter and regularization
process in the system. Since different electrochemical processes function, respectively. The regularization parameter λ plays a
have different time constants, DRT is able to separate, vital role in the solution. A large regularization parameter leads
visualize, and quantify the various underlying electrochemical to an oversmooth fitting result with peaks merged so that the
processes in the system. It could be considered as an upgraded independent processes cannot be separated properly. With λ
successor to CNLS analysis.74 decreasing, the smooth and wide peaks gradually become sharp
3.2.1. Tikhonov Regularization. In principle, the DRT and narrow. The overlapped characteristic peaks are also
method is used to analyze the capacitance reactance part of separate from each other. However, an excessively small λ can
EIS. Therefore, the pristine impedance spectrum should go cause an overfitting problem. Recently, the L-curve method,86
through a data-preprocessing procedure to remove the hyperparameter optimization,87 particle swarm optimization
inductive reactance and diffusion sections. One can use the algorithm,88 and adaptive multiparameter regularization
inductor and Warburg element in ECM to fit the superhigh- approach89 have been developed to optimize the regularization
frequency inductance and low-frequency diffusion process and parameter. Comparing the impedances reconstructed from the
then subtract the as-obtained value of the series inductance combination of inversion of DRT and ECM with the original
and Warburg element from the raw data to get the data is also an important validation step.
preprocessed data for further DRT analysis. In contrast to
ECM, DRT is a model-free approach. It uses an infinite series
of RC elements and ohmic resistance to fit the impedance (18)
spectra. Based on the assumption, the impedance could be
expressed as eq 14, where Rp represents the total polarization (19)
resistance and g(τ) represents the distribution function of
relaxation times. As the frequencies are often displayed
logarithmically, a variable substitution shown in eq 16 is
(20)
used to transform eq 14 to eq 17. Thus, the issue is to calculate
the continuous function G(τ) on the basis of the finite 3.2.2. Peak Identification. Based on these algorithms, DRT
impedance data. It is the so-called “ill-posed problem”, which profiles with several peaks distributed over the time domain
means that G(τ) has more than one possible solution and is could be acquired. Then the most important and difficult
very sensitive to data noise. procedure is to assign DRT peaks to specific physicochemical
J. Phys. Chem. C 2023, 127, 4465−4495
processes. When the battery chemistry is known and clear or interfacial resistance of SEI, and charge-transfer resistance of
the peak characteristics in DRT are simple, comparing the time anode and cathode. Moreover, the activation energy of
constants of peaks with similar references could realize the different processes could be calculated based on the temper-
peak identification. Since the time constant of a specific ature-dependent EIS and eq 21, known as the Arrhenius law
process is determined by the kinetic characteristic, it would
vary within a relatively narrow range in the same battery
chemistry. The variation may be attributed to the difference of (21)
the composition, purity, and structure of electrode materials or where Rn, A, Ea, r, and T represent the resistance of a random
electrolyte. However, the method could not take effect when process in batteries, pre-exponential factor, activation energy,
the battery chemistry is complicated and unclear or there are gas constant, and temperature.
too many peaks. In that case, special strategies should be 3.2.3. DRT for Degradation Mechanism Diagnosis. DRT is
employed to separate the information on anode and cathode a better tool to do battery aging diagnosis than the ECM
from the full cell’s EIS. method. As it could deconvolute the impedance of the anode
Assembling anode and cathode half-cells to compare with and cathode from total EIS spectra of the battery, a
full-cell is an available way to decouple their effects on a full degradation mechanism level diagnosis could be achieved.
cell’s DRT profile. However, the lithium counter electrode After peak identification, the complicated EIS plot could be
would exert a negative influence on peak identification, as the deconvoluted to several individual peaks belonging to a certain
peaks derived from SEI and the charge-transfer process of the kinetic process. The integral areas of peaks represent the
lithium counter electrode may overlap other peaks. In contrast resistance of corresponding processes. By monitoring the
to the half-cell method, assembling the symmetric cells of variation of the peaks, the contribution of different degradation
cathode and anode is a feasible way to avoid the influence of mechanisms on battery aging could be quantified. Iurilli et al.
lithium counter electrode.90−93 In addition, introducing a compared commercial NMC/graphite full cells and anode/
reference electrode to construct a three-electrode system is also cathode half-cells at the beginning and end of life to build a
a common strategy to directly obtain the impedance spectra of relationship between the impedance and different degradation
anode and cathode. In a recent review, Xiao et al. mechanisms.90 Except the peak with τ < 10−3 s assigned to
comprehensively summarized the design parameters and electric and magnetic effects which shows no contribution to
typical applications of reference electrodes in LIBs.94 The degradation, the changes of other five peaks in the full cell’s
materials, manufacturing, placement, and geometry of the DRT profiles are assigned to the different degradation
reference electrode could influence the obtained results.95 In mechanisms, including SEI growth/Li plating (10−3 s < τ <
order to minimize the artifacts, the settings of the reference 10−2 s), cathode particle cracking (two peaks, 10−2 s < τ < 10−1
electrode were carefully rationalized. Compared to a symmetric s and 10−2 s < τ < 10−1 s), graphite degradation (1 s < τ < 10
cell, the biggest advantage of three-electrode system is that EIS s), and kinetic slow down (τ > 10 s). Zhou et al. combined
measurement could be operated at in situ conditions. However, DRT with a physical-based model to give a quantitative
there are still two limitations to the method. First, the stability analysis of the contribution from different sources on power
of the reference electrode is not good enough to support long- fade of LIBs aged under high-temperature cycling.98 The
term use. Thus, it is still a challenge to achieve in situ EIS evolution of the polarization resistance of different processes
measurement using the three-electrode system during battery during battery aging is distinctly provided, which finds that the
aging. In addition, the configuration of the three electrode low-frequency diffusion polarization played a dominant role
system for a large-capacity cell is still a difficulty, as the small and accounted for 48% of the total growth in the polarization
reference electrode can cause a local current heterogeneity. resistance. Moreover, the simulated voltage profile based on
Besides the above strategies, controlling the measured DRT results and eq 22 fits the experimental data well.
conditions, such as temperatures or SOC, to distinguish
peaks based on their different responsiveness is also a feasible
method. Some of the processes in LIBs are temperature- (22)
dependent, while others are SOC-dependent. Some processes
even occur only under specific testing conditions. Combining where Ut and U0 represent the terminal voltage and initial
the physical meaning of processes with DRT peaks would help open-circuit voltage and I is the discharge current. The voltage
to realize accurate peak identification. For example, it is known loss of ohmic resistance, an individual polarization process and
that lithium−sulfur (Li−S) batteries have a complex cell SOC change are calculated as IR0,
chemistry due to the existence of polysulfides. Soni et al.
combined a symmetric-cell strategy with a condition-controlled and , respectively.
strategy to achieve an unambiguous identification of the eight Although DRT possesses great advantages in EIS data
peaks in the DRT plot.96 Since polysulfides form gradually interpretation and battery-aging diagnosis, there are some
during the first discharge and appear at a specific voltage, the issues that should be paid attention to when using DRT. Due
peaks from polysulfide can be determined by comparing cells to its high-resolution, a small noise or error in the primary EIS
in a fresh state or with different SOCs. Zhu et al. used DRT to data could lead to a large deviation in DRT and
analyze the EIS of commercial NMC-NCA/graphite batteries misunderstanding about physicochemical processes in bat-
at different temperatures ranging from −5 to +25 °C. The teries. Thus, the quality of the primary data must be validated
resistance of the charge-transfer reaction hidden by SEI at by K−K relations before further analysis. In addition, the
room temperature is able to be distinguished at low pseudopeak formed during DRT calculation is another thorny
temperature.97 According to the different temperature depend- problem that may lead to incorrect modeling results. Wang et
ency of time constants and resistances, the four peaks in DRT al. discussed the types, origins, and factors of pseudopeaks in
are assigned to particles/current collector contact resistance, DRT.99 The pseudopeaks could be sorted into three types: (1)
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 5. (a−c) Schematic diagram of the interfacial reactions in different electrodes and their electrochemical models. Reprinted from ref 107.
Copyright 2015 American Chemical Society. (d) Typical physicochemical processes in a LIBs electrode and the physics-based transmission line
model. Reprinted with permission from ref 101. Copyright 2021 Elsevier.
artificial peaks due to an overfitting problem; (2) missing peaks sections, the electron conduction and ion diffusion in the
due to an underfitting problem, and (3) shifted peaks. The current collector/electrolyte interface, porous electrode/
regularization parameter, uniformity of peak intensity, and electrolyte interface, and the porous separator/electrolyte
impedance model are related to the generation of pseudopeaks. interface. In addition, the model could be simplified to the
Simulating impedance spectra from integer to fractional-order de Levie model by removing C04 and C14 when the diffusion of
models, which means splitting the whole spectrum, may be a electrolyte in the pores and separator could be neglected.
good way to deal with the pseudopeak issue. However, there is Moškon et al. discussed the calculative process of resistance in
still lack of a simple and effective method or standard to the TLM.101 Several commercial and open-access software
evaluate the degree of overfitting in DRT. To sum up, further programs have also been developed to help achieve the
development of new algorithms to improve the accuracy of simulation based on TLM.102−104
DRT, decrease the artifacts caused by artificially choosing TLM possesses significant advantages in assessing the
fitting parameters, and even achieve a one-click modeling resistance of ionic diffusion (Rion) in porous electrodes and
procedure plays a crucial role in better application of DRT. extracting the parameters of ionic conductivity in the electrode.
3.3. Transmission Line Model. TLM is a physical model Therefore, it is a powerful tool to investigate the effects of
used to treat the impedance issue of the porous electrode. For structural decomposition during battery aging, such as changes
a simple electrode/electrolyte interface, the electrochemical of pore volume and tortuosity in the electrode. Nara et al.
behavior is described by a series-connected RC-element and compared the fitting results of a conventional Randles circuit
resistance of solution (Figure 5a). However, there is an and TLM in LIBs, which found that TLM exhibited better
assumption of the circuit that the interface could be regarded fitting accuracy.105 Moreover, Rion could reflect the volume
as a uniform and continuous plane, which is not always the change of active materials during aging effectively. Bernard et
case for LIBs. TLM considers the entire interface of the porous al. used TLM to evaluate the influence of electrochemically
electrode as the continuum of an infinite conjoint small blocking pores during cycling and calendar aging of LIBs. A
interface. The small interface contains three physicochemical noticeable polarization of anode at low frequency after aging
processes, including electron conduction in electrode, ion indicates that the decreased porosity caused by SEI growth and
conduction in electrolyte, and interface reaction. Figures 5b,c gas bubble formation also plays a major role in battery
shows the corresponding TLMs of different interface reactions aging.106
(nonfaradaic processes or faradaic processes). Each RC
element simulates the electrochemical reaction occurring on 4. APPLICATION OF EIS TO LIB’S AGING STUDY
the nanometers of the total pore surface.100 Moreover, the RC 4.1. EIS Variations at Different Aging Conditions. In
element has different time constants, suggesting the different this section, the effects of calendar aging conditions (SOC,
time scale characteristics of the pore surface. Therefore, TLM temperature, battery chemistry) and accelerated stresses
can estimate the depth of electrochemical accessibility and the (temperature, charge/discharge rate, overcharging, etc.) on
electrochemical accessible surface area of the porous electrode battery aging behaviors and the corresponding impedance
or thick electrode. Figure 5d exhibits the schematic of typical evolution are systematically overviewed.
processes during Li+ intercalation and the TLM with the 4.1.1. Calendar Aging. Calendar aging study is a critical but
corresponding physical meaning. According to the interfaces in often overlooked issue in battery research. There are only
the electrode, the schematic could be divided into three about 500 reports in the latest 5 years when using the term
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 6. Calendar aging conditions: (a) Temperature: variation of EIS of calendar-aged cells over storage time. Changes in microstructures and
XRD patterns of the cathode material. Reprinted with permission from ref 109. Copyright 2021 Elsevier. (b) SOC: charge and discharge profiles of
the calendar-aged cells. Nyquist plots for disassembled anodes and cathodes and simulation results. Reprinted with permission from ref 112.
Copyright 2010 Elsevier. (c) Cathode: variation of capacity and internal resistance of different batteries stored at various SOC and temperatures.
Reprinted with permission from ref 114. Copyright 2016 The Electrochemical Society. (d) Further cycling: discharge curves and EIS fitting results
of the fresh cells and calendar-aged cells before and after cycling. Reprinted with permission from ref 117. Copyright 2019 The Electrochemical
Society.
“calendar” and “battery” to search on Web of Science. deficiency in the lattice of the calendar-aged cathode material,
Calendar aging means storage of LIBs without working. indicating that the degradation of the cathode is the main
Since there is no electrochemical reaction happening, the degradation mechanism. Schmitt et al. studied the effects of
performance degradation is caused by side reactions in the temperature on the calendar aging behavior of NMC batteries
battery chemistry. To be specific, the main degradation by storing them at different temperatures (0, 20, and 45 °C)
mechanisms include electrolyte consumption and decomposi- for 470 days.111 The Rohm and Rp change little at 0 °C storage
tion, SEI and CEI growth, transition-metal dissolution, and gas but are linearly enhanced at high temperature. Moreover, the
generation. The storing temperature, initial SOC, and battery slopes show a positive correlation with temperature. Sabet et
chemistry significantly influence the calendar aging.108 al. harnessed DRT to track the impedance evolution of
Although calendar aging is different from cycling aging, the commercial NMC cells stored at high temperature (35 and 45
basic usage of EIS to make battery aging diagnoses is the same. °C) and SOC (80% and 100%) for 700 days.110 It is found that
Both ECM and DRT are able to interpret the EIS evolution to RSEI and Rct increase by 5.14 and 3.18 times under the
unravel the internal change and degradation mechanisms of conditions of 80% SOC and 45 °C. The rise of Rct derives from
LIBs. the CEI growth and transition-metal dissolution, which is
From the perspective of an application scenario, investigat- proved by the change of time constant and post-mortem
ing the calendar aging behavior of LIBs at high SOC and high physical analysis. Li et al. explored the influence of SOC on
temperature has more practical significances. Du et al. calendar aging by storing LCO batteries at 55 °C and different
evaluated the calendar degradation behavior of LiNiO2-based SOC for 100 days.112 As shown in Figure 6b, the capacity loss
LIBs by storing them at 80 °C and 100% SOC for different gradually increases with the rise of SOC from 0 to 75%, while a
durations.109 As shown in Figure 6a, both Rohm and RSEI show capacity plunge appears at 100% SOC. It is found that both
little change after 45 days of storage, while Rct increases up to RSEI and Rohm of the anode show an approximate linear
10 times more than the fresh cell. The post-mortem analyses increase from 0 to 100% SOC, corresponding to the growth of
by SEM and XRD exhibit a cracked microstructure and lithium a new SEI layer. As a comparison, Rct of the cathode shows a
J. Phys. Chem. C 2023, 127, 4465−4495
slow rise at low SOC but dramatically increases at 100% SOC. a 1× higher RSEI and a 6× higher Rct than the battery stored at
The post-mortem analysis indicates that the high reactivity of 60 °C. The XRD, SEM, and XPS results indicate that cation
LCO at high SOC causes TM dissolution, CEI formation, and disorder and thickened SEI as well as CEI contribute to the
pore clogging. The research indicates that SEI growth impedance change of calendar aging. As the Rct increase
dominates the calendar aging at low SOC, while degradation dominates the impedance increase of the full cell, the
of cathode dominates the high SOC condition. polarization caused by cathode kinetics slowing down is
In addition to the calendar aging conditions, cathode identified as the primary effect of calendar aging on the further
materials also play important roles in batteries’ stability of cycling life of LIBs. In another work, the time span of calendar
batteries in calendar aging. Geisbauer et al. conducted a aging is also proven to significantly influence the batteries’
comparative study on the calendar aging behavior of five cycling performance.119 The batteries stored at 80 °C for 7
different cathode materials by relating the capacity decay to the days show little influence on the cycling performance.
resistance increase of full cells. It is found that NMC and NCA However, the cycling aging of a battery is accelerated when
show a great resistance increase at high temperature and SOC, the storage time is prolonged to 15 days. Thus, the test
while LFP exhibits the best stability with little resistance protocol of the combination of calendar aging conditions and
increase at different conditions.113 Keil et al. also studied the cycling aging conditions is necessary since it can lead to
impacts of SOC and temperature on the calendar aging of considerably different results. Moreover, the influence of
three kinds of commercial LIBs with NMC, NCA, and LFP calendar aging on battery cycling performance appears more
cathodes.114 As shown in Figure 6c, the internal resistance of in Rct than RSEI. The calendar aging could accelerate the
NMC and NCA increases slowly at low SOC (<60%) but degradation of active material, while the SEI grown during
rapidly enhances at high SOC, while the resistance of LFP calendar aging would not accelerate the SEI growth during
shows little variation at different SOC. The structural cycling aging.
decomposition of the cathode and side reactions between In summary, calendar aging-induced degradation mecha-
the cathode and electrolyte in NMC and NCA batteries is nisms mainly include SEI growth, electrolyte consumption, and
much more severe than in LFP batteries. Kassem et al. stored structural decomposition of positive electrode. Enhancing the
LFP batteries under different temperatures (35, 45, and 60 °C) temperature and SOC could markedly accelerate calendar
and SOC (30%, 65%, 100%) for 8 months. It is found that aging, as a high temperature speeds the rates of side reactions
elevated temperature exerts a more serious negative effect on up, and a high SOC increases the activity of electrodes. The
capacity than high SOC.115 Moreover, the Rp and Rohm impedance variations under different calendar aging conditions
significantly increase after storage, while RW only slightly are concluded in Table 2. The difficulties of calendar aging
increases, indicating that LLI (SEI growth) instead of LAMPE study are not the impedance data interpretation or degradation
is the main source of capacity fade. Nevertheless, the mechanism diagnosis but sufficient space and channel
contribution of RSEI and Rct on the increase of Rp is still resources and the protocols for calendar tests and reference
ambiguous because the time constants of Li+ diffusion in SEI performance tests. In a recent commentary, Dufek et al.
and the charge-transfer reaction of LFP are so close that they provided strategies to handle the issues and guidance for data
are hard to separate by ECM. To figure out the issue, Zheng et collection and test protocols of calendar aging study.120
al. disassembled the calendar aged LFP cell (55 °C, 100% Moreover, coupling EIS with machine learning may be a
SOC, 10 months) and reassembled half cells to investigate the promising method for calendar aging prediction.
impedance change of the cathode and anode.116 The Rct of the 4.1.2. Normal Accelerated Aging Protocols. In general,
cathode changes little, while the RSEI and Rct of the anode LIBs possess years of lifespan under normal working
significantly increase. The post-mortem analysis shows that Fe conditions. To assess the quality of LIBs, accelerated aging is
deposition and SEI growth at elevated temperature and high a critical issue to measure batteries’ lifespan in a relatively short
SOC causes the loss of cyclable or“active”lithium and capacity time span. By employing accelerated stresses, such as high
decay. temperatures, high charge/discharge rates, overcharging, over-
Assessment of the influence of calendar aging on the further discharging, the end of life could be arrived in advance.
use (cycling aging) of LIBs is an important issue in practical However, the battery aging characteristics can dramatically
application. But few studies have paid attention to it due to the deviate from the case under normal working conditions,
difficulty in setting a uniform test protocol. Mendoza et al. especially under harsh accelerated conditions. Based on the
compared the cycling performance of the fresh and calendar- types of stresses, accelerated aging could be divided into
aged LFP cells at −5 °C.117 As shown in Figure 6d, the normal and abnormal accelerated aging. The former should
calendar-aged cell possesses a characteristic of Li stripping in fulfill the prerequisite that the dominant degradation
the discharge curves, accompanied by a more significant mechanisms or modes change little compared with normal
impedance rise than the fresh cell. The results indicate that the working conditions, and the deviation part could be quantified
calendar aging process aggravates the Li plating of LFP and compensated in the lifespan calculation. Since EIS could
batteries in low-temperature operation. Zhang et al. combined make an in situ quantitative degradation mechanism diagnosis,
calendar aging and cycling aging to investigate the degradation it is a necessary technique to assess the accelerated aging
mechanisms and quantify the effects of the stored procedure conditions.
on cycling life.118 The NMC batteries are first stored at The high temperature accelerates the kinetic processes while
different temperatures (25, 60, 70 °C) and 100% SOC for 15 also increasing the speed of the side reactions. Jalkanen et al.
days and then cycled under normal conditions. It is found that used EIS to explore the impedance evolution of commercial
the temperature of calendar aging exerts a great influence on NMC pouch cells at different cycling temperatures (room
the further cycling performance. The batteries without storage temperature, 45 °C, and 45 °C charge/65 °C discharge).123 As
and stored at 60 °C show close enhancements of RSEI and Rct shown in Figure 7a, the capacity of different batteries drops to
after the cycling test, while the battery stored at 70 °C exhibits 80% of the nominal capacity after 2600, 2000, and 800 cycles
J. Phys. Chem. C 2023, 127, 4465−4495
for the three cycling conditions, respectively. Both the Rohm
109
112
111
110
121
122
115
116
118
ref
and Rct of all batteries show continuous upward tendencies,
corresponding to continuous electrolyte consumption and
SEI growth dominates the low SOC condition, while degradation
cation disorder at high temperature storage degrades the cycling

SEI growth is the dominant cause, while kinetic slowing down of
crack-induced cathode degradation. The RSEI of cells at room
LLI induced by SEI growth is the main cause, while LAM of

temperature and 45 °C first increases and then changes little,
SEI growth, electrolyte consumption, Fe dissolution, and

SEI growth, TM dissolution, cation-mixing, CEI growth
while the RSEI of cells at 45/65 °C surges at around 500 cycles
TM dissolution, CEI growth, structural decomposition

and keeps growing. The drastic impedance rise at 45/65 °C is
Particle crack, SEI growth, electrolyte consumption attributed to Li plating caused by excessive cycling, electrolyte
the cathode and anode is the minor reason

drying, and the growth of a thick and resistive SEI layer.
Degradation mechanism
Particle crack, TM dissolution, CEI growth
of cathode dominates the high SOC case
Strehle et al. studied the degradation of the NMC cathode

under 45 °C cycling tests, which found that the increase of Rct
was double that of the ambient temperature conditions.124 It
results from the faster kinetics for transition-metal dissolution
cathode is the minor reason
and rearrangement, oxygen removal, and reconstructed surface
performance of batteries
of the cathode at the higher temperature. Zhu et al. found that
high temperature accelerated the local overcharge reaction and
Table 2. Calendar Aging Tests and Corresponding Impedance Change and Degradation Mechanism Diagnosis of Different LIBs
uneven degradation of NCA cathode, as Rct of the aged

cathode from different locations exhibited a 10-fold differ-
deposition
ence.125 High temperature accelerates battery aging by

increasing the speed of SEI growth at the anode side, as well
as the particle crack and structural decomposition at the
cathode side.
Rct and RSEI significantly enhance, while the anode contribution is larger than that of the cathode
Only Rct shows a clear increase after storage. The aged cells at 70 °C show a significant impedance
RSEI and Rct of the anode linearly increase with SOC, while Rct of the cathode suddenly increases
Enhancing the charge and discharge rates is also a common

Rohm hardly changes while Rct of graphite and NMC increases significantly. Moreover, a new
method to accelerate aging, although their accelerated effects

Rp and Rohm increase faster at high temperature and SOC. SOC only shows a distinctive
Both RSEI and Rp show a similar linear upward tendency while the slope increases with
and mechanisms are quite different. Raj et al. investigated how

charge current (from 1C to 8C) and varying SOC ranges
(from 10% to 100%) affected the cycling performance of NMC
batteries.126 As shown in Figure 7b, the high charge rate with
RSEI and Rct increase by 5.14 and 3.18 times under 80% SOC and 45 °C
wide SOC range (4C/60% and 8C/30%) leads to serious Li

semiarc, derived from CEI, appears at the EIS of degraded cathode
plating, rapid increase of impedance, and capacity decay, while

RSEI significantly increases, while Rct of the cathode hardly changes
rise in further cycling tests compared to the aged cells at 60 °C
most cases show quasi-linear capacity decay and impedance

rise caused by SEI growth. Thus, the impedance vs cycle
number baseline (red dashed line in Figure 7b) is constructed
Impedance change
to assess the accelerated aging conditions of whether Li plating

Rohm, RSEI hardly change, while Rct increases 10 times
has happened or not. Similarly, Mussa et al. found Li plating,

impedance rise, and capacity plunge at 3C charging.127 But
when charging at 4C, the impedance of the cathode and anode
is smaller than the conditions with lower charge rates, while
accelerated effect at high temperature
the direct-current resistance of the full cell is much larger than

them. The phenomenon results from serious gas evolution due
to SEI and electrolyte decomposition. It could be determined
that the battery would quickly degrade to end-of-life with a
totally different aging process and degradation mechanism if
the charge rate is too high. Zheng et al. used varying discharge
at 100% SOC
rates from 0.5 to 3C to accelerate the aging of the NMC

temperature
batteries.128 As shown in Figure 7c, the capacity loss

accelerates when the discharge rate increases to 3C, although
all of them show a linear degradation tendency without distinct
capacity plunge. The Nyquist plots of the batteries show a
time (days)
gradual increase of different impedances at the low discharge

Storage
rates, while Rct dramatically increases at a 3C discharge rate.

100
470
700
133
392
870
300
45
15
The results indicate that the main causes for capacity

attenuation changes from SEI growth and electrolyte
100
100
100
consumption to structural damage of active materials with

SoC
(%)
100
100
100
0−
0−
50,
80
90
50
discharge rate elevated. Zhu et al. also observed a faster

increase in Rct and a moderate change in RSEI at high discharge
Temp (°C)
80
55
0−45
35, 45
75
25, 40, 60
55
0−60
25, 60, 70
rates.129 Obviously, enhancing the discharge rates would not

lead to such a dramatic change of degradation mechanism as
enhancing the charge rate does.
Moreover, coupling high temperatures with high charge/
chemistry
Li(Ni,M)
Battery
discharge rates is a promising way to be applied in accelerated

NMC
NMC
NMC
NMC
LMO
aging protocols for fast lifespan measurement. Since the high

LCO
O2
LFP
LFP
temperature could accelerate the kinetics of processes, it

J. Phys. Chem. C 2023, 127, 4465−4495
Figure 7. Normal accelerated aging stresses: (a) Temperature: variation of capacity, EIS, Rohm, RSEI, and Rct of batteries during cycling at different
temperatures. Reprinted with permission from ref 123. Copyright 2015 Elsevier. (b) Charge rate: variation of capacity and impedance of batteries
during cycling at different charge rates and SOC ranges. Reprinted with permission from ref 126. Copyright 2020 The Electrochemical Society. (c)
Discharge rate: variation of capacity and EIS of batteries after cycling at different discharge rates; schematic diagram of the change of degradation
mechanisms. Reprinted with permission from ref 128. Copyright 2022 John Wiley and Sons. (d) Coupling of high temperature and high charge
rate: variation curve of capacity retention and impedance (R1 and R2 represent Rohm and Rp, respectively) of batteries during cycling at different
temperatures and charge rates. Reprinted with permission from ref 132. Copyright 2022 Elsevier.
expands the boundaries of available charge/discharge rates that be found that enhancing the charge rate has a larger
fulfill the requirements for unchangeable degradation modes. accelerated effect than enhancing the discharge rate, as a
Qu et al. evaluated the coupling effects of high charging rate high charge rate could lead to Li plating. Actually, it is also an
(1, 3, 5C) to high temperature (25, 40 °C) on battery aging.130 accelerated strategy to artificially let Li plating occurs. The
The capacity decay of batteries at high charge rate and 25 °C is available methods include high charge rate, overcharging, and
quicker than that of batteries at high charge rate and 40 °C. temperature control. Rangarajan et al. operated cycling aging
The ICA and post-mortem analysis prove that Li plating occurs on commercial LCO batteries at different temperatures (−5,
at 3C/25 °C, 5C/25 °C, and 5C/40 °C cycling conditions. 22, and 40 °C), which found that the cells cycled at −5 °C lost
The impedance evolutions of batteries aged at 5C/40 °C show 60% capacity in 15 cycles, and the cells cycled at 40 °C
that RSEI and Rct first change slightly and then increase exhibited a hard knee point over 70 cycles, followed with swift
significantly after 400 cycles, which corresponds to the Li capacity decay.133 Both of the aging conditions lead to Li
plating-induced knee point. In addition, EIS could also be used plating, and the mechanisms are different. In the low
to evaluate the accelerated effects of coupling high temper- temperature condition, Li plating occurs at the initial cycle
atures with frequency regulation protocols based on pulsed due to the slow intercalation kinetics. The anode impedance
operations.131 Wu et al. used coupling accelerated conditions shows a continuous plating-induced rise resulting from the
of different temperatures and 30 C pulse charging to study the reaction between plated Li and the electrolyte, while at the
interplay of them.132 As shown in Figure 7d, the battery high temperature, the anode impedance shows a slight increase
capacity plunges under low temperature due to Li plating. during 30 to 70 cycles and surges after 70 cycles. The increased
When the temperature is gradually increased, the evolutions of SEI resistance in the first stage causes a higher overpotential,
impedance show two stages, where Rp first slightly enhances resulting in Li plating that in turn accelerates porosity
and then drastically increases near the knee point. The two- reduction and electrolyte consumption, further leading to a
stage aging behavior are attributed to SEI growth and Li plating-induced rise of anode impedance. Carter et al.
plating, respectively. investigated the influence of the directionality of the
4.1.3. Abnormal Accelerated Aging Protocols. At interelectrode thermal gradient on the degradation mode by
abnormal accelerated aging conditions, the aging processes employing a ±2 °C temperature gap between the cathode and
are considerably deviated from normal cycling aging. anode side.134 As shown in Figure 8a, the batteries lose 77%
Investigating these situations could help us understand the capacity in 20 cycles when the anode is warmer than the
battery failure mechanisms and suppress the unexpected cathode and 100% capacity when the cathode is warmer than
abnormal accelerated aging during battery working. It could the anode. The temperature gradient causes the imbalance of
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 8. Abnormal accelerated aging stresses. (a) Li plating: cycling profiles and EIS of batteries after cycling. Schematic diagram of aging
mechanisms for different thermal gradient conditions. Reprinted with permission from ref 134. Copyright 2021 Elsevier. (b) Overcharge: variation
of capacity and EIS of batteries under overcharging cycling. Reprinted with permission from ref 135. Copyright 2020 Elsevier. (c) Overdischarge:
variation of discharge and charge curves, EIS of full cell, anode, and cathode during 10% overdischarging cycles. Reprinted with permission from ref
138. Copyright 2014 The Electrochemical Society. (d) Pressure: correlation of SOH and internal resistance with pressure. Reprinted with
permission from ref 139. Copyright 2021 Elsevier. Schematic illustration of the influence of pressure on impedance. Reprinted with permission
from ref 140. Copyright 2022 Elsevier.
kinetics between cathode and anode. Based on the EIS aging mechanism at overcharge conditions. Similarly, Togasaki
degradation mode diagnoses, the higher temperature in the et al. also found a two-stage aging process of NCA batteries at
anode side exacerbates the normal degradation modes without overcharge conditions (SOC ≈ 107%).136 In addition, they
changing their proportion severely, while the higher temper- used temperature-dependent EIS to separate the charge
ature in the cathode side accelerates Li plating, leading to a transfer resistance of the anode and cathode, which found
capacity plunge. Though Li plating could significantly that the metal deposition on the anode could cause significant
accelerate battery aging, the degradation modes seriously change of resistance and capacitance of the anode. Liu et al.
deviate from the normal conditions. It is still hard to control Li used a 10% overcharge condition to accelerate the aging of
plating in full cells by controlling the aging conditions. commercial LFP batteries, which also found the LAM of
Overcharging is supposed to cause Li plating, as well. But cathode dominated the aging mechanisms.137
the capacity of an anode is always higher than the cathode in Overdischarging is another strategy to accelerate cycling
commercial batteries. The N/P ratio makes Li plating not so aging. It can cause irreversible decomposition of the SEI and
easy to occur at a normal charge rate. Liu et al. applied slight generation of gases, followed with the formation of a new SEI
overcharge stress (SOC ≈ 118%) to commercial NCA at the exposed fresh graphite surface. Liu et al. used EIS to
batteries, which led to 70% capacity decay in 50 cycles.135 investigate the failure of commercial LFP cells under
As shown in Figure 8b, the aging process could be divided into overdischarging conditions.138 As shown in Figure 8c, the
three stages by tracking the impedance evolution and batteries possessing 80% capacity retention after 1600 cycles
quantifying the proportion of different degradation modes rapidly failed after 49, 7, 3, and 1 cycles under 5%, 10%, 15%,
based on the basis of eq 12. In the first stage, the degradation and 20% overdischarge/normal charge cycling, respectively. It
mode has no change to the normal conditions with a slow is found that Rohm shows an upward tendency during aging,
increase of all the resistance. In the latter two stages, the while RSEI remained unchanged or fluctuated in the initial
tolerance of cathode to high voltage gradually becomes weaker, several cycles and suddenly increased before failed. The rise of
and finally structure damage occurs. Thus, the proportion of Rohm results from the corrosion of the current collector and
LAM exhibits an accelerated rising tendency, which is the main continuous electrolyte consumption. The evolution of RSEI
J. Phys. Chem. C 2023, 127, 4465−4495
Table 3. Accelerated Aging Tests and Corresponding Impedance Change and Degradation Mechanism Diagnosis of Different LIBs
Battery
Accelerated Stress chemistry Experimental conditions Acceleration effects Impedance change Degradation mechanism ref
Temperature NMC/ RT, 45, 45C/65D °C, 1C The EOL decreases from 2600 to 800 cycles Rohm and Rct keep a linear rise tendency, while RSEI surges after 500 SEI growth, electrolyte drying, Li 123
Gr charge/discharge cycles plating
NMC/ RT, 45 °C, 0.5C charge/ The EOL decreases from 1000 (79% retention) to The increase of Rct is double that of RT conditions and also much TM dissolution, oxygen removal, 124
Gr discharge 700 cycles (67%) quicker than Rohm and RSEI. reconstructed surface
Charge rate NMC/ 30 °C, 1−8C charge, 0.2C The cells exhibited a quasi-linear decay, but the Rp first increases linearly and surges near the knee point SEI growth (linear degradation 126
Gr discharge extreme high rate causes the capacity plunge region), Li plating (knee point)
NMC/ 34 °C, 1−4C charge, 0.25C The 3C charge rate causes a capacity plunge, while The Rp of the cathode and anode at 3C is the largest, while the Li plating (3C), gas evolution (4C) 127
Gr discharge 4C causes a quick failure direct current resistance of full cells at 4C is the largest
Discharge rate NMC/ 25 °C, 0.5C charge, 0.5−3C The batteries show a linear capacity decay with slopes All impedance increases gradually at low C, while Rct dramatically SEI growth (0.5−2C), structural 128
Gr discharge increasing with discharge rates enhance at 3C decomposition of cathode (3C)
Coupling condition NMC/ 25, 40 °C, 1−5C charge, 1C High temperature expands the available rates without RSEI and Rct show an accelerated increase tendency at high T and SEI growth, Li plating 130
(high T, high C) Gr discharge Li plating (20 °C 3C and 45 °C 5C). high C
The Journal of Physical Chemistry C
Overdischarge LFP/Gr 105−120% DOD The batteries rapidly fail in 7 cycles under 10% RSEI fluctuated in the initial cycles and suddenly increased before Electrolyte consumption, SEI 138
overdischarge failing, while Rct of cathode hardly changes decomposition/reformation
Li plating NMC/ temperature gradients (±2 °C) The batteries quickly fail in 20 cycles When anode is warmer, Rohm and Rp increase by more than 2 times Exacerbate normal degradation 134
Gr between cathode and anode modes
When cathode is warmer, Rohm and Rp increase more than 50 times Li plating, SEI growth
LCO/Gr −5, 40 °C, 1C charge/ Low T causes quick failure in 15 cycles, while high T Rp of anode shows a quasi-linear increase at low T, while that Li plating, SEI growth 132
discharge leads to capacity plunge at 70 cycles shows a drastic rise near the knee point at high T
Overcharge NMC/ 118% SOC The batteries show a hard knee point at 30 cycles and Rohm and Rct show a sudden rise at knee point, while RSEI hardly Particle crack and structural 135
Gr then rapidly failure changes decomposition of cathode
NCA/Gr 107% SOC The batteries show a hard knee point at 100 cycles Rct of both cathode and anode significantly increase around the TM dissolution, particle crack and 136
and then rapidly failure knee point structural decomposition of
4482
cathode
Pressure NMC/ 0.66−1.98 MPa The pressure at 1.32 MPa extends battery life, while Rohm and Rp of cathode and anode increase with pressure changed, SEI growth, kinetic retardation, 141
Gr too high or low pressure accelerates battery aging while the increase of Rw of the full cell is ascribed to graphite surface side-reactions
pubs.acs.org/JPCC
Article
https://doi.org/10.1021/acs.jpcc.3c00033
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 9. Schematic graph of impedance techniques for Li plating detection. (a) DRT results of batteries after the different number of cycles at 0
°C. Reprinted with permission from ref 145. Copyright 2021 Elsevier. (b) Schematic current and voltage profiles and the evolution of EIS during
the voltage relaxation process. Reprinted with permission from ref 147. Copyright 2016 Elsevier. (c) variation of Ri over SOC, while the inflection
point in the profile indicates the onset of Li plating. Reprinted with permission from ref 148. Copyright 2022 Elsevier. (d) Schematic illustration of
RSEI evolution during fast charging for the operando determination of onset Li plating. Reprinted with permission from ref 73. Copyright 2021
Elsevier. (e) Schematic illustration of dynamic capacitance measurement method for the operando determination of onset Li plating. Reprinted with
permission from ref 151. Copyright 2022 Elsevier.
implies the decomposition and reformation processes of SEI, typically lead to rapid battery failures with different dominant
while the sudden increase is attributed to the largely forming degradation mechanisms. EIS could not only assess the
nonconductive components. Moreover, with the impedance of boundary conditions of the normal accelerated aging protocols
the anode and cathode separated by a three-electrode system, but also unravel the main cause and contributions of different
it is found that the impedance increase of the anode dominates degradation mechanisms on battery failure.
the impedance increase of the full cell, indicating that the 4.2. Degradation Mechanism Analysis. This section
change of SEI is the main degradation mechanism under focuses on the impedance methods for investigating two main
overdischarging condition. degradation mechanisms, Li platin, and SEI growth. The newly
Besides the electrochemically accelerated stress, external developed impedance techniques for detecting the onset
pressure could also influence the lifespan of LIBs. Deich et al. boundaries of Li plating are summarized. The SEI growth
exerted different external pressures on LIBs under cycling monitoring based on the impedance evolutions during battery
aging.139 As shown in Figure 8d, the external pressure is aging are also introduced as well as the physics-based SEI
positively and linearly correlated to the capacity fading rate and impedance model.
the impedance rising rate. However, Mussa et al. reported a 4.2.1. Li Plating. Li plating usually occurs when LIBs are
“V” type relationship between battery lifespan and external charging at harsh operation conditions (such as fast charging,
pressure. The batteries show the longest cycling lifespan and charging at low temperature, or overcharging) or undergo
smallest change of Rohm and Rp after cycling at 1.32 MPa while long-term cycles. When Li plating seriously occurs, the plated
further increasing the pressure results in a faster rate of Li metal could puncture the SEI layer and react with the
impedance rise and capacity decay. Li et al. systematically electrolyte to form a new SEI layer, leading to LLI and
concluded the influence of the pressure on battery irreversible capacity loss. Moreover, since Li stripping during
impedance.140 The appropriate pressure could help achieve the discharge process generally happens near the surface of the
the optimal porosity and good contact between the electrode anode, the tip of the Li dendrite would split away and form
and electrolyte with decreases in Rion, Rohm, and Rct, while an electrical isolated “dead Li”. The detrimental Li plating can not
excessively high pressure would reduce the contact area, only result in capacity fade but may also cause serious safety
leading to an accelerated increase of Rion and Rct. hazards. Therefore, developing a nondestructive, real-time way
The accelerated stresses could exert different effects on the to monitor and quantify Li plating is of vital importance to
battery aging process, leading to different degradation modes develop methods to prevent it, improve LIBs’ safety in use, and
and impedance evolutions, which are summarized in Table 3. predict the RUL of LIBs. There have been many physical
The normal accelerated aging conditions, such as coupled high characterizations and electrochemical methods used to observe
charge/discharge rate and temperature, could linearly accel- and quantify Li plating. Tian et al. compared the advantages
erate the capacity fade with little degradation mechanism and disadvantages of different detection methods for Li plating,
change, so as to estimate the batteries’ normal lifespan in a which showed that EIS has no advantage over quantitative-
short time. The abnormally accelerated aging conditions ness.142 Although EIS has been used to characterize Li plating
J. Phys. Chem. C 2023, 127, 4465−4495
for a long time, most of the earlier studies were qualitative due than Y2 in the midfrequency range in the cells aged at −10 °C,
to the difficulty of extracting the information on Li plating. which was not observed in the cells aged at 25 °C.54 The
Both Li plating and Li intercalation are Faradaic processes in phenomenon is attributed to the Li plating on the anode. Due
the EIS test. They are a pair of competing reactions convoluted to the different sensitivity of higher harmonics to Li plating,
in the semicircle of the charge transfer process, leading to great the intensity changes of Y3:Y2 at 50 Hz is chosen as an
difficulty to separate them. indicator for Li plating, though the physical and quantitative
Petzl et al. employed a low-temperature (−22 °C) cycling relations between them are hard to build up. Koseoglou et al.
aging study on commercial LFP batteries to cause Li plating, employed continuous linear DEIS during the fast-charging
which was proven by post-mortem characterization.143 The procedure with a narrow frequency range (3.2 kHz to 1 Hz),
impedance evolution shows that the RSEI and Rct have hardly short time (9 s), and small SOC variation (<1.5%) for each
changed, while Rohm increase slightly, owing to the electrolyte impedance test.150 The operando impedance exhibits an abrupt
consumption caused by the reaction between plated lithium decrease at high SOC, indicating the onset of Li plating. Brown
and electrolyte. Schuster et al. used EIS to investigate the et al. used linear DEIS and DRT to study the impedance
nonlinear aging characteristics of LIBs, which found that Li change of Li plating during fast charging in three-electrode
plating could occur after a long-term duration at moderate NMC cells.73 Despite the decrease in Rct, a significant increase
working conditions and cause the knee point.144 The increase in RSEI is observed, as shown in Figure 9d. The inflection point
of Rp in the midfrequency range is regarded as the signal of the is considered as the onset of Li plating. The plating detection
occurrence of Li plating. However, it is hard to distinguish the method is cross-validated by ex situ mass titrations of Li metal,
information on Li plating from Li intercalation by ECM. EIS which proves to possess high Li detection sensitivity (<0.6% of
accompanied by DRT is more effective in this issue. Chen et al. the graphite capacity). Xu et al. developed an operando Li
used DRT to analyze the EIS of NMC cells cycled at normal plating detection and quantification method based on single-
and low temperature. As shown in Figure 9a, Li plating causes frequency DEIS:151 Since Li plating causes a big enhancement
the shift of the Rct peak to high frequency, accompanied by a of electrochemical active surface area (ECSA) so as to increase
decrease in peak intensity.145 The results indicate that Li the electrochemical double layer capacitance (EDLC), the
plating could lead to a rise of the time constant and a decrease EDLC on the surface of graphite anode could be used as a
of resistance of the charge-transfer process. Thus, the shift of quantitative indicator for Li plating. The characteristic
the DRT peak of the charge-transfer process is considered as frequency of the charge-transfer process of the graphite
an indicator for the onset of Li plating. Gargh et al. used DRT anode is first determined as 15 Hz and used in single-
to study the impedance change of NMC/graphite cells during
frequency DEIS to observe the dynamic capacitance variation
fast charging,146 which found that the impedance of batteries
during charging. As shown in Figure 9e, the capacitance of the
with Li plating significantly reduced after charging and then
graphite anode shows a significant linear rising tendency after
increased to the baseline value during the relaxation period.
charging to a point, which is regarded as the onset of Li plating.
The authors ascribe the phenomenon to the damage of the SEI
Moreover, a linear relationship between the amount of plated
layer by Li plating and the reformation of SEI during the
relaxation period, but it should relate to the Li plating on the Li and capacitance evolution is constructed to achieve the
anode and reintercalation during relaxation. quantitatively study about the inactive Li, which is further
Combining voltage relaxation with EIS is an effective validated by TGC method. It is an innovative impedance-based
strategy to detect Li plating. It is known that the plated Li method that not only detects the onset of Li plating but also
could undergo chemical transformation to intercalated Li quantifies the inactive Li, though there is still room for
during the relaxation process. Schindler et al. investigated the improvement in quantification capability on low-content plated
impedance change during the relaxation time by conducting an Li.
EIS test for every 10 min after plating.147 As shown in Figure In summary, the effects of Li plating on the electrochemical
9b, the Rct gradually enhances with an increse in relaxation impedance (EIS) of the negative electrodes such as graphite
time, indicating that the plated Li reduces the anodic charge electrodes can be summarized as follows: (1) exposing newly
transfer resistance, while the reintercalation process of the plated Li to the electrolyte leads to the growth of new SEI
plated Li during relaxation enhances it. Similarly, Xu et al. also layer, accompanied with the rise of RSEI; (2) the resistance of
used the voltage relaxation process by employing a transient charge transfer process should decrease, while the double layer
relaxation process during charging to observe the change in iR capacitance would increase; (3) the time constant of the
drop.148 The resistance Ri defined from the iR drop is used as charge transfer process may change, corresponding to the shift
an indicator for the onset of Li plating since it would drop of peak in DRT. Based on these characteristics, different
suddenly once Li plating occurs (Figure 9c). Koleti et al. impedance-based methods have been developed and are shown
controlled the time of the transient relaxation process based on in Table 4. Generally, the impedance of the charge-transfer
the frequency of charge transfer reaction, which was able to process consists of two different but tightly related and
test 85% of the total change of voltage profile in 3 s.149 The convoluted processes, Li intercalation and Li plating. Both of
impedance at the chosen frequency is obtained by an them are faradaic reactions with close time constants. Once Li
intermittent charging procedure interrupted by a relaxation plating occurs, the activation energy of Li deposition will
process for every 1% SOC. The point where the impedance remarkably decrease compared with the nucleation process,
significantly decreases during charging is considered as the making it much easier to occur than intercalation in kinetic.
onset of Li plating. Thus, the Rct of Li plating is much smaller than that of Li
In addition to classical EIS, DEIS techniques show some intercalation. Based on the discussion, the change of peak
potential in Li plating detection. NFRA is able to identify the intensity and time constant of charge transfer process in DRT
information on Li plating in Y3 harmonic spectroscopy. should be attributed to variable dominance between the two
Harting et al. found that Y3 showed a more significant enhance processes. Nevertheless, the modeling of the anode’s EIS after
J. Phys. Chem. C 2023, 127, 4465−4495
Li plating and the quantification of plated Li by impedance
151
145
150
149
148
147
Ref
51
73
with low detection limit are still big challenges.
4.2.2. SEI Formation and Growth. Since Peled put forward
Quantification
Measurement
Operando and the formation of SEI on Li for the first time in 1979,152 SEI has
condition
drawn great attention and wide applications by battery
Operando
Operando
In situ
In situ
In situ
In situ
In situ
researchers. It was proposed as an ionic conductive and
electronic isolated passivation layer, which is formed on the
surface of Li and the graphite anode once the anode contacts
Onset
√
√
with electrolyte.153 SEI plays a crucial role in determining the
×
×
The changes of Rct and its time constant reflect the transformation battery performance, as it could protect the anode and prevent
The plated Li could decrease the Rohm of graphite anode so as to
The chemical intercalation during the voltage relaxation decreases

The higher harmonics show different sensitivity to Li plating, while
and form new SEI, leading to a rise of RSEI with SOC increasing
The onset of Li plating causes a rapid increase in ECSA and the
With a certain length of relaxation time, the estimated resistance

When Li plating occurs, the plated Li would break the SEI layer
cause the drop of Ri obtained from the transient state analysis

it from further reacting with the electrolyte. Despite its positive
could reflect the impedance of high and midfrequency range.

of charge transfer reaction from Li intercalation to plating
effects, it could consume lithium inventory and electrolyte,
impede the diffusion, and enhance the resistance of the
electrodes. The continuous growth of the SEI is a main
Li plating could decrease the Rct of graphite anode
degradation mechanism in battery aging, when the thickness

rise in double layer capacitance consequently
the amount of plated Li on graphite surface

The Li plating could decrease Rohm and Rct
and resistance gradually increase in the process. EIS is one of
the most powerful tools to characterize SEI due to its
Y3 increases more significantly than Y2
Mechanism
advantages of nondestructivity, in situ measurement, and easy

accessibility.154 The impedance and time constant of SEI could
be easily obtained from EIS plot by ECM or DRT. At present,
it has been commonly used to characterize SEI performance,
evaluate novel electrolyte,155 study SEI formation,156−158 and
monitor SEI evolution.159
As the growth of the SEI is one of the main degradation
mechanisms of LIBs, the impedance change could reflect the
evolution of the SEI during battery aging. Steinhauer et al. used
in situ EIS and DRT methods to monitor the SEI formation
process by carrying out EIS measurements at equidistant
The emergence of an upward trend of Cdl is considered as the
The emergence of an upward trend of RSEI is considered as the
The sudden decrease of the estimated impedance indicates the
The emergence of a downward trend of Ri is considered as the

The rise of Y3/Y2 ratio at midfrequency range is considered as
onset of Li plating, with the normalized intensities used to
The emergence of a downward trend of Rct is considered as
voltage intervals of 0.1 V during the first two cycles.160 As

The increase of Rohm and Rct during the voltage relaxation
The decrease of Rct of the anode and its shift in DRT is
shown in Figure 10a, the interfacial impedance showed a

maximum in the voltage region between 0.8 and 0.3 V vs Li/
Li+ in the first lithiation and decreased in the second lithiation,
indicating that SEI completely formed in the first cycle. Qu et
al. compared the impedance and DRT evolutions between two
Signal of Li plating
considered as the signal of Li plating
indicate the occurrence of Li plating

quantified the amount of inactive Li
cells by using baseline electrolyte (1.0 M LiPF6 in EC/EMC 3/

7) and improved electrolyte (1.0 M LiPF6 in EC/EMC/DMC
5/30/65 with 1% VC and 0.5% LiBOB) during 100 cycles at
Table 4. Different Impedance-Based Methods for Li Plating Detection
60 °C.161 The cell with improved electrolyte shows a much

occurrence of Li plating
the onset of Li plating
a signal of Li plating
smaller impedance rise and better cycling performance. The

onset of Li plating
onset of Li plating
results are ascribed to a dense SEI layer, determined by

comparing the CPE values of the initial EIS after formation
cycle since low capacitance indicates low surface area and
porosity. Moreover, the EIS and physical characterization
coupled strategy play a vital role in relating SEI impedance to
its composition. Wang et al. associated in situ EIS and DRT
Single-frequency DEIS is used to monitor the change
Monitoring the change of impedance during voltage

Classical EIS combined with DRT is used to unravel
drop in the periodical transient relaxation during

of Z″ at 15 Hz so as to monitor the capacitance
The continuous DEIS is employed to monitor the
relaxation process with a specific relaxation time
Classical EIS is applied continuously in parallel to
with XPS to study the formation and evolution of SEI, which

Quantifying the impedance (Ri) based on the iR
The continuous DEIS is utilized to monitor the
A high amplitude (1.5 C) alternating current is
evolution of midfrequency impedance during
found that the high-frequency and midfrequency semiarcs

the impedance change caused by Li plating
RSEI evolution during fast charging process
employed to obtain the higher harmonics
derived from the inorganic and organic components,

respectively.159 In Figure 10b, the resistance of organic layer
first increases and then decreases, accompanying with the
Required testing
change during charging process
the voltage relaxation process
increase of the resistance of inorganic layer. The evolution

indicates that organic components could transform to
inorganic components in the first 50 cycles. The rupture-
fast charging process
charging procedure
reformation mechanism is proposed to explain the increase of

RSEI. Xu et al. combined time-of-flight secondary ion mass
spectroscopy (TOF-SIMs) with EIS to study the influence of
dissolved Ni2+ on the composition, structure and impedance
evolutions of SEI (Figure 10c).162
The thickness of the SEI is supposed to be a good indicator
for battery states during aging.163 However, it is very hard to
estimation
DEIS+DRT
Impedance
Method
EIS+DRT
achieve an in situ observation on SEI, needless to say, through

VR+EIS
iR drop
NFRA
DCM
in situ monitoring the evolution of SEI thickness during aging.

DEIS
EIS seems to be a promising indirect method because it can

J. Phys. Chem. C 2023, 127, 4465−4495
Figure 10. (a) EIS and corresponding DRT of a graphite anode during the first and second lithiation. Reprinted with permission from ref 160.
Copyright 2017 Elsevier. (b) Differential Nyquist top diagram of the cell during the 50th cycle in cathodic and anodic processes. Reprinted with
permission from ref 159. Copyright 2022 Elsevier. (c) TOF-SIMS depth profiles of selected secondary ion fragments on the surface of the E-EL
electrode. The EIS profiles of electrodes stored in the pure electrolyte and the electrolyte with the addition of Ni2+. Reprinted with permission from
ref 162. Copyright 2022 John Wiley and Sons. (d) Transmission line model and typical corresponding impedance spectrum with denoted main
processes. Reprinted with permission from ref 168. Copyright 2019 American Chemical Society.
obtain the resistance and capacitance of SEI nondestructively. (24)

Nevertheless, it is still a thorny issue to transfer its electrical
property to thickness due to its inhomogeneous structure. If SEI film is generally recognized as a complex structure
the structure of SEI is assumed to be homogeneous, and the consisting of an inner, thin, and compact layer close to the
ionic resistivity and dielectric permittivity are unvaried during particle surface of about a few nanometers and an outer, thick,
aging, the SEI thickness (dSEI) can be calculated based on the porous layer exposed to electrolyte. Based on this under-
capacitance formula of plate capacitor and resistance formula standing, several physics-based models have been constructed
shown in eqs 23 and 24, where ε0, εr, σ, and A represent to describe the relations between SEI growth and impedance
permittivity of vacuum, relative permittivity, ionic conductivity, variation. Ploehn et al. developed a diffusion-controlled model
and surface area, respectively. However, the assumption which predicts that SEI thickness increases linearly with the
apparently deviates from the realistic conditions. The square root of time.164 Safari et al. developed the porous SEI
impedance of SEI is determined not only by the thickness model to evaluate the effects of SEI porosity on capacity decay
but also the chemical composition and morphology. Lu et al. and interfacial resistance increase.165 Chen et al. proposed a
found that the inorganic components were less capacitive and microstructure-aware impedance model to predict the effect of
more conductive compared to organic species.158 The the electrode microstructure on SEI thickness and interfacial
impedance of inorganic components and organic components resistance.166 Single et al. modeled the SEI as a porous surface
could be highly different due to their different structure, film and developed a physics-based model based on a
porosity, and dielectric constant. Hence, eq 23 needs to be thermodynamically consistent theory for electrolyte trans-
rectified based on the SEI structure and growth law. port.167 The model introduced two parameters, porosity and
tortuosity, to rectify the resistance formula in order to describe
the porous SEI structure. Though the proposed analytical
(23) impedance expressions approximately are in accord with the
J. Phys. Chem. C 2023, 127, 4465−4495
Figure 11. Three categories of feature extraction methods provide the data-driven sources for machine learning-based battery state prediction.
Reprinted with permission from ref 180. Copyright 2022 Elsevier.
results of ECM, the model is overparametrized with usually estimated by ampere hour counting and the OCV
experimentally inaccessible parameters. Tailian et al. built a methods. However, the ampere hour counting method suffers
SEI impedance model by using a RC-element and transmission from the problem of accumulating errors. The OCV method,
line model to simulate the composite layer and porous layer, although accurate, requires a long time, which is not suitable
respectively (Figure 10d), which could achieve a good fitting under working conditions. Since the impedance at specific
effect for further quantitative analysis.168 Recently, Santos et al. frequency ranges are SOC dependent, it can also be used to
combined feedback and multifrequency alternating-current estimate the SOC.172 In addition, as it is fairly difficult to
modes of scanning electrochemical microscopy (SECM) to measure the aging behavior under random accelerated
evaluate quantitatively the local and in situ electronic and ionic conditions, constructing an impedance-based model for SOH
properties of SEI in LIBs for the first time, respectively.169 The estimation is of vital importance. Generally, the performance of
method has potential in calculating the critical parameters of data-driven models highly depends on the ML algorithms,
SEI, such as resistivity and dielectric permittivity. quality, and quantity of data and feature extraction. The
In summary, although EIS is a proven technique in commonly used algorithms contain Gaussian process regres-
characterizing the electrical properties of SEI, the thorny sion, support vector regression, random forest regression, deep
issue is to quantitatively assess the thickness evolution of the neural networks, and so on. All of these methods have their
SEI during battery aging by monitoring impedance change. To advantages and disadvantages, as discussed more comprehen-
achieve this work, the resistivity and permittivity of individual sively elsewhere.173−175 This review mainly pays attention to
SEI components or organic and inorganic layers are necessary different categories of initial data, feature extraction methods,
parameters. Moreover, in-depth understanding of the structure and their effects on model performance.
and composition variation during SEI growth is required to There are usually three categories of feature extraction
build SEI models. methods used in machine learning, i.e., the broadband EIS
4.3. Machine Learning for Battery States Estimation. without any feature engineering, model parameter based on
In recent years, machine learning (ML) has played a more and ECM, and single-frequency impedance feature (Figure 11).
more instrumental role in LIBs and other energy storage Zhang et al. used over 20000 impedance spectra of commercial
applications.171 As EIS could reflect the internal information LCO/graphite batteries at different operation conditions as the
on LIBs and highly correlate with battery states, it could be database and employed Gaussian process regression to predict
used as the input data in machine learning to get data-driven RUL of batteries without knowing its past operating
models for SOC, SOH estimation, and RUL prediction. SOC, histories.22 The EIS spectra are directly taken as inputs. In
defined as the ratio of current capacity to nominal capacity, is this way, the spectral features used to predict degradation are
J. Phys. Chem. C 2023, 127, 4465−4495
Table 5. Impedance-Based Data-Driven Models for SOH, SOC Estimation, and RUL Prediction
Battery Feature ML
chemistry Experimental conditions Pristine Data Number of features extraction method Target Error Ref
45 mAh 25, 35, 45 °C, 1C charge-2C Collect EIS every 120 (Z′ and Z″ of all Automatic GRP RUL R2 = 0.88, 0.81, 0.72 22
LCO discharge (1CC-2CD) even cycles frequency points) for 25, 35, 45 °C
Automatic DNN RUL 0.60% MAPE 182
Automatic SVR- SOH R2= 0.92 183
RBF
2.5 Ah 25−60 °C, 5−100% SOC Collect EIS every 4 (Rohm, Rp, Cdl, RW) ECM ANN SOC 1.2% RMSE 177
LFP 10% SOC (Manually)
2 Ah 1CC-1.5CD Collect EIS every 4 (Rohm, RSEI, Rct, ECM GA-BP SOH 1.6% RMSE 1.7% 176
NMC 200 cycles terminal voltage) (Manually) MAPE
3.5 Ah 25, 35, 45 °C, 0.5CC-1CD Collect EIS every 4 (Z′ and Z″ of 0.25 Fix-frequency LR SOH 1.0% RMSE 179
NMC 50 cycles and 398 Hz) (Manually) SVM SOH 0.64% RMSE
GRP SOH 0.50% RMSE
2.75 Ah 25, 35 °C, 0.5CC-1CD, 0.5CC- Collect EIS every 122 (Z′ and Z″ of all Automatic GRP SOH 1.2% RMSE 0.83% 180
NCA NEDC 25 cycles frequency points) MSD
5 (Rohm, RSEI, CSEI, Rct, ECM GRP SOH 0.93% RMSE 0.80%
Cdl) (Manually) MSD
6 (Z′ and Z″ of 1, 5, Fix-frequency GRP SOH 0.90% RMSE 0.67%
and 10 Hz) (Manually) MSD
automatically determined, which shows the most predictive in frequencies (0.25 and 398 Hz) exhibited the highest
the low-frequency region. However, directly using the whole correlation coefficient with the capacity fade.179 Therefore,
EIS as input without artificially selecting features could lead to they used the real and imaginary part of the two single-
some problems. First, there are more than 100 input values for frequency impedance, i.e., four parameters, as features
a single Nyquist plot. The huge amounts of data would lead to combined with different machine learning algorithms, includ-
a long training time and low efficiency of the models. ing linear regression (LR), support vector machine (SVM),
Moreover, the data quality could seriously influence feature and Gaussian process regression (GPR). The best model
selection and model accuracy since the features are completely exhibited the lowest RMSE of 0.5% using the GPR algorithms.
based on the algorithm and raw EIS data. Thus, a great deal of Jiang et al. conducted a comparative study of the three
EIS data are necessary to decrease the error caused by data categories of representative features, including the broadband
quality. EIS feature, model parameter feature, and fixed-frequency
Extracting the information inside EIS by ECM is a good impedance feature, with Gaussian process regression to figure
approach to improve the accuracy and efficiency of the ML out which type exhibits the best prediction performance.180 It
model, as the ECM could reflect the battery states with fewer is found that the model using the proper fixed-frequency
parameters. Liu et al. used the second-order Randles circuit to impedance as features showed the best estimation accuracy,
fit EIS data, while the three as-obtained model parameters, efficiency, and confidence. The RMSE, mean standard
Rohm, RSEI, and Rct, were further used as inputs for neural deviation (MSD), and training time of single-frequency
network.176 By conducting the genetic algorithm to optimize impedance model are 0.93%, 0.66%, 0.45 s, while the values
the back-propagation neural network, the training model could of pristine EIS data-based model and ECM parameter-based
predict SOH based on the three ECM parameters, with both model are 1.2%, 0.81%, 0.87s and 0.93%, 0.80%, 0.45 s,
mean absolute error (MAE) and root-mean-square error respectively. Moreover, by choosing different parameter groups
(RMSE) near 1.6%. Moreover, it is found that the MAE as inputs, a five parameters group, including Rohm, RSEI, CSEI,
rapidly increases to 4.8% when using the data with SOH > 80% Rct, and Cdl shows a better model performance than the ten-
as training data and SOH < 80% as testing data, indicating that parameters group containing all parameters of ECM and the
the comprehensiveness and representativeness of the input three-parameters group containing resistance only. Never-
data could markedly influence the model performance. Pizarro theless, the fixed-frequency impedances chosen in this
et al. also used four ECM-derived parameters as input data and contribution are 1, 5, and 10 Hz, whose positions are all
SOC as target to build a model based on an artificial neural located at the charge transfer process, leading to similar model
network, which achieved the SOC estimation with 1.2% performance when changing the combination of them. There is
RMSE.177 still room for improvement in the representativeness of the
Different from the ECM parameter-based feature, fix- chosen frequency.
frequency impedance-based feature is to select the impedance Besides only using impedance as inputs in the data-driven
at a specific frequency that correlates tightly to capacity or ML model, combining EIS with voltage profile and relevant
SOC. The mechanism is similar to single-frequency DEIS. As differential curves may further improve the model accuracy. Li
the processes related to the degradation mechanism possess et al. integrated impedance data and charging data as basic data
time constants, selecting the impedances at the corresponding and conducted a cuckoo search algorithm to build up an online
frequency could reflect the effects of the degradation degradation diagnosis framework.181 By collecting the field
mechanism on capacity decay. Locorotondo et al. found that data, the model can accurately identify eight parameters, where
there may be a certain frequency point that can be considered five parameters are impedance-related parameters and the
as a reliable indicator of SOH through data clustering.178 Zhu others are the OCV-related parameters, including remaining
et al. first investigated the relations between battery capacity capacity and stoichiometric parameters at the end of charge
and impedance, which found that the impedance at two and end of discharge. The identified model parameters could
J. Phys. Chem. C 2023, 127, 4465−4495
estimate both capacity decay and power decay with MAE plating in real-time battery working, among which Rct, Cdl, and
values lower than 0.5% and 1.5%, respectively. RSEI could be considered as the indicators in different
In summary, impedance-based models derived from ECM strategies. Lastly, an impedance-based data-driven model by
parameters and single-frequency impedance show better machine learning for SOC, SOH estimation, and RUL
estimation performance, as the artificial selected features prediction is introduced from the perspective of input data
could reflect the battery states well. Table 5 concludes the and feature extraction. The ECM parameters and fix-frequency
existing studies that use impedance for SOC and SOH impedance methods for feature extraction show better
estimation. Nevertheless, some issues need to be considered. estimation performance of the models.
First, although a single-frequency impedance-based method There are several challenging issues for developing advanced
exhibits the best estimation performance, it highly depends on EIS techniques for operando estimation of the battery state and
the representativeness of the chosen frequencies on the specific safety boundary, as listed below:
degradation modes. Once the frequencies lose their (1) Develop operando methods to detect Li plating with a
representativeness, the error may become huge. Both the low detection limit in the LIBs working at normal conditions.
aging level and aging protocol of the batteries can also lead to a Currently, most of the reported methods detect the onset
change of degradation mechanism and time constant variation. boundary of Li plating at low temperature, extremely fast
The operation conditions of aging protocols, such as charge/ charging, or overcharging, which possess a distinct and
discharge rate and temperature, are rarely considered in the considerable amount of plated Li. However, if the plated Li
present ML models. Second, DRT has hardly been used for is little and reversible, it will be difficult to detect it at the
machine learning so far, owing to the considerable difficulty in beginning stage of Li plating occurrence under normal
data importing. Since it is a continuous function over time operation conditions. Detection of residual Li plating during
domain, how to filter the valuable data and discrete data from normal operation mode as early as possible and repair the state
DRT is a thorny problem. Still, there is no doubt that DRT has of LIBs is of vital practical importance. Moreover, non-
huge potential in this application. Third, combining the voltage destructively quantifying the amount of plated Li in the aged
profile with impedance data for machine learning may achieve cells is also a challenging and meaningful task for safety issues.
a better estimation performance. The voltage profile reflects (2) Achieve SEI thickness monitoring by operando EIS
the thermodynamic properties of electrode materials, while the during battery work or aging research. EIS could extract the
EIS reflects the kinetic properties. The combination of them variations of the resistance and capacitance of SEI layer during
could act as the “fingerprint” of battery states. In addition, battery aging. However, it is hard to build a robust relationship
ORP-EIS is supposed to be a good method for SOC estimation between the electric parameters and accurate SEI thickness.
since a continuous function of impedance to SOC could be Sole use of EIS results is not sufficient for solving these tough
obtained. issues. A combination of EIS and in situ physical character-
ization should be a necessary and promising way to bridge the
5. SUMMARY AND PERSPECTIVES contribution between the electric parameters and the thickness
In summary, this review comprehensively introduces the and composition of SEI layer, especially their dynamic
principles, measurements, validation, modeling of EIS, and relationships.
recent progress of its applications in LIBs’ degradation and (3) Combine feature-extracted impedance data with the
aging research. For classical EIS, the measurement should fulfill aging conditions to construct ML-based model for predicting
the linearity, stability, and causality conditions to guarantee the battery aging behaviors, including cycling aging and calendar
quality of the data. Since EIS modeling highly depends on the aging. Since battery aging conditions significantly influence the
reliability of raw data, data validation by K−K relations should EIS evolutions, ML-models with these relations input could
be a standard checkpoint before publication of impedance broaden the applicability of the model and reduce the
data. In addition, several typical dynamic EIS techniques are experiments. Currently, ML is rarely used in calendar aging
introduced, respectively. Furthermore, the commonly used research. Considering the length of time needed in calendar
modeling methods such as ECM, DRT, and TLM and their aging studies, ML will be an indispensable tool in the future
usage on general battery aging diagnosis are introduced and research. Moreover, how to introduce DRT results into ML-
compared. Compared with the priori method ECM, DRT based model is also a challenge so as to realize a higher
could be regarded as an upgraded substitute with a stronger accuracy, since they contain more time-scale information than
quantitative ability to achieve degradation mechanism conventional methods.
diagnosis, while TLM is a physics-based model that has a (4) Develop credible operando EIS measuring methods for
great advantage in analyzing porous electrodes and the large-size batteries or battery pack. Although the reported
influence of porosity variation on battery degradation. impedance techniques are mature in small-size batteries for
The recent progress of EIS’s application in LIB’s degradation laboratory research, it is not easy to obtain comprehensive but
and aging research is summarized in three aspects. First, the operando EIS or onboard impedance spectrum of large-size
EIS variations at different aging conditions of calendar aging batteries or even battery pack due to the cost limitation and
and accelerated aging are systematically overviewed. EIS could technical difficulty. The impedance estimation methods still
not only diagnose the degradation mechanisms of LIBs at exhibit a relatively ambiguous accuracy, which require further
corresponding aging conditions, but it is also a good way to improvements in measurement and algorithms.
determine the boundary conditions of normal accelerated (5) Develop efficient algorithms and fitting evaluation
aging protocols for fast battery life test. Second, the newly methods for DRT. Admittedly, DRT is a powerful tool for
developed impedance techniques in detecting Li plating and EIS modeling and in-depth mechanism research. However, the
monitoring SEI growth, the two main degradation mecha- existence of artificial parameters and flexibility of modeling
nisms, are elaborately reviewed. Several Operando DEIS process could cause errors and misunderstandings in the
techniques show great potential in detecting the onset of Li vertical comparison of battery aging research. Thus, it is urgent
J. Phys. Chem. C 2023, 127, 4465−4495
to improve the algorithms and develop a better fitting program, D-M-D Dissolution-membrane-deposition
even a “one-click” fitting program without artificial parameters, EIS Electrochemical impedance spectroscopy
to achieve higher accuracy and efficiency. ECM Equivalent circuit model
Anyway, although EIS could provide abundant information f (Hz) Frequency
about battery states such as State of Charge (SOC) or State of GPR Gaussian process regression
Health (SOH) during battery aging, as EIS is mostly electrical- Gr Graphite
based data which is not difficult to be deconvoluted using a LCO LiCoO2
simple equivalent circuit model, it is necessary to combine EIS LFP LiFePO4
with other physical and electrochemical characterization LR Linear regression
methods to calibrate the data quantitatively and obtain a LLI Loss of lithium inventory
cross-validated conclusion rather than using EIS alone. We LAM Loss of active materials
expect that EIS could push and contribute our understanding LTV Linear time-varying
and insight into battery aging behavior to a deeper level once ML Machine learning
these problems are worked out. MAE Mean absolute error
MSD Mean standard deviation
■ AUTHOR INFORMATION
Corresponding Author
NLTV
NFRA
Nonlinear time-varying
Nonlinear frequency response analysis
Yong Yang − State Key Laboratory for Physical Chemistry of NMC LiNixCoyMnzO2
Solid Surfaces, Collaborative Innovation Center of Chemistry NCA LiNixCoyAlzO2
for Energy Materials and Department of Chemistry, College ORP-EIS Odd random phase electrochemical impedance
of Chemistry and Chemical Engineering and School of spectroscopy
Energy, Xiamen University, Xiamen 361005, People’s RT Room temperature
Republic of China; orcid.org/0000-0002-9928-7165; RUL Remaining useful life
Email: [email protected] RMSE Root means square error
Rohm (Ω) Ohmic resistance
Authors RSEI (Ω) Ion diffusion resistance of SEI layer
Wenxuan Hu − State Key Laboratory for Physical Chemistry Rct (Ω) Charge transfer resistance
of Solid Surfaces, Collaborative Innovation Center of Rion (Ω) Ion diffusion resistance in pores
Chemistry for Energy Materials and Department of RW (Ω) Warburg resistance
Chemistry, College of Chemistry and Chemical Engineering, Rp (Ω) Polarization resistance; the real part of impedance
Xiamen University, Xiamen 361005, People’s Republic of between the Rohm and RW
China; orcid.org/0000-0003-1222-459X SVM Support vector machine
Yufan Peng − State Key Laboratory for Physical Chemistry of SEI Solid electrolyte interface
Solid Surfaces, Collaborative Innovation Center of Chemistry SOC State of charge
for Energy Materials and Department of Chemistry, College SOH State of health
of Chemistry and Chemical Engineering, Xiamen University, TLM Transmission line model
Xiamen 361005, People’s Republic of China TM Transition metal
Yimin Wei − Contemporary Amperex Technology Co., Ltd., Yn n-th harmonic (n = 2, 3, 4...)
Ningde 352100, China Z′ (Ω) Impedance of real part
Z″ (Ω) Impedance of imaginary part
Complete contact information is available at:
τ (s) Time constant
https://pubs.acs.org/10.1021/acs.jpcc.3c00033
Notes
The authors declare no competing financial interest.
■ REFERENCES
(1) Palacin, M. R.; de Guibert, A. Why do batteries fail? Science
2016, 351 (6273), 1253292.
■ ACKNOWLEDGMENTS
This research was financially supported by the National Key
(2) Li, W.; Li, M.; Hu, Y.; Lu, J.; Lushington, A.; Li, R.; Wu, T.;
Sham, W. K.; Sun, X. Synchrotron-Based X-ray Absorption Fine
Structures, X-ray Diffraction, and X-ray Microscopy Techniques
R&D Program of China (Grant No. 2021YFB2401800). Applied in the Study of Lithium Secondary Batteries. Small Methods
2018, 2, 1700341.
■
ANN
NOMENCLATURE
Artificial neural network
(3) Han, B.; Zou, Y. C.; Xu, G. Y.; Hu, S. G.; Kang, Y. Y.; Qian, Y.
X.; Wu, J.; Ma, X. M.; Yao, J. Q.; Li, T. T.; et al. Additive Stabilization
of SEI on Graphite Observed Using Cryo-Electron Microscopy.
CPE Constant phase element Energy Environ. Sci. 2021, 14, 4882−4889.
CNLS Complex nonlinear least-squares (4) Fang, C.; Li, J.; Zhang, M.; Zhang, Y.; Yang, F.; Lee, J. Z.; Lee,
CEI Cathode electrolyte interface M. H.; Alvarado, J.; Schroeder, M. A.; Yang, Y.; et al. Quantifying
CL Conductivity loss inactive lithium in lithium metal batteries. Nature 2019, 572, 511−
CSEI (F)Capacitance of SEI layer 515.
(5) Tao, M.; Xiang, Y.; Zhao, D.; Shan, P.; Sun, Y.; Yang, Y.
Cdl (F) Electric double-layer capacitance Quantifying the Evolution of Inactive Li/Lithium Hydride and Their
dSEI (nm)
Thickness of SEI layer Correlations in Rechargeable Anode-free Li Batteries. Nano Lett.
DRT Distribution of relaxation times 2022, 22 (16), 6775−6781.
DEIS Dynamic electrochemical impedance spectroscopy (6) Tao, M.; Xiang, Y.; Zhao, D.; Shan, P.; Yang, Y. Protocol for
DNN Deep neural network quantifying inactive lithium in anode-free lithium batteries by mass
DCM Dynamic capacitance measurement spectrometry titration. Commun. Mater. 2022, 3, 50.
J. Phys. Chem. C 2023, 127, 4465−4495
(7) Xiang, Y.; Tao, M.; Zhong, G.; Liang, Z.; Zheng, G.; Huang, X.; (26) Lu, Y.; Zhao, C.; Huang, J.; Zhang, Q. The Timescale
Liu, X.; Jin, Y.; Xu, N.; Armand, M.; et al. Quantitatively analyzing the Identification Decoupling Complicated Kinetic Processes in Lithium
failure processes of rechargeable Li metal batteries. Sci. Adv. 2021, 7, Batteries. Joule 2022, 6, 1172−1198.
eabj3423. (27) Gaddam, R. R.; Katzenmeier, L.; Lamprecht, X.; Bandarenka, A.
(8) Wandt, J.; Jakes, P.; Granwehr, J.; Eichel, R. A.; Gasteiger, H. A. S. Review on physical impedance models in modern battery research.
Quantitative and Time-resolved Detection of Lithium Plating on Phys. Chem. Chem. Phys. 2021, 23, 12926−12944.
Graphite Anodes in Lithium Ion Batteries. Mater. Today 2018, 21, (28) McCarthy, K.; Gullapalli, H.; Ryan, K. M.; Kennedy, T.
231−240. Review�Use of Impedance Spectroscopy for the Estimation of Li-ion
(9) Chen, X.; Hu, Y.; Li, S.; Wang, Y. X.; Li, D.; Luo, C.; Xue, X.; Battery State of Charge, State of Health and Internal Temperature. J.
Xu, F.; Zhang, Z.; Gong, Z. L.; et al. State of Health (SoH) Estimation Electrochem. Soc. 2021, 168, No. 080517.
and Degradation Modes Analysis of Pouch NMC532/Graphite Li-ion (29) Wang, X.; Wei, X.; Zhu, J.; Dai, H.; Zheng, Y.; Xu, X.; Chen, Q.
Battery. J. Power Sources 2021, 498, 229884. A Review of Modeling, Acquisition, and Application of Lithium-Ion
(10) Campbell, I. D.; Marzook, M.; Monica, M.; Offer, G. J. How Battery Impedance for Onboard Battery Management. eTransportation
Observable Is Lithium Plating? Differential Voltage Analysis to 2021, 7, 100093.
Identify and Quantify Lithium Plating Following Fast Charging of (30) Osaka, T.; Mukoyama, D.; Nara, H. Review�Development of
Cold Lithium-Ion Batteries. J. Electrochem. Soc. 2019, 166 (4), 725− Diagnostic Process for Commercially Available Batteries, Especially
739. Lithium Ion Battery, by Electrochemical Impedance Spectroscopy. J.
(11) Adam, A.; Knobbe, E.; Wandt, J.; Kwade, A. Application of the Electrochem. Soc. 2015, 162 (14), 2529−2537.
Differential Charging Voltage Analysis to Determine the Onset of (31) Meddings, N.; Heinrich, M.; Overney, F.; Lee, J. S.; Ruiz, V.;
Lithium-plating During Fast Charging of Lithium-ion Cells. J. Power Napolitano, E.; Seitz, S.; Hinds, G.; Raccichini, R.; Gaberšcě k, M.;
Sources 2021, 495, 229794. et al. Application of electrochemical impedance spectroscopy to
(12) He, J.; Bian, X.; Liu, L.; Wei, Z.; Yan, F. Comparative study of commercial Li-ion cells: A review. J. Power Sources 2020, 480, 228742.
curve determination methods for incremental capacity analysis and (32) Hirschorn, B.; Tribollet, B.; Orazem, M. E. On Selection of the
state of health estimation of lithium-ion battery. J. Energy Storage Perturbation Amplitude Required to Avoid Nonlinear Effects in
2020, 29, 101400. Impedance Measurements. Isr J. Chem. 2008, 48, 133−142.
(13) Smith, A. J.; Burns, J. C.; Trussler, S.; Dahn, J. R. Precision (33) Esteban, J. M.; Orazem, M. E. On the Application of the
Measurements of the Coulombic Efficiency of Lithium-ion Batteries Kramers-Kronig Relations to Evaluate the Consistency of Electro-
and of Electrode Materials for Lithium-ion Batteries. J. Electrochem. chemical Impedance Data. J. Electrochem. Soc. 1991, 138, 67−76.
(34) Schonleber, M.; Klotz, D.; Ivers-Tiffee, E. A Method for
Soc. 2010, 157, 196−202.
Improving the Robustness of linear Kramers-Kronig Validity Tests.
(14) Burns, J. C.; Stevens, D. A.; Dahn, J. R. In-situ Detection of
Electrochim. Acta 2014, 131, 20−27.
Lithium Plating Using High Precision Coulometry. J. Electrochem. Soc.
(35) Hahn, M.; Schindler, S.; Triebs, L. C.; Danzer, M. A. Optimized
2015, 162, 959−964.
Process Parameters for a Reproducible Distribution of Relaxation
(15) Gaberšcě k, M. Understanding Li-based battery materials via
Times Analysis of Electrochemical Systems. Batteries 2019, 2, 5.
electrochemical impedance spectroscopy. Nat. Commun. 2021, 12,
(36) Battistel, A.; La Mantia, F. On the Physical Definition of
6513.
Dynamic Impedance: How to Design an Optimal Strategy for Data
(16) Hahn, M.; Schindler, S.; Triebs, J. C.; Danzer, M. A. Optimized
Extraction. Electrochim. Acta 2019, 304, 513−520.
process parameters for a reproducible distribution of relaxation times (37) Itagaki, M.; Kobari, N.; Yotsuda, S.; Watanabe, K.; Kinoshita,
analysis of electrochemical systems. Batteries 2019, 5, 43. S.; Ue, M. In situ electrochemical impedance spectroscopy to
(17) Schönleber, M.; Ivers-Tiffée, E. The Distribution Function of investigate negative electrode of lithium-ion rechargeable batteries.
Differential Capacity as a new tool for analyzing the capacitive J. Power Sources 2004, 135, 255−261.
properties of Lithium-Ion batteries. Electrochem. Commun. 2015, 61, (38) Itagaki, M.; Kobari, N.; Yotsuda, S.; Watanabe, K.; Kinoshita,
45−48. S.; Ue, M. LiCoO2 electrode/electrolyte interface of Li-ion
(18) Quattrocchi, E.; Wan, T. H.; Curcio, A.; Pepe, S.; Effat, M. B.; rechargeable batteries investigated by in situ electrochemical
Ciucci, F. A general model for the impedance of batteries and impedance spectroscopy. J. Power Sources 2005, 148, 78−84.
supercapacitors: The non-linear distribution of diffusion times. (39) Huang, J.; Zhang, J. B.; Li, Z.; Song, S. L.; Wu, N. N. Exploring
Electrochim. Acta 2019, 324, 134853. Differences between Charge and Discharge of LiMn2O4/Li Half-cell
(19) Lyu, N.; Jin, Y.; Xiong, R.; Miao, S.; Gao, J. Real-time with Dynamic Electrochemical Impedance Spectroscopy. Electrochim.
overcharge warning and early thermal runaway prediction of Li-ion Acta 2014, 131, 228−235.
battery by online impedance measurement. IEEE Trans. Ind. Electron. (40) Itagaki, M.; Honda, K.; Hoshi, Y.; Shitanda, I. In-situ EIS to
2022, 69 (2), 1929−1936. Determine Impedance Spectra of Lithium-Ion Rechargeable Batteries
(20) Iurilli, P.; Brivio, C.; Wood, V. On the use of electrochemical During Charge and Discharge Cycle. J. Electroanal. Chem. 2015, 737,
impedance spectroscopy to characterize and model the aging 78−84.
phenomena of lithium-ion batteries: a critical review. J. Power Sources (41) Huang, J.; Ge, H.; Li, Z.; Zhang, J. B. Dynamic Electrochemical
2021, 505, 229860. Impedance Spectroscopy of a Three-Electrode Lithium-Ion Battery
(21) Qu, D.; Wang, G.; Kafle, J.; Harris, J.; Crain, L.; Jin, Z.; Zheng, during Pulse Charge and Discharge. Electrochim. Acta 2015, 176,
D. Electrochemical Impedance and its Applications in Energy-Storage 311−320.
Systems. Small Methods 2018, 2, 1700342. (42) Watanabe, H.; Omoto, S.; Hoshi, Y.; Shitanda, I.; Itagaki, M.
(22) Zhang, Y.; Tang, Q.; Zhang, Y.; Wang, J.; Stimming, U.; Lee, A. Electrochemical Impedance Analysis on Positive Electrode in
Identifying degradation patterns of lithium ion batteries from Lithium-Ion Battery with Galvanostatic Control. J. Power Sources
impedance spectroscopy using machine learning. Nat. Commun. 2021, 507, 230258.
2020, 11, 1706. (43) Huang, J.; Li, Z.; Zhang, J. Dynamic electrochemical impedance
(23) Wang, S.; Zhang, J.; Gharbi, O.; Vivier, V.; Gao, M.; Orazem, spectroscopy reconstructed from continuous impedance measurement
M. E. Electrochemical impedance spectroscopy. Nat. Rev. Methods of single frequency during charging/discharging. J. Power Sources
Primers 2021, 1, 41. 2015, 273, 1098−1102.
(24) Vivier, V.; Orazem, M. E. Impedance Analysis of electro- (44) Pintelon, R.; Louarroudi, E.; Lataire, J. Detecting and
chemical system. Chem. Rev. 2022, 122, 11131−11168. Quantifying the Nonlinear and Time-Variant Effects in FRF
(25) Ciucci, F. Modeling electrochemical impedance spectroscopy. Measurements Using Periodic Excitations. IEEE Trans. Instrum.
Curr. Opin. Electrochem. 2019, 13, 132−139. Meas. 2013, 62 (12), 3361−3373.
J. Phys. Chem. C 2023, 127, 4465−4495
(45) Fasmin, F.; Srinivasan, R. Review�Nonlinear Electrochemical Lithium-Ion Batteries during Fast Charging. J. Power Sources 2022,
Impedance Spectroscopy. J. Electrochem. Soc. 2017, 164 (7), 443−455. 544, 231852.
(46) Wilson, J. R.; Schwartz, D. T.; Alder, S. B. Nonlinear (63) Halleman, N.; Widanage, W. D.; Zhu, X. H.; Moharana, S.;
Electrochemical Impedance Spectroscopy for Solid Oxide Fuel Cell Rashid, M.; Hubin, A.; Lataire, J. Operando Electrochemical
Cathode Materials. Electrochim. Acta 2006, 51, 1389−1402. Impedance Spectroscopy and its Application to Commercial Li-Ion
(47) Koster, D.; Du, G. Q.; Battistel, A.; Mantia, F. L. Dynamic Batteries. J. Power Sources 2022, 547, 232005.
Impedance Spectroscopy Using Dynamic Multi-Frequency Analysis: (64) Gaberšcě k, M. Impedance spectroscopy of battery cells: Theory
A Theoretical and Experimental Investigation. Electrochim. Acta 2017, versus experiment. Curr. Opin. Electrochem. 2022, 32, 100917.
246, 553−563. (65) Edge, J. S.; O’kane, S.; Prosser, R.; Kirkaldy, N. D.; Patel, A. N.;
(48) Erinmwingbovo, C.; Siller, V.; Nunez, M.; Trocolo, R.; Brogioli, Hales, A.; Ghosh, A.; Ai, W. L.; Chen, J. Y.; Yang, J.; et al. Lithium Ion
D.; Morata, A.; Mantia, D. L. Dynamic Impedance Spectroscopy of Battery Degradation: What You Need to Know. Phys. Chem. Chem.
LiMn2O4 Thin Films Made by Multi-Layer Pulsed Laser Deposition. Phys. 2021, 23, 8200−8221.
Electrochim. Acta 2020, 331, 135385. (66) O’Kane, S.; Ai, W. L.; Madabattula, G.; Alonso, D. A.; Timms,
(49) Sacci, R. L.; Seland, F.; Harrington, D. A. Dynamic R.; Sulzer, V.; Edge, J. S.; Wu, B.; Offer, G. J.; Marinescu, M. Lithium-
Electrochemical Impedance Spectroscopy for Electrocatalytic Reac- Ion Battery Degradation: How to Model it. Phys. Chem. Chem. Phys.
tions. Electrochim. Acta 2014, 131, 13−19. 2022, 24, 7909−7922.
(50) Harting, N.; Wolff, N.; Roder, F.; Krewer, U. Nonlinear (67) Jorcin, J. B.; Orazem, M. E.; Pebere, N.; Tribollet, B. CPE
Frequency Response Analysis (NFRA) of Lithium-Ion Batteries. Analysis by Local Electrochemical Impedance Spectroscopy. Electro-
Electrochim. Acta 2017, 248, 133−139. chim. Acta 2006, 51, 1473−1479.
(51) Harting, N.; Wolff, N.; Krewer, U. Identification of Lithium (68) Van Havenbergh, K.; Turner, S.; Driesen, K.; Bridel, J.-S.; Van
Plating in Lithium-Ion Batteries using Nonlinear Frequency Response Tendeloo, G. Solid−Electrolyte Interphase Evolution of Carbon-
Analysis (NFRA). Electrochim. Acta 2018, 281, 378−385. Coated Silicon Nanoparticles for Lithium-Ion Batteries Monitored by
(52) Wolff, N.; Harting, N.; Heinrich, M.; Krewer, U. Nonlinear Transmission Electron Microscopy and Impedance Spectroscopy.
frequency response analysis on lithium-ion batteries: Process Energy Technol. 2015, 3, 699−708.
identification and differences between transient and steady-state (69) Pastor-Fernández, C.; Uddin, K.; Chouchelamane, G. H.;
behavior. Electrochim. Acta 2018, 260, 614−622. Widanage, W. D.; Marco, J. A Comparison between Electrochemical
(53) Harting, N.; Wolff, N.; Roder, F.; Krewer, U. State-of-Health Impedance Spectroscopy and Incremental Capacity-Differential
Diagnosis of Lithium-Ion Batteries Using Nonlinear Frequency Voltage as Li-ion Diagnostic Techniques to Identify and Quantify
Response Analysis. J. Electrochem. Soc. 2019, 166 (2), 277−285. the Effects of Degradation Modes within Battery Management
(54) Harting, N.; Schenkendorf, R.; Wolff, N.; Krewer, U. State-of- Systems. J. Power. Sources 2017, 360, 301−318.
Health Identification of Lithium-Ion Batteries Based on Nonlinear (70) Xiong, R.; Pan, Y.; Shen, W.; Li, H.; Sun, F. Lithium-ion battery
Frequency Response Analysis: First Steps with Machine Learning. aging mechanisms and diagnosis method for automotive applications:
Appl. Sci. 2018, 8, 821. Recent advances and perspectives. Renew. Sustain. Energy Rev. 2020,
(55) Murbach, M. D.; Hu, V. W.; Schwartz, D. T. Nonlinear 131, 110048.
(71) Teliz, E.; Zinola, C. F.; Diaz, V. Identification and
Electrochemical Impedance Spectroscopy of Lithium-Ion Batteries:
Quantification of Aging Mechanisms in Li-ion Batteries by Electro-
Experimental Approach, Analysis, and Initial Findings. J. Electrochem.
chemical Impedance Spectroscopy. Electrochim. Acta 2022, 426,
Soc. 2018, 165 (11), 2758−2765.
140801.
(56) Murbach, M. D.; Schwartz, D. T. Extending Newman’s Pseudo-
(72) Xu, H. Y.; Han, C.; Li, W. T.; Li, H. Y.; Qiu, X. P.
Two-Dimensional Lithium-Ion Battery Impedance Simulation Ap-
Quantification of Lithium Dendrite and Solid Electrolyte Interphase
proach to Include the Nonlinear Harmonic Response. J. Electrochem.
(SEI) in Lithium-Ion Batteries. J. Power Sources 2022, 529, 231219.
Soc. 2017, 164 (11), 3311−3320. (73) Brown, D. E.; McShane, E. J.; Konz, Z. M.; Knudsen, K. B.;
(57) Liebhart, B.; Satzke, M.; Komsiyska, L.; Endisch, C. Application McCloskey, B. D. Detecting Onset of Lithium Plating During Fast
of Nonlinear Impedance Spectroscopy for the Diagnosis of Lithium- Charging of Li-ion Batteries Using Operando Electrochemical
Ion Battery Cells under Various Operating Conditions. J. Power Impedance Spectroscopy. Cell Rep. Phys. Sci. 2021, 2, 100589.
Sources 2020, 480, 228673. (74) Boukamp, B. A. Distribution (Function) of Relaxation Times,
(58) Van Gheem, E. V.; Pintelon, R.; Vereecken, J.; Schoukens, J.; Successor to Complex Nonlinear Least Squares Analysis of Electro-
Hubin, A.; Verboven, P.; Blajiev, O. Electrochemical impedance chemical Impedance Spectroscopy? J. Phys. Energy 2020, 2,
spectroscopy in the presence of non-linear distortions and non- No. 042001.
stationary behaviour. Electrochim. Acta 2004, 49 (26), 4753−4762. (75) Tuncer, E.; Gubanski, S. M. Using the Monte Carlo Method to
(59) Hallemans, N.; Pintelon, R.; Zhu, X.; Collet, T.; Claessens, R.; Obtain Relaxation Time Distribution and Comparing Non-linear
Wouters, B.; Hubin, A.; Lataire, J. Detection, Classification, and Spectral Function Fits. IEEE Trans. Dielectr. Electr. Insul. 2001, 8 (3),
Quantification of Nonlinear Distortions in Time-Varying Frequency 310−320.
Response Function Measurements. IEEE Trans. Instrum. Meas. 2021, (76) Boukamp, B. A. Fourier Transform Distribution Function of
70, 6500814. Relaxation Times; Application and Limitations. Electrochim. Acta
(60) Hallemans, N.; Pintelon, R.; Gheem, E. V.; Collet, T.; 2015, 154, 35−46.
Claessens, R.; Wouters, B.; Ramharter, K.; Hubin, A.; Lataire, J. Best (77) Tesler, A. B.; Lewin, D. R.; Baltianski, S.; Tsur, Y. Analyzing
Linear Time-Varying Approximation of a General Class of Nonlinear Results of Impedance Spectroscopy Using Novel Evolutionary
Time-Varying Systems. IEEE Trans. Instrum. Meas. 2021, 70, Programming Techniques. J. Electroceramics 2010, 24 (4), 245−260.
6503814. (78) Hershkovitz, S.; Tomer, S.; Baltianski, S.; Tsur, Y. ISGP:
(61) Hallemans, N.; Pintelon, R.; Zhu, X.; Collet, T.; Havigh, M. D.; Impedance spectroscopy analysis using evolutionary programming
Wouters, B.; Revilla, R. I.; Claessens, R.; Ramharter, K.; Hubin, A.; procedure. ECS Trans. 2011, 33 (40), 67−73.
Lataire, J. Trend Removal in Measurements of Best Linear Time- (79) Saccoccio, M.; Wan, T. H.; Chen, C.; Ciucci, F. Optimal
Varying Approximations − with Application to Operando Electro- Regularization in Distribution of Relaxation Times Applied to
chemical Impedance Spectroscopy. IEEE Trans. Instrum. Meas. 2022, Electrochemical Impedance Spectroscopy: Ridge and Lasso Regres-
71, 6501711. sion Methods - A Theoretical and Experimental Study. Electrochim.
(62) Zhu, X. H.; Halleman, N.; Wouters, B.; Claessens, R.; Lataire, Acta 2014, 147, 470−482.
J.; Hubin, A. Operando Odd Random Phase Electrochemical (80) Wan, T.; Saccoccio, M.; Chen, C.; Ciucci, F. Influence of the
Impedance Spectroscopy as a Promising Tool for Monitoring Discretization Methods on the Distribution of Relaxation Times
J. Phys. Chem. C 2023, 127, 4465−4495
Deconvolution: Implementing Radial Basis Functions with DRTtools. (98) Zhou, X.; Huang, J.; Pan, Z.; Ouyang, M. Impedance
Electrochim. Acta 2015, 184, 483−499. Characterization of Lithium-Ion Batteries Aging under High Temper-
(81) Quattrocchi, E.; Wan, T.; Belotti, A.; Kim, D.; Pepe, S.; Kalinin, ature Cycling: Importance of Electrolyte-Phase Diffusion. J. Power
S. V.; Ahmadi, M.; Ciucci, F. The deep-DRT: A deep neural network Sources 2019, 426, 216−222.
approach to deconvolve the distribution of relaxation times from (99) Wang, J.; Huang, Q.; Li, W.; Wang, J.; Bai, Y.; Bai, Y.; Zhao, Y.;
multidimensional electrochemical impedance spectroscopy data. Li, X.; Zhang, J. Insight into the Origin of Pseudo Peaks Decoded by
Electrochim. Acta 2021, 392, 139010. the Distribution of Relaxation Times/Differential Capacity Method
(82) Quattrocchi, E.; Py, B.; Maradesa, A.; Meyer, Q.; Zhao, C.; for Electrochemical Impedance Spectroscopy. J. Electroanal. Chem.
Ciucci, F. Deconvolution of Electrochemical Impedance Spectroscopy 2022, 910, 116176.
Data Using the Deep-Neural-Network-Enhanced Distribution of (100) Qu, D.; Ji, W.; Qu, H. Probing process kinetics in batteries
Relaxation Times. Electrochim. Acta 2023, 439, 141499. with electrochemical impedance spectroscopy. Commun. Mater. 2022,
(83) Liu, J.; Cuicci, F. The Gaussian process distribution of 3, 61.
relaxation times: A machine learning tool for the analysis and (101) Moškon, J.; Gaberšcě k, M. Transmission line models for
prediction of electrochemical impedance spectroscopy data. Electro- evaluation of impedance response of insertion battery electrodes and
chim. Acta 2020, 331, 135316. cells. J. Power Sources Adv. 2021, 7, 100047.
(84) Maradesa, A.; Py, B.; Quattrocchi, E.; Ciucci, F. The (102) ZView 3.5, Impedance/gain Phase Graphing and Analysis
probabilistic deconvolution of the distribution of relaxation times Software, Scribner Ass. Inc., http://www.scribner.com/software/68-
with finite Gaussian processes. Electrochim. Acta 2022, 413, 140119. general-electrochemistr376-zview-for-windows/ (accessed February 9,
(85) Wang, J.; Huang, Q.; Li, W.; Wang, J.; Zhuang, Q.; Zhang, J. 2023).
Fundamentals of distribution of relaxation times for electrochemical (103) Python-assisted physics-based electrochemical impedance
impedance spectroscopy. J. Electrochem. 2020, 26 (5), 607−627. spectroscopy. https://github.com/thuylinhpham/PyPhyEIS (ac-
(86) Paul, T.; Chi, P.; Wu, P.; Wu, M. Computation of Distribution cessed February 9, 2023).
of Relaxation Times by Tikhonov Regularization for Li Ion Batteries: (104) CNLS (Complex Nonlinear Least Squares), Immittance,
Usage of L-Curve Method. Sci. Rep. 2021, 11, 12624. Inversion, and Simulation Fitting Programs, J. R. MacDonald, https://
(87) Zhu, D.; Ma, T. C.; Yang, Y. B. Optimization and Application jrossmacdonald.com/levmlevmw/ (accessed February 9, 2023).
of the Distribution of Relaxation Times Based on Characteristic (105) Nara, H.; Mukoyama, D.; Yokoshima, T.; Momma, T.; Osaka,
Frequency Resolution and Hyperparameters. J. Power Sources 2022, T. Impedance Analysis with Transmission Line Model for Reaction
545, 231955. Distribution in a Pouch Type Lithium-Ion Battery by Using Micro
(88) Wang, C. L.; Zhu, G. Y.; Zhang, P. B.; Fang, X. H. Optimization Reference Electrode. J. Electrochem. Soc. 2016, 163, 434−441.
Procedures for the Inversion of Impedance Spectra to the Distribution (106) Bernard, P.; Martinez, H.; Tessier, C.; Garitte, E.; Franger, S.;
of Relaxation Times. J. Electroanal. Chem. 2022, 911, 116199. Dedryvere, R. Role of Negative Electrode Porosity in Long-Term
(89) Zic, M.; Pereverzyev, S.; Subotic, V.; Pereverzyev, S. Adaptive Aging of NMC//Graphite Li-Ion Batteries. J. Electrochem. Soc. 2015,
Multiparameter Regularization Approach to Construct the Distribu- 162, 7096−7103.
(107) Ogihara, N.; Itou, Y.; Sasaki, T.; Takeuchi, Y. Impedance
tion Function of Relaxation Times. GEM-Int. J. Geomath. 2020, 11
Spectroscopy Characterization of Porous Electrodes under Different
(1), 2.
Electrode Thickness Using a Symmetric Cell for High-Performance
(90) Iurilli, P.; Brivio, C.; Wood, V. Detection of Lithium-Ion Cells’
Lithium-Ion Batteries. J. Phys. Chem. C 2015, 119, 4612−4619.
Degradation through Deconvolution of Electrochemical Impedance
(108) Dubarry, M.; Qin, N.; Brooker, P. Calendar aging of
Spectroscopy with Distribution of Relaxation Time. Energy Technol.
commercial Li-ion cells of different chemistries − A review. Curr.
2022, 10, 2200547.
Opin. Electrochem. 2018, 9, 106−113.
(91) Momma, T.; Yokoshima, T.; Nara, H.; Gima, Y.; Osaka, T. (109) Du, Y.; Fujita, K.; Shironita, S.; Sone, Y.; Hosono, E.; Asakura,
Distinction of impedance responses of Li-ion batteries for individual D.; Umeda, M. Capacity fade characteristics of nickel-based lithium-
electrodes using symmetric cells. Electrochim. Acta 2014, 131, 195− ion secondary battery after calendar deterioration at 80 °C. J. Power
201. Sources 2021, 501, 230005.
(92) Gordon, I. J.; Genies, S.; Larbi, G. S.; Boulineau, A.; Daniel, L.; (110) Shafiei Sabet, P.; Warnecke, A. J.; Meier, F.; Witzenhausen,
Alias, M. Original implementation of Electrochemical Impedance H.; Martinez-Laserna, E.; Sauer, D. U. Non-invasive Yet Separate
Spectroscopy (EIS) in symmetric cells: Evaluation of post-mortem Investigation of Anode/Cathode Degradation of Lithium-Ion
protocols applied to characterize electrode materials for Li-ion Batteries (Nickel−Cobalt−Manganese vs. Graphite) due to Accel-
batteries. J. Power Sources 2016, 307, 788−795. erated Aging. J. Power Sources 2020, 449, 227369.
(93) Schmidt, J. P.; Chrobak, T.; Ender, M.; Illig, J.; Klotz, D.; Ivers- (111) Schmitt, J.; Maheshwari, A.; Heck, M.; Lux, S.; Vetter, M.
Tiffee, E. Studies on LiFePO4 as cathode material using impedance Impedance Change and Capacity Fade of Lithium Nickel Manganese
spectroscopy. J. Power Sources 2011, 196, 5342−5348. Cobalt Oxide-Based Batteries During Calendar Aging. J. Power Sources
(94) Xiao, Y.; Xu, R.; Yan, C.; Huang, J. Q.; Zhang, Q.; Ouyang, M. 2017, 353, 183−194.
G. A Toolbox of Reference Electrodes for Lithium Batteries. Adv. (112) Li, J.; Zhang, J.; Zhang, X.; Yang, C.; Xu, N.; Xia, B. Study of
Funct. Mater. 2022, 32, 2108449. the storage performance of a Li-ion cell at elevated temperature.
(95) Hoshi, Y.; Narita, Y.; Honda, K.; Ohtaki, T.; Shitanda, I.; Electrochim. Acta 2010, 55, 927−934.
Itagaki, M. Optimization of reference electrode position in a three- (113) Geisbauer, C.; Wohrl, K.; Koch, D.; Wihelm, G.; Schneider,
electrode cell for impedance measurements in lithium-ion recharge- G.; Schweiger, H. G. Comparative Study on the Calendar Aging
able battery by finite element method. J. Power Sources 2015, 288, Behavior of Six Different Lithium-Ion Cell Chemistries in Terms of
168−175. Parameter Variation. Energies 2021, 14, 3358.
(96) Soni, R.; Robinson, J. B.; Shearing, P. R.; Brett, D. J.; Rettie, A. (114) Keil, P.; Schuster, S. F.; Wihelm, J.; Travi, J.; Hauser, A.; Karl,
J.; Miller, T. S. Lithium-Sulfur Battery Diagnostics through R. C.; Jossen, A. Calendar Aging of Lithium-Ion Batteries. I. Impact of
Distribution of Relaxation Times Analysis. Energy Storage Mater. the Graphite Anode on Capacity Fade. J. Electrochem. Soc. 2016, 163
2022, 51, 97−107. (9), 1872−1880.
(97) Zhu, J.; Knapp, M.; Liu, X.; Yan, P.; Dai, H.; Wei, X.; (115) Kassem, M.; Bernard, J.; Revel, R.; Pelissier, S.; Duclaud, F.;
Ehrenberg, H. Low-Temperature Separating Lithium-Ion Battery Delacourt, C. Calendar Aging of a graphite/LiFePO4 Cell. J. Power
Interfacial Polarization Based on Distribution of Relaxation Times Sources 2012, 208, 296−305.
(DRT) of Impedance. IEEE Trans. Transport. Electrific. 2021, 7 (2), (116) Zheng, Y.; He, Y.; Qian, K.; Li, B.; Wang, X.; Li, J.; Miao, C.;
410−421. Kang, F. Effects of state of charge on the degradation of LiFePO4/
J. Phys. Chem. C 2023, 127, 4465−4495
graphitebatteries during accelerated storage test. J. Alloys Compd. thermal gradients in lithium-ion batteries dictates diverging
2015, 639, 406−414. degradation modes. Cell Rep. Phys. Sci. 2021, 2, 100351.
(117) Mendoza-Hernandez, O. S.; Hosono, E.; Asakura, D.; (135) Liu, J.; Duan, Q.; Ma, M.; Zhao, C.; Sun, J.; Wang, Q. Aging
Matsuda, H.; Sayoko, S.; Umeda, M.; Sone, Y. Impact of calendar Mechanisms and Thermal Stability of Aged Commercial 18650
degradation on the Performance of LiFePO4-Graphite Li-Ion Cells Lithium Ion Battery Induced by Slight Overcharging Cycling. J. Power
during Charge-Discharge Cycling at − 5°C. J. Electrochem. Soc. 2019, Sources 2020, 445, 227263.
166 (15), 3525−3530. (136) Togasaki, N.; Yokoshima, T.; Oguma, Y.; Osaka, T. Prediction
(118) Zhang, L.; Liu, J.; Du, L.; Fan, P.; Xu, X.; Ma, Y.; Zuo, P.; Qu, of Overcharge-Induced Serious Capacity Fading in Nickel Cobalt
B.; Yin, G.; Fu, Q.; et al. LiNi0.5Co0.2Mn0.3O2/Graphite batteries Aluminum Oxide Lithium-Ion Batteries Using Electrochemical
storing at high temperature: Capacity fading and raveling of aging Impedance Spectroscopy. J. Power Sources 2020, 461, 228168.
mechanisms. J. Power Sources 2021, 496, 229858. (137) Liu, Y. D.; Xie, J. Failure Study of Commercial LiFePO4 Cells
(119) Zhang, L.; Liu, J.; Fan, P.; Du, L.; Ma, Y.; Qu, B.; Yin, G.; Fu, in Overcharge Conditions Using Electrochemical Impedance Spec-
Q.; Yang, F.; et al. Unraveling the effect of short-term high- troscopy. J. Electrochem. Soc. 2015, 162 (10), 2208−2217.
temperature storage on the performance and thermal stability of (138) Liu, Y.; Liu, Q.; Li, Z.; Ren, Y.; Xie, J.; He, H.; Xu, F. Failure
LiNi0.5Co0.2Mn0.3O2/Graphite Battery. J. Power Sources 2020, 459, Study of Commercial LiFePO4 Cells in over-Discharge Conditions
227842. Using Electrochemical Impedance Spectroscopy. J. Electrochem. Soc.
(120) Dufek, E. J.; Tanim, T. R.; Chen, B. R.; Kim, S. Battery 2014, 161 (4), 620−632.
Calendar Aging and Machine Learning. Joule 2022, 6, 1363−1367. (139) Deich, T.; Storch, M.; Steiner, K.; Bund, A. Effects of module
(121) Jaini, R. R.; Setzler, B. P.; Fuller, T. F. Investigating Changes stiffness and initial compression on lithium-ion cell aging. J. Power
in the Morphological Structure of High-Temperature, Calendar-Aged Sources 2021, 506, 230163.
Li-Ion Cells. J. Electrochem. Soc. 2018, 165 (13), A3125. (140) Li, R.; Li, W.; Singh, A.; Ren, D.; Hou, Z.; Ouyang, M. Effect
(122) Yoshida, Y.; Takahashi, M.; Morikawa, S.; Ihara, C.; of external pressure and internal stress on battery performance and
Katsukawa, H.; Shiratsuchi, T.; Yamaki, J. Degradation Mechanism lifespan. Energy Storage Mater. 2022, 52, 395−429.
and Life Prediction of Lithium-Ion Batteries. J. Electrochem. Soc. 2006, (141) Mussa, A. S.; Klett, M.; Lindberg, G.; Lindstrom, R. W. Effects
153 (6), 576−582. of external pressure on the performance and ageing of single-layer
(123) Jalkanen, K.; Karppinen, J.; Skogstrom, L.; Laurila, T.; Nisula, lithium-ion pouch cells. J. Power Sources 2018, 385, 18−26.
M.; Vuorilehto, K. Cycle aging of commercial NMC/graphite pouch (142) Tian, Y.; Lin, C.; Li, H.; Du, J.; Xiong, R. Detecting undesired
cells at different temperatures. Appl. Energy 2015, 154, 160−172. lithium plating on anodes for lithium-ion batteries − A review on the
(124) Strehle, B.; Friedrich, F.; Gasteiger, H. A. A Comparative in-situ methods. Applied Energy 2021, 300, 117386.
Study of Structural Changes during Long-Term Cycling of NCM-811 (143) Petzl, M.; Kasper, M.; Danzer, M. A. Lithium plating in a
commercial lithium-ion battery�A low-temperature aging study. J.
at Ambient and Elevated Temperatures. J. Electrochem. Soc. 2021, 168,
Power Sources 2015, 275, 799−807.
No. 050512.
(144) Schuster, S. F.; Bach, T.; Elena, F.; Muller, J.; Brand, M.; Sextl,
(125) Zhu, X.; Revilla, R. I.; Jaguemont, J.; Mierlo, J. V.; Hubin, A.
G.; Jossen, A. Nonlinear Aging Characteristics of Lithium-Ion Cells
Insights into Cycling Aging of LiNi0.80Co0.15Al0.05O2 Cathode Induced
under Different Operational Conditions. J. Energy Storage 2015, 1,
by Surface Inhomogeneity: A Post-mortem Analysis. J. Phys. Chem. C
44−53.
2019, 123, 30046−30058.
(145) Chen, X.; Li, L.; Liu, M.; Huang, T.; Yu, A. Detection of
(126) Raj, A.; Rodrigues, M. F.; Abraham, D. P. Rate-Dependent
Lithium Plating in Lithium-ion Batteries by Distribution Relaxation
Aging Resulting from Fast Charging of Li-Ion Cells. J. Electrochem. Times. J. Power Sources 2021, 496, 229867.
Soc. 2020, 167, 120517. (146) Gargh, P.; Sarkar, A.; Lui, Y.; Shen, S.; Hu, C.; Hu, S.;
(127) Mussa, A. S.; Liivat, A.; Marzano, F.; Klett, M.; Philippe, B.; Nlebedim, I. C.; Shrotriya, P. Correlating Capacity Fade with Film
Tengstedt, C.; Lindbergh, G.; Edstrom, K.; Lindstrom, R. W.; Svens, Resistance Loss in Fast Charging of Lithium-Ion Battery. J. Power
P. Fast-charging effects on ageing for energy-optimized automotive Sources 2021, 485, 229360.
LiNi1/3Mn1/3Co1/3O2/graphite prismatic lithium-ion cells. J. Power (147) Schindler, S.; Bauer, M.; Petzl, M.; Danzer, M. A. Voltage
Sources 2019, 422, 175−184. relaxation and impedance spectroscopy as in-operando methods for
(128) Zheng, H.; Li, X.; Huang, Z.; Yang, X.; Deng, J.; Zhang, G.; the detection of lithium plating on graphitic anodes in commercial
Tang, X. Investigation of the Performance and Recession Mechanisms lithium-ion cells. J. Power Sources 2016, 304, 170−180.
of High-Nickel Ternary Lithium-Ion Batteries Under Artificial Aging (148) Xu, L.; Yang, Y.; Xiao, Y.; Cai, W.; Yao, Y.; Chen, X.; Yan, C.;
Discharge Rates. Energy Technol. 2022, 10, 2200600. Yuan, H.; Huang, J. In-situ Determination of Onset Lithium Plating
(129) Zhu, J.; Su, P.; Darma, M. S.; Hua, W.; Mereacre, L.; Liu- for Safe Li-ion Batteries. J. Energy Chem. 2022, 67, 255−262.
Theato, X.; Heere, M.; Sorensen, D. R.; Dai, H.; Wei, X.; et al. (149) Koleti, U. R.; Dinh, T. Q.; Macro, J. A New On-Line Method
Multiscale investigation of discharge rate dependence of capacity fade for Lithium Plating Detection in Lithium-Ion Batteries. J. Power
for lithium-ion battery. J. Power Sources 2022, 536, 231516. Sources 2020, 451, 227798.
(130) Qu, J.; Jiang, Z.; Zhang, J. Investigation on lithium-ion battery (150) Koseoglou, M.; Tsioumas, E.; Ferentinou, D.; Jabbour, N.;
degradation induced by combined effect of current rate and operating Papagiannis, D.; Mademlis, C. Lithium Plating Detection Using
temperature during fast charging. J. Energy Storage 2022, 52, 104811. Dynamic Electrochemical Impedance Spectroscopy in Lithium-Ion
(131) Leonardi, S. G.; Aloisio, D.; Brunaccini, G.; Stassi, A.; Ferraro, Batteries. J. Power Sources 2021, 512, 230508.
M.; Antonucci, V.; Sergi, F. Investigation on the ageing mechanism (151) Xu, L.; Xiao, Y.; Yang, Y.; Yang, S.; Chen, X.; Xu, R.; Yao, Y.;
for a lithium-ion cell under accelerated tests: The case of primary Cai, W.; Yan, C.; Huang, J. Q.; et al. Operando Quantified Lithium
frequency regulation service. J. Energy Storage 2021, 41, 102904. Plating Determination Enabled by Dynamic Capacitance Measure-
(132) Wu, Y.; Long, X.; Lu, J.; Wu, Y.; Zhou, R.; Liu, L. Effect of ment in Working Li-Ion Batteries. Angew. Chem., Int. Ed. 2022, 61,
temperature on the high-rate pulse charging of lithium-ion batteries. J. e202210365.
Electroanal. Chem. 2022, 922, 116773. (152) Peled, E. The Electrochemical Behavior of Alkali and Alkaline
(133) Perumaram Rangarajan, S.; Barsukov, Y.; Mukherjee, P. P. Earth Metals in Nonaqueous Battery Systems�The Solid Electrolyte
Plating Energy as a Universal Descriptor to Classify Accelerated Cell Interphase Model. J. Electrochem. Soc. 1979, 126 (12), 2047−2051.
Failure under Operational Extremes. Cell Rep. Phys. Sci. 2022, 3, (153) Aurbach, D.; Levi, M. D.; Gamulski, K.; Markovsky, B.; Salitra,
100720. G.; Levi, E.; Heider, U.; Heider, L.; Oesten, R. Capacity fading of
(134) Carter, R.; Kingston, T. A.; Atkinson, R. W.; Parmananda, M.; LixMn2O4 spinel electrodes studied by XRD and electroanalytical
Dubarry, M.; Fear, C.; Mukherjee, P. P.; Love, C. T. Directionality of techniques. J. Power Sources 1999, 81−82, 472−479.
J. Phys. Chem. C 2023, 127, 4465−4495
(154) Wu, J. X.; Ihsan-Ul-Haq, M.; Chen, Y. M.; Kim, J. K. (172) Xu, J.; Mi, C.; Cao, B.; Cao, J. A new method to estimate the
Understanding solid electrolyte interphases: Advanced character- state of charge of lithium-ion batteries based on the battery
ization techniques and theoretical simulations. Nano Energy 2021, 89, impedance model. J. Power Sources 2013, 233, 277−284.
106489. (173) Luo, K.; Chen, X.; Zheng, H.; Shi, Z. A Review of Deep
(155) Sun, C.; Ji, X.; Weng, S.; Li, R.; Huang, X.; Zhu, C.; Xiao, X.; Learning Approach to Predicting the State of Health and State of
Deng, T.; Fan, L.; Chen, L.; et al. 50C Fast-Charge Li-Ion Batteries Charge of Lithium-Ion Batteries. J. Energy Chem. 2022, 74, 159−173.
using a Graphite Anode. Adv. Mater. 2022, 34, 2206020. (174) Shu, X.; Shen, S.; Shen, J.; Zhang, Y.; Li, G.; Chen, Z.; Liu, Y.
(156) Solchenbach, S.; Huang, X. Y.; Pritzl, D.; Landesfeind, J.; State of Health Prediction of Lithium-Ion Batteries Based on Machine
Gasteiger, H. A. Monitoring SEI Formation on Graphite Electrodes in Learning: Advances and Perspectives. iScience 2021, 24, 103265.
Lithium-Ion Cells by Impedance Spectroscopy. J. Electrochem. Soc. (175) Sui, X.; He, S.; Vilsen, S. B.; Meng, J. H.; Teodorescu, R.;
2021, 168, 110503. Stroe, D. I. A Review of Non-Probabilistic Machine Learning-Based
(157) Ehteshami, N.; Ibing, L.; Stolz, L.; Winter, M.; Paillard, E. State of Health Estimation Techniques for Lithium-Ion Battery.
Applied Energy 2021, 300, 117346.
Ethylene Carbonate-Free Electrolytes for Li-Ion Battery: Study of the
(176) Liu, J.; Duan, Q.; Qi, K.; Sun, J.; Wang, Z.; Wang, Q. Capacity
Solid Electrolyte Interphases Formed on Graphite Anodes. J. Power
Fading Mechanisms and State of Health Prediction of Commercial
Sources 2020, 451, 227804. Lithium-Ion Battery in Total Lifespan. J. Energy Storage 2022, 46,
(158) Lu, P.; Li, C.; Schneider, E. W.; Harris, S. J. Chemistry, 103910.
Impedance, and Morphology Evolution in Solid Electrolyte (177) Pizarro-Carmona, V.; Cortes-Carmona, M.; Palma-Behnke, R.;
Interphase Films during Formation in Lithium Ion Batteries. J. Phys. Calderon-Munoz, W.; Orchard, M. E.; Estevez, P. An Optimized
Chem. C 2014, 118, 896−903. Impedance Model for the Estimation of the State-of-Charge of a Li-
(159) Wang, P.; Yan, D.; Wang, C. Y.; Ding, H.; Dong, H.; Wang, J.; Ion Cell: The Case of a LiFePO4. Energies 2019, 12, 681.
Wu, S.; Cui, X.; Li, C.; Zhao, D.; et al. Study of the formation and (178) Locorotondo, E.; Cultrera, V.; Pugi, L.; Berzi, L.; Pierini, M.;
evolution of solid electrolyte interface via in-situ electrochemical Lutzemberger, G. Development of a Battery Real-Time State of
impedance spectroscopy. Appl. Surf. Sci. 2022, 596, 153572. Health Diagnosis based on Fast Impedance Measurements. J. Energy
(160) Steinhauer, M.; Risse, S.; Wagner, N.; Friedrich, K. A. Storage 2021, 38, 102566.
Investigation of the Solid Electrolyte Interphase Formation at (179) Zhu, J.; Zhang, Q.; Mereacre, L. D.; Wang, X.; Jiang, B.; Dai,
Graphite Anodes in Lithium-Ion Batteries with Electrochemical H.; Wei, X.; Knapp, M.; Ehrenberg, H. Alternating Current
Impedance Spectroscopy. Electrochim. Acta 2017, 228, 652−658. Impedance Probing Capacity of Lithium-Ion Battery by Gaussian
(161) Qu, H.; Zhang, X.; Ji, W.; Zheng, D.; Qu, D. Impedance Process Regression. Energy Technol. 2022, 10, 2200437.
Investigation of the High Temperature Performance of the Solid- (180) Jiang, B.; Zhu, J.; Wang, X.; Wei, X.; Shang, W.; Dai, H. A
Electrolyte-Interface of a Wide Temperature Electrolyte. J. Colloid comparative study of different features extracted from electrochemical
Interface Sci. 2022, 608, 3079−3086. impedance spectroscopy in state of health estimation for lithium-ion
(162) Xu, H.; Li, Z.; Liu, T.; Han, C.; Guo, C.; Zhao, H.; Li, Q.; Lu, batteries. Appl. Energy. 2022, 322, 119502.
J.; Amine, K.; Qiu, X. Impacts of Dissolved Ni2+ on the Solid (181) Li, W.; Chen, J.; Quade, K.; Luder, D.; Gong, J.; Sauer, D. U.
Electrolyte Interphase on a Graphite Anode. Angew. Chem., Int. Ed. Battery degradation diagnosis with field data, impedance-based
2022, 61, e202202894. modeling and artificial intelligence. Energy Storage Mater. 2022, 53,
(163) Broussely, M.; Herreyre, S.; Biensan, P.; Kasztejna, P.; 391−403.
Nechev, K.; Staniewicz, R. J. Aging Mechanism in Li Ion Cells and (182) Babaeiyazdi, I.; Rezaei-Zare, A.; Shokrzadeh, S. Transfer
Learning with Deep Neural Network for Capacity Prediction of Li-ion
Calendar Life Predictions. J. Power Sources 2001, 97−98, 13−21.
Batteries Using EIS Measurement. IEEE Trans. Transport. Electrific.
(164) Ploehn, H. J.; Ramadass, P. L.; White, R. E. Solvent Diffusion
doi: 20221 (accessed February 9, 2023).
Model for Aging of Lithium-Ion Battery Cells. J. Electrochem. Soc.
(183) Penjuru, N. M. H.; Reddy, G. V.; Nair, M. R.; Sahoo, S.;
2004, 151 (3), 456−462. Mayank; Jiang, J.; Ahmed, J.; Wang, H.; Roy, T. Machine Learning
(165) Safari, M.; Morcrette, M.; Teyssot, A.; Delacourt, C. Aided Predictions for Capacity Fade of Li-Ion Batteries. J. Electrochem.
Multimodal Physics-Based Aging Model for Life Prediction of Li- Soc. 2022, 169, No. 050535.
Ion Batteries. J. Electrochem. Soc. 2009, 156 (3), 145−153.
(166) Chen, C.; Mukherjee, P. P. Probing the Morphological
Influence on Solid Electrolyte Interphase and Impedance Response in
Intercalation Electrodes. Phys. Chem. Chem. Phys. 2015, 17, 9812.
(167) Single, F.; Horstmann, B.; Latz, A. Theory of Impedance
Spectroscopy for Lithium Batteries. J. Phys. Chem. C 2019, 123,
27327−27343.
(168) Drvaric̆ Talian, S.; Bobnar, J.; Sinigoj, A. R.; Humar, I.;
Gaberscek, M. Transmission Line Model for Description of the
Impedance Response of Li Electrodes with Dendritic Growth. J. Phys.
Chem. C 2019, 123, 27997−28007.
(169) Santos, C. S.; Botz, A.; Bandarenka, A. S.; Ventosa, E.;
Schuhmann, W. Correlative Electrochemical Microscopy for the
Elucidation of the Local Ionic and Electronic Properties of the Solid
Electrolyte Interphase in Li-Ion Batteries. Angew. Chem., Int. Ed. 2022,
61, e202202744.
(170) Teo, L.; Subramanian, V.; Schwartz, D. Dynamic Electro-
chemical Impedance Spectroscopy of Lithium-ion Batteries: Revealing
Underlying Physics through Efficient Joint Time Frequency
Modeling. J. Electrochem. Soc. 2021, 168, No. 010526.
(171) Jiang, B.; Gent, W.; Mohr, F.; Das, S.; Berliner, M. D.;
Forsuelo, M.; Zhao, H.; Attia, P. M.; Grover, A.; Herring, P. K.; et al.
Bayesian Learning for Rapid Prediction of Lithium-Ion Battery-
Cycling Protocols. Joule 2021, 5, 3187−3203.
J. Phys. Chem. C 2023, 127, 4465−4495

Application of Electrochemical Impedance Spectroscopy To Degradation and Aging Research of Lithium-Ion Batteries

Uploaded by

Copyright:

Available Formats

Application of Electrochemical Impedance Spectroscopy To Degradation and Aging Research of Lithium-Ion Batteries

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Application of Electrochemical Impedance Spectroscopy To Degradation and Aging Research of Lithium-Ion Batteries

Uploaded by

Copyright:

Available Formats

pubs.acs.

Application of Electrochemical Impedance Spectroscopy to

Cite This: J. Phys. Chem. C 2023, 127, 4465−4495 Read Online

ACCESS Metrics & More Article Recommendations

ABSTRACT: An in-depth understanding of battery degradation and aging in-Operando not

1. INTRODUCTION understanding of the degradation mechanisms of LIBs, which

© 2023 American Chemical Society https://doi.org/10.1021/acs.jpcc.3c00033

This review aims at providing a guideline for choosing and (1)

battery aging. Second, there is no one-to-one correspondence

for the three cycling conditions, respectively. Both the Rohm

SEI growth dominates the low SOC condition, while degradation

cation disorder at high temperature storage degrades the cycling

LLI induced by SEI growth is the main cause, while LAM of

SEI growth, electrolyte consumption, Fe dissolution, and

TM dissolution, CEI growth, structural decomposition

the cathode and anode is the minor reason

of cathode dominates the high SOC case

Strehle et al. studied the degradation of the NMC cathode

uneven degradation of NCA cathode, as Rct of the aged

ence.125 High temperature accelerates battery aging by

Enhancing the charge and discharge rates is also a common

method to accelerate aging, although their accelerated effects

and mechanisms are quite different. Raj et al. investigated how

wide SOC range (4C/60% and 8C/30%) leads to serious Li

plating, rapid increase of impedance, and capacity decay, while

rise in further cycling tests compared to the aged cells at 60 °C

most cases show quasi-linear capacity decay and impedance

to assess the accelerated aging conditions of whether Li plating

has happened or not. Similarly, Mussa et al. found Li plating,

the direct-current resistance of the full cell is much larger than

rates from 0.5 to 3C to accelerate the aging of the NMC

batteries.128 As shown in Figure 7c, the capacity loss

gradual increase of different impedances at the low discharge

rates, while Rct dramatically increases at a 3C discharge rate.

The results indicate that the main causes for capacity

consumption to structural damage of active materials with

discharge rate elevated. Zhu et al. also observed a faster

rates.129 Obviously, enhancing the discharge rates would not

discharge rates is a promising way to be applied in accelerated

aging protocols for fast lifespan measurement. Since the high

temperature could accelerate the kinetics of processes, it

Li plating and the quantification of plated Li by impedance

drawn great attention and wide applications by battery

The plated Li could decrease the Rohm of graphite anode so as to

The chemical intercalation during the voltage relaxation decreases

With a certain length of relaxation time, the estimated resistance

cause the drop of Ri obtained from the transient state analysis

could reflect the impedance of high and midfrequency range.

degradation mechanism in battery aging, when the thickness

the amount of plated Li on graphite surface

advantages of nondestructivity, in situ measurement, and easy

The emergence of an upward trend of RSEI is considered as the

The sudden decrease of the estimated impedance indicates the

The emergence of a downward trend of Ri is considered as the

The emergence of a downward trend of Rct is considered as

voltage intervals of 0.1 V during the first two cycles.160 As

shown in Figure 10a, the interfacial impedance showed a

considered as the signal of Li plating