Angewandte

Eine Zeitschrift der Gesellschaft Deutscher Chemiker Chemie www.angewandte.de

Akzeptierter Artikel

Titel: Understanding and Calibration of Charge Storage Mechanism in

Cyclic Voltammetry Curves

Autoren: Xiangjun Pu, Dong Zhao, Chenglong Fu, Zhongxue Chen,

Shunan Cao, Chunsheng Wang, and Yuliang Cao

Dieser Beitrag wurde nach Begutachtung und Überarbeitung sofort als

"akzeptierter Artikel" (Accepted Article; AA) publiziert und kann unter
Angabe der unten stehenden Digitalobjekt-Identifizierungsnummer
(DOI) zitiert werden. Die deutsche Übersetzung wird gemeinsam mit der
endgültigen englischen Fassung erscheinen. Die endgültige englische
Fassung (Version of Record) wird ehestmöglich nach dem Redigieren
und einem Korrekturgang als Early-View-Beitrag erscheinen und kann
sich naturgemäß von der AA-Fassung unterscheiden. Leser sollten
daher die endgültige Fassung, sobald sie veröffentlicht ist, verwenden.
Für die AA-Fassung trägt der Autor die alleinige Verantwortung.

Zitierweise: Angew. Chem. Int. Ed. 10.1002/anie.202104167

Link zur VoR: https://doi.org/10.1002/anie.202104167

 Angewandte Chemie 10.1002/ange.202104167

((Catch Phrase)) DOI: 10.1002/anie.201((will be filled in by the editorial staff))

Understanding and Calibration of Charge Storage Mechanism in Cyclic

Voltammetry Curves
Xiangjun Pu,† ‡ Dong Zhao,† Chenglong Fu,† Zhongxue Chen,*† Shunan Cao,*† Chunsheng
Wang,§ and Yuliang Cao*‡

Abstract: Noticeable pseudo-capacitance behaviour out of charge in grid-scale energy storage systems (MW level) due to their
storage mechanism (CSM) has attracted intensive studies because it advantages either in cost or safety.2-9
can provide both high energy density and large output power. High power density and high energy density as two guiding
Although cyclic voltammetry is recognized as the feasible features of electrochemical devices are often ascribed to capacitors
electrochemical technique to determine it quantitatively in the and batteries, respectively. Capacitive charge storage through non-

Accepted Manuscript
previous works, the results are inferior due to uncertainty in the faradaic process offers many desirable properties compared with
definitions and application conditions. Herein, three successive conventional batteries, including charging within seconds, long-term
treatments, including de-polarization, de-residual and de-background, cycling stability and the capability to deliver up to ten times more
as well as a non-linear fitting algorithm are employed for the first time power than batteries, but its energy density is substantially lower, by
to calibrate the different CSM contribution of three typical cathode at least one order of magnitude, than batteries associated with faradaic
materials, LiFePO4, LiMn2O4 and Na4Fe3(PO4)2P2O7, and achieve reactions.10-12 Electrode materials at nano-scale can bridge the
well-separated physical capacitance, pseudo-capacitance and power/energy gap between traditional double layer electric capacitors
diffusive contributions to the total capacity. This work can eliminate and typical rock-chair type batteries, due to a large fraction of Li+
misunderstanding concepts and correct ambiguous results of the intercalation-deintercalation reaction taking place at the near surface.
pseudo-capacitance contribution, and recognize the essence of CSM 11, 13 In this process, like capacitor, the reaction is kinetically facile

in electrode materials. and diffusion is not the rate-determining step (RDS), thus we call this
portion of capacity as pseudo-capacitance.13-16 The determination of
Introduction: The successful commercialization of rechargeable pseudo-capacitance was firstly proposed by Dunn et al. in 2010 based
batteries has been making a profound impact on modern industrial on cyclic voltammetry (CV) measurement on ordered mesoporous α-
society and our daily life from many aspects. As the representative, MoO3 with iso-oriented nanocrystalline walls.10 They proposed that
lithium-ion battery (LIB) is the most state-of-art technology widely for a redox reaction limited by semi-infinite diffusion, the peak
used in portable electronics (W level) and electrical vehicles (KW current i varies with v1/2 (v is the scan rate), while for a capacitive
level).1 Besides LIBs, various emerging prototypes based on process limited by charge-transfer process, it varies with v. In other
intercalation chemistry, for instance sodium-ion battery (SIB), are words, the current i at a fixed potential (V) originates from the
also under extensive researches and seeking for potential applications combination of two separate mechanisms: capacitive effects (k1v) and
diffusion-controlled insertion (k2v1/2) according to:
V ) k1ν + k2ν 0.5
I (= ( Eq. 1-1)
† ν 0.5 k1ν 0.5 + k2
I (V ) /= ( Eq. 1-2 )
[ ] X. J. Pu, D. Zhao, C. L. Fu, Z. X. Chen, S, A, Cao
Key Laboratory of Hydraulic Machinery Transients, Ministry According to these equations, the accumulated pseudo-capacitance
of Education
can be calculated by fitting the two constants k1 and k2 at every
School of Power and Mechanical Engineering
Wuhan University, Wuhan, China potential. Since then, this method is widely adopted by hundreds of
E-mail: [email protected], [email protected]. researchers on various kinds of electrode materials for LIBs17-19,
SIBs20-21 and many other prototypes batteries. 22-25 However, when
‡
[ ] X. J. Pu, Y. L. Cao we directly use Eq. 1-2 to calibrate the pseudo-capacitance, here
Hubei Key Laboratory of Electrochemical Power Sources Li2MnO4 is taken as an example (Figure S1), some confusing issues
College of Chemistry and Molecular Sciences arise as listed below.
Wuhan University, Wuhan, China.
1) The ohmic resistance is ignored, leading to the deviation of
E-mail: [email protected].
redox peak potentials between experimental data and fitting results.
§ This phenomenon often appears in the previous reports, 10, 20-21, 23
[ ] C. S. Wang
because Eq. 1-2 is only applicable on the conditions with the absence
Department of Chemical and Biomolecular Engineering
University of Maryland, College Park, Maryland, USA
of ohmic polarization.
2) The residual current is not considered when the scan direction is
reversed, leading to the fitted current even beyond the experimental
We wish to thank the National Natural Science Foundation of value during the CV scope. The reason behind is that there is a
China (21875171, 21972108, U20A20249), the National Key significant oxidative current when reduction process begins and vice
Research Program of China (2016YFB0901500), the versa. Although some researchers have noticed this error,25 no
Intergovernmental International Science and Technology efficient method was proposed to address this problem.
Innovation Cooperation Project (2019YFE010186), the Hubei
3) The double-layer capacitance is not specifically calculated but
Natural Science Foundation (2020CFB771)
is included into pseudo-capacitance current, rendering the over
Supporting information for this article is available on the exaggeration of pseudo-capacitance. In details, the current in CV test
WWW under http://www.angewandte.org or from the at a certain potential is composed of three rather than two
author. components: physical capacitance (double-layer capacitance) current,
pseudo-capacitance current (surface-confined redox reactions) and

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

diffusive current (bulk redox reactions in electrodes).26 As can be Eq. 4 also indicates a feasible use of an exponential curve to simulate
seen from Eq. 1-1, the physical capacitance is not distinguished from the residual current to remove it:
pseudo-capacitance. = I1 exp(a1Vcathodic ) + b1 (5 -1)
4) The two constants k1 and k2 at every potential are obtained by
linear fitting of Eq. 1-2 at one potential point by another, which is in =I2 exp(a2Vanodic ) + b2 (5 - 2)
the need of repeatedly calculating several hundred even several Where I 1 = The residual current during cathodic process.
thousand times when fitting the potential window. I 2 = The residual current during anodic process.
Herein, three successive treatment steps on CV curves, including
de-polarization, de-residual, de-background, are firstly introduced to a1 , b1 and a2 , b2 are four variables to determine.
calibrate physical capacitance, pseudo-capacitance and diffusive I 'de − residual , cathodic = I original , cathodic - I1 (6 -1)
contribution separately. Furthermore, different from conventional
I 'de − residual , anodic = I original , - I2 (6 - 2)
works to obtain the constants k1 and k2 at each potential by linear anodic

fitting tediously and repeatedly, we propose for the first time a novel Where I 'de − residual , cathodic = The cathodic current after residual calibration
non-linear fitting algorithm and by one-step matrix treatment I original , cathodic = The original cathodic current
calculation with MATLAB, from which all k1 and k2 values could be
obtained simply and efficiently. I 'de − residual , anodic = The anodic current after residual calibration
I original , anodic = The original anodic current

Accepted Manuscript
De-background (Second step of current calibration)
Brief on principles of electrochemistry and As stated above, the current in CV test at a certain potential is
mathematics composed of three rather than two components: physically capacitive
current, pseudo-capacitance current and diffusive current.26 However,
De-polarization (Potential calibration) as demonstrated in Eq. 1, the physical capacitance is not
With the increase of scan rate in CV method, the oxidation peaks distinguishable from pseudo-capacitance, so that we rewrite Eq. 1
into the three components that will be discussed in detail in Figure 4b
would inevitably shift positively while reduction peaks move
as the double-layer current ( k1ν
'
), pseudo-capacitance current ( k1''ν )
negatively due to irreversible polarization. Equivalently speaking,
there is a potential deviation from equilibrium potential caused by and diffusive current ( k2ν 0.5 ) as follow,
irreversible resistance (R), which is brought mainly by independent I ' = k1ν + k2ν 0.5 = k1ν
'
+ k1''ν + k2ν 0.5 ( Eq. 7)
ohmic resistance (the detailed explanation and analysis can be found
Then the following steps are to determine the capacity of double-
in Section 3.1). To eliminate the potential polarization, we need to
layer capacitance and then the second step is to calibrate the current
find the dominating R value and then restore the potential by Equation
according to,
2:
' I ''de − background , cathodic = I 'de − residual , cathodic + k1ν
'
= k1''cν + k2 cν 0.5 ( Eq. 8 − 1)
= Voriginal − R * I original ( Eq. 2)
Vcalibrated
'
I ''de − background , anodic = I 'de − residual , anodic − k1ν
'
= k1''aν + k2 aν 0.5 ( Eq. 8 − 2)
Where, Vcalibrated = Calibrated voltage
Where I ''de − background , cathodic = The cathodic current after de-background
Voriginal = Original ( experimental ) voltage
calibration
I original = Original ( experimental ) current
I ''de − background , anodic = The anodic current after de-background calibration
After the removal of irreversible potential shift, all the potentials
k1' , k1''c and k2 c are three variables in cathodic process
would be calibrated to the electrochemical potential, which apply to
the universal current-potential relationship as below, k1' , k1''a and k2 a are three variables in anodic process
 C (0, t )
 C (0, t ) 
 The double-layer current follows the relationship as,
= i i0  o ∗ exp(−α fnη ) − R ∗ exp[(1 − α ) fnη )] ( Eq.3) dV '

 C o C R 
 dQ
= * dt i *
i= ( Eq. 9)
d
η =E-E eq ν
Thus, the double-layer current at various scan rates changes as,
Where CO* and CR* is the concentration of oxidative and reductive dQd dQd
= i = ( Eq.10)
state in bulk ( solid ) solution, respectively; CO (0, t )and CR (0, t ) is dt dV '/ ν
the concentration of oxidative and reductive state at the reaction Where Qd = The double - layer capacitance
interface, respectively; n = electrons transferred per formula; i = The double - layer current
f = F / RT . V ' = The calibrated voltage after de-polarization
C (0, t ) C (0, t ) ν = The scan rate
When 0.9 ≤ o ∗ and R ∗ ≤ 1.1 , corresponding to a non-
Co CR Since this part of capacity is highly reversible and suffer from
diffusion-controlled condition, Eq.3 could be simplified as the well- negligible decay, Eq. 10 implies that the double-layer currents at two
known Butler-Volmer Equation, different scan rates have the following relationship:
i = i0 {exp(−α fnη ) − exp[(1 − α ) fnη )]} ( Eq.4) i1 / ν 1 = i2 / ν 2 ( Eq.11)
De-residual calibration (First step of current calibration) which can be used to compare and calculate the double-layer capacity
As shown in Eq. 3, there is always a residual current arising from at different scan rates. With the acquisition of three independent
reverse reaction, that is, a reductive current during anodic process components through the above three key steps, we are also able to
with an oxidative current during cathodic process. This residual restore the detailed results by adding the ohmic polarization in the
current that will be discussed in detail in Figure 3a would be even experimental CV curves.
larger when the scan rate increases because more abundant active
material is left and unable to receive or release electrons, leading to Non-linear fitting
the sacrifice of capacity, as well as significant oxidative and reductive From Eq. 8-1 and Eq. 8-2, a pair of k1 and k2 should be calculated
ends above zero benchmark (seeing in Figure 1a, b and c). Eq. 3 and at every potential point. When calculating N points of potential in a

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

certain range, we must spend a lot of time to repeatedly calculate them α nF β nF

= ic i 0 [exp( ηc ) - exp(- ηc )] ( Eq. 14 − 2)
by N*2 times of linear fitting. Instead, we propose a new non-linear RT RT
solution of batch treatment, assuming the final calibrated current where α = 1- β
(after de-residual and de-background), scan rate and the variables as ηc = ϕe, cathode - ϕcathode
three matrices of Y , X and G , respectively. Through calculating G
mathematically, we can get all variables at every potential easily, and The third case is much more straightforward, as all batteries have
further integrate them to obtain the final physical capacitance, an intrinsic ohmic resistance to charge flow, resulting in cell potential
pseudo-capacitance and diffusive contribution separately. loss. The ohmic polarization is ascribed to the electrical resistance in
the cell configuration, including electrolyte especially from the non-
aqueous electrolyte, electrode, separator, and other components. In
this case, the potential shift can be described as:
Results and Discussions
∆V = iRΩ ( Eq. 15)
With the increase of scan rate in CV test, the oxidation peaks would The total resistance (Rt) is the sum of the three parts:
inevitably shift positively while reduction peaks move negatively due Rt = Re + Rd + RΩ ( Eq. 16)
to polarization, equivalently caused by a non-ignorable resistance As we can see in Eq. 13, Eq. 14-1 and Eq. 14-2, the relationship
between potential shift and current is nonlinear, in other words, there

Accepted Manuscript
between potential and current. The experimental potential also known
as terminal potential is affected by electromotive force, diffusional, exists nonlinear diffusion resistance (Rd) or electrochemical
electrochemical and ohmic polarizations as follows: resistance (Re), however, Eq.15 clearly indicates that ohmic
V = ϕcathode - ϕ anode - iR resistance (RΩ) is linear. To get a deeper insight into their effects on
CV shape and to extend the application scope, we firstly select several
= (ϕe , cathode -ηcathode ) - (ϕe , anode +η anode ) - iR ( Eq.12)
typical electrode materials for LIBs and for emerging prototype SIBs.
= E - (ηcathode + η anode ) - iR Here ‘typical’ means they should be the common candidates, undergo
Where, V = terminal voltage single-phase or two-phase electrochemical reaction, and have single
E = electromotive force or multiple redox peaks. Following these principles, we have chosen
i = net current of circuit three representatives, i.e. LiFePO4 with single peak undergoing two-
ϕe, cathode = equilibrium potential of cathode phase process,27 LiMn2O4 for commercial LIBs with multiple peaks
(LixMn2O4, 0＜x≤1)28-29, and Na4Fe3(PO4)2P2O7 for primary SIBs
ϕe, anode = equilibrium potential of anode with multiple peaks30 to conduct our calibration. In addition, scan rate
ηcathode = overpotential of cathode also be regarded as a factor, thus low and high scan rates have been
η anode = overpotential of anode applied to these samples as well.
All parameters have the same meaning as after unless
pointed out
In detail, if the whole reaction rate is controlled by diffusion
resulting from the sluggish ion diffusion in electrolyte or electrode,
which cannot meet electrons’ motions in the outer loop, the equation
of potential deviation and current in this situation can be expressed
as:
RT i
ϕcathode (or anode) =ϕe, cathode (or anode) + ln(1 − ) ( Eq. 13 − 1)
nF id
RT i
ηcathode (or anode) = ln(1 − ) ( Eq. 13 − 2)
nF id
Where, i d = diffusion-limited current of cathode (or anode)
R = 8.314 J / ( mol  K ) , ideal gas constant
T = temperature in Kelvins
n = the number of electrons transferred in the cell reaction
F = the Faraday constant
All parameters have the same meaning as after unless
specifically pointed out Figure 1. (a), (b), (c) original cyclic voltammetry; (d), (e), (f) linear
In second case, if the whole reaction rate is predominately relationship of anodic/cathodic peaks; (g), (h), (i) depolarized cyclic
determined by electrochemical process (electrons transfer process), voltammetry of LiFePO4, LiMn2O4 and Na4Fe3(PO4)2P2O7, respectively.
the potential change significantly affects the activation energy of
redox reaction. For anodic reaction (charge process on cathode or
discharge process on anode), it gives: Shown in Figure 1a, b and c, no matter for LiFePO4 (single peak,
β nF α nF with low scan rates of 0.2-0.8 mV s-1), LiMn2O4 (multiple peaks, with
ia i 0 [exp( η a ) - exp(- η a )] ( Eq. 14 − 1)
RT RT low scan rates of 0.2-1.0 mV s-1) or Na4Fe3(PO4)2P2O7 (multiple
where i 0 = exchange current density peaks, with high scan rates of 0.5-10 mV s-1), all the three original
β = transfer coefficient cyclic voltammetry curves are inclined, and this trend would be
ηa = ϕcathode - ϕe, cathode amplified with increasing scan rate, indicating a tremendous potential
shift due to one or more kinds of the above mentioned polarizations
And for cathodic reaction (discharge process on cathode or charge
that makes the equilibrium potential difficult to be distinguished.
process on anode),
Fortunately, the conspicuous redox peaks could give us some clues.
At the peak point, the oxidative or reductive process reaches the

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

maximal degree, if the reaction is limited by diffusion resistance or

electrochemical resistance, the potential-current shape should be
nonlinear. However, to our greatest surprise when we depict the
current and potential value of peak point at different scan rates in
Figure 1d, e and f, a significant linear relationship appears, suggesting
the total resistance is mainly contributed by ohmic resistance. We can
also get hint that a steeper slope means increased ohmic resistance,
which provides a simple and feasible way to directly compare the
resistance of different electrochemical systems in the absence of
exhausted measurement.
To further evaluate the impact weight on peak shifts, we conduct a
linear fitting between current and potential, and the fitting results of
the three samples are listed in Table 1. It’s worthwhile to mention the
correlation coefficient (R) is the mostly used parameter by many
researchers to evaluate the linear relationship, here it can be defined

Accepted Manuscript
as:
cov( I p ,V p ) E[( I p − EI p )(V p − EV p )]
= R = ( Eq.17)
DI p • DV p DI p • DV p
Figure 2. (a) 200 fitting lines and (b) the corresponding correlation
Where the Ip and Vp represents the current and potential values at coefficient of the I/ν0.5 -ν0.5 relation at a narrow region after only de-
five CV peaks, respectively. However, here we prefer to recommend polarization. (c) 200 fitting lines and (d) the corresponding correlation
coefficient of the I/ν0.5 -ν0.5 relation at a wide region after both de-
the use of R2, also known as coefficient of determination (COD), as
polarization and de-residual.
the statistical measure to qualify the linear regression. The reason
behind is that it could give a firm evidence whether there is a strong
linear reliance between current and potential at peak points because After depolarization, we used Eq. 1-2 to verify the feasibility of
this is the core foundation of our method, for example, R=0.95 only our method. The concrete operations listed in Figure S2 show a
equals to COD=0.9, but COD=0.982 (the lower limiting value in selected region equally divided into N-1 intervals with N
Figure 1e) equals to R=0.991. interpolating points. Then a successive linear fitting is repeatedly
There are some important points in Table 1 worthy to be mentioned: conducted and correlation coefficient is employed as a parameter to
(i) For a certain sample, the unchanged slope values during either check the effectiveness. As illustrated in Figure 2a and b, when the
anodic or cathodic process suggest the ohmic resistance is a constant. narrow region of a = 4.05, b = 4.30; c = 4.30, d = 4.05 is chosen, a
(ii) For a certain sample, the two linear fitting lines are approximately satisfactory linear effect can be found at each potential, because the
parallel to each other, which is another strong evidence that the total correlation coefficient is very close to 1.0 for oxidative process and -
resistance is dominated by ohmic resistance, because it is independent 1.0 for the reduced segment, which prove the validity of our
on neither electrons’ receiving nor electrons release process. (iii) The depolarization calibration. However, when extended the region to a =
intercept and slope values are both positive, indicating that they have 3.75, b = 4.30; c = 4.30, d = 3.75, as shown in Figure S3, a poor [abs
the same influence on CV shapes (causing the positive shift of (correlation coefficient) << 1.0] even reverse result (-1.0 for oxidative
oxidation segment and negative shift of reduction segment). (iv) The while 1.0 for the reductive segment) would appear, suggesting some
intercept value is one magnitude lower than slope value. Further inferiorities in our method.
calculation on the ratio of 1-RΩ/Rt indicates that the intercept value Before proceeding to the next step, two issues need to be addressed
(corresponding to the nonlinear Re and Rd) accounts for less than ~7% clear. The first is why here we use R rather than the above mentioned
of total resistance (Rt) and could be ignored in data calibration R2? The reason behind this is that the reverse correlation coefficient
compared to the slope value (corresponding to the linear RΩ). (v) the close to -1.0 at oxidative segment will give us false guidelines and
result provides a hint that this ratio would become smaller with a wide make chaos [(-1)2→1.0]. The other one is why we still have a better
scan range from LiFePO4, to Na4Fe3(PO4)2P2O7. In other words, correlation coefficient in the narrow region in terms of inferiority.
ohmic polarization is the main inducement responsible for potential The answer is that the disturbance from the inferiority is negligible in
shift and would become predominate over electrochemical and this area, and we will further explain below.
diffusional polarizations when scan rate increases. This change trend As stated above, some active materials are unable to release/accept
is easy to understand as shown in Eq.15 due to the appearance of electrons when the scan direction is reversed especially under high
rapidly rising current at high scan rate. scan rate, in other words, strong residual current is included in the
After depolarization of potential calibration found in Figure 1g, h observed current curve. As a result, the anodic current starts with a
and i, the CV shapes can be corrected according to Eq. 2 and become negative value while cathodic curve begins at positive region, which
symmetric to the potential axis at peak positions as well as other goes against its definition and causes poor or even reverse correlation
potential values. In short, the offset of potentials in CV shapes is coefficient in Figure S3. Therefore, it is still crucial to calibrate the
mainly caused by ohmic polarization, and we can correct it effectively deviation of experimental currents caused by residual current, and we
regardless of single or double-phase electrochemical process, the define this operation to eliminate the adverse influence as de-residual.
peak numbers and scan rates. To address the dilemma, the mirror lines of extended trend lines
beyond the scan scope are further stimulated, which degrades to zero
at ∞ in theory, however, it will soon reach I = 0 in experiment when
the two ends of CV region applied. As demonstrated in Figure 3a with
𝜈𝜈 = 0.6 mV/s as an illustration, according to Eq. 3 and Eq.4, we
could introduce two exponential curves for each scan rate, I1 for

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

cathodic and I2 for anodic would be the residual currents to further layer physical capacitive near square can be found, and the current
adjust the CV shapes following the relationship described in Eq. 6-1 undergoes negligible change within the whole potential window. This
and Eq. 6-2. We could also find that disturbance from the residual reminds us that in CV images we could adopt the current before
current is negligible in this area, which is the reason responsible for significant redox reaction to approximately estimate the double layer
why there is still a great linear relationship in Figure 2b even without current. Note we are not using any data in Figure S4 during the
de-residual process. following analysis. We have considered the current resulted from
physical adsorption, that is, the pure physical contribution 𝑘𝑘1′ 𝜈𝜈 at each
scan rate as an average value before the current changes sharply. The
validity of this assumption is checked by Eq.11. Figure S5 shows the
theoretical and experimental double-layer capacitances of LMO, and
Figure 4a describes the linear relationship between them. The
excellent COD validates the correctness of our method and implies
the negligible contribution of adsorption current.
After the three key steps including de-polarization, de-residual and
de-background calibration, the following step is to repeatedly conduct

Accepted Manuscript
hundreds or even thousands of linear regressions between the revised
Figure 3. (a) The principle to de-residual at 𝜈𝜈 = 0.6 mV/s and (b) the
potential and refreshed current. However, we propose a new method
depolarized CV shapes after further de-residual.
of matrix process based on non-linear fitting. Here we will give an
elaborate description of principles about it and we hope this may be

Table 1. Details of the fitted intercept and slope of peak current and peak potential for various samples (A=anodic, C=cathodic).

Scan range R2 (COD) Slope Intercept 1-RΩ/Rt

Sample
mV s-1 A C A C A C A C
LiFePO4 0.2-0.6 0.999 0.997 43.34 42.56 3.46 3.38 0.073 0.074
Li2MnO4 0.2-1.0 0.998 0.991 58.75 69.70 4.14 3.98 0.054 0.066
Na4Fe3(PO4)2P2O7 0.5-10 0.997 0.995 97.23 106 3.29 2.83 0.034 0.026
an alternative but simplified as well as effective method in the future.
Table 2. Parameter values of exponential decay curves for different At any certain potential Vn, the pseudo-capacity and diffusive redox
scan rates. capacity could have originated simultaneously, we can consider the
Scan rate(mV/s) a b c d mathematical relationship between them by rewriting Eq. 8-1 of
0.2 3.6867 57816.40 4.3823 84571.67 anodic process as:
0.4 3.7020 13113.35 4.3793 20632.62 I jn - k1' n v j = k1''n v j + k2 n v1/j 2 ( Eq.18 − 1)
0.6 3.6948 3786.15 4.3696 5655.98
= y jn I nj - k1' n v j ( Eq.18 − 2)
0.8 3.7049 1492.39 4.3493 2229.18
1.0 3.7211 739.58 4.3251 1133.67 k1''n v j + k2 n v1/j 2
y jn = ( Eq.18 − 3)
The two parameters of every exponential curve are calculated with Where n means the nth point of potential while j represents the jth
the two groups of data (the first and the last of each CV segment), and cycle scan, when j varies from 1 to 5 we have,
the details are exhibited in Table 2. When the depolarized curves are = y1n k1''n v11/ 2 + k2 n v11/ 2
further adjusted by de-residual process, a refreshed CV shapes in
y2 n k1''n v2 + k2 n v1/2 2
=
Figure 3b are brought out, in which all the residual currents are
eliminated. Thus, at the beginning of anodic and cathodic process, y3n k1''n v3 + k2 n v31/ 2
= ( Eq.19)
every curve starts at zero. Then we could further conduct the linear ''
y4 n k v + k v
= 1n 4
1/ 2
2n 4
regression based on Eq. 1-2 to verify the effectiveness of our de-
y5 n k1''n v5 + k2 n v51/ 2
=
residual treatment. As shown in Figure 2c and d, within the same
extended region applied above, all the correlation coefficients are Rewrite it into matrix form,
elevated to above 0.95, reinforcing the rationality and preciseness of  y1n   ν 1 ν 11/ 2 
   1/ 2 
the de-residual calculation.  y2 n  ν 2 ν 2   k '' 
The de-polarization treatment would eliminate the potential offset  y3n  = ν 3 ν 31/ 2   1n  ( Eq.20)
mainly caused by ohmic resistance while de-residual calibration     k
1/ 2  2 n 
 y4 n  ν 4 ν 4 
would alleviate the experimental current from residual reaction.  y  ν ν 1/ 2 
However, as can be seen from Eq. 1-1 and Eq. 1-2, the current  5n   5 5 

corresponding to double layer capacity in CV curve is included in When we extend the n from 1 to N, we have,
pseudo-capacitance current, which will lead to overestimation of  y11 y12 ⋅ ⋅ ⋅ y1N   ν 1 ν 11/ 2 
pseudo-capacitance contribution. To exclude it from the total current,    1/ 2 
 y21 y22 ⋅ ⋅ ⋅ y2N  ν 2 ν 2   k '' k '' ⋅ ⋅ ⋅ k '' 
we have rewritten the Eq. 1-2 into Eq. 7 to determine the double-  y31 y32 ⋅ ⋅ ⋅ y3N  = ν 3 ν 31/ 2   11 12 1N
( Eq.21)

layer current (𝑘𝑘1′ 𝜈𝜈), pseudo-capacitance current (𝑘𝑘1′′ 𝜈𝜈) and diffusive     k k22 ⋅ ⋅ ⋅ k2 N 
1/ 2  21
current (𝑘𝑘2 𝜈𝜈 0.5 ) separately. As reasonably, double layer current 𝑘𝑘1′ 𝜈𝜈  y41 y42 ⋅ ⋅ ⋅ y4N  ν 4 ν 4 
 y y ⋅⋅⋅ y  ν ν 1/ 2 
is a constant originated from the period before the potential reaches  51 52 5N   5 5 

the equilibrium value, and remain unchanged during the whole Which can be further denoted as:
potential window. To approve this inference, we replace LMO with Y=XG ( Eq.22)
active carbon and pair it with the same binder and Super P as well as
the same ratio. The result is shown in Figure S4, a significant double-

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

Where Y , X and G are the above three matrixes successively. the contribution of the three components while Figure 4d is the
Then we can obtain any k1n'' and k2n at each potential by solving the restored double-layer (physical), pseudo and diffusive current in the
above matrix equation, but X is not square matrix thus there exists no experimental CV image when we applied polarization to them. All
inverse matrix (the pseudo-inverse matrix is not discussed here). the corresponding contribution details after de-residual and in initial
Mathematically we can construct a square matrix by multiplying the CV curves can be found in Figure S6 and Figure S7, respectively. We
transposed matrix X T to its left to obtain a square matrix X T X . To can find that the diffusive ratio is a little larger in oxidative part than
note, since the row number of X is far larger than its column number that of in reductive part. To unveil this puzzle, we have calculated the
(j >2), X T X won’t generate an unreliable singular matrix, thus Eq.22 diffusion coefficients of LMO, LFP and NFPP, and the satisfactory
has a unique solution. Consequently, G can be figured out by, fitting relationships are illustrated in Figure S8, Figure S9 and Figure
S10 successively. Further calculations show the apparent diffusion
X T Y = X T XG ( Eq.23 - 1)
T -1 T
coefficients of LMO are 9.52*10-11 and 7.02 *10-11 cm2/s for
(X X) X Y = G (Eq.23 - 2) oxidative peaks and reductive peaks, respectively. This result is in
Where coincidence with Figure 4b and d that the diffusive current is getting
ν 1 ν 11/ 2  smaller in cathodic segment. In other words, there is a slightly
  increased resistance for Li ions to insert into compared to extract from
ν 2 ν 21/ 2 

Accepted Manuscript
the crystal lattice. Please note the comparison between the two
X = ν 3 ν3 1/ 2 
  different samples like LFP and NFPP is meaningless because of the
ν 4 ν 41/ 2  significant difference in electrochemical activity. It’s also easy to find
ν ν 51/ 2 
 5 in Figure 4c that the contribution of Pseudo part is overwhelming
X T is the transposed matrix of X and (X T X)-1 is the inverse while diffusive component is decreasing when scan rate increases,
matrix of X T X . and in some report, the authors found the Pseudo contribution even
above 90%,25 The reason behind this can be explained by the
diffusion formula,
r 2 = t * Dion ( Eq.24)
Where r is the average diffusion distance, t is the diffusion time,
and Dion is the diffusion coefficient. As proved by the above
calculation, Dion can be roughly regarded as a constant. The three
charge storage mechanisms are schematically shown in Figure 5.

Figure 4. (a) The comparison of theoretical and experimental double-

layer capacitance of LMO. (b) The calibrated double-layer (physical),
pseudo and diffusive current in the CV image after de-residual at 0.4
mV/s. (c) Details of the three components contribution at various scan
rates. (d) The restored double-layer (physical), pseudo and diffusive
current in the experimental CV image at 0.4 mV/s. Figure 5. Schematic illustration of the three charge storage means at
low and high scan rate

Following these algorithms, we can conduct nonlinear fitting, and

then will obtain the value of G by solving the above matrix. Since In essence, both diffusive and Pseudo capacity are originated from
 k ''  Faradic reactions, which is primarily distinct from double layer
the any row  1n  represents the corresponding constants of capacitance where charge is stored through physical interaction at
 k2 n  rough surface. In meantime, the only difference between the diffusive
(ν j ν 1/j 2 ) at the scan rate j at the nth point, thus the capacitive and and the Pseudo portion is the length of reaction route caused by
diffusive contribution at point n, scan rate j are the results of the first electrons-ions imbalance. Since CV undergoes potentiodynamic scan
 k ''  process, the diffusion time t will be smaller when scan rate v increased,
and second term of (ν j ν 1/j 2 )  1n  , respectively. The mathematical therefore a smaller diffusive r or decreasing diffusive component will
 k2 n  be expected. In contrast, because the Pseudo current is kinetically
data processing method is essentially a widely accepted least squares facile and diffusion is not the rate-determining step, the increased
method and the correctness is also proved in the Supporting scan rate doesn’t have significant influence on the Pseudo capacity.
Information. Equivalently, Li or Na ions are unable to diffuse into the deep part of
Figure 4b is the calibrated double-layer (physical), pseudo and
a bulk material due to the limited time at high rate, leading to the
diffusive current in the CV after de-residual at 0.4 mV/s while Figure
capacity decay as well as the relatively increased Pseudo contribution
4d is the restored details by adding the ohmic polarization in the
to the total capacity in Figure 4c.
original experimental CV image. The Figure 4b is used to determine

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

To extend the application scope of our method, by following the [7] Pu, X.; Wang, H.; Yuan, T.; Cao, S.; Liu, S.; Xu, L.; Yang, H.; Ai, X.;
same calibration steps and calculating algorithm, we also specify the Chen, Z.; Cao, Y., Energy Stor. Mater. 2019, 22, 330-336.
[8] Yuan, T.; Wang, Y.; Zhang, J.; Pu, X.; Ai, X.; Chen, Z.; Yang, H.; Cao,
capacitive contribution for single-peak electrode material LiFePO4 Y., Nano Energy 2019, 56, 160-168.
(undergoing two-phase process) at low scan rate ranges, which has [9] Pu, X.; Wang, H.; Zhao, D.; Yang, H.; Ai, X.; Cao, S.; Chen, Z.; Cao, Y.,
small destructiveness thus a more complete CV images are obtained. Small 2019, 15, 1805427.
The results of calibrated double-layer (physical), pseudo and [10] Brezesinski, T.; Wang, J.; Tolbert, S. H.; Dunn, B., Nat. Mater. 2010, 9,
146-151.
diffusive current in the CV after de-residual at various scan rates can [11] Simon, P.; Gogotsi, Y., Nat. Mater. 2008, 7, 845-854.
be found in Figure S11 while Figure S12 shows the restored details [12] Lin, T.; Chen, I. W.; Liu, F.; Yang, C.; Bi, H.; Xu, F.; Huang, F., Science
in the original experimental CV image. And for Na4Fe3(PO4)2P2O7, a 2015, 350, 1508-1513.
[13] Simon, P.; Gogotsi, Y.; Dunn, B., Science 2014, 343, 1210-1211.
typical cathode material for sodium ion batteries, the feasibility of our
[14] Gogotsi, Y.; Penner, R. M., ACS Nano 2018, 12, 2081-2083.
method under much more complicated conditions (CV with multiple [15] Simon, P.; Gogotsi, Y., Acc. Chem. Res. 2013, 46, 1094-1103.
peaks and in an extended scan range) is also tested in Figure S13. The [16] Jian, Z.; Raju, V.; Li, Z.; Xing, Z.; Hu, Y.-S.; Ji, X., Adv. Funct. Mater.
green marked areas are responding to the pseudo current and blue 2015, 25, 5778-5785.
curves means the calibrated and restored diffusive current. After all [17] Zhang, L.; Zhao, K.; Luo, Y.; Dong, Y.; Xu, W.; Yan, M.; Ren, W.; Zhou,
L.; Qu, L.; Mai, L., ACS Appl. Mater. Interfaces 2016, 8, 7139-7146.
the rational steps, a remarkable and ordered image from the three [18] Lesel, B. K.; Ko, J. S.; Dunn, B.; Tolbert, S. H., ACS Nano 2016, 10,

Accepted Manuscript
components could be obtained in Figure S6 and Figure S11, but in the 7572-7581.
CVs with the ohmic polarization in Figure S7 and Figure S12, a [19] Chen, Z.; Augustyn, V.; Jia, X.; Xiao, Q.; Dunn, B.; Lu, Y., ACS Nano
2012, 6, 4319-4327.
distorted and disordered results would appear.
[20] Chen, M.; Chen, L.; Hu, Z.; Liu, Q.; Zhang, B.; Hu, Y.; Gu, Q.; Wang, J.
L.; Wang, L. Z.; Guo, X.; Chou, S. L.; Dou, S. X., Adv. Mater. 2017, 29,
1605535.
Conclusion and outlook [21] Ma, X.; Wu, X.; Shen, P., ACS Appl. Energy Mater. 2018, 1, 6268-6278.
[22] Yang, M.; Kong, Q.; Feng, W.; Yao, W., Carbon 2021, 176, 71-82.
[23] Zhang, Y.; Li, T.; Cao, S.-a.; Luo, W.; Xu, F., Chem. Eng. J. 2020, 387,
We firstly propose a method including three key steps in terms of 124125.
de-polarization, de-residual, de-background to precisely calibrate the [24] Niu, Y.; Zhang, Y.; Xu, M., J. Mater. Chem. A 2019, 7, 15006-15025.
pseudo-capacitance contribution in CV curves. In addition, we also [25] Huang, M.; Wang, X.; Meng, J.; Liu, X.; Yao, X.; Liu, Z.; Mai, L., Nano
Energy 2020, 77, 105069.
provide an alternative non-linear fitting algorithm based on matrix [26] Jiang, Y.; Liu, J., Energy Environ. Mater. 2019, 2, 30-37.
manipulating to facilitate the accuracy and efficiency. Based on these [27] Orikasa, Y.; Maeda, T.; Koyama, Y.; Murayama, H.; Fukuda, K.; Tanida,
fundamentals, we successively conducted matrix process on LiMn2O4 H.; Arai, H.; Matsubara, E.; Uchimoto, Y.; Ogumi, Z., Chem. Mater. 2013, 25,
undergoing single-phase process with multiple peaks and low scan 1032-1039.
[28] Amatucci, G.; Tarascon, J. M., J. Electrochem. Soc. 2002, 149, K31.
rates, LiFePO4 undergoing two-phase process with single peak and [29] Hosono, E.; Kudo, T.; Honma, I.; Matsuda, H.; Zhou, H., Nano Lett. 2009,
low scan rates, Na4Fe3(PO4)2P2O7 undergoing single-phase process 9, 1045-1051.
with multiple peaks and high scan rates to obtain the physical [30] Kim, H.; Park, I.; Lee, S.; Kim, H.; Park, K.-Y.; Park, Y.-U.; Kim, H.;
capacitance, pseudo-capacitance and diffusive contribution Kim, J.; Lim, H.-D.; Yoon, W.-S.; Kang, K., Chem. Mater. 2013, 25, 3614-
3622.
separately. During the calibrating process, electrochemical analysis,
coefficient of determination (R2), correlation coefficient (R),
comparison of theoretical and experimental double-layer capacitance,
fitting of diffusion coefficient and the diffusion formula are employed
one by another as threshold requirements to analyze and check the
calibration validity. This work will help to eliminate ambiguous
concepts and understandings in the previous reports while provide a
more rational and reliable method to calculate the pseudo-capacitance
contribution.

Received: ((will be filled in by the editorial staff))

Published online on ((will be filled in by the editorial staff))

Keywords: Charge storage mechanism, Physical-capacitance,

Pseudo-capacitance, Calibration, Non-linear fitting

[1] Zhang, J.; Cui, C.; Wang, P.-F.; Li, Q.; Chen, L.; Han, F.; Jin, T.; Liu, S.;
Choudhary, H.; Raghavan, S. R.; Eidson, N.; von Cresce, A.; Ma, L.; Uddin,
J.; Addison, D.; Yang, C.; Wang, C., Energy Environ. Sci. 2020, 13, 2878-
2887.
[2] Chen, Z.; Yuan, T.; Pu, X.; Yang, H.; Ai, X.; Xia, Y.; Cao, Y., ACS Appl.
Mater. Interfaces 2018, 10, 11689-11698.
[3] Yuan, T.; Zhang, J.; Pu, X.; Chen, Z.; Tang, C.; Zhang, X.; Ai, X.; Huang,
Y.; Yang, H.; Cao, Y., ACS Appl. Mater. Interfaces 2018, 10, 34108-34115.
[4] Li, H.; Liu, S.; Wang, H.; Wang, B.; Sheng, P.; Xu, L.; Zhao, G.; Bai, H.;
Chen, X.; Cao, Y.; Chen, Z., Acta Phys. Chim. Sin. 2019, 35, 0001-0009.
[5] Pu, X.; Rong, C.; Tang, S.; Wang, H.; Cao, S.; Ding, Y.; Cao, Y.; Chen,
Z., Chem. Commun. 2019, 55, 9043-9046.
[6] Chen, L.; Wu, H.; Wang, H.; Chen, L.; Pu, X.; Chen, Z., Electrochim. Acta
2019, 307, 224-231.

This article is protected by copyright. All rights reserved.

 Angewandte Chemie 10.1002/ange.202104167

Table of contents

Well-separated physical capacitance, pseudo-capacitance and diffusive capacity are achieved from the CV curves of typical
electrode materials for metal-ions batteries after three successive treatments including de-polarization, de-residual and de-
background as well as non-linear fitting calculation, offering a more rational and reliable method to calculate the pseudo-
capacitance contribution.

Accepted Manuscript

This article is protected by copyright. All rights reserved.