Impact of temperature on calendar ageing of

Lithium-ion battery using incremental capacity analysis

Matthieu Maures, Yuanci Zhang, Cyril Martin, Jean-Yves Deletage,
Jean-Michel Vinassa, Olivier Briat

To cite this version:

Matthieu Maures, Yuanci Zhang, Cyril Martin, Jean-Yves Deletage, Jean-Michel Vinassa, et al..
Impact of temperature on calendar ageing of Lithium-ion battery using incremental capacity analysis.
Microelectronics Reliability, 2019, 100-101, �10.1016/j.microrel.2019.06.056�. �hal-02506186�

HAL Id: hal-02506186

https://hal.science/hal-02506186
Submitted on 21 Dec 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.

Distributed under a Creative Commons Attribution - NonCommercial| 4.0 International

License
Version of Record: https://www.sciencedirect.com/science/article/pii/S0026271419304731
Manuscript_90a82f420290748c988901dae139f3dc

Impact of temperature on calendar ageing of Lithium-ion

battery using incremental capacity analysis
M. Mauresa,*, Y. Zhanga, C. Martina, J.-Y. Delétagea, J.-M. Vinassaa, O. Briata
a
Univ. Bordeaux, CNRS Bordeaux INP, IMS, UMR 5218, F-33400 Talence, France

Abstract

In this paper, a calendar ageing model featuring time and temperature dependence of a Lithium-ion battery
is proposed. The ageing procedure is done at a high state of charge (SoC) of 95% and the tested temperatures
range from -20°C to 55°C. Electrochemical methods such as incremental capacity analysis (ICA) and differential
voltage analysis (DVA) are used to quantify the evolution of the most important degradation modes such as
conductivity loss (CL), loss of active material (LAM) and loss of lithium inventory (LLI). While the
quantification methods used here are extracted from previous power cycling studies, their validity on calendar
ageing data is verified, and a model of each degradation mode evolution versus time including temperature
parametrization is proposed.

1. Introduction temperature and the current draw.

To quantify and model the SoH and its
Since the emergence of electric vehicles (EVs) to dependence on the batteries' working conditions,
answer the negative effects of CO2 emissions on ageing tests must be performed. Usually, those can
climate change and global warming, the Lithium-ion be done in several ways: abuse test, power cycling,
battery market has tripled, and the EV market is set and calendar ageing.
to reach up to 60% of the car market by 2050 [1,2]. Abuse tests are described in [6] and consist in tests
Meanwhile, more electric aircrafts (MEAs) are also that are outside of the manufacturer recommended
increasingly using batteries to power their on-board specifications. Power cycling can be done with an
systems and regulate the voltage demand [3]. idealized current profile, at a given ambient
However, multiple drawbacks remain with this temperature and C-rate [5,7], or based on the
technology and prevent its early adoption in those repetition of a current profile which is representative
applications. In particular, EVs present range, of the application. The C-rate is a current unit based
charging time and lifespan (of the battery packs) on the capacity of the cell and the desired time of
issues compared to internal combustion engine discharge. For example, a 1C current will discharge
(ICEs) vehicles. MEAs on the other hand face the battery in 1h, while a C/20 current will discharge
significant challenges regarding the safety and the it in 20h.
state of health (SoH) of their batteries, since they can Finally, calendar ageing consists in studying the
face important temperature variations from -20°C to battery degradation when it is not under load.
55°C. Degradations of the battery are often described as
The SoH of a battery indicates how close it is to combination of cycling and calendar ageing.
its end of life, when a specific failure criterion is Extensive work on calendar ageing has been made
met. Depending on energy and power requirements by [9,10] at temperatures above 25°C and 40°C
of the target application, the SoH can be defined as respectively, while [11] studied the effect of calendar
SoHE and SoHP respectively [4,5]. The SoHE ageing on supercapacitors. Ref. [12] studied the
corresponds to the ratio of the current maximal identification of calendar degradation through ICA,
capacity over the initial maximal capacity, while the although did not provide quantification of the
SoHP corresponds to the rise in internal resistance of different degradation modes that can be observed.
the battery, which is correlated to more Joule losses In this paper, the quantification of the evolution of
hence less usable power for the application. Both three degradation modes is presented. As this was
SoH are affected by the chemistry of the cell, the already performed on cycling tests at 25°C by [5], a

__________________________________________
* Corresponding author. [email protected]
Tel: +33 (5) 40 00 26 58

© 2019 published by Elsevier. This manuscript is made available under the CC BY NC user license
https://creativecommons.org/licenses/by-nc/4.0/
key contribution here is to test it during calendar
max − max
ageing, and over a large temperature range. Since the
= 2
Arrhenius law usually describes the effect of
max
temperature on degradation mechanisms, its validity
as an acceleration factor of degradation modes is also
tested, and an alternative is proposed.
|max | − |max |
= 3
This paper is divided as following: firstly, a
reminder of the different degradation modes |max |
occurring in lithium-ion batteries, and the
formulation of their quantifications based on where:
literature models and incremental capacity and - GCL is the relative evolution of the degradation
differential voltage analysis (ICA and DVA through conductivity loss,
respectively) are presented. Secondly, the - GLAM is the relative evolution of the degradation
experimental protocol and the associated results of through loss of active material of the entire battery,
the calendar ageing are shown. Thirdly, a - GLLI is the relative evolution of the degradation
temperature study of the degradation modes and the through loss of lithium inventory,
validity of the Arrhenius law is questioned. Finally, - OCV (V) is the battery open circuit voltage,
the conclusion is drawn. - Q (Ah) is the capacity of the battery. In this
paper, the discharged capacity will be used,
2. Degradation evolutions formulations - V (V) is the voltage of the battery,
- n is the index of the check-up at which the
The degradation of batteries can be summarized quantification of degradations is made.
in four categories: causes, degradation mechanisms,
degradation modes and effects on the battery [13]. 3. Experimentations
Since degradation mechanisms are numerous
and attributed to internal parasitic chemical To compute each degradation growth from ICA
reactions, it can be tricky to quantify each of them and DVA, ageing tests were performed.
without opening the battery. They can however be
grouped in three distinct degradation modes [5]: 3.1 Protocol of experiments
conductivity loss (CL), caused by the cracking and
the dissolution of the current collectors and their All the tests were performed at the Cacyssée
binder, loss of active material (LAM), which results platform at the IMS laboratory. This platform is
from the consumption of electrode material in designed to conduct different types of tests on
chemical side reactions, and loss of lithium inventory storage devices such as supercapacitors or Lithium-
(LLI), when the Lithium reacts with the electrolyte ion batteries. The platform includes several climatic
to form solid electrolyte interphase (SEI), or when, at chambers as well as numerous electrochemical
lower temperatures, lithium plating of the negative stations and measurement devices that are used to
electrode occurs. perform impedance spectroscopies, power and
Previous work [5,14] showed it is possible to thermal cycling, or calendar ageing of the storage
quantify the relative evolution of each degradation systems.
mode using incremental capacity analysis (ICA) and Here, six Samsung INR18650-25R5 Lithium-ion
differential voltage analysis (DVA). Each method batteries were used in this protocol. They feature
consists in performing a charge or a discharge with Nickel-Cobalt-Aluminum (NCA) cathodes and
specific conditions then deriving the capacity (ICA) Graphite anodes, and a commercial capacity of
or the voltage (DVA) and plotting it versus the 2500mAh. The batteries were held in place using
voltage (ICA) or the capacity (DVA). This usually BioLogic BH-1i holders and placed inside Climats
leads to peaks and valleys (as seen in Fig. 1.). These EX5413TA climatic chambers.
analyses can be done on characterization curves The temperatures were recorded using Type K
(check-ups) performed regularly during an thermocouples taped onto the surface of the batteries.
accelerated ageing test. Electrical measurements were performed using the
From there, three formulas can be extracted to four-wires method, connected to a BioLogic BCS-
quantify those evolutions [8]: 815 device remotely programmed on a computer
through the EC-Lab software. Data points were
max − max measured every 30 seconds or every 10mV voltage
= 1
max variation.
 The ageing procedure can be summarized in the
Table 1 below.
Table 1. Fig. 1. Check-up (top) and 95% SoC set-up
Calendar ageing procedure. procedures (bottom).
Step Procedure Then, a full discharge/charge ③/④ to measure
the capacity, 30min of OCV ⑤, a discharge ⑥ until
1 Initial Check-up (see Fig. 1. below). the battery reach 3.6V (its nominal voltage) and 1h
2 The battery SoC is set to 95%. of OCV ⑦. A resistance measurement ⑧ using ±1C
30s pulses is then done. After 15min of OCV ⑨, the
3 The battery is put to test temperature. battery is fully discharged ⑩ then charged ⑪ for
4 Calendar ageing for a set amount of time. 30min (no CV phase here). The battery then rests for
a few hours ⑫ until they are charged back ⑬ to
5 Test stops. The battery waits for 4h at Tamb ≈ 24°C. 100% SoC, relax for 30min ⑭, then discharge ⑮ for
6 Check-up (same procedure as in step 1). 3min (~5% of the capacity). This procedure provides
a lot of information although, in this paper, all of the
7 Repeat steps 2 to 7 until the battery dies.
information is extracted from the discharge only.
After the initial check-up, the batteries are
Unless specified, all charges/discharges are 1C labelled based on their ageing temperature. Two
CCCV/1C CC respectively. This means charges are batteries are chosen for each temperature to ensure
separated in two phases: a constant current phase repeatability. The characteristics of the batteries are
(CC) until it reaches a given voltage (here 4.2V), summed-up in Table 2 below.
then a constant voltage phase (CV), where the It should be noted that, for different
current drawn by the battery will slowly decrease temperatures, not all check-ups were done at the
until it reaches an end criterion (a C/20 current here). same time. Since it was initially assumed batteries
Discharges on the other end never feature a CV would age the slowest at 25°C, check-ups on SA09
phase. Each check-up (see Fig. 1. below, where the and SA10 were performed every 1000h on average,
circled numbers refer to the procedures described while they were performed every 500h on average on
hereafter) consists of a charge ① from 95% (or 50% the others.
for the initial check-up, as newly received fresh
batteries usually have a ~50% initial SoC) to 100% Table 2.
SoC, 30min of OCV ②. Batteries characteristics and test conditions.
Label Initial Capacity (mAh) Test Temperature (°C)

SA09 2508.41 25

SA10 2499.28 25

SA17 2483.94 -20

SA18 2459.35 -20

SA21 2463.03 55

SA22 2454.83 55

3.2 Experimental results

In order to calculate the evolution of each

degradation mode, discharge, IC, and DV curves are
required. First, the voltage and capacity values of the
batteries operating at the same temperature are
averaged together. Then, the IC and DV curves are
computed in MATLAB using the "diff" function. To
avoid parasitic peaks and quantification noise, both
the voltage and the capacity are filtered using a
Fourier transform and a numerical low-pass filter.
Those operations result in the three set of curves
represented in Fig. 2. To improve readability, only
curves averaged on SA09 and SA10 are presented. In
this case relatively monotonous behaviours with
respect to ageing time can be seen, although this is
not necessarily the case at other temperatures, as
explained in part 4.
It should also be noted that the computation of
LAM degradation using equation (2) is peak
sensitive. In particular, each peak in the IC curve
represents a chemical reaction, more specifically a
phase transition of an electrode active material [16].
In our case, the peak around 3.55V is used to
compute LAM. As it can be seen in Fig. 2 and Fig. 3,
the peak position does not move significantly with Fig. 2. Discharge curves (top), IC curves (middle) and DV
ageing temperature. This is on par with previous curves (bottom) averaged on SA09 and SA10 at 25°C, for
work [5,12] where this peak is used for SoH multiple calendar ageing times.
estimation.

Fig. 3. IC curves for calendar ageing done at -20°C (top)

and 55°C (bottom).

4. Time and temperature models of degradations

4.1 Degradation evolutions time dependence

After calculating degradation evolutions using

equations (1), (2) and (3), and plotting them against
the square root of the calendar ageing time, linear
behaviours can be seen (see Fig. 4.). In particular,
squared correlation coefficients for each degradation
and temperature are shown in Table 3.
They are relatively high (often higher than 0.85),
except at -20°C, particularly for the CL degradation
mode. Multiple factors can explain this: first, no
break-in has been performed on the batteries.
Usually, break-in consists in fully
charging/discharging the batteries three times, using
C/2 CCCV/C/2 CC procedures respectively, and
leads to increase in capacity of the battery. In our
case, this leads to dips (sometimes negative) in the
first values, which wouldn't happen otherwise.
Furthermore, the computation of the CL
degradation evolution is normally based on pseudo-
OCV values (voltage values from low C-rate Fig. 4. Growths of CL (top), LAM (middle) and LLI
discharges, below C/10). Here, check-ups are (bottom) against the square root of time.
performed at 1C. To compute GCL, the voltage at the
end of the first relaxation period is used. This value
CL growth also features relatively important error bars, indicating
however is close enough from its pseudo-OCV
it is explained that improvements in battery
counterpart [16] to justify its use here.
manufacturing processes limit this degradation.

Table 3.
R² between linear models and data points.
T (°C) 25°C -20°C 55°C

R²CL 0.88 0.016 0.95

R²LAM 0.98 0.80 0.85

R²LLI 0.99 0.85 0.96

4.2 Temperature parametrization

For each degradation, the linear model can be

written as:

= √ 4

where logarithms of ACL, ALAM, and ALLI can all be

plotted against the inverse temperature. As seen in
Fig. 5. below, LAM and LLI coefficients can be
modelized by an Arrhenius law, hence:
$
# %
= !"
&' ( √ 5

where kB is the Boltzmann constant, and the G0

parameters and Ea activation energies are summed-
up in the Table 4 below. This is interesting since,
usually, the Arrhenius law serves as a degradation
mechanism model. Degradation modes are a
composition of multiple mechanisms with different
properties that do not necessarily add up into another
Arrhenius model. Mathematically, this can be
explained if a degradation mechanism predominates
the others. This is also supported by [10] where
similar equations can be found to directly model the
capacity fade, but at higher temperatures (between
40°C and 60°C). There, the capacity fade is assumed
to be directly correlated to LLI, which is itself
associated with SEI formation and follows an
Arrhenius law. However, at lower temperatures, SEI
formation is not necessarily the predominant
degradation mechanism, in particular at -20°C where
lithium plating occurs, yet Fig. 5 suggests the
Arrhenius law remains valid nonetheless.
Meanwhile, on the same figure, the CL mode
presents worse linearity than the others (R² = 0.942,
compared to 0.999 for LAM and 0.997 for LLI),
however, when plotting ACL against the inverse
temperature (see Fig. 5), the linearity is improved Fig. 6. New model of the linear coefficient between the
CL degradation mode and the square root of the time.
significantly (R² > 0.999), and can be described by
the following proposed model:
Conclusion
+,
= ! *1 − 0√ 6
-. / An extension to temperature-dependent ageing
models was presented. Three degradation modes
This formalism was chosen to allow comparison known as conductivity loss, loss of active material,
between the two different model, where the G0 and and loss of lithium inventory were chosen, and,
Ea parameters in (6) are presented in the Table 4 based on previous work, their identification and
below. quantification through ICA and DVA was shown.
Although the formulations of the degradation modes
Table 4. evolutions were extracted from power cycling
Model parameters for each degradation. literature tests, they were proven to be adequate for
calendar ageing tests. An ageing model of each
CL (6) LAM (5) LLI (5)
degradation mode featuring temperature
G0 (%/h0.5) 0.0474 11.7 189 parametrization was then presented, and although the
Arrhenius law proved to be sufficient for accurate
Ea (zJ) 3.48 17.5 30.7 representation of the LAM and LLI degradation
modes, a new simple model was proposed for the
CL.
This model could be improved by performing a
new ageing test with more data points in order to
increase the correlation between the degradation
growth and the square root of time. The new test
could also use more batteries to check statistical
variations.
Future work should seek for a correlation
between each of these degradation modes and their
main effects: capacity loss and power fade. This
would be especially useful to create an ageing model
of the internal resistance of the battery, which is
heavily correlated to the CL degradation mode.
Fig. 5. Arrhenius law of the linear coefficient between
References
each degradation and the square root of the time. This
model does not seem suitable for the CL mode.
[1] C. Pillot, aabc Europe, 2017.
[2] T. Gnann, T. S. Stephens, Z. Lin, P. Plötz, C. Liu,
and J. Brokate, Renewable and Sustainable
Energy Reviews, vol. 93, pp. 158–164, May 2018.
[3] R. Mishra and R. Saxena, 2017 7th International
Conference on Power Systems, ICPS 2017, pp.
 702–707, 2018.
[4] D. Andre, C. Appel, T. Soczka-Guth, and D. U.
Sauer, Journal of Power Sources, vol. 224, pp.
20–27, 2013.
[5] C. Pastor-Fernández, W. D. Widanage, G. H.
Chouchelamane, and J. Marco, 6th Hybrid and
Electric Vehicles Conference (HEVC 2016), 2016.
[6] V. Ruiz, A. Pfrang, A. Kriston, N. Omar, P. Van
den Bossche, and L. Boon-Brett, Renewable and
Sustainable Energy Reviews, vol. 81, May 2017,
pp. 1427–1452, 2018.
[7] C. R. Birkl et al., Journal of The Electrochemical
Society, vol. 164, no. 12, pp. A2644–A2653,
2017.
[8] C. Pastor-Fernández, K. Uddin, G. H.
Chouchelamane, W. D. Widanage, and J. Marco,
Journal of Power Sources, vol. 360, pp. 301–318,
2017.
[9] A. Eddahech, O. Briat, E. Woirgard, and J. M.
Vinassa, Microelectronics Reliability, vol. 52, no.
9–10, pp. 2438–2442, 2012.
[10] S. L. Hahn, B. Obry, M. Storch, R. Swaminathan,
K. P. Birke, and J. Bandlow, Journal of Power
Sources, vol. 400, pp. 402–414, July 2018.
[11] O. Briat, J. M. Vinassa, N. Bertrand, H. El Brouji,
J. Y. Delétage, and E. Woirgard, Microelectronics
Reliability, vol. 50, no. 9–11, pp. 1796–1803,
2010.
[12] Y. C. Zhang, O. Briat, J. Y. Delétage, C. Martin,
N. Chadourne, and J. M. Vinassa,
Microelectronics Reliability, vol. 88–90, pp.
1231–1235, July 2018.
[13] C. R. Birkl, M. R. Roberts, E. McTurk, P. G.
Bruce, and D. A. Howey, Journal of Power
Sources, vol. 341, pp. 373–386, 2017.
[14] M. Dubarry, C. Truchot, and B. Y. Liaw, Journal
of Power Sources, vol. 219, pp. 204–216, 2012.
[15] M. Dubarry et al., Journal of Power Sources, vol.
196, no. 23, pp. 10328–10335, 2011.
[16] P. Keil and A. Jossen, Journal of Energy Storage,
vol. 6, pp. 125–141, 2016.