Journal of Energy Storage 6 (2016) 125–141

Contents lists available at ScienceDirect

Journal of Energy Storage

journal homepage: www.elsevier.com/locate/est

Charging protocols for lithium-ion batteries and their impact on cycle

life—An experimental study with different 18650 high-power cells
Peter Keil* , Andreas Jossen
Chair of Electrical Energy Storage Technology, Technical University of Munich (TUM), Arcisstr. 21, 80333 Munich, Germany

A R T I C L E I N F O A B S T R A C T

Article history: This paper presents an overview on charging strategies for lithium-ion batteries. Moreover, a detailed
Received 4 October 2015 assessment of charging strategies is performed, based on an extensive experimental study with three
Received in revised form 14 February 2016 different cell types.
Accepted 15 February 2016
The experimental results reveal that the impact of charging currents and charging voltages on cycle life
Available online xxx
can vary markedly among different lithium-ion batteries. In general, the cycle life is influenced more by
high charging currents than by high discharging currents. Different boost charging protocols have
Keywords:
disclosed that high charging currents can deteriorate cycle life not only at high state of charge (SoC), but
Charging protocols
Fast charging
also at very low SoC. Our investigations on pulse charging show that lithium-ion cells withstand charging
Boost charging pulses of high current or high voltage without any deterioration in cycle life, when the duration of the
Pulse charging pulses remains short and the mean current and voltage values are considerably lower. For pulses of less
Battery aging than 1 s, cycle life has been similar for pulsed and continuous charging with the same mean charging
Cycle life currents and identical cycle depths.
This paper also presents the impact of charging currents and charging voltages on capacity utilization,
charging time, and efficiency to support the development process of optimized charging protocols for
practical applications.
ã 2016 Elsevier Ltd. All rights reserved.

1. Introduction this, several alternative charging protocols can be found in

Lithium-ion batteries provide higher energy and power
densities than other commercial rechargeable battery technolo-
gies. Thus, they are used in various mobile applications, such as
notebooks, cellular phones, cordless tools, and electric vehicles. To
maximize battery life, the methods of operation have to be
optimized. The optimization potential for the discharging process-
es is usually very limited, as the discharging depends largely on the
users’ ways of operating the devices. The charging processes,
however, can be influenced substantially by the manufacturers: by
implementing a specific charging protocol, adjustments between
charging time, capacity utilization, and cycle life can be realized.
The standard charging protocol for lithium-ion batteries is
constant current constant voltage (CCCV) charging. In addition to
Abbreviations: BC, boost charging; CC, constant current; CCCV, constant current
constant voltage; CCPC, constant current + pulsed charging; CV, constant voltage;
EIS, electrochemical impedance spectroscopy; EoL, end of life; LFP, LiFePO4, lithium
iron phosphate; MSCC, multistage constant current; OCV, open circuit voltage; PC,
pulse charging; SoC, state of charge.
* Corresponding author. Fax: +49 89 289 26968.
E-mail address: peter.keil@tum.de (P. Keil).
literature. Section 2 will provide an overview on the different
categories of charging protocols and their specific characteristics.
Many of the alternative charging protocols claim to increase the
amount of charge stored, improve efficiency, or reduce the
charging duration without deteriorating cycle life. However, there
are often only few measurement data presented to verify these
claims. Moreover, the experimental procedures vary among the
different publications, which impedes a direct comparison. As the
experiments are usually performed with only one specific cell type,
no interdependencies between charging protocol and cell type are
revealed. Some charging protocols are even derived solely from
simulation without any experimental validation. Consequently, it
is difficult to compare the performance of the various charging
protocols and assess them for specific practical applications.
To overcome these shortcomings, we present an extensive
experimental study on charging protocols: for three high-power
18650 lithium-ion cell types from different manufacturers and
with different cell chemistries, several charging protocols are
examined with key parameters varied. Moreover, we have
conducted a cycle life study, where the lithium-ion cells have
been cycled with various charging protocols under identical
environmental conditions. This enables an objective comparison
126 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
and evaluation of the different charging strategies. It also reveals
the crucial parameters for optimizing the charging procedure for
lithium-ion batteries.
2. Charging protocols for lithium-ion batteries
In general, optimized charging procedures aim to provide a
short charging time, a good capacity utilization, and a high energy
efficiency, while maintaining a long cycle life [1,2]. Before
introducing the different categories of charging protocols, the
basic limitations for charging lithium-ion batteries are presented
as described in Ref. [3]: the charging process of lithium-ion cells is
mainly limited by two factors: lithium plating on the anode and
oxidation of the electrolyte solution due to high potentials at the
cathode [4,5]. Both undesired side reactions lead to an irreversible
loss of cyclable lithium. Moreover, they consume electrolyte
components and promote the growth of resistive surface layers
[4,6,7].
Lithium plating describes the reduction of Li+ ions, which are
dissolved in the electrolyte, to metal lithium at the surface of the
anode’s active material. This reactions takes place instead of the
regular intercalation of lithium into the lattice structure of the
active material [8]. It can originate from limitations in charge
transfer or lithium solid diffusion [9,10]. Lithium plating can occur,
when the anode potential drops below the standard potential of
Li+/Li [9]. Some of the plated lithium later reacts irreversibly with
the electrolyte and forms insoluble side products [6,7]. Graphite
anodes, which are used in most lithium-ion cells, are very prone to
lithium plating due to their low equilibrium potential, especially at
high state of charge (SoC) [11]. As a general trend, lithium plating
increases with higher SoC, higher charging current, and reduced
temperature [9,12]. Moreover, the intercalation of lithium into the
graphite anode causes volume changes and mechanical stress,
which can lead to battery aging [13,14]. As shown in Ref. [15], aged
cells can become more susceptible to lithium plating. All in all, the
charging currents for graphite-based lithium-ion cells are mainly
limited by the intercalation kinetics at the anode [16].
The charging voltage is limited by the oxidation of electrolyte
solvents, which occurs at high cathode (over)potentials [4,5].
Overcharging a lithium-ion cell promotes heat generation and
causes irreversible damage to the cathode’s crystallographic
structure, when the cathode material is completely delithiated
Fig. 1. Schematic illustration of different charging protocols: (a) CCCV, (b) MSCC, (c) CCPC, (d) PC, (e) BC, and (f) charging with a predefined voltage trajectory.
[17]. This leads to further oxidative side-reactions that can entail
gas evolution, overpressure inside the cell, an opening of the cell’s
safety vent, and leakage of electrolyte. As the organic electrolytes
of lithium-ion batteries are highly flammable, this can cause a fire
or lead to an explosion of the cell [18]. Thus, complying with the
maximum cell voltage, specified by the manufacturer, is essential.
Consequently, any charging procedure for lithium-ion batteries
has to consider these fundamental limitations to achieve safe
operation and good cycle life. The following subsections present
different categories of charging protocols for lithium-ion batteries
and their specific characteristics.
2.1. Constant current constant voltage (CCCV) charging
The standard charging protocol for lithium-ion cells is CCCV
charging [19]. Fig. 1a illustrates the two phases of CCCV charging:
at first, the cell is charged with a constant current Ich, until the cell
voltage reaches the specified charging voltage Vch. Then, the cell
voltage is kept constant at Vch, entailing a continuous reduction of
the charging current. This constant voltage (CV) phase is
terminated, when the charging current drops below a predefined
threshold value Iend or when a predefined maximum charging time
tmax is exceeded. The speed of the charging process is mainly
influenced by Ich; the capacity utilization is determined by Vch and
Iend. As demonstrated in Ref. [20], higher charging voltages and
higher charging currents can deteriorate cycle life considerably.
Thus, it is necessary to choose Ich and Vch appropriately to
minimize lithium plating and electrolyte decomposition.
Setting Iend = Ich results in sole constant current (CC) charging,
which is considered as a special case of CCCV charging within this
paper. Due to the missing CV phase, the capacity utilization for sole
CC charging is generally lower. High capacity utilization with sole
CC charging can only be achieved with very low charging currents.
As this extends charging times massively, CCCV charging is usually
preferred to CC charging.
2.2. Multistage constant current (MSCC) charging
In MSCC charging (Fig. 1b), the CV phase of the CCCV protocol is
replaced by a series of CC periods with monotonic decreasing
charging currents (Ich1 > Ich2 > . . . > IchN) [21,22]. Each time the cell
voltage reaches the charging voltage Vch, the charging current is
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
reduced to the next level. The charging process is terminated,
when Vch is reached at the lowest current stage IchN. The MSCC
protocol reduces implementation costs, as no additional circuitry
or software algorithms for voltage regulation are required. The
speed of the charging process and the capacity utilization can be
adjusted by the same means as for CCCV charging. Due to the
stepwise reduced charging current, MSCC charging is somewhat
slower than CCCV charging with the same initial and final charging
currents (Ich1 = Ich,IchN = Iend).
2.3. Pulsed charging
As an alternative to CC or CV charging, several pulse charging
profiles can be found in literature. They base on periodic changes in
amperage and/or direction [23]. The charging current can be
reduced, interrupted, or replaced by short discharging pulses for a
certain period of time. In this paper, two categories of pulsed
charging protocols are distinguished: charging protocols where
the CV phase from CCCV charging is replaced by pulsed charging
and charging protocols consisting solely of pulsed charging.
2.3.1. Constant current + pulsed charging (CCPC)
In CCPC charging (Fig. 1c), pulsed charging replaces the CV
phase of CCCV charging [24]. The lithium-ion cell is charged with a
constant current Ich, until a predefined voltage level Vswitch is
reached. Then, the charging mode changes to pulsed charging:
charging pulses of amperage Ich and duration tpulse are applied to
the cell. After each pulse, the current flow is interrupted, until the
cell voltage has dropped below Vfloat again. Usually, a minimum
pause length tpause,min is also specified. In most implementations,
Vswitch and Vfloat are identical. With increasing SoC, the pauses
become longer. Typical values for minimum pulse and pause
lengths lie between 0.1 s and 0.5 s [25]. The charging procedure is
terminated, when the pause after a charging pulse exceeds a
predefined time span tpause,max.
This charging protocol can be implemented at very low cost, as
it requires neither voltage control nor variable charging currents.
Capacity utilization and cycle life depend on Vfloat and tpause,max:
setting Vfloat close to the maximum voltage specification (Vmax) of
the lithium-ion cell and increasing tpause,max maximizes capacity
utilization. However, this can deteriorate cycle life, as the charging
pulses lead to an exceedance of Vmax for a certain period of time.
2.3.2. Pulse charging (PC)
In addition to CCPC protocols, there are also the PC protocols,
where the entire charging procedure is performed by charging
pulses [1,23]. Fig. 1d illustrates such a PC profile, where the current
alternates between Ihigh and Ilow. The time period ttotal = thigh + tlow
and the duty cycle D = thigh/ttotal define the shape of the pulse
profile. All parameters can also change during the charging
process. Reaching a predefined charging voltage Vend terminates
the charging procedure. The combination of Ihigh and Vend
determines the capacity utilization. In Ref. [23], pulse frequencies
range between 100 Hz and 0.2 Hz.
The motivation of PC protocols is to reduce Li+ gradients and
eliminate concentration polarization [1]. Some studies report
beneficial effects of PC for lithium-ion batteries, such as reduced
diffusion resistance, better active material utilization, improved
cycle life, and reduced charging time, as a CV phase becomes
redundant [1,5,10]. Other studies, however, showed that PC has no
or even detrimental effects on the performance and cycle life of
lithium-ion batteries [23,26]. Numerical simulations based on a 1D
physicochemical model indicated that pulse charging with fixed
frequency and fixed duty cycle does not have any advantages over
CC charging with the same mean charging current [10,27]. Another
study showed that for an identical mean current, losses rise and
127
efficiency decreases with an increasing deviation of the charging
current from a CC profile [23]. In Refs. [27,28], specific PC profiles
with low or decreasing charging currents are presented that reduce
battery aging. However, it is not disclosed whether the improved
cycle life is a benefit from PC or simply a result of the reduced mean
charging current or reduced cycle depth of these profiles.
Furthermore, no long-term cycle life experiments were performed
in these studies.
2.4. Boost charging (BC)
BC is derived from CCCV charging and features an additional CC
or constant power interval at the beginning of the charging process
[29]. This additional boost interval shall reduce charging time
without deteriorating cycle life, as the batteries are less susceptible
to lithium plating at lower SoC. In Fig. 1e, the cell is charged with a
high current Iboost in the beginning, until a substantial amount of
charge is transferred into the cell. The boost interval is either
limited by a time value tboost or a maximum voltage Vboost. It can
also be implemented as a CCCV period defined by these three
parameters. After the boost interval, the charging procedure
changes to a conventional CCCV protocol with the parameters Ich,
Vch, and Iend. The charging speed can be adapted by Iboost and tboost.
Moreover, the same rules for adjusting capacity utilization and
charging speed as for CCCV charging apply here.
An experimental study showed that a lithium-ion cell can be
charged substantially faster with a BC protocol, while providing the
same cycle life as a cell that is charged with a standard CCCV
protocol without the high boost current in the beginning [29].
Another study also attributes short charging time and good cycle
life to a BC protocol [30]. However, none of the studies compared
the BC protocol with a fast-charging CCCV protocol where
Ich = Iboost. Hence, it remains unclear whether reducing the
charging current at higher SoC is beneficial for the cells’ cycle
life or whether the cells used in these studies were not susceptible
to high charging currents and could have been charged with a high-
current CCCV protocol as well.
2.5. Other charging protocols
This last section gives an overview on some more exotic
charging protocols proposed in literature. It also comprises
adaptive procedures, which adjust the charging currents depen-
dent on the properties of the lithium-ion cell at its actual condition.
The main characteristics of these charging strategies are presented
in the following paragraphs.
Several charging procedures are based on varying-current
profiles, which start with high charging currents and decrease
currents with increasing SoC. Optimization methods are used to
achieve fast charging without exceeding certain voltage,
temperature, stress, or concentration limitations [31–33]. The
charging profiles are either determined experimentally [34] or
derived from simulation models to obtain the maximum
charging current which is applicable without provoking lithium
plating [9,35–38]. Such protocols have hardly been used in
practical applications, as determining the maximum charging
currents appears to be challenging, since these currents vary
considerably with temperature and degradation of the cell. As
precise information about the actual polarization or concentra-
tion gradients inside the cell are necessary, additional state
estimation routines are required to determine internal variables
of the cell [39], which are also influenced by the short-term and
long-term load history.
Moreover, there are some charging protocols featuring a lower
charging current in the beginning of the charging process, as the
internal resistances generally have the highest values at low SoC. A
128 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141

period with a low charging current or steadily rising current is used

to minimize losses at very low SoC (Fig. 1f) [4,40].
Furthermore, Ref. [40] proposes a charging protocol that is no
longer based on predefined charging currents, but uses a voltage
trajectory. This trajectory, which is the result of charging a new cell
with a reference current profile, remains constant over the entire
battery life. Thus, charging currents decrease with proceeding
capacity fade, while the charging time remains constant.
Many of the studies presenting more exotic charging strategies
provide only few experimental data for validation. Furthermore,
there are also various investigations on charging lithium-ion
batteries which are solely based on simulation and present no
experimental validation for the charging process (e.g., [41–44]).

3. Experimental

Although there are a wide variety of publications on charging

protocols for lithium-ion batteries, it is hardly possible to find
experimental data on performance and cycle life under compara-
ble experimental conditions. As such data is essential for an
objective assessment and comparison of charging protocols, we Fig. 2. (a) Discharging curves for low-current discharging with 200 mA. (b)
have performed an extensive experimental study with a represen- Differential voltage spectra with characteristic graphite peak highlighted.

tative set of charging protocols and different lithium-ion cells. This

section introduces the lithium-ion cells and charging protocols
examined in our study. Furthermore, the test procedure is
described.
3.1. Lithium-ion cells
In our experimental study, high-power lithium-ion cells with a
nominal capacity CN of about 1 Ah have been examined. Different
cell types have been considered by selecting three 18650 cell
models from different manufacturers and with different cell
chemistry to identify interdependencies between cell types and
charging protocols. Table 1 lists characteristic properties of the
three cell models. For the cell voltages, explicit maximum and
minimum values were specified in the datasheets. For the charging
currents, however, no maximum values ware specified in the
datasheets; only standard charging currents were provided. For
cell models B and C, the datasheets also disclosed a fast charging
current.
The three cell models all contain graphite-based anodes but
different cathode materials. Fig. 2a depicts low-current discharg-
ing curves for the three cell models, which reveal the differences in
the voltage characteristics for the LFP-based cell model C: cell
model C exhibits a relatively constant potential of about 3.3 V over
a wide SoC range; the cell voltage changes considerably only at
very high and very low SoC. This results from the rather constant
open circuit voltage (OCV) curve, typical for LiFePO4 (LFP)-based
lithium-ion cells [45]. At a discharged capacity of about 0.29 Ah,
there is a small potential step, which is typical for graphite and
leads to a characteristic peak in the spectrum obtained by
differential voltage analysis [46], depicted in Fig. 2b. This peak
indicates a degree of lithiation of approximately 50% of a graphite
anode [47]. Since this peak is located at about 30% of the total
capacity for all three cell models examined, they all exhibit a
similar oversizing of the graphite anode: the anodes are
completely delithiated at the end of the discharging process (0%
SoC) and half lithiated at about 70% SoC; Hence, the fully lithiated
anodes would correspond to 140% SoC, which means that the
anodes are all oversized by about 40%.
3.2. Charging protocols
For the experimental investigations on different charging
strategies, a representative set of charging protocols was
composed. This set comprises CCCV, CCPC, PC, and BC protocols.
MSCC charging was not taken into further consideration, as it does
not differ considerably from CCCV charging. From the more exotic
charging strategies, the following two basic principles were
incorporated in adapted BC protocols: Using a lower charging
current in the beginning; and applying higher boost currents at
specific SoC intervals. For the sake of comparability, identical

Table 1
Characteristic properties of the examined lithium-ion cell models.

Model A Model B Model C

Manufacturer Sanyo Sony A123
Model UR18650SA US18650VT1 APR18650M1A
Nominal capacitya CN 1.25 Ah 1.1 Ah 1.1 Ah
Maximum voltagea Vmax 4.2 V 4.1 V 3.6 V
Minimum voltagea Vmin 2.75 V 2.5 V 2.0 V
Discharging currenta Imax 10 A 10 A 30 A
Standard charging currenta Istandard 0.9 A 1A 1.5 A
Fast charging current Ifast not provided 4A 4A
Anode materialb Graphite Graphite Graphite
Cathode materialb LiMn2O4 + LiNiCoMnO2 LiNiCoMnO2 + LiCoO2 LiFePO4
a
Values from datasheets.
b
Information from suppliers, not verified.
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141 129

Table 2 voltages from Vmax 500 mV to Vmax + 50 mV with a charging

Parameter combinations for the CCCV charging protocols used in the cycle life
current of 3 A.
study.

Protocol ID Ich Vch Iend tmax 3.2.2. Pulsed charging protocols

Current variations 1A CCCV 1A Vmax 100 mA – One CCPC protocol and two PC protocols were defined. To
3A CCCV 3A Vmax 100 mA – achieve good comparability with the baseline 3A CCCV protocol,
5A CCCV 5A Vmax 100 mA –
the same (mean) charging currents were used: for the CCPC
Voltage variations 3A CCCV +50 mV 3A Vmax + 50 mV 100 mA –
protocol, Ich was set to 3 A; for both PC protocols, the charging
3A CCCV 3A Vmax 100 mA – currents Ihigh = 5 A and Ilow = 1 A were used. Moreover, rectangular
3A CCCV 50 mV 3A Vmax 50 mV 100 mA – current pulses with a constant duty cycle D = 50% were specified in
3A CCCV 100 mV 3A Vmax 100 mV 100 mA – the PC protocols to identify the direct impact of charging pulses.
3A CCCV 200 mV 3A Vmax 200 mV 100 mA –
This avoids distortions due to different mean charging currents
originating from reduced charging currents or increased pause
lengths. Two pulse frequencies in the middle of the frequency
range examined in Ref. [23] were considered in our PC protocols:
(mean) charging currents were used wherever applicable. Fur-
1 Hz and 25 Hz. As Ref. [23] had observed that the impact of the
thermore, parameter variations were performed for the different
pulse frequency on the cell performance differed above and below
charging protocols to identify the key parameters for optimizing
approximately 10 Hz, one pulse frequency was defined below and
the charging process.
one above that frequency value. Table 3 lists the parameters of the
In our test protocols, all currents were defined as absolute
CCPC protocol and Table 4 lists the parameters of the two PC
ampere values and not as C-rates, which means in proportion to
protocols. To obtain a good capacity utilization, Vswitch and Vfloat
the cell’s nominal capacity. C-rates are beneficial, when comparing
were set only 50 mV below Vmax for the CCPC protocol and Vend was
cells with similar electrode design but different sizes, because in
set 50 mV above Vmax for the PC protocols. Higher voltage values
such cases, C-rates lead to similar current densities and stress for
were not chosen, as this would prolong the time periods in which
the electrode materials. In our study, however, cells of similar size
the cell voltage exceeds Vmax during and after the charging pulses.
but different electrode design, i.e., cathode material, tab positions,
The CCPC charging ends as soon as the pause interval is longer than
porosity, and electrode thickness, are compared. In that case,
10 times the duration of the charging pulse. This was set in
C-rates are not able to guarantee similar current densities and
accordance to Ref. [25], where the phase of pulsed charging with
stress for the electrode materials. Furthermore, in practical
increasing pause intervals ends, when the mean charging current
applications using cells of a specific form factor, such as the
has dropped below 10% of its initial value.
18650 format, the number of cells in the battery system is often
To examine the capacity utilization of the CCPC and PC
predetermined by the construction space available. Consequently,
protocols, the charging pulses have always been followed by a
the load per cell is also predefined and, thus, independent from the
CV charging period identical to that of the baseline 3A CCCV
cell model and its capacity. As battery aging often increases with
charging protocol (Vch = Vmax, Iend = 0.1 A). This reveals the amount
higher currents [48,49], cells with a higher capacity would get
of capacity remaining unused, which is required for an assessment
penalized in the cycle life study for their higher charge content, if
of the capacity utilization. Moreover, it ensures similar cycle
the currents were defined in proportion to the capacity. For these
depths for the cycle life investigations.
reasons, we expect identical absolute ampere values to provide a
more representative comparison for the different 18650 cells
3.2.3. BC protocols
examined.
In the category of BC, three protocols were designed to cover the
impacts of high charging currents at the very beginning and at two
3.2.1. CCCV protocols
somewhat later stages of the charging process. Again, current
As the CCCV protocol is the standard charging protocol for
values from the CCCV protocols were reused for the BC protocols:
lithium-ion batteries, it serves as a baseline in our study. For all
Iboost = 5 A and Ich = 1 A. This provides information about the impact
three cell models examined our study, the CCCV protocol is the
of mean and maximum charging current on the cycle life of
charging procedure recommended by the manufacturer. Extensive
lithium-ion cells. The duration of the boost interval was set to 300 s
parameter variations were performed for the charging current Ich
for the cell models B and C, and to 360 s for cell model A, due to the
and the charging voltage Vch. Table 2 lists all combinations
about 20% higher capacity. Hence, about 40% of the cell’s nominal
investigated in the cycle life study, comprising three charging
capacity CN is charged during the boost interval. In total, the mean
currents from 1 A to 5 A and five charging voltages between
charging current of the CC and boost phases lies below 3 A. As Ref.
200 mV below and 50 mV above the cell’s maximum voltage
[29] showed that exceeding Vmax by 100 mV during the boost
specification Vmax (see Table 1). As no explicit maximum charging
interval did not have any negative impact on cycle life, Vboost is also
current was provided in the cells’ datasheets, we chose a maximum
slightly increased to Vmax + 50 mV in our cycle life study. In order to
charging current of 5 A for our study, which lies somewhat above
investigate the impact of high charging currents at different SoC
the fast charging currents specified in two of the three datasheets.
intervals, two variations of the protocol were defined, where the
In addition to the parameter combinations listed in Table 2, the
boost interval starts after charging 10% or 20% of CN with the low
impact of Vch on capacity utilization and charging duration has
charging current Ich. Table 5 summarizes the parameters of the
been examined by charging a new and an aged cell to 12 different
three BC protocols.

Table 3 Table 4
Parameter set for the CCPC protocol. Parameter sets for the PC protocols.

Protocol ID Ich Vswitch Vfloat tpulse tpause, tpause,max Protocol ID Ihigh Ilow D f thigh tlow Vend
min
PC 25 Hz 5A 1A 50% 25 Hz 20 ms 20 ms Vmax + 50 mV
CCPC 3 A Vmax 50 mV Vmax 50 mV 400 ms 100 ms 4000 ms PC 1 Hz 5A 1A 50% 1 Hz 500 ms 500 ms Vmax + 50 mV
130 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141

Table 5
Parameter sets for the BC protocols.

Protocol ID Ich Iboost Vboost tboosta AhStartBoost Vch Iend

BC 0–40% 1A 5A Vmax + 50 mV 360 s or 300 s 0 Ah Vmax 100 mA
BC 10–50% 1A 5A Vmax + 50 mV 360 s or 300 s 0.1 CN Vmax 100 mA
BC 20–60% 1A 5A Vmax + 50 mV 360 s or 300 s 0.2 CN Vmax 100 mA
a
The longer boost interval is applied for cell model A due to its higher capacity.

3.3. Test procedure
The experimental investigations on charging protocols for
lithium-ion batteries have been performed with a BaSyTec CTS
battery test system. As the capacity and internal resistances of a
lithium-ion cell depend notably on temperature, thermal cham-
bers have been used to establish a constant environmental
temperature of 25  C. In this paper, the focus is on charging at
this moderate temperature. Since lithium plating is sensitive to
temperature, the impact of higher and lower operating temper-
atures on charging lithium-ion batteries will be investigated in a
separate study. The following sections describe the cycle life test
procedure, comprising the repetitive cycling and the degradation
monitoring by checkups and electrochemical impedance spectros-
copy (EIS).
3.3.1. Repetitive cycling
To determine the impact of different charging strategies on
cycle life, the following repetitive cycling procedure has been
performed with the various charging protocols defined above:
each cell is discharged with a constant current Idis = 3 A until the
cell’s minimum voltage Vmin (see Table 1) is reached. After a pause
of 15 min, the cell is charged again with its assigned charging
protocol. When the charging process is completed, there is another
pause of 15 min, before the next discharging and charging
sequence begins.
In a preliminary inspection, the cycling procedure was executed
with a higher Idis of 5 A and shorter pauses of only 10 min. As the
cell temperatures were not able to relieve to environmental
temperature during the pauses, a lower Idis and longer pauses were
defined in the main study. A comparison of the results from the
preliminary study and the main study is also shown in this paper,
as this reveals the impact of higher discharging currents on the
cycle life of the cells.
The circle markers indicate ohmic behavior, which means that
Im{Z} is zero and Z consists only of a real part. The value of this real
part is named Ri and is influenced by the electrolyte conductivity
and resistances from metal conductors, e.g., the current collectors.
In the cycle life study, Ri is evaluated to identify electrolyte
degradation, as Ri rises with decreasing electrolyte conductivity
[50,51]. Ri is computed by interpolation between the two
neighboring frequency points above and below the x-axis.
The lower part of the spectrum represents inductive behavior,
which usually remains unchanged with aging. The upper part of
the spectrum comprises a distorted semicircle and a sloping line at
the right end. The semicircle covers effects from passivation layers
and charge transfer resistances at both electrodes as well as double
layer capacities [52,53]. The sloping line in the low-frequency
domain represents limitations in mass transport due to diffusion
processes [54]. The square markers indicate the cell impedance at
10 mHz. The real part of this impedance is named Rac,10 mHz and is
used in the cycle life study as a more comprehensive indicator for
changes in the internal resistances, as it also covers effects from
charge transfer and diffusion.
Cell models A and B exhibit a more pronounced inductive
behavior than cell model C, since the imaginary parts of their
impedances at 10 kHz reach down to about 55 mV. This indicates
differences in the geometric design of the jelly roll. In particular,
the position of the tabs, connecting the current collector foils and
the poles of the cell, varies. A higher inductance correlates with a
longer distance through the spiral windings of the jelly roll
between both tabs [55]. Opening the cells has revealed that cell
models A and B have the negative tab at the outermost winding
and the positive tab in the inner third of the jelly roll. By contrast,
cell model C has both tabs in the middle of the jelly roll. Hence,
there is only a very short distance between both tabs, which
explains the substantially lower Ri: according to the computations

3.3.2. Checkups
Before the cycling procedure and after every 100 cycles of
discharging and charging, each cell has performed a checkup. The
checkups are used to determine the remaining capacity with the
same measurement procedure for all cells. The remaining capacity
is measured by CC discharging with 1 A after charging the cell with
the 1A CCCV protocol and a pause of 15 min. In the end of the
checkup, the cell is charged to 50% of its actual capacity with a
constant current of 1 A. After every second checkup, the cells
perform an EIS measurement at this SoC, before returning to the
repetitive cycling procedure again.

3.3.3. Electrochemical impedance spectroscopy

To monitor changes of the internal resistances, EIS measure-
ments are conducted after every second checkup, i.e., every
200 cycles. The EIS measurements are performed as galvanostatic
measurements over a frequency range from 10 kHz to 10 mHz with
a Gamry G750 galvanostat/potentiostat at 50% SoC. The excitation
current was defined with Iac,RMS = 100 mA and no offset (Idc = 0 mA).
Fig. 3 shows the impedance spectra of the three different cell
models in new condition. Cell model C exhibits the lowest, whereas
cell model B exhibits the highest internal resistances.
Fig. 3. Impedance spectra of new cells at 50% SoC, ranging from 10 kHz to 10 mHz.
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
of Refs. [56,57], two tabs in the middle of the jelly roll lead to
substantially lower ohmic resistances than two tabs at opposite
ends of the electrodes.
3.3.4. End of life
The aging of lithium-ion batteries usually comprises the loss of
capacity and the reduction in performance due to increasing
internal resistances. The end of life (EoL) for a battery is often
defined as a capacity loss of more than 20% or an increase of the cell
resistance by more than 100% [58,59].
For each combination of cell type and charging protocol
examined in the cycle life study, a separate new cell has been used.
Preliminary inspections revealed that the cell-to-cell variations of
the mass-produced cells were low and could be neglected
compared to the differences in cycle life caused by different
charging protocols. The target number of discharging and charging
cycles was 1200. For those cells that have not reached 1200 cycles,
the test procedure was terminated at a capacity loss greater than
30%, which means considerably beyond the conventional EoL.
4. Results
The results of our experimental investigations on charging
protocols for lithium-ion batteries provide information on
charging time, capacity utilization, and efficiency for different
CCCV, CCPC, PC, and BC protocols. Moreover, the cycle life study
with up to 1200 discharging and charging cycles discloses the
impact of the charging protocol on battery aging. As all data are
obtained under identical environmental conditions, they enable an
objective comparison and assessment of different charging
strategies for lithium-ion batteries. Moreover, some results from
preliminary inspections are also presented, as they allow separat-
ing the impact of high discharging currents and high charging
currents on cycle life.
4.1. Impact of high discharging currents
To analyze the impact of high discharging currents and high
charging currents on cycle life, data from a preliminary inspection
with Idis = 5 A and results from the main study with Idis = 3 A are
compared. From both studies, the capacity fade for 500 cycles of
the 3A CCCV and the 5A CCCV protocol are depicted in Fig. 4. This
reveals different aging characteristics for the three cell models:
Cell model A exhibits faster aging with 5 A charging than with 3 A
charging, independent from the discharging current (Fig. 4a). Cell
model B shows almost no dependency on the charging and
discharging currents at all (Fig. 4b). Cell model C exhibits a
substantially lower aging for charging with 3 A, which is again
independent from the discharging current (Fig. 4c). Only for 5 A
charging, which causes a massive degradation compared to 3 A
charging, considerable differences in the aging progress can be
observed for cell model C. This effect of a faster capacity fade for
discharging with 3 A instead of 5 A has been repeatable. The reason
for the accelerated capacity fade, however, remains unclear.
Comparing the measurement data from both protocols reveals
an about 1% smaller depth of discharge for discharging with 5 A
and an about 5  C higher cell temperature at the end of the
discharging process. The impact of the beginning of a high-current
charging process will be further examined in the boost charging
section.
All in all, the different combinations of discharging and
charging currents show that if the cycle life of a lithium-ion cell
is affected by amperage, the charging current has the main impact
on cycle life.
131
4.2. CCCV protocols
As CCCV charging is the standard strategy for charging lithium-
ion batteries, comprehensive investigations with several variations
of CCCV protocols have been performed to identify the effects of
charging current and charging voltage on the performance of CCCV
charging.
4.2.1. Variation of the charging current
The main parameter to adjust the charging speed of CCCV
charging is the charging current Ich. In our study, its impact on
charging time, capacity utilization, and cycle life has been
analyzed.
4.2.1.1. Charging time. For charging currents of 1 A, 3 A, and 5 A,
Fig. 5 depicts the voltage and capacity curves during CCCV charging
for new cells. Table 6 discloses the explicit duration of the CC phase
and the entire CCCV charging process for the different charging
currents and cell models. The difference of both time values
represents the duration of the CV phase. For cell models A and C,
most of the capacity is charged within the CC phase; the CV phase
lasts only about 10–15 min. Cell model B, however, exhibits a
longer CV phase, which lasts between 25 min and 35 min. For all
three cell models, the total charging duration can be shortened by
about 50%, when Ich is increased from 1 A to 3 A. A further increase
of Ich to 5 A yields a considerable acceleration only for cell models A
and C: for these cells, the charging duration can be further reduced
by about one third. For cell model B, however, the dominant CV
phase prevents a further acceleration of the charging process with
Ich = 5 A compared to Ich = 3 A, as only about half of the capacity can
be charged during the constant current phase.
4.2.1.2. Capacity utilization. Fig. 5d–f shows the charged amount of
ampere-hours over time for the three charging currents. The
explicit values for the CC phase and the entire charging process are
also listed in Table 6. For all cell models, the differences in charged
capacity lie below 0.5%. This indicates that the capacity utilization
for CCCV charging is rather independent from the charging current
for all three high-power lithium-ion cells investigated, when the
values for Vch and Iend are identical.
4.2.1.3. Cycle life. The impact of Ich on battery aging is depicted in
Fig. 6. For all cell models examined, the capacity fade accelerates
with higher Ich. However, the extent of this acceleration varies
substantially among the cell models.
For cell model A (Fig. 6a), higher charging currents reduce cycle
life substantially: a capacity loss of 20%, indicating the EoL, is
reached after less than 600 cycles for the 5A CCCV protocol, after
about 800 cycles for the 3A CCCV protocol, and after more than
1000 cycles for the 1A CCCV protocol. Hence, reducing Ich can
prolong the cycle life markedly for this cell model.
Fig. 4. Cycle life for different combinations of moderate (3 A) and high (5 A)
charging currents Ich and discharging currents Idis.
132 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141

Fig. 5. Charging progress for new cells charged with CCCV protocols with different charging currents: (a–c) cell voltage, (d–f) charged capacity.

In contrast to cell model A, the cycle life of cell model B exhibits
almost no dependency on the charging current (Fig. 6b). Compar-
ing the 5A CCCV protocol and the 1A CCCV protocol reveals that the
difference in remaining capacity is only about 2–3 percentage
points after 1200 cycles. The difference in cycles, when reaching
the EoL, amounts only to about 100 cycles, which is less than 10% of
the cells’ cycle life.
The LFP-based cell model C provides the best cycle life in our
study (Fig. 6c). For the 1A CCCV protocol, about 97% of the initial
capacity is still available after 1200 cycles. With higher charging
currents, however, battery aging aggravates: The 3A CCCV protocol
reduces the capacity available after 1200 cycles to about 95%;
increasing Ich even further to 5 A leads to a pronounced
degradation and the EoL is reached within less than 700 cycles.
These results indicate that lithium plating increases considerably
with the higher charging currents and causes a massive degrada-
tion for 5 A charging.
4.2.2. Variation of the charging voltage
The second parameter of the CCCV protocol varied in our study
is Vch. A lower value for Vch is often used to reduce cell aging related
to elevated voltage potential. This section shows how capacity
utilization, duration of the CC and CV phase, and cycle life are
affected by Vch.
4.2.2.1. Capacity utilization. For all three cell models, the capacity
utilization in relation to Vch has been examined with a new and an
aged cell. For the aged cell, a specimen from the repetitive cycling
procedure was selected which had a capacity loss between 20% and
25%. Each cell has been charged to 12 different Vch values with an
identical Ich of 3 A and Iend of 100 mA. Fig. 7a–c illustrates the
capacity utilization for the six cells examined. For cell models A and
B, there is a clear decrease of capacity utilization with reduced Vch:
lowering Vch by 100 mV reduces the capacity utilization between
10% and 20%. Regarding the charged ampere-hours during the CC
phase only, an almost linear relationship with Vch can be observed.
The CV phase, however, exhibits certain variations and does not
monotonically increase or decrease with Vch. These variations are
rather similar for new and aged cells. For aged cells, a certain
increase of the ampere-hours charged during the CV phase can be
observed.
The LFP-based cell model C exhibits a special characteristic in
capacity utilization: due to the flat OCV curve of LFP-based cells,
only minor variations in capacity utilization occur between 3.4 V
and 3.65 V charging voltage. Yet, the ratio between the ampere-
hours charged in the CC phase and in the CV phase varies, as lower
charging voltages lead to earlier CV phases, but slower reductions
of the charging current. Below 3.4 V, a steep decrease in capacity
utilization can be observed. For Vch = 3.2 V or lower, the available
capacity approaches zero.
4.2.2.2. Charging time. The charging times for different charging
voltages are depicted in Fig. 7d–f. For all three cell models, the
relation between Vch and the total charging time is not monotonic.

Table 6
Duration and charged capacity of the CC phase and the total charging procedure for different CCCV protocols and cell models.

Cell type Charging protocol CapCC (Ah) Captotal (Ah) tCC (min) ttotal (min)
Model A 1A CCCV 1.194 1.263 71.6 81.1
3A CCCV 1.048 1.266 21.0 35.4
5A CCCV 0.909 1.261 10.9 26.0

Model B 1A CCCV 0.824 0.983 49.4 77.2

3A CCCV 0.649 0.986 13.0 46.6
5A CCCV 0.494 0.985 5.9 41.6

Model C 1A CCCV 0.988 1.017 59.3 66.1

3A CCCV 0.955 1.017 19.1 26.4
5A CCCV 0.825 1.018 9.9 19.7
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
Fig. 6. Cycle life for CCCV protocols with different charging currents.
For aged cells, the duration of the CC phase decreases, whereas the
duration of the CV phase increases notably. For cell models A and B,
the total charging times of a new and an aged cell are rather similar.
Cell model C exhibits prolonged CV phases for charging voltages
that lie only slightly above the LFP plateau voltage of 3.3 V. For
Vch = 3.4 V, the CV phase covers about 95% of the total charging time
and the charging time is about twice the time of Vch = 3.6 V (Fig. 7f).
This reveals that the CV phase can become longer with lower Vch. In
contrast to cell models A and B, the total charging duration of cell
model C shortens with aging, as the duration of the CV phase does
not increase notably with aging.
4.2.2.3. Cycle life. For the investigations on cycle life, five 3A CCCV
protocols with different charging voltages are compared. Fig. 8
illustrates the impact of Vch on cycle life. The markers indicate the
checkups, which have been performed every 100 cycles. As
capacity utilization and, thus, cycle depth varies with Vch, the
capacity fade is plotted versus equivalent full cycles to provide a
better comparability. Equivalent full cycles express the charge
throughput as multiples of the cell’s nominal capacity CN. As
expected, the capacity fade generally increases with higher Vch.
However, the sensitivity of the different cell models’ aging
behavior to Vch varies markedly.
For cell model A (Fig. 8a), only about 100 equivalent full cycles
more can be achieved at the EoL when lowering the charging
voltage by 100 mV, i.e., from 4.2 V to 4.1 V. A reduction of 200 mV is
necessary to obtain a considerable improvement in cycle life.
However, this correlates to a reduction of usable capacity of about
Fig. 7. (a–c) Capacity utilization and (d–f) charging duration for CCCV protocols with different charging voltages (Ich = 3 A, Iend = 100 mA).
133
30% (see Fig. 7a). Overcharging by 50 mV yields only a minor
decrease in cycle life.
A greater sensitivity to Vch is observed for cell model B (Fig. 8b):
Lowering Vch by 100 mV, i.e., from 4.1 V to 4.0 V, reduces the
capacity fade after 800 equivalent full cycles by more than 5
percentage points. The capacity degradation for Vch = Vmax 50 mV
illustrates that the first 50 mV of voltage reduction have a higher
impact on enhancing cycle life than the second 50 mV. A reduction
of Vch by 200 mV to 3.9 V has not decreased battery aging any
further. Overcharging with the 3A CCCV + 50 mV protocol accel-
erates aging notably for this cell model: The slight exceedance of
the nominal charging voltage Vmax by 50 mV already increases the
capacity fade considerably.
The highest sensitivity against overcharging can be observed for
cell model C (Fig. 8c). Charging to 50 mV above Vmax aggravates
aging and cycle life diminishes massively. For charging voltages up
to Vmax, cell model C has revealed the best cycle life of all three cell
models. Although the capacity utilization of this cell model does
not vary notably for voltage reductions up to 200 mV (see Fig. 7c), a
certain improvement in cycle life can be observed when lowering
Vch by 50 mV or 100 mV. When reducing Vch by 200 mV to 3.4 V,
cycle life further improves and approaches the low aging of the 1A
CCCV protocol, as 96.9% of the initial capacity is still available after
the entire test procedure with 1200 cycles. Fig. 9 discloses that this
lowering of Vch entails a substantial reduction of the charging
current due to the early onset of the CV phase. Thus,
the improvements in cycle life can also be a result of the
lower charging currents at higher SoC. The charging time of the
134 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
Fig. 8. Cycle life for 3A CCCV protocols with different charging voltages. Equivalent full cycles express the charge throughput as multiples of the cell’s nominal capacity.
3A CCCV 200 mV protocol, which lies above 60 min, is also rather
similar to the 1A CCCV protocol.
4.3. Pulsed charging
To elaborate on the advantages and disadvantages of pulsed
charging, one CCPC and two PC charging protocols have been
examined. Capacity utilization, charging time, and cycle life are
evaluated.
4.3.1. CCPC protocol
The CCPC protocol is rather similar to the 3A CCCV protocol, as
both feature an identical CC phase in the beginning. Hence,
differences in capacity utilization, charging time, and cycle life
origin from the subsequent pulse phase.
4.3.1.1. Capacity utilization. For new cells, our CCPC protocol leads
to a capacity utilization of 92% for cell model A, 82% for cell model
B, and 98% for cell model C. As the CCPC protocol features no
reduction of the current amplitude, the charging voltage Vfloat is
always set below the maximum voltage specification Vmax to avoid
damage due to the high overpotentials during the charging pulses.
This voltage reduction leads to a lower capacity utilization. Cell
model B suffers from its high internal resistances, which create
higher overpotentials and, thus, lead to an earlier end of the
charging process. The high capacity utilization of model C
correlates to the specific OCV characteristic of LFP-based cells,
where a considerable voltage rise occurs only at the very end of the
charging process.
4.3.1.2. Charging time. The charging process of the CCPC protocol
is identical to the 3A CCCV protocol during the CC phase; the
charging progress of the subsequent pulse phase is somewhat
slower than that of a CV phase. Hence, after the CC phase, the CCPC
protocol always leads to a longer charging time than the 3A CCCV
protocol for the same amount of ampere-hours charged.
4.3.1.3. Cycle life. To obtain comparable cycle depths in the cycle
life study, the CCPC protocol contains an additional CV phase at the
end of each cycle. This provides a similar capacity utilization and,
thus, a similar cycle depth as for the CCCV protocols. The cycle lives
for CCPC charging are illustrated in Fig. 10, together with previous
results from CCCV protocols with different charging currents and
the results from PC protocols, which will be explained in the next
section. For the CCPC protocol, the cycle life of all three cell models
is rather similar to 3A CCCV charging, which features the same
charging current during the CC phase. This demonstrates that the
repetitive exceedance of Vmax by up to 70 mV during the charging
pulses has not led to accelerated aging. These observations differ
from the results in Section 4.2.2, where the higher charging voltage
in the 3A CCCV + 50 mV protocol accelerates aging massively.
4.3.2. PC protocols
The second category of charging protocols utilizing pulsed
charging is represented by the PC protocols, where the entire
charging procedure is performed by pulsed charging with high and
low amperage. As the aging of cell model B is rather independent
from the charging currents, the PC protocols have only been
applied to cell models A and C.
4.3.2.1. Capacity utilization & charging time. The capacity
utilization for the PC protocols lies notably below the CCCV and
CCPC protocols: It amounts to about 80% for cell model A and 95%
for cell model C, with negligible differences between 1 Hz and
25 Hz pulse frequency. Due to the high overpotentials during the
Ihigh pulses, the charging voltage Vend = Vmax + 50 mV is reached
already at a lower SoC. The charging speed is similar to the 3A CCCV
protocol, which features a charging current identical to the mean
charging current of the PC protocols.
4.3.2.2. Cycle life. As for the CCPC protocol, a CV phase is appended
to compensate for the lower capacity utilization; this ensures a
comparable cycle depth in the cycle life study. The cycle life for the
two PC protocols is also shown in Fig. 10a and c. The comparison
with CCCV protocols with different charging currents reveals that
the cycle lives of the PC protocols are similar to the 3A CCCV
protocol. Only a somewhat faster capacity fade can be observed for
the lower frequency of 1 Hz.
4.4. Boost charging
Another way of charging with changing charging currents is
investigated with the BC protocols. Again, only cell models A and C
have been investigated due to model B’s low sensitivity of cycle life
on amperage.
Fig. 9. Charging currents for LFP-based cell model C for 3A CCCV protocols with
different charging voltages.
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
Fig. 10. Cycle life for the CCPC and PC protocols in comparison to CCCV protocols with different charging currents.
4.4.1. Capacity utilization
Aside from the 5 A boost interval, the BC protocols in our study
are identical to the 1A CCCV protocol. Consequently, the capacity
utilization is similar to the CCCV protocols. As the 1A CCCV protocol
causes only minor aging, the low charging currents during the CC
phases in our BC protocols allow to identify the impact of the boost
interval on battery aging at the different SoC intervals.
4.4.2. Charging time
As a result of the low charging current during the CC phases, the
charging times are rather high: With a total charging time of about
57 min for cell model A and about 48 min for cell model C, the
charging times of new cells lie almost precisely in the middle
between the 1A CCCV and the 3A CCCV protocol (see Table 6).
4.4.3. Cycle life
Fig. 11 shows the cycle life for the BC protocols and discloses
differences in cycle life depending on the location of the boost
interval. For cell model A (Fig. 11a), charging a completely
discharged cell with the high boost current leads to a fast capacity
degradation, close to the capacity fade for the 5A CCCV protocol.
Starting the boost interval at 10% CN or 20% CN results in a better
cycle life, which lies in the region of the 3A CCCV protocol. Cell
model C also exhibits a dependence of cycle life on the location of
the boost interval (Fig. 11b): the boost intervals starting at 0% CN
and 10% CN provoke a considerably higher battery aging than the
boost interval at 20% CN. The latter case leads to a cycle life
comparable to the 3A CCCV protocol.
Although the overall charging time lies in the middle between
the 1A CCCV and the 3A CCCV protocol, the cycle life is at best in the
range of the 3A CCCV protocol for both cell models A and C. This
reveals a notable impact of the high-current boost interval on cycle
life, especially at very low SoC.
Fig. 11. Cycle life for the BC protocols with different locations of the boost interval in
comparison to CCCV protocols with different charging currents. Cell model B has not
been examined due its low sensitivity of cycle life on amperage.
135
4.5. Further results
In the previous sections, capacity utilization, charging time, and
cycle life have been presented for the different charging protocols
separately. In this section, the changes in cell impedance, identified
by EIS measurements, are presented for all protocols together.
Moreover, energy efficiency is analyzed for CCCV and PC protocols
to compare the losses for continuous and pulsed charging.
4.5.1. EIS measurements
The EIS measurements, conducted every 200 cycles, reveal
changes in the cells’ impedances. Fig. 12 contains three Nyquist
plots illustrating the changes in the impedance spectrum from a
new to an aged cell beyond the EoL. The three pairs of impedance
spectra demonstrate that the ohmic resistances of all cell models
increase with cell aging, as this causes the right-shift of the
impedance curves. Moreover, cell models A and B exhibit a
substantial enlargement of the semicircle in the mid-frequency
range. All cell models also exhibit an extension of the sloping line
in the low-frequency domain. Only in the high-frequency domain,
no differences are observed.
For all three cell models, the cells exhibit a similar correlation
between capacity fade and resistance increase, although they have
been charged with different charging protocols. Fig. 13 illustrates
the correlations between capacity fade and increase of Ri and
Rac,10 mHz. As they are rather similar for the different charging
protocols, only extreme cases are labelled in the six subplots. For all
cell models, Ri has increased by less than 3 mV, when the capacity
loss amounts to 20% (Fig. 13a–c). This corresponds to relative
increases of Ri of only about 10%, when the EoL capacity criterion is
met. For cell model C, also the increase of Rac,10 mHz, which
comprises changes of ohmic, charge transfer, and diffusion
resistances, lies below 10% (Fig. 13f). For cell models A and B, a
more pronounced rise of Rac,10 mHz can be observed as a result of the
enlarged diameter of the mid-frequency semicircle (Fig. 13d and e).
This leads to relative increases of Rac,10 mHz between 30% and 40%,
when the EoL criterion of 20% capacity loss is met. Consequently,
all cells examined have reached the EoL by the capacity criterion
and not by a resistance criterion.
Fig. 13 also shows some outliers, which do not follow precisely
the general trend between capacity loss and resistance increase.
Charging cell model A with the 5A CCCV protocol or charging cell
model B with the 3A CCCV + 50 mV protocol leads to a somewhat
slower rise of Rac,10 mHz in relation to the capacity fade than for the
other protocols (Fig. 13d, e). This is a result of the disproportion-
ately fast capacity degradation of these protocols. For charging cell
model A with the 1A CCCV protocol, a step in Rac,10 mHz after
1000 cycles can be observed. This is a side effect resulting from an
interruption of the testing procedure of this cell. After the step, the
slope of the resistance rise is again similar to the other cells. For cell
136 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141

Fig. 12. Changes of the cells’ impedance spectra with aging. For the aged cells, impedance spectra from the cycle life study were taken which correspond to a capacity loss of
20% or somewhat more. The circle markers indicate Ri, the square markers Rac,10 mHz.

model C, charging with the 5A CCCV or the 3A CCCV + 50 mV
protocol provokes fastest aging. However, even for an aged cell, the
LFP-based cell model C exhibits the lowest resistances Ri and
Rac,10 mHz in our study.
4.5.2. Energy efficiency
To identify the impact of the charging current on the losses in
the cycling process, the energy efficiency is evaluated for CCCV
protocols with different charging currents. Moreover, analyzing
the energy efficiency of the PC protocols allows a comparison of
continuous and pulsed charging. The energy efficiencies are
computed from data from the cycle life testing, where fully-
charged cells are discharged with a constant current of 3 A and
charged again with their assigned charging protocols. The
efficiency values represent the entire losses during one cycle of
discharging and charging. As the discharging procedure is identical
for all cells, the differences in efficiency of the new cells can be
attributed directly to the different charging protocols.
Fig. 14 shows the efficiency values for CCCV protocols with
different charging currents and for the PC protocols. For new cells,
the efficiency values range between 91.4% and 94.5% for cell model
A (Fig. 14a), between 89.2% and 92.5% for cell model B (Fig. 14b),
and between 89.2% and 92.3% for cell model C (Fig. 14c). The three
CCCV protocols depicted in Fig. 14 illustrate that the efficiency
decreases with higher charging currents. As expected, the losses
increase about linearly with the current. The PC protocols reveal
similar efficiencies for both pulse frequencies of 1 Hz and 25 Hz,
which lie about 0.5 percentage points below the efficiency of the
3A CCCV protocol.
For cell models A and B, the energy efficiencies of aged cells
beyond the EoL have also been evaluated. The number of cycles
before the efficiency evaluation has been chosen in such a way that
the cells exhibit a comparable capacity loss of about 25%. For cell
model C, most cells have not reached the EoL within the 1200 cycles
of the cycle life study. To indicate these cells in Fig. 14c, their bars
are colored in gray.
The orange bars demonstrate that the aged cells beyond EoL
exhibit an energy efficiency that lies 5–6 percentage points below
the efficiency in new condition. This means that the heat generated
during charging and discharging increases considerably for aged
cells, which is a result of higher internal cell resistances. The
orange bars also show for cell models A and B that the differences
in energy efficiency between the different charging protocols
remain rather constant with battery aging.
5. Discussion
Ideal charging protocols for lithium-ion batteries shall
maintain a long cycle life while providing good capacity
utilization, fast charging times, and high efficiency. The impact

Fig. 13. Correlations between capacity loss and (a–c) Ri and (d–f) Rac,10 mHz, depicted for all charging protocols together. Deviating protocols labeled only. The identical scaling
enables a direct comparison of the internal resistances of the different high-power 18650 cell models.
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
Fig. 14. Energy efficiency of one cycle of discharging with a constant current of 3 A and charging with different CCCV and PC protocols. Orange bars indicate cells beyond the
EoL, whereas gray bars represent cells which have not reached the EoL after 1200 cycles. (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)
of the charging protocols on these criteria is discussed in the
following sections.
5.1. Cycle life
Although the aging behavior of the three cell models examined
varies substantially, clear aging trends can be identified for the
different cell models. The design of our cycle life study with only
one sample per combination of cell model and charging protocol
has been sufficient to visualize these aging trends, as the effects
from the different charging protocols appear larger than potential
cell-to-cell variations. In the following sections, the impact of the
charging currents and charging voltages of the various charging
protocols on cycle life is explained.
5.1.1. Impact of the charging current
Charging a lithium-ion battery with high currents can deterio-
rate its cycle life by provoking lithium plating. This can be observed
clearly for cell models A and C, where the comparison of CCCV
protocols with different charging currents has revealed a lower
cycle life for a higher charging current. Especially the 5A CCCV
protocol, which exceeds the fast charging specification provided
for cell model C, has led to rapid degradation.
5.1.1.1. Cell model B’s low sensitivity to high charging
currents. However, there are also cell models, such as cell
model B, which are not susceptible to high charging currents.
For this cell model, a charging current above the fast charging
current in the datasheet can be used without negative effects on
cycle life. As cell model B’s cycle life has revealed almost no
dependency on the charging current of different CCCV protocols,
lithium plating appears to be no main driver of battery aging for
this cell model. Hence, the internal resistances at the anode must
be rather low, as no lithium plating and, thus, no voltage drops
below 0 V vs. Li/Li+ seem to occur. As this cell model exhibits the
highest cell resistances in our study, the internal resistances have
to originate largely from the cathode. Several studies have reported
substantially higher cathode resistances than anode resistances in
lithium-ion cells containing LiNiCoMnO2 or LiCoO2 cathode
materials [60–62]. Such a combination of high cathode
resistances and low anode resistances can help to prevent
lithium plating: as the high overall resistance of the cell
generates high overpotentials at the terminals, there is an early
transition from the CC to the CV phase. Since the CV phase leads to
137
a reduction of the charging current, the probability of lithium
plating is automatically reduced. For cell model B, such an effect
can be clearly observed, as only half of the cell’s capacity can be
charged in the CC phase of the 5A CCCV protocol. This means that,
on the one hand, cell model B is less suitable for fast charging due
to its pronounced CV phase, but on the other hand, the early CV
phase is beneficial for minimizing lithium plating.
5.1.1.2. Partial charging with high current. For cell models A and C,
which are more susceptible to lithium plating, BC protocols have
been used to investigate whether charging only part of the cell’s
capacity with a high charging current can be beneficial for cycle
life. The BC protocols also address the question whether the cells’
sensitivity to high charging currents varies with SoC. For boost
intervals starting at an SoC below 20%, accelerated aging has been
observed. This demonstrates that an empty cell cannot
automatically tolerate high charging currents, even if the cell
voltage remains considerably below the Vmax specification during
the boost interval. Further investigations are necessary to
determine whether the accelerated aging is a result from
increased lithium plating due to higher cell resistances at low
SoC or caused by another mechanism, such as high volume changes
of the anode material in almost completely delithiated condition
[63]. Although the overall charging time of the BC protocols lies in
the middle between the 1A CCCV and the 3A CCCV protocol, the
cycle life is at best in the range of the 3A CCCV protocol for both cell
models A and C. Hence, there is still a disproportionately high
impact of the boost interval on cycle life. This confirms that
continuous charging with high charging currents accelerates aging
considerably for cell models A and C, even if only 40% of CN is
charged during the boost interval.
5.1.1.3. Pulsed charging versus continuous charging. Cell models A
and C have also been examined with the PC protocols to investigate
the impact of high charging currents used for charging pulses. Both
PC protocols have led to similar aging as the 3A CCCV protocol. In
Fig. 10c, the aging behavior of cell model C demonstrates clearly
that the aging for the PC protocols is not an average of the aging
velocities of 1 A charging and 5 A charging. In fact, the aging is
rather similar to continuous charging with the same mean current.
For the pulse frequency of 25 Hz, no differences between
continuous and pulsed charging can be observed. This supports
the assumption that the double-layer capacities of the electrodes
buffer most of the high-frequent load dynamics, whereas only
138 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
frequencies below about 10 Hz lead to higher charge transfer and,
thus, increased aging [64]. In terms of battery aging, the cells
exhibit a certain low-pass behavior, where the degradation is
determined by the average value of the pulsed charging currents.
All in all, the lithium-ion cells examined can tolerate high charging
currents with no reductions in cycle life, when the high-current
pulses persist only for short times of less than 1 s and the mean
charging current is substantially lower. Moreover, no
improvements in cycle life can be observed for our PC protocols,
which do not decrease pulse amplitudes or increase pause lengths
at high SoC and also have the same cycle depth as continuous
charging with CCCV protocols.
5.1.2. Impact of the charging voltage
The variation of the charging voltage has also led to differences
in cycle life. Improvements due to lower charging voltages can be
attributed to two effects: on the one hand, the cell potential
remains lower, which reduces electrolyte oxidation, and on the
other hand, the charging current during CCCV charging is reduced
earlier, which helps to avoid lithium plating.
Whereas cell model A shows only minor changes in cycle life
related to Vch, cell model B, which has not been susceptible to
lithium plating, exhibits a considerable dependence of cycle life on
Vch. Overcharging by 50 mV has decreased cycle life markedly; a
reduction of Vch by up to 100 mV has improved cycle life
considerably. These effects can be attributed to increased or
reduced electrolyte decomposition and increased or reduced
damage to the crystallographic structure of the cathode material.
5.1.2.1. Specific behavior of LFP-based cell model C. For the LFP-
based cell model C, which exhibits a superior cycle life,
electrolyte oxidation due to high cell voltages can be
considered as only a minor driver of battery aging, as the
maximum cell voltage of 3.6 V lies substantially below the other
cell models. Yet, a clear dependency of cycle life on Vch can be
observed also for this cell model, although its capacity utilization
is rather independent from Vch. This can be attributed to the
impact of Vch on lithium plating: The 3A CCCV protocols with
different charging voltages depicted in Fig. 9 show that a lower Vch
leads to an earlier onset of the CV phase. This entails an earlier
reduction of the charging current at higher SoC, which reduces
lithium plating. The results from cell model C demonstrate that
reducing the charging current at high SoC can be an effective
measure for improving the cycle life of lithium-ion cells
susceptible to lithium plating under a predefined charging
current. However, cell model C’s rapid degradation for the 3A
CCCV + 50 mV protocol cannot be explained solely by increased
lithium plating due to a somewhat later reduction of the charging
current. It rather reveals the small overcharge reserve of LFP-
based cells. Unlike cells with other cathode materials, which still
contain 10–50% of the intercalated lithium at 100% SoC, an LFP
cathode is almost completely delithiated at 100% SoC [17,18]. Thus,
a small overcharging can already cause irreversible damage to the
crystallographic structure of the cathode.
5.1.2.2. Pulsed charging versus continuous charging. A certain
overcharging might also occur for the CCPC and the PC
protocols, when the charging pulses cause an exceedance of
Vmax at higher SoC. Yet, this has not led to a notable impact on cycle
life. Again, the lithium-ion cells exhibit a low-pass filtering
behavior for the charging pulses with lengths below 1 s. Hence,
it is rather the average than the peak value of the charging voltage
or current that determines the aging. This explains why no
pronounced electrolyte oxidation or lithium plating has occurred
as a result of the charging pulses from the CCPC and PC protocols.
5.2. Capacity utilization
The results of our experimental investigations have shown that
capacity utilization depends strongly on the charging protocol and
the specified charging voltages. For the high-power lithium-ion
cells examined, capacity utilization can be regarded as unaffected
by the charging current of CCCV charging. The protocols based on
pulsed charging have led to substantially lower capacity utiliza-
tion, as the charging process has been terminated earlier to prevent
damage due to the high overpotentials during the charging pulses
with constant amplitudes. An optimal capacity utilization can only
be achieved with a low charging current at the end of the charging
process. Only for the LFP-based cells, a good capacity utilization
between 95% and 98% can be achieved also with the CCPC and PC
protocols, as the cell voltage rises considerably only at very high
SoC. Our experiments have also demonstrated that lowering Vch
gradually reduces capacity utilization for most lithium-ion cells.
Only for the cells with a flat OCV curve, capacity utilization remains
rather unchanged over a wide range of charging voltages. For these
cells, changing the end condition, such as Iend, is the more effective
way to adjust capacity utilization, e.g. to perform a partial
charging.
5.3. Charging time
Our study has confirmed that the mean charging current is the
main parameter to increase the charging speed. However, this
often affects cycle life considerably. For practical applications, a
compromise between charging time and cycle life has to be found.
As cell model B has demonstrated, not every cell model can be fast
charged: Due to the high internal resistances, the reduction of the
charging current for the 5A CCCV protocol has already begun at
about 50% SoC, which decreases the charging speed considerably.
Hence, low internal resistances are necessary for short charging
times, as they prevent early current reductions.
Lower charging voltages have not always reduced the charging
time, as only the CC phase shortens, while the CV phase can also
prolong substantially. This has been observed especially for the
LFP-based cells due to their specific OCV characteristic. For aged
cells, charging times often do not decrease notably, as only the CC
phase shortens, whereas the CV phase becomes longer. This has to
be considered, when estimating the charging times for aged
battery systems.
5.4. Internal resistances
The internal resistances of lithium-ion cells generally rise with
aging. The shift of the impedance spectra to the right, representing
higher ohmic resistances Ri, indicates a reduced electrolyte
conductivity [51], which is caused by side reactions consuming
electrolyte. The mid-frequency semicircle has enlarged substan-
tially for cell models A and B and has remained rather stable for cell
model C. In general, growing semicircles in the mid-frequency
domain are attributed to increasing charge transfer resistances.
However, they cannot be linked explicitly to a specific degradation
process, because they can be caused by several aging mechanisms.
In literature, common explanations are increasing passivation
layers on both electrodes, decreasing intercalation kinetics, and
reduced active surface areas due to pore clogging and particle
disconnection [65,66]. For LiCoO2, LiMn2O4, and LiNiCoMnO2
cathodes, substantial impedance increases in the mid-frequency
range have been reported [50,61]. In contrast, LFP is known as a
very stable cathode material lasting up to several thousands of
cycles with only minor impedance increase [67]. For graphite
anodes, growing SEI passivation layers and particle contact loss are
supposed to increase impedances in the mid-frequency domain
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
[52]. As all cells contain graphite anodes, but only cell models A and
B exhibit a considerable growth of mid-frequency semicircles, it is
concluded that these impedance changes origin mainly from the
cathode. In the low-frequency domain, all cells exhibit an
extension of the sloping line at the right end of the impedance
spectrum. This reveals higher diffusion resistances, which can
origin from slower mass transport in the electrolyte as well as in
the solid electrode materials [60].
In our experimental study, the increases of the internal
resistances have only ranged between 10% and 40%, when the
EoL capacity criterion of 20% capacity loss has been reached. The
EIS results have revealed that the correlation between resistance
rise and capacity fade has been rather similar for all cells of the
same model, although cycled with different charging protocols.
However, a precise estimation of the remaining capacity based on a
measured cell resistance remains challenging: as the increases of
the resistances have been comparatively low, small variations in
the resistance values lead to high deviations in the estimated
capacity. This inhibits a robust and reliable capacity estimation. For
the high-power cells models examined, the EoL has always been
reached by meeting the capacity criterion and not by a doubling of
cell resistances. Hence, the available capacity is the more
representative property for assessing the aging condition of the
cells.
5.5. Energy efficiency
A high efficiency is beneficial, as this reduces the energy losses
and, thus, the thermal load on the cells. Hence, less cooling is
required in practical applications. Comparing the energy efficien-
cies of CCCV protocols with different charging currents has
revealed a rather linear increase of the losses with higher
charging currents. Pulsed charging has generated only somewhat
higher losses than continuous charging. When designing cooling
systems for batteries, it also has to be considered that the losses
increase with aging. The energy efficiencies of aged cells have
demonstrated that the losses still increase rather linearly with the
current.
5.6. Recommendations for cycle life studies
Our experimental cycle life study on charging protocols for
lithium-ion batteries has shown that a sophisticated study design
is essential for separating the effects of different parameters on
the performance of charging protocols. A similar capacity
utilization is essential to identify the specific advantages or
disadvantages in cycle life of different charging protocols and not
just the side-effects of varying cycle depths. Moreover, similar
mean charging currents should be used when comparing pulsed
and continuous charging, as this avoids side-effects from varying
charging currents due to modifications in the pulse profile.
Furthermore, the cells should also be examined with high-current
CCCV protocols, as a precise knowledge of the cells’ vulnerability
to lithium plating is essential, especially for the assessment of
fast-charging protocols.
6. Conclusions
In this paper, an overview on charging protocols for lithium-ion
batteries has been provided. Our extensive experimental study,
comprising three high-power 18650 cell types from different
manufacturers and with different cell chemistries, has enabled an
objective comparison of the performance of different charging
protocols and their impact on cycle life. The experimental results
have shown that the impact of charging currents and charging
139
voltages on cycle life can vary markedly among different lithium-
ion batteries.
6.1. Impact of high charging and discharging currents
In general, the cycle life has been influenced more by high
charging currents than by high discharging currents. Fast charging,
resulting in high charging currents, deteriorates cycle life
considerably, when it provokes lithium plating. The boost charging
protocols, where only part of the battery’s capacity is charged with
a high charging current, have still exhibited disproportionately
high degradation attributable to the high-current boost intervals.
Moreover, this study has revealed that high charging currents can
deteriorate cycle life not only at high SoC, but also at very low SoC.
In other words, a completely discharged battery cannot automati-
cally tolerate the highest charging currents.
6.2. Pulsed charging versus continuous charging
Our investigations on pulsed charging have shown that it is
mainly the mean value of current or voltage that determines
battery degradation. Lithium-ion cells can withstand high pulse
currents or peak voltages without deteriorating cycle life, when the
duration of the pulses remains short (here: <1 s) and the mean
current and voltage values are considerably lower. No advantages
have been observed for pulsed charging compared to continuous
charging in our experimental study, where the mean charging
currents and the cycle depths have been identical. Capacity
utilization and efficiency have even been lower for pulsed
charging. All in all, the conventional CCCV protocol is an excellent
starting basis for an optimized charging method for lithium-ion
batteries. Pulse charging can be beneficial, when higher losses are
desired, e.g., for heating up a battery at cold temperatures [68]. In
such cases, the heating effect can be intensified by using pulse
charging with discharging pulses, since this increases losses for the
same mean charging current, because some of the capacity has to
be charged several times due to the repeated discharging pulses.
6.3. Cycle life
As the aging behavior of different lithium-ion cells can vary
considerably, a precise knowledge of the cells, e.g., by experimental
examination, is necessary for developing an optimal charging
strategy. For lithium-ion batteries susceptible to lithium plating,
the charging currents should be reduced, especially at high SoC. A
lower charging voltage also helps to reduce the charging currents
at high SoC to decrease the probability of lithium plating. For
batteries where cycle life is deteriorated markedly by high
voltages, a reduction of the charging voltage is essential for
maximizing cycle life. However, this generally leads to a lower
capacity available.
6.4. Capacity utilization and charging time
Capacity utilization and charging time have also been analyzed
for the different charging protocols. Maximum capacity utilization
can only be obtained with a low charging current at the end of the
charging process. The charging time, which is mainly determined
by the mean charging current, has revealed no monotonic
relationship to the charging voltage. For aged cells, the charging
time has remained almost constant for the non-LFP cells, although
the capacity has decreased considerably.
A special behavior has been observed for the LFP-based cells
due to the flat OCV curve: Reducing the charging voltage does not
lower the capacity utilization. It rather causes an earlier reduction
of the charging current, which prolongs the charging time
140 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
considerably. For the other, non-LFP cells, capacity utilization has
decreased by about 10–20% per 100 mV reduction of the charging
voltage.
6.5. Internal resistances and energy efficiency
For the three cell types examined, the correlation between
resistance rise and capacity fade has been rather similar for the
different charging protocols investigated. Moreover, the capacity
fade has been the dominant aging effect limiting the useful life of
the batteries.
The efficiencies of the lithium-ion cells have decreased rather
linearly with the charging current. In general, the energy efficiency
decreases with aging due to higher losses caused by the increased
internal resistances. This has to be considered when designing
cooling systems for lithium-ion battery systems.
6.6. Practical applications
For practical applications requiring fast charging, cells with low
internal resistances should be used. Low cell resistances delay
reductions of the charging currents due to voltage limitations to a
late stage of the charging process. Moreover, the cells must have an
appropriate cell design, e.g., low electrode thicknesses and
sufficient porosity, to achieve fast charging times [68] and
minimize the susceptibility to lithium plating (see Refs. [69,70]).
Using cells with a lithium-titanate anode instead of a graphite
anode allows even higher charging currents, as this anode material
is less vulnerable to lithium plating due to a higher potential versus
Li+/Li [71]. However, this also leads to a lower energy content for
such cells. If a high energy content and, thus, a high capacity
utilization is required, it is important to select cells that can
withstand the maximum cell voltage without accelerated degra-
dation.
For practical battery systems, it is most important to select a
well-suited cell type. For such cells, a CCCV charging protocol with
an appropriate charging current and charging voltage will provide
a good overall performance. When the cycle life is affected by
lithium plating, a reduction of the charging current at high SoC
should be considered to achieve an ideal compromise between fast
charging and long cycle life. Since lithium plating increases with
lower operating temperatures, the impact of temperature on
charging lithium-ion batteries will be investigated in a subsequent
study.
References
[1] J. Li, E. Murphy, J. Winnick, P.A. Kohl, The effects of pulse charging on cycling
characteristics of commercial lithium-ion batteries, J. Power Sources 102
(2001) 302–309.
[2] E. Prada, D. Di Domenico, Y. Creff, J. Bernard, V. Sauvant-Moynot, F. Huet,
Simplified electrochemical and thermal model of LiFePO4-graphite Li-ion
batteries for fast charge applications, J. Electrochem. Soc. 159 (2012) A1508–
A1519.
[3] P. Keil, A. Jossen, Aging of lithium-ion batteries in electric vehicles: impact of
regenerative braking, 28th International Electric Vehicle Symposium (EVS28),
Goyang, 2015.
[4] S.S. Zhang, The effect of the charging protocol on the cycle life of a Li-ion
battery, J. Power Sources 161 (2006) 1385–1391.
[5] P.E. de Jongh, P.H.L. Notten, Effect of current pulses on lithium intercalation
batteries, Solid State Ionics 148 (2002) 259–268.
[6] P. Arora, R.E. White, M. Doyle, Mathematical modeling of the lithium
deposition overcharge reaction in lithium-ion batteries using carbon-based
negative electrodes, J. Electrochem. Soc. 146 (1999) 3543–3553.
[7] Z. Li, J. Huang, B. Yann Liaw, V. Metzler, J. Zhang, A review of lithium deposition
in lithium-ion and lithium metal secondary batteries, J. Power Sources 254
(2014) 168–182.
[8] V. Zinth, C. von Lüders, M. Hofmann, J. Hattendorff, I. Buchberger, S. Erhard, J.
Rebelo-Kornmeier, A. Jossen, R. Gilles, Lithium plating in lithium-ion batteries
at sub-ambient temperatures investigated by in situ neutron diffraction, J.
Power Sources 271 (2014) 152–159.
[9] N. Legrand, B. Knosp, P. Desprez, F. Lapicque, S. Raël, Physical characterization
of the charging process of a Li-ion battery and prediction of Li plating by
electrochemical modelling, J. Power Sources 245 (2014) 208–216.
[10] B.K. Purushothaman, U. Landau, Rapid charging of lithium-ion batteries using
pulsed currents, J. Electrochem. Soc. 153 (2006) A533–A542.
[11] M. Petzl, M.A. Danzer, Nondestructive detection characterization, and
quantification of lithium plating in commercial lithium-ion batteries, J. Power
Sources 254 (2014) 80–87.
[12] S.S. Zhang, K. Xu, T.R. Jow, Study of the charging process of a LiCoO2-based Li-
ion battery, J. Power Sources 160 (2006) 1349–1354.
[13] M. Ecker, N. Nieto, S. Käbitz, J. Schmalstieg, H. Blanke, A. Warnecke, D.U. Sauer,
Calendar and cycle life study of Li(NiMnCo)O2-based 18650 lithium-ion
batteries, J. Power Sources 248 (2014) 839–851.
[14] I. Laresgoiti, S. Käbitz, M. Ecker, D.U. Sauer, Modeling mechanical degradation
in lithium ion batteries during cycling: solid electrolyte interphase fracture, J.
Power Sources 300 (2015) 112–122.
[15] S.F. Schuster, T. Bach, E. Fleder, J. Müller, M. Brand, G. Sextl, A. Jossen, Nonlinear
aging characteristics of lithium-ion cells under different operational
conditions, J. Energy Storage 1 (2015) 44–53.
[16] Y. Yamada, M. Yaegashi, T. Abe, A. Yamada, A superconcentrated ether
electrolyte for fast-charging Li-ion batteries, Chem. Commun. (Camb.) 49
(2013) 11194–11196.
[17] D. Doughty, E.P. Roth, A general discussion of Li ion battery safety, Electrochem.
Soc. Interface 21 (2012) 37–44.
[18] M. Brand, S. Gläser, J. Geder, S. Menacher, S. Obpacher, A. Jossen, D. Quinger,
Electrical safety of commercial Li-ion cells based on NMC and NCA technology
compared to LFP technology, 27th International Electric Vehicle Symposium
(EVS27), Barcelona, 2013.
[19] D.U. Sauer, BATTERIES | charge–discharge curves, Encyclopedia of
Electrochemical Power Sources, Elsevier, 2009, pp. 443–451.
[20] S.S. Choi, H.S. Lim, Factors that affect cycle-life and possible degradation
mechanisms of a Li-ion cell based on LiCoO2, J. Power Sources 111 (2002) 130–
136.
[21] Y.-F. Luo, Y.-H. Liu, S.-C. Wang, Search for an optimal multistage charging
pattern for lithium-ion batteries using the Taguchi approach, TENCON 2009—
IEEE Region 10 Conference, Singapore, 2009.
[22] Y.-H. Liu, C.-H. Hsieh, Y.-F. Luo, Search for an optimal five-step charging pattern
for Li-ion batteries using consecutive orthogonal arrays, IEEE Trans. Energy
Convers. 26 (2011) 654–661.
[23] F. Savoye, P. Venet, M. Millet, J. Groot, Impact of periodic current pulses on Li-
ion battery performance, IEEE Trans. Ind. Electron. 59 (2012) 3481–3488.
[24] G. Pistoia, Battery Operated Devices and Systems: From Portable Electronics to
Industrial Products, Elsevier, Amsterdam, London, 2009.
[25] Linear Technology, LTC 4052-4.2—Lithium-ion Battery Pulse Charger with
Overcurrent Protection, Datasheet, cds.linear.com/docs/en/datasheet/
405242f.pdf, (accessed September, 15).
[26] H.Z.Z. Beh, G.A. Covic, J.T. Boys, Effects of pulse and DC charging on lithium iron
phosphate (LiFePO4) batteries, IEEE Energy Conversion Congress and
Exposition (ECCE), Denver, USA, 2013, pp. 315–320.
[27] B.K. Purushothaman, P.W. Morrison, U. Landau, Reducing mass-transport
limitations by application of special pulsed current modes, J. Electrochem. Soc.
152 (2005) J33.
[28] J. Fan, S. Tan, Studies on charging lithium-ion cells at low temperatures, J.
Electrochem. Soc. 153 (2006) A1081–A1092.
[29] P.H.L. Notten, J.H.G. Op het Veld, J. van Beek, Boostcharging Li-ion batteries: a
challenging new charging concept, J. Power Sources 145 (2005) 89–94.
[30] D. Anseán, M. González, J.C. Viera, V.M. García, C. Blanco, M. Valledor, Fast
charging technique for high power lithium iron phosphate batteries: a cycle
life analysis, J. Power Sources 239 (2013) 9–15.
[31] M. Torchio, N.A. Wolff, D.M. Raimondo, L. Magni, U. Krewer, B. Gopaluni, J.
Paulson, R.D. Braatz, Real-time model predictive control for the optimal
charging of a lithium-ion battery, American Control Conference (ACC),
Chicago, 2015, pp. 4536–4541.
[32] B. Suthar, P.W.C. Northrop, R.D. Braatz, V.R. Subramanian, Optimal charging
profiles with minimal intercalation-induced stresses for lithium-ion batteries
using reformulated pseudo 2-dimensional models, J. Electrochem. Soc. 161
(2014) F3144–F3155.
[33] R. Methekar, V. Ramadesigan, R.D. Braatz, V.R. Subramanian, Optimum
charging profile for lithium-ion batteries to maximize energy storage and
utilization, ECS Trans. 25 (2010) 139–146.
[34] G. Sikha, P. Ramadass, B.S. Haran, R.E. White, B.N. Popov, Comparison of the
capacity fade of Sony US 18650 cells charged with different protocols, J. Power
Sources 122 (2003) 67–76.
[35] S. Tippmann, D. Walper, L. Balboa, B. Spier, W.G. Bessler, Low-temperature
charging of lithium-ion cells part I: electrochemical modeling and
experimental investigation of degradation behavior, J. Power Sources 252
(2014) 305–316.
[36] R. Klein, N.A. Chaturvedi, J. Christensen, J. Ahmed, R. Findeisen, A. Kojic,
Optimal charging strategies in lithium-ion battery, American Control
Conference (ACC), San Francisco, 2011, pp. 382–387.
[37] J. Remmlinger, S. Tippmann, M. Buchholz, K. Dietmayer, Low-temperature
charging of lithium-ion cells part II: model reduction and application, J. Power
Sources 254 (2014) 268–276.
[38] B. Suthar, D. Sonawane, R.D. Braatz, V.R. Subramanian, Optimal low
temperature charging of lithium-ion batteries, IFAC-PapersOnLine 48 (2015)
1216–1221.
 P. Keil, A. Jossen / Journal of Energy Storage 6 (2016) 125–141
[39] B. Suthar, V. Ramadesigan, S. De, R.D. Braatz, V.R. Subramanian, Optimal
charging profiles for mechanically constrained lithium-ion batteries, Phys.
Chem. Chem. Phys. 16 (2014) 277–287.
[40] Z. Guo, B.Y. Liaw, X. Qiu, L. Gao, C. Zhang, Optimal charging method for lithium
ion batteries using a universal voltage protocol accommodating aging, J. Power
Sources 274 (2015) 957–964.
[41] X. Hu, S. Li, H. Peng, F. Sun, Charging time and loss optimization for LiNMC and
LiFePO4 batteries based on equivalent circuit models, J. Power Sources 239
(2013) 449–457.
[42] I. Abou Hamad, M.A. Novotny, D.O. Wipf, P.A. Rikvold, A new battery-charging
method suggested by molecular dynamics simulations, Phys. Chem. Chem.
Phys. 12 (2010) 2740–2743.
[43] J.-W. Lee, Y.K. Anguchamy, B.N. Popov, Simulation of charge–discharge cycling
of lithium-ion batteries under low-earth-orbit conditions, J. Power Sources
162 (2006) 1395–1400.
[44] A. Hoke, A. Brissette, K. Smith, A. Pratt, D. Maksimovic, Accounting for lithium-
ion battery degradation in electric vehicle charging optimization, IEEE J.
Emerg. Sel. Top. Power Electron. 2 (2014) 691–700.
[45] M.M. Doeff, Battery cathodes, in: R.J. Brodd (Ed.), Batteries for Sustainability,
Springer, New York, 2013, pp. 5–49.
[46] I. Bloom, A.N. Jansen, D.P. Abraham, J. Knuth, S.A. Jones, V.S. Battaglia, G.L.
Henriksen, Differential voltage analyses of high-power, lithium-ion cells, J.
Power Sources 139 (2005) 295–303.
[47] M. Dubarry, C. Truchot, B.Y. Liaw, Synthesize battery degradation modes via a
diagnostic and prognostic model, J. Power Sources 219 (2012) 204–216.
[48] G. Ning, B. Haran, B.N. Popov, Capacity fade study of lithium-ion batteries
cycled at high discharge rates, J. Power Sources 117 (2003) 160–169.
[49] I. Belharouak, Y.-K. Sun, J. Liu, K. Amine, Li(Ni1/3Co1/3Mn1/3)O2 as a suitable
cathode for high power applications, J. Power Sources 123 (2003)
247–252.
[50] D. Zhang, B. Haran, A. Durairajan, R. White, Y. Podrazhansky, B. Popov, Studies
on capacity fade of lithium-ion batteries, J. Power Sources 91 (2000) 122–129.
[51] J. Huang, H. Ge, Z. Li, J. Zhang, An agglomerate model for the impedance of
secondary particle in lithium-ion battery electrode, J. Electrochem. Soc. 161
(2014) E3202–E3215.
[52] J. Vetter, P. Novák, M.R. Wagner, C. Veit, K.-C. Möller, J.O. Besenhard, M. Winter,
M. Wohlfahrt-Mehrens, C. Vogler, A. Hammouche, Ageing mechanisms in
lithium-ion batteries, J. Power Sources 147 (2005) 269–281.
[53] M. Levi, K. Gamolsky, D. Aurbach, U. Heider, R. Oesten, On electrochemical
impedance measurements of LixCo0.2Ni0.8O2 and LixNiO2 intercalation
electrodes, Electrochim. Acta 45 (2000) 1781–1789.
[54] A. Jossen, Fundamentals of battery dynamics, J. Power Sources 154 (2006)
530–538.
[55] P.J. Osswald, S.V. Erhard, A. Noel, P. Keil, F.M. Kindermann, H. Hoster, A. Jossen,
Current density distribution in cylindrical Li-ion cells during impedance
measurements, J. Power Sources (2016), doi:http://dx.doi.org/10.1016/j.
jpowsour.2016.02.070.
141
[56] Y.-S. Chen, K.-H. Chang, C.-C. Hu, T.-T. Cheng, Performance comparisons and
resistance modeling for multi-segment electrode designs of power-oriented
lithium-ion batteries, Electrochim. Acta 55 (2010) 6433–6439.
[57] J.N. Reimers, Analytic theory for foil impedance in spiral wound cell
geometries, J. Power Sources 262 (2014) 429–443.
[58] E. Sarasketa-Zabala, I. Gandiaga, L.M. Rodriguez-Martinez, I. Villarreal,
Calendar ageing analysis of a LiFePO4/graphite cell with dynamic model
validations: towards realistic lifetime predictions, J. Power Sources 272 (2014)
45–57.
[59] R. Spotnitz, Simulation of capacity fade in lithium-ion batteries, J. Power
Sources 113 (2003) 72–80.
[60] J. Huang, Z. Li, J. Zhang, S. Song, Z. Lou, N. Wu, An analytical three-scale
impedance model for porous electrode with agglomerates in lithium-ion
batteries, J. Electrochem. Soc. 162 (2015) A585–A595.
[61] B. Stiaszny, J.C. Ziegler, E.E. Krauß, J.P. Schmidt, E. Ivers-Tiffée, Electrochemical
characterization and post-mortem analysis of aged LiMn2O4–Li
(Ni0.5Mn0.3Co0.2)O2/graphite lithium ion batteries. Part I: cycle aging, J. Power
Sources 251 (2014) 439–450.
[62] J. Song, Two- and three-electrode impedance spectroscopy of lithium-ion
batteries, J. Power Sources 111 (2002) 255–267.
[63] H. He, C. Huang, C.-W. Luo, J.-J. Liu, Z.-S. Chao, Dynamic study of Li intercalation
into graphite by in situ high energy synchrotron XRD, Electrochim. Acta 92
(2013) 148–152.
[64] M. Uno, K. Tanaka, Influence of high-frequency charge–discharge cycling
induced by cell voltage equalizers on the life performance of lithium-ion cells,
IEEE Trans. Veh. Technol. 60 (2011) 1505–1515.
[65] Y. Zhang, C.-Y. Wang, Cycle-life characterization of automotive lithium-ion
batteries with LiNiO[sub 2] cathode, J. Electrochem. Soc. 156 (2009) A527.
[66] D.P. Abraham, S.D. Poppen, A.N. Jansen, J. Liu, D.W. Dees, Application of a
lithium–tin reference electrode to determine electrode contributions to
impedance rise in high-power lithium-ion cells, Electrochim. Acta 49 (2004)
4763–4775.
[67] J. Groot, M. Swierczynski, A.I. Stan, S.K. Kær, On the complex ageing
characteristics of high-power LiFePO4/graphite battery cells cycled with high
charge and discharge currents, J. Power Sources 286 (2015) 475–487.
[68] M.F. Hasan, C.-F. Chen, C.E. Shaffer, P.P. Mukherjee, Analysis of the implications
of rapid charging on lithium-ion battery performance, J. Electrochem. Soc. 162
(2015) A1382–A1395.
[69] C.-K. Park, Z. Zhang, Z. Xu, A. Kakirde, K. Kang, C. Chai, G. Au, L. Cristo, Variables
study for the fast charging lithium ion batteries, J. Power Sources 165 (2007)
892–896.
[70] H. Zheng, J. Li, X. Song, G. Liu, V.S. Battaglia, A comprehensive understanding of
electrode thickness effects on the electrochemical performances of Li-ion
battery cathodes, Electrochim. Acta 71 (2012) 258–265.
[71] K. Zaghib, M. Dontigny, A. Guerfi, P. Charest, I. Rodrigues, A. Mauger, C.M.
Julien, Safe and fast-charging Li-ion battery with long shelf life for power
applications, J. Power Sources 196 (2011) 3949–3954.