Access to this full-text is provided by IOP Publishing.
Content available from Journal of The Electrochemical Society
This content is subject to copyright. Terms and conditions apply.
Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017) A1361
Oxygen Release and Its Effect on the Cycling Stability of
LiNixMnyCozO2(NMC) Cathode Materials for Li-Ion Batteries
Roland Jung,a,b,∗,zMichael Metzger,a,∗Filippo Maglia,bChristoph Stinner,b
and Hubert A. Gasteigera,∗∗
aChair of Technical Electrochemistry, Department of Chemistry and Catalysis Research Center,
Technische Universit¨
at M¨
unchen, Garching, Germany
bBMW AG, Munich, Germany
Layered LiNixMnyCozO2(NMC) is a widely used class of cathode materials with LiNi1/3Mn1/3 Co1/3O2(NMC111) being the
most common representative. However, Ni-rich NMCs are more and more in the focus of current research due to their higher
specific capacity and energy. In this work we will compare LiNi1/3Mn1/3Co1/3 O2(NMC111), LiNi0.6Mn0.2 Co0.2O2(NMC622),
and LiNi0.8Mn0.1 Co0.1O2(NMC811) with respect to their cycling stability in NMC-graphite full-cells at different end-of-charge
potentials. It will be shown that stable cycling is possible up to 4.4 V for NMC111 and NMC622 and only up to 4.0 V for NMC811.
At higher potentials, significant capacity fading was observed, which was traced back to an increase in the polarization of the
NMC electrode, contrary to the nearly constant polarization of the graphite electrode. Furthermore, we show that the increase in the
polarization occurs when the NMC materials are cycled up to a high-voltage feature in the dq/dV plot, which occurs at ∼4.7 V vs.
Li/Li+for NMC111 and NMC622 and at ∼4.3 V vs. Li/Li+for NMC811. For the latter material, this feature corresponds to the H2
→H3 phase transition. Contrary to the common understanding that the electrochemical oxidation of carbonate electrolytes causes the
CO2and CO evolution at potentials above 4.7 V vs. Li/Li+, we believe that the observed CO2and CO are mainly due to the chemical
reaction of reactive lattice oxygen with the electrolyte. This hypothesis is based on gas analysis using On-line Electrochemical Mass
Spectrometry (OEMS), by which we prove that all three materials release oxygen from the particle surface and that the oxygen
evolution coincides with the onset of CO2and CO evolution. Interestingly, the onsets of oxygen evolution for the different NMCs
correlate well with the high-voltage redox feature at ∼4.7 V vs. Li/Li+for NMC111 and NMC622 as well as at ∼4.3 V vs. Li/Li+
for NMC811. To support this hypothesis, we show that no CO2or CO is evolved for the LiNi0.43Mn1.57O4(LNMO) spinel up to
5 V vs. Li/Li+, consistent with the absence of oxygen release. Lastly, we demonstrate by the use of 13C labeled conductive carbon
that it is the electrolyte rather than the conductive carbon which is oxidized by the released lattice oxygen. Taking these findings into
consideration, a mechanism is proposed for the reaction of released lattice oxygen with ethylene carbonate yielding CO2,CO,and
H2O.
© The Author(s) 2017. Published by ECS. This is an open access article distributed under the terms of the Creative Commons
Attribution Non-Commercial No Derivatives 4.0 License (CC BY-NC-ND, http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial reuse, distribution, and reproduction in any medium, provided the original work is not changed in any
way and is properly cited. For permission for commercial reuse, please email: oa@electrochem.org. [DOI: 10.1149/2.0021707jes]
All rights reserved.
Manuscript submitted January 16, 2017; revised manuscript received April 13, 2017. Published May 2, 2017. This article is a version
of Paper 39 from the New Orleans, Louisiana, Meeting of the Society, May 28-June 1, 2017.
Li-Ion batteries have recently been used as power supply for elec-
tric vehicles (EVs). In order to penetrate the mass market, a significant
reduction in costs and further performance improvements have to be
achieved to realize a longer driving range of EVs.1The latter highly
depends on the choice of the cathode active material, for which several
potential materials exist,2of which layered lithium nickel manganese
cobalt oxide (LiNixMnyCozO2, NMC) is one of the most promising
class of cathode materials.3This is due to the high specific capacity
and good stability of the layered structure which changes its volume
by less than 2% during Li insertion/extraction.4–6Due to the sloped
voltage profile of NMC, a higher capacity and energy density can be
achieved when the upper cutoff voltage is increased.7–9Even though
the theoretical capacity of NMC is as high as ∼275 mAh/gNMC, not
all of the lithium can be extracted due to structural instabilities oc-
curring when an exceedingly large fraction of lithium is removed.9,10
Additionally, the sloped voltage profile requires very high voltages
to achieve complete removal of lithium, which in turn can lead to
electrolyte oxidation, surface film formation, and transition metal dis-
solution, ultimately diminishing the cycling stability.11–17 For these
reasons, the operating potential of NMC based cathode materials is
nowadays limited to ∼4.3 V, restricting their capacities to much below
their theoretical values.5In order to improve the accessible capacity at
reasonable upper cutoff voltages, Ni-rich NMCs (Ni-content >> Mn-
and Co-content) recently became the focus of interest, as more lithium
can be extracted from their structure within the same voltage window.
Therefore, they provide larger specific capacities and energy densities,
which are crucial for a longer driving range of electric vehicles.2,18,19
∗Electrochemical Society Student Member.
∗∗Electrochemical Society Fellow.
zE-mail: roland.jung@tum.de
So far, however, Ni-rich NMCs suffer from a shorter lifetime due to
a faster capacity fading compared to LiNi1/3Mn1/3Co1/3 O2(NMC111),
the most common NMC material with a nickel:manganese:cobalt ra-
tio of 1:1:1.18–20 For example, Noh et al. reported initial capacities
at a 0.1 C-rate and an end-of-charge potential of 4.3 V vs. Li/Li+
of 203 mAh/g and 163 mAh/g for LiNi0.8Mn0.1 Co0.1O2(NMC811)
and NMC111, respectively.18 Unfortunately, the capacity retention for
NMC811 in NMC-Li cells after 100 cycles at a 0.5 C-rate was only
70% compared to 92% for NMC111.18 Furthermore, it was shown that
layered oxides with rising Ni-contents are thermally less stable.18,21 At
temperatures ≥170◦C, the bulk materials undergo a two-phase tran-
sition from their layered structure to a spinel structure and eventually
to a rock-salt structure, both of which are accompanied by release
of lattice oxygen.18,21–27 For materials aged under battery operating
conditions (<60◦C), the formation of disordered spinel and rock-salt
type phases was reported to happen on the particle surface with the
bulk structure remaining intact, i.e., remaining in the rhombohedral
structure as reported for NMC,9LiNi0.8Co0.2 O2,28,29, LiNiO2,30 ,and
NCA (LiNi0.80Co0.15 Al0.05O2).31 Even though it was not explicitly
shown in these latter reports, the observed phase transitions suggest a
release of oxygen from the particle surface, which was also pointed
out in the reports by Abraham et al., Muto et al., and Hwang et al.28–31
So far, a release of oxygen from the oxide lattice under battery op-
erating conditions was shown only for overlithiated NMC materials
(Li1+x(Ni,Mn,Co)1-xO2), in which lithium additionally occupies the
transition metal layers.32–36 The exception is a recent publication by
Gu´
eguen et al., who showed oxygen evolution during battery cycling
for NMC111.37
In this study, we will compare two Ni-rich NMCs, namely
LiNi0.6Mn0.2 Co0.2O2(NMC622) and LiNi0.8 Mn0.1Co0.1 O2(NMC811)
to LiNi1/3Mn1/3 Co1/3O2(NMC111) with respect to their cycling
A1362 Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017)
stability in full-cells with graphite anodes. Through an evaluation
of the anode and cathode polarization in a three-electrode set-up, we
conclude that the capacity fading at high voltages is due to the NMC
electrode rather than the graphite electrode. By means of On-line
Electrochemical Mass Spectrometry (OEMS) we prove that at high
degrees of delithiation all three NMC materials release oxygen already
at room temperature. The onset of the oxygen evolution corresponds
well with the onset of the formation of CO2and CO, which is typically
assigned to electrochemical electrolyte oxidation, raising the question
whether the evolution of O2actually causes the observed CO2and CO
evolution. This question as well as the consequences of the oxygen
release on the polarization and the cycling stability will be discussed
with the experimental findings presented in this work.
Experimental
Electrode Preparation.—Layered NMC and spinel LNMO elec-
trodes were prepared by dispersing the active material particles
(LiNi1/3Mn1/3 Co1/3O2(NMC111), LiNi0.6 Mn0.2Co0.2 O2(NMC622),
LiNi0.8Mn0.1 Co0.1O2(NMC811) or LiNi0.43 Mn1.57O4(LNMO), all
from Umicore, Belgium) (91.5 %wt), conductive carbon (Super
C65, Timcal, Switzerland) (4.4 %wt) and polyvinylidene fluoride
binder (PVDF, Kynar HSV 900, Arkema, France) (4.1 %wt)inN-
methylpyrrolidone (NMP, anhydrous, 99.5%, Sigma-Aldrich). The
slurry was mixed in a planetary mixer (Thinky, USA) at 2000 rpm for
2×5 minutes. In between the two runs the slurry was ultrasonicated
for 10 minutes in an ultrasonic bath. The resulting ink was spread
onto aluminum foil (thickness 18 μm, MTI Corporation, USA) us-
ing a gap bar coater (RK PrintCoat Instruments, UK). For OEMS
measurements, the ink was coated onto a H2013 polyolefin separator
(Celgard, USA) or a stainless steel mesh (316 grade, 26 μmaper-
ture, 25 μm wire diameter, The Mesh Company, UK) to allow for
a short diffusion time of the evolved gases to the head-space of the
OEMS cell and to the capillary leading to the mass spectrometer.38,39
NMC622 electrodes containing 13C-labeled carbon (99 %atom 13 C,
Sigma-Aldrich, Germany) were prepared with the same composition
as the ones containing Super C65, however, due to strong agglomera-
tion of the carbon, the ink was prepared in a ball mill (Pulverisette 7,
Fritsch, Germany) using zirconia balls with a diameter of 10 mm at
180 rpm for 60 minutes. After drying at 50◦C, electrodes were punched
and dried overnight at 120◦C (if coated on aluminum or stainless steel
mesh) and at 95◦C (if coated on H2013 separator) under dynamic
vacuum in a glass oven (drying oven 585, B¨
uchi, Switzerland) and
transferred to a glove box (O2and H2O<0.1 ppm, MBraun, Germany)
without exposure to ambient air.
The graphite electrodes were composed of graphite (MAG-D20,
Hitachi), Super C65 (Timcal, Switzerland), sodium carboxymethyl-
cellulose (Na-CMC, Dow Wolff Cellulosics) and styrene-butadiene
rubber (SBR, JSR Micro) at a weight ratio of 95.8:1:1:2.2. The slurry
was prepared by dispersing graphite, Super C65 and Na-CMC in
highly pure water (18 Mcm, Merck Millipore, Germany) using a
planetary mixer (Thinky, USA; at 2000 rpm for 30 minutes). The
slurry was ultrasonicated for 10 minutes in an ultrasonic bath. SBR
was added to the slurry and mixed at 500 rpm for 2 minutes. The
ink was coated onto copper foil (thickness 12 μm, MTI Corporation,
USA) using a gap bar coater (RK PrintCoat Instruments, UK). The
coating was dried at 50◦C in air, punched out, dried overnight at 120◦C
under vacuum in a glass oven (B¨
uchi oven, s. above) and transferred
to a glove box without exposure to ambient air.
The specific surface areas of the NMC and LNMO were deter-
mined by BET, using an Autosorb iQ nitrogen gas sorption analyzer
(Quantachrome Instruments, USA). The determined BET areas of
these materials are 0.26 m2/g, 0.35 m2/g, 0.18 m2/g, and 0.23 m2/g
for NMC111, NMC622, NMC811, and LNMO, respectively.
Electrochemical characterization.—The electrochemical charac-
terization of NMC was performed in Swagelok T-cells which were as-
sembled in an argon filled glove box (O2and H2O<0.1 ppm, MBraun,
Germany), with NMC as working electrode (10 mm diameter)
and graphite as counter electrode (11 mm diameter). The areal mass
loading of the NMC electrodes was 15.5 ±1 mg/cm2and the one of the
graphite electrodes was adapted according to the mass loading of the
NMC electrodes and their specific capacities at the various cutoff volt-
ages, aiming to achieve a constant balancing factor. The areal capacity
of the anode (in mAh/cm2) was 1.2-fold oversized compared to the
cathode (referenced to the reversible capacities of NMC and graphite
at a 1 C-rate; if referenced to 0.1 C, the anode is roughly 1.1-fold
oversized). To monitor the potential of both the NMC cathode and the
graphite anode, a lithium reference electrode (thickness 0.45 mm, bat-
tery grade foil, 99.9 %, Rockwood Lithium, USA) was used. Two glass
fiber separators (glass microfiber filter, 691, VWR, Germany) punched
to a diameter of 11 mm were used between working and counter
electrode, and one at the reference electrode (diameter of 10 mm).
80 μL of LP57 electrolyte (1 M LiPF6in EC:EMC 3:7 wt/wt,
<20 ppm H2O, BASF, Germany) were used between working and
counter electrode and 40 μL were added to the reference electrode
side. The cells were cycled in a climate chamber (Binder, Germany) at
25◦C with a battery cycler (Series 4000, Maccor, USA). All cells were
cycled 300 times at 1 C with a lower cutoff of 3 V and an upper cutoff
of 4.2, 4.4, or 4.6 V for NMC111 and NMC622, and 4.0, 4.1, 4.2 V
for NMC811. Prior to cycling, the formation of the cells was done
with 2 cycles at 0.1 C in the voltage range between 3 V and 4.2 V.
If the upper cutoff was >4.2 V, the first cycle after formation, i.e.,
the third cycle of the cell was also done at 0.1 C to the speci-
fied upper cutoff. For upper cutoff voltages <4.2 V, i.e., for the
NMC811 cells also the upper cutoff during formation was adapted
to this voltage. The C-rate was referenced to the approximate re-
versible capacity of the NMC at 1 C: i) 140, 160, and 180 mAh/g
for NMC111 at upper cutoff voltages of 4.2, 4.4, and 4.6 V,
respectively; ii) 160, 180, and 200 mAh/g for NMC622 at upper
cutoff voltages of 4.2, 4.4, and 4.6 V, respectively; and, iii) 130, 150,
and 170 mAh/g for NMC811 at cutoff voltages of 4.0, 4.1, and 4.2 V,
respectively. The charge was done in constant current-constant voltage
(CCCV) mode with a current limitation corresponding to C/20, while
the discharge was done in constant current (CC) mode. Two cells were
built for each combination of NMC material and cutoff voltage and
the error bars in the figures represent the standard deviation from two
cells for each combination.
For recording dq/dV plots, NMC-graphite full-cells were assem-
bled as described above and were cycled in a climate chamber (Binder,
Germany) at 25◦C with a battery cycler (Series 4000, Maccor, USA).
The formation of the cells was done at 0.1 C (two cycles) in the volt-
age range between 3 V and 4.2 V. In the third cycle, the cutoff voltage
was increased to 4.8 V. The lower cutoff was kept constant at 3 V. The
dq/dV plot of the third cycle will be shown in the Results section.
On-line electrochemical mass spectrometry (OEMS).—Two dif-
ferent types of OEMS experiments were designed and performed with
either NMC or spinel LNMO (diameter 15 mm) as working electrode,
and either metallic lithium or graphite as counter electrode. With
metallic lithium as counter electrode (lithium metal foil, diameter
16 mm, thickness 0.45 mm, battery grade foil, 99.9 %, Rockwood
Lithium, USA), two H2013 polyolefin separators (diameter 28 mm,
Celgard, USA) and 120 μL of LP57 electrolyte (1 M LiPF6in
EC:EMC 3:7 wt/wt, <20 ppm H2O, BASF, Germany) were employed.
The cells were charged up to 5 V at a 0.05 C-rate (referenced to the
theoretical capacities of NMC111, NMC622, NMC811, and LNMO
of 277.8 mAh/g, 276.5 mAh/g, 275.5 mAh/g, and 147 mAh/g, respec-
tively). The loadings of the cathodes were 15.8 mg/cm2(NMC111),
15.5 mg/cm2(NMC622), 15.0 mg/cm2(NMC811), and 17.5 mg/cm2
(LNMO). All electrodes were coated on H2013 separator (Celgard,
USA).
With graphite as the counter electrode (diameter 16 mm, see upper
section for details on the type of graphite), two glass fiber separators
(diameter 28 mm, glass microfiber filter, 691, VWR, Germany) and
400 μLof1.5MLiPF
6in ethylene carbonate (EC, BASF, Germany)
were employed. The mixture of EC with LiPF6is a liquid at room tem-
perature due to the melting point depression caused by the addition
Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017) A1363
of LiPF6. The cells were cycled 4 times at a 0.2 C-rate (referenced
to the above given theoretical capacities of the NMC materials) in
the voltage range 2.6–4.8 V for NMC111 and NMC622 and from
2.6–4.4 V for NMC811. The loadings of the cathode active ma-
terial were 9.4 mg/cm2(NMC111), 11.4 mg/cm2(NMC622), and
9.3 mg/cm2(NMC811) and the electrodes were coated on stainless
steel mesh (see above for details). The graphite counter electrode was
capacitively 1.4-fold oversized (referenced to the theoretical capaci-
ties of NMC and graphite).
All cells were assembled in a glove box with argon atmosphere
(O2and H2O<0.1 ppm, MBraun, Germany). The cells were placed
in a climate chamber at 25◦C (Binder, Germany) and connected to the
potentiostat (Series G300 potentiostat, Gamry, USA) and the mass
spectrometer system, which has been described in detail elsewhere.39
The cells were held at OCV for 4 h before starting the above described
protocols. The gas evolution during the OCV and the charging/cycling
period was recorded by OEMS. All mass signals were normalized to
the ion current of the 36Ar isotope to correct for fluctuations of pressure
and temperature. Conversion of the ion currents to concentrations was
done for O2,CO
2,H
2,C
2H4, and CO using calibration gases (Ar with
2000 ppm each of H2,O
2,C
2H4,andCO
2as well as Ar with 2000
ppm each of H2,O
2, CO, and CO2; Westfalen, Germany) and based
on a cell volume of 9.5 cm3. Unlike all other gases quantified in this
work, CO does not have a unique m/z channel. Therefore the amount
of CO was determined on channel m/z =28, which was corrected
for the fractions of C2H4and CO2both of which give additional
intensity to channel m/z =28. The fraction of the signal on channel
m/z =28 stemming exclusively from CO was therefore calculated
as the total signal on channel m/z =28 subtracted by 1.51 times the
signal on channel m/z =26 and 0.14 times the signal on channel
m/z =44, with the factors 1.51 and 0.14 being the fractions of C2H4
and CO2on channel m/z =28 compared to their unique signals on
m/z =26 (C2H4) and m/z =44 (CO2), respectively. These factors were
determined by flowing pure C2H4and CO2through the OEMS cell
and recording the resulting signals originating from the pure gases.40
Results
Electrochemical cycling of NMC-graphite cells.—Figure 1a
shows the cycling stability of NMC111-graphite full-cells with differ-
ent upper cutoff voltages of 4.2, 4.4, and 4.6 V. The cells have a very
stable cycling performance for upper cutoff voltages of 4.2 V and 4.4 V
(black and gray lines), however, cycling to 4.6 V leads to a fast ca-
pacity fading (light gray line), which is in agreement with previous
reports in the literature.9,11,13 While the error bars are hardly visibly
for cutoff voltages ≤4.4 V and at low cycle numbers at a cutoff of
4.6 V, the error bars at higher cycle numbers significantly increase,
which is due to the delayed onset of the so-called rollover-failure for
the two cells. This failure mechanism was described previously by
Dubarry et al. and Burns et al. and was shown to be due to growing
kinetic resistances or more generally an impedance buildup.41,42 In
our data the increasing polarization stems almost exclusively from
the NMC cathodes, which will be discussed below. The coulombic
efficiencies (right axis in Figure 1a) for cells cycled to 4.2 V and
4.4 V are in average >99.9%, indicating the absence of major side
reactions. When the end-of-charge voltage is increased to 4.6 V, the
coulombic efficiency decreases to ∼99.6% (before the onset of the
rollover-failure), reflecting an increasing loss of cyclable lithium. A
further decrease of the coulombic efficiency is observed at the on-
set of the rollover-failure. On the other hand, any increase in the
polarization during cell discharge can be monitored by plotting the
charge-averaged mean discharge voltage, defined as:
¯
Vdischarge =∫Vdischarge ·dqdischarge /∫dqdischarge [1]
As the cells were cycled with a lithium reference electrode,
¯
Vdischarge can be determined independently for the NMC111 cathode
(≡¯
Vcathode
discharge) and the graphite anode (≡¯
Vanode
discharge) for each end-of-
charge voltage as a function of the cycle number, which is depicted in
Figure 1b by the solid and dashed lines, respectively. While the energy
Figure 1. (a) Specific discharge capacity and coulombic efficiency of
NMC111-graphite cells and (b) charge-averaged mean discharge voltage (s.
Eq. 1) of the NMC111 cathode (≡¯
Vcathode
discharge; solid lines) and the graphite anode
(≡¯
Vanode
discharge; dashed lines) vs. cycle number in LP57 electrolyte (1 M LiPF6
in EC:EMC 3:7) operated with different upper cutoff voltages (4.2 V, 4.4 V,
4.6 V) and a constant lower cutoff voltage of 3.0 V. Formation was done at
a rate of 0.1 C. Cycling was performed at 1 C and 25◦C. For each condition,
two independent cells were run and the data in the figure always represent
the average of two cells (the error bars in (a) represent the standard deviation
between the two cells).
fading of the cells is further detailed in the Discussion section, it may
be noted here that the discharge energy for each cycle corresponds
to the product of capacity and ¯
Vdischarge =¯
Vcathode
discharge −¯
Vanode
discharge.Un-
der conditions where the loss of cyclable lithium is the only aging
mechanism, i.e., in the absence of an impedance buildup, ¯
Vcathode
discharge for
cathode active materials with a strongly sloping charge/voltage curve
like NMC would be expected to gradually increase with the number
of cycles. This can indeed be seen when cycling with an upper cutoff
potential of 4.2 V (solid black line in Figure 1b). On the other hand,
when impedance buildup becomes dominant, ¯
Vcathode
discharge decreases with
the number of cycles, as can be seen when the upper cutoff poten-
tial reaches 4.6 V (solid light gray line in Figure 1b). Interestingly,
the charge-averaged mean discharge voltages of the graphite anodes
(¯
Vanode
discharge) remain fairly constant over the complete number of cycles,
even at high end-of-charge voltages. This suggests that a crucial con-
tributing factor for the fast capacity fading of the NMC111-graphite
cells at an upper cutoff of 4.6 V is a strong impedance buildup on
the NMC111 cathode rather than on the graphite anode. In fact, pre-
vious reports in the literature showed a drastic rise of the low fre-
quency semicircle in the impedance spectra of NMC111-graphite11
and NMC442-graphite cells,43,44 which was attributed to the positive
electrode. Later, Petibon et al. showed that the increase of impedance
in NMC442-graphite cells operated at high cutoff potentials, indeed
stems from the positive electrode, proven by using symmetric cells.45
Even though these results are consistent with our observations on
the charge-averaged mean discharge voltage (Figure 1b) one has
to be careful since an additive-containing electrolyte was used in
A1364 Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017)
Figure 2. (a) Specific discharge capacity and coulombic efficiency of
NMC622-graphite cells and (b) charge-averaged mean discharge voltage (s.
Eq. 1) of the NMC622 cathode (≡¯
Vcathode
discharge; solid lines) and the graphite anode
(≡¯
Vanode
discharge; dashed lines) vs. cycle number in LP57 electrolyte (1 M LiPF6
in EC:EMC 3:7) operated with different upper cutoff voltages (4.2 V, 4.4 V,
4.6 V) and a constant lower cutoff voltage of 3.0 V. Formation was done at
a rate of 0.1 C. Cycling was performed at 1 C and 25◦C. For each condition,
two independent cells were run and the data in the figure always represent
the average of two cells (the error bars in (a) represent the standard deviation
between the two cells).
References 43–45, which likely causes a different surface film forma-
tion and impedance. A detailed discussion about the reason for the rise
in the polarization of NMC111 with upper cutoff potential is given in
the Discussion section.
Figure 2a shows the cycling stability of NMC622-graphite cells.
Similar to the case of the NMC111-graphite cells also NMC622-
graphite cells can be cycled stably to upper cutoff voltages of 4.2 V
and 4.4 V with excellent coulombic efficiencies of >99.9%, whereas
at an upper cutoff potential of 4.6 V, the capacity fades rapidly and
the coulombic efficiency decreases to ∼99.6% (before the rollover-
failure), as was observed for NMC111. In analogy to the cells with
NMC111, the occurrence of a rollover-failure41,42 at 4.6 V cutoff indi-
cates growing polarization and causes the large error bars at high cycle
numbers as described above. Also with respect to the mean discharge
voltages, NMC622 (s. Figure 2b) is very similar to NMC111: ¯
Vcathode
discharge
slightly increases with cycle number for 4.2 V cutoff voltage, remains
essentially constant for 4.4 V cutoff voltage, and decreases rapidly at
4.6 V cutoff voltage, proving a continuous impedance growth of the
cathode active material which causes the rollover-failure; on the other
hand, ¯
Vanode
discharge remains essentially constant, independent of the cutoff
voltage.
Figure 3a displays the cycling performance of LiNi0.8Mn0.1 Co0.1O2
(NMC811)-graphite cells. Due to the less stable cycling behavior of
NMC811, the upper cutoff voltages were limited to 4.0 V, 4.1 V, and
4.2 V. It can be observed that the NMC811 only performs fairly stable
with a coulombic efficiency >99.9%, when the upper cutoff voltage
is set to 4.0 V. For cutoff potentials of 4.1 V and 4.2 V, poor cycling
Figure 3. (a) Specific discharge capacity and coulombic efficiency of
NMC811-graphite cells and (b) charge-averaged mean discharge voltage (s.
Eq. 1) of the NMC811 cathode (≡¯
Vcathode
discharge; solid lines) and the graphite anode
(≡¯
Vanode
discharge; dashed lines) vs. cycle number in LP57 electrolyte (1 M LiPF6
in EC:EMC 3:7) operated with different upper cutoff voltages (4.0 V, 4.1 V,
4.2 V) and a constant lower cutoff voltage of 3.0 V. Formation was done at
a rate of 0.1 C. Cycling was performed at 1 C and 25◦C. For each condition,
two independent cells were run and the data in the figure always represent
the average of two cells (the error bars in (a) represent the standard deviation
between the two cells).
stability is observed. In order to aid the comparison between the dif-
ferent NMCs, the capacity retentions measured between the 5th and
the 300th cycles at a 1C-rate for all cells presented in Figures 1–3are
summarized in Table I. Stable cycling with capacity retentions ≥90 %
is possible for NMC111 and NMC622 up to 4.4 V and for NMC811
only up to 4.0 V, whereby its capacity retention is still clearly lower
than that for the cells with NMC111 and NMC622 cycled to 4.4 V. It
is interesting to note that the measured specific capacity of NMC811
at a 4.2 V cutoff is similar to the one of NMC622 at 4.4 V (see values
in parentheses in Table I), however, with the latter one still having a
stable cycling performance. The impact of the different cutoff volt-
ages on the specific energy of the cells will be picked-up again in
the Discussion section. The coulombic efficiencies for the NMC811-
graphite cells are >99.9% at 4.0 V cutoff potential, and even at 4.1 V
and 4.2 V, where pronounced capacity fading is observed, their
coulombic efficiency remains at ∼99.9%, i.e., similar to that of the
NMC111 and NMC622 cells at 4.4 V. The fact that the latter display
substantially lower capacity fading suggests that its origin must be an
enhanced cathode and/or anode impedance growth.
The mean discharge voltages versus cycle number of the NMC811
cathodes and the graphite anodes are shown in Figure 3b. Different
from the constant or even slightly increasing ¯
Vcathode
discharge values with cycle
number observed for NMC111 and NMC622 at 4.2 V cutoff potential,
the NMC811 cells display a continuously decreasing ¯
Vcathode
discharge value,
even at the lowest cutoff potential of 4.0 V. At 4.2 V cutoff, ¯
Vcathode
discharge
drops as rapidly with cycle number for NMC811 as in the case of
Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017) A1365
Table I. Measured capacity retentions between the 5th and 300th cycle of the NMC-graphite cells shown in Figures 1–3. The values in brackets
are the specific capacities in units of mAh/gNMC of the 5th and the 300th cycles.
4.0 V 4.1 V 4.2 V 4.4 V 4.6 V
NMC111 - - 93% (140.2 →130.0) 94% (162.8 →153.2) 42% (183.4 →77.1)
NMC622 - - 95% (155.4 →147.3) 94% (177.8 →166.8) 39% (203.1 →79.9)
NMC811 90% (131.9 →118.1) 77% (149.3 →114.4) 66% (172.5 →114.7) - -
4.6 V for NMC111 and NMC622, indicating that the observed strong
cathode impedance growth sets in at ∼0.4 V lower cutoff potentials
for NMC811. On the other hand, the mean discharge potentials for
the graphite anode in the NMC811-graphite cells ( ¯
Vanode
discharge)behave
similarly as for the NMC111 and the NMC622 cells, showing negligi-
ble increase with cycle number for all cutoff potentials. In summary,
the observed capacity decay at >4.0 V cutoff potential for NMC811
full-cells and at >4.4 V for NMC111 and NMC622 full-cells seems to
be largely related to the onset of a strong cathode impedance growth
(i.e., a strong fading of ¯
Vcathode
discharge) above these cutoff potentials.
The above results clearly demonstrate a similarity between
NMC111 and NMC622, but a big difference to NMC811 with re-
spect to the onset of the cathode impedance growth. To investigate
the origin of this difference and to find the reason for the instability
occurring for NMC111 and NMC622 at 4.6 V and for NMC811 at
4.1–4.2 V, a dq/dV plot of the delithiation and lithiation of the three
NMC materials in NMC-graphite cells of the 3rd cycle is depicted in
Figure 4. The voltage region up to 3.8 V is very similar for all three
NMC compositions, with two anodic peaks between 3.4 V and 3.8 V.
While the first one originates from the lithium intercalation into the
graphite anode, the second one stems from the phase transition from
a hexagonal to a monoclinic (H1 →M) lattice of the NMC.18,46–49
In the region >3.8 V, it becomes very obvious that the dq/dV curve
for the NMC811 cell deviates substantially from that of the NMC111
and NMC622 cells. In particular, NMC811 has a small anodic fea-
ture at ∼3.95 V and a large anodic peak at ∼4.15 V, both of which
are absent for the other NMCs. The first one belongs to the M →
H2 phase transition and the latter one corresponds to the H2 →H3
phase transition as was reported before for LiNiO246–48 and Ni-rich
NMC18,49 materials. In contrast, for NMCs with Ni-contents <80%
the M →H2 and H2 →H3 phase transitions have not been reported.
Accordingly, for NMC111 and NMC622 such distinct features are not
Figure 4. Differential capacity vs. cell voltage of NMC-graphite cells
recorded at a 0.1 C-rate (3rd cycle). The vertical dotted lines mark the up-
per cutoff voltages, which were chosen for the cells in Figures 1–3. The peaks
are assigned to their corresponding phase transitions with H1, H2 and H3 rep-
resenting the three hexagonal phases and M the monoclinic one. C6→LiCx
indicates the lithiation of graphite.
observed. However, for NMC622 a broad peak around 4.1 V is visible,
which might indicate an M →H2 phase transition. For both NMC111
as well as NMC622, a clear redox peak is observed at 4.6 V, which
could correspond, in analogy to NMC811, to a H2 →H3 phase tran-
sition or could also indicate an oxygen redox feature, a process which
has been suggested for Li2Ru1-ySnyO350 and Li2IrO351 by Tarascon’s
group and was investigated theoretically using DFT.52,53 The vertical
dotted lines mark the upper cutoff voltages which were chosen for the
cells presented in the Figures 1–3. Note that up to the onset of the H2
→H3 phase transition of NMC811 at >4.0 V and up to the onset of
the redox feature at >4.4 V of NMC111 and NMC622, the capacity
retention of the materials is very stable. In other words, stable cycling
was only possible if the cutoff voltage was below the onset of the last
peak in the dq/dV plot. The early onset of the H2 →H3 transition at
>4.0 V (NMC811) explains why NMC811 cannot be cycled stably at
>4.0 V cutoff voltages, whereas NMC111 and NMC622 cells show
an excellent performance at potentials as high as 4.4 V. The detri-
mental effect of the H2 →H3 phase transition was already described
before for LiNiO2and NMC811 and was explained by a significant
reduction of the unit-cell volume upon this phase transition, which we
will critically review in the Discussion section.18,47
Gas analysis of NMC-Li and LNMO-Li half-cells by OEMS.—
Figure 5shows the results of On-line Electrochemical Mass Spectrom-
etry (OEMS) measurements with NMC-Li and LNMO-Li half-cells.
For these experiments, metallic lithium was chosen as a counter-
electrode in order to achieve a stable reference potential. Figure 5a
displays the voltage profiles of NMC111 (black), NMC622 (red),
NMC811 (green) as well as LNMO (blue) upon the first charging
from OCV to 5 V at a 0.05 C-rate and 25◦C as a function of the state-
of-charge (SOC) (note that 100% SOC is defined as the removal of
all lithium from the cathode materials; s. Experimental section). The
three lower panels show the total moles of evolved gas, normalized to
the BET surface area of the cathode active material (CAM) in units of
μmol/m2
CAM for O2(Figure 5b), CO2(Figure 5c), and CO (Figure 5d).
Note that normalization of the gassing data to the BET surface area
is meant to account for the differences in the available surface area
for electrochemical oxidation reactions. Figure 5b demonstrates that
for all three NMC compositions a release of oxygen can be detected
near a state-of-charge of ∼80–90%, corresponding to onset potentials
for O2evolution of ∼4.3 V vs. Li/Li+(or ∼4.2 V cell voltage in
a full-cell vs. graphite) for NMC811 and of ∼4.7 V vs. Li/Li+(or
∼4.6 V cell voltage in a full-cell vs. graphite) for NMC111 and
NMC622 (this will be seen more clearly later, when discussing Figure
6). The observed onset for O2evolution on NMC111 at ∼80% SOC
during electrochemical delithiation (s. Figure 5) is in surprisingly good
agreement with the observed onset for oxygen loss upon the chemical
delithiation of NMC111 (with NO2BF4), which was found to initiate
at a lithium content corresponding to ∼75% SOC.6The scatter in the
reported O2concentration of ca. ±0.5 μmolO2/m2CAM for NMC111
and NMC622 and of ca. ±1μmolO2/m2CAM for NMC811 corresponds
to our experimental error in quantifying the O2concentration of ca.
±10 ppm. As was already reported previously,54 no O2evolution is
observed for the LNMO half-cell up to 5.0 V.
At roughly the same potentials at which the evolution of O2initi-
ates, a strong increase of the CO2(Figure 5c) and the CO (Figure 5d)
evolution rates (i.e., an increase in the slope of the lines) is observed
for all NMC materials. Here it should be noted that the more gradual
increase of the CO2concentration (Figure 5c) starting at low SOC
A1366 Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017)
Figure 5. (a) Cell voltage vs. specific capacity of NMC-Li cells using
NMC111 (black), NMC622 (red), NMC811 (green), and LNMO (blue). The
cells contain 120 μL LP57 electrolyte (1 M LiPF6in EC:EMC 3:7) and Celgard
H2013 separators. The total moles of evolved gases in the OEMS cell, nor-
malized by the cathode active material (CAM) BET area versus the theoretical
state-of-charge (SOC) is shown for (b) O2,(c)CO
2,and(d)CO.
values is believed to be due to the electrooxidation of Li2CO3im-
purities (reported to occur in the potential range above ∼3.7 V55 to
∼4.0 V12) and possibly of transition metal carbonates in the first
charge depicted in Figure 5. In this case, one would expect that the
CO2evolution at low voltages (i.e., below ∼4.2 V) would be absent in
the second charge, which indeed is the case (see discussion of Figures
7–9). In addition, the fact that the evolution of CO does not occur
until the onset of O2evolution (s. Figure 5d) is consistent with the
assumption that the CO2evolution at lower potentials is due to the
oxidation of carbonate impurities. The very similar onsets of O2,CO
2,
Figure 6. (a) Specific differential capacity vs. cell voltage of the NMC-Li
cells shown in Figure 5. (b) Evolution of O2as a function of the cell voltage.
The OEMS data are smoothed, baseline corrected, and converted into units of
[μmol/m2
NMC].
and CO evolution raise the question whether the formation of CO2
and CO at higher potentials is only due to the electrooxidation of the
electrolyte and/or the conductive carbon on the cathode surface, or if it
is linked to the release of highly reactive oxygen (e.g., atomic oxygen
or singlet oxygen) from the NMC lattice and its subsequent chemical
reaction with electrolyte and/or conductive carbon to CO and CO2.
We will present a detailed answer to this fundamental question in the
Discussion section, and first present the other experimental results.
Again, in contrast to the data shown for the NMC materials, no CO
evolution is observed for the LNMO half-cell up to 5.0 V, and only
minor amounts of CO2(∼10 μmol/m2
CAM) are formed at ∼15% SOC
(corresponding to ∼4.5 V), which are likely due to the oxidation of
low amounts of carbonate impurities on the surface of LNMO. This
is at variance with Luo et al.,35 who observed the formation of CO2
on LNMO surfaces above 4.75 V (at room temperature), which they
suggested to be due to the electrooxidation of electrolyte. While we
cannot explain this discrepancy, we do not observe any significant
CO/CO2formation on LNMO at >4.7 V and up to 5.0 V (i.e., after
the initial formation from presumably surface impurities), so that we
believe that the electrochemical oxidation of the electrolyte and/or the
carbon support is negligible on LNMO surfaces up to 5.0 V at 25◦C
(this is consistent with our previous OEMS study54). The fact that
hardly any gas evolution is observed at operating voltages as high as
5 V for LNMO but that significant CO/CO2formation is observed for
the NMC materials at >4.2 V, supports the hypothesis that the CO
and CO2evolution is at least partially a consequence of the release
of reactive oxygen from the NMC lattice. A catalytic effect of Ni
or Co on the electrolyte oxidation also appears unlikely, as the gas
evolution for NMC111 and NMC622 shows great similarity, although
the materials differ in both the Ni and the Co content. More clearly,
a catalytic effect of Ni species can be ruled out due to their presence
in LNMO, which evidently shows insignificant gas evolution up to
5.0 V.
Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017) A1367
Figure 7. (a) Cell voltage vs. time of a NMC111-graphite cell over four
charge/discharge cycles at 0.2 C and 25◦C between 2.6 and 4.8 V, in a
cell containing 400 μLof1.5MLiPF
6in ethylene carbonate (EC), glass-
fiber separators and 16.69 mg NMC111. (b) Evolution of CO2(dark blue),
H2(green), C2H4(orange), CO (blue), and O2(black, 10-fold magnified)
as a function of time. Solid lines indicate the gases stemming from the
NMC electrode and dashed lines from the graphite electrode; gas concen-
trations are referenced to the NMC BET area (left y-axis) and to the sum of
graphite and conductive carbon BET area (right x-axis). The OEMS data are
smoothed, baseline corrected, and converted into units of [μmol/m2
NMC]and
[μmol/m2
C].
In order to better visualize the gas evolution at the H2 →H3
phase transition for NMC811 and at the redox feature at ∼4.6 V for
NMC111 and NMC622, which were shown to be detrimental for the
cycling stability, the charging curves of the NMC materials in Figure 5
are now plotted in their dq/dV representation and the corresponding O2
evolution data are shown as a function of the potential (see Figure 6).
The observed peaks in the specific differential capacity vs. voltage plot
(Figure 6a) are in good agreement with the features observed in Figure
4(note that the positive shift of ∼0.1 V in the peak positions in Figure
6a is due to the fact that in Figure 4the full-cell potential is plotted,
whereas in Figure 6a the potential is plotted vs. Li). Figure 6b depicts
the O2evolution and demonstrates clearly that the onset potential of
O2evolution fits very well to the H2 →H3 phase transition (NMC811)
and to the redox feature at ∼4.6 V (NMC111 and NMC622), indicating
that the release of oxygen is not related to a specific potential, but is
rather depending on the occurrence of this very last peak in the dq/dV
plot.
Gas analysis of NMC-graphite full-cells by OEMS.—In order to
investigate if oxygen release occurs only in the first cycle or also in
the subsequent ones, the gas evolution was measured for all three
NMC materials cycled four times in a full-cell setup at 0.2 C vs.
a graphite anode. In order to avoid signal fluctuations (i.e., on the
oxygen channel m/z =32) coming from the transesterification of
the linear carbonate EMC,40,56–58 the LP57 electrolyte is replaced
by 1.5 M LiPF6in EC for these full-cell experiments. Additionally,
due to the low vapor pressure of EC, the background signals from
the electrolyte decrease by two orders of magnitude, leading to an
improved signal to noise ratio in the mass spectrometer.59 Since we
are particularly interested in the oxygen release occurring at the last
peak in the dq/dV plot (see Figure 4and Figure 6), the upper cutoff
potentials were 4.8 V for NMC111 and NMC622, and 4.4 V for
NMC811 (compare to features in Figure 4and Figure 6a). The first
four cycles of the NMC111-graphite cell are depicted in Figure 7a
Figure 8. (a) Cell voltage vs. time of a NMC622-graphite cell over four
charge/discharge cycles at 0.2 C and 25◦C between 2.6 and 4.8 V, in a cell
containing 400 μLof1.5MLiPF
6in ethylene carbonate (EC), glassfiber
separators and 20.23 mg NMC622. (b) Evolution of CO2(dark blue), H2
(green), C2H4(orange), CO (blue), and O2(black, 10-fold magnified) as a
function of time. Solid lines indicate the gases stemming from the NMC
electrode and dashed lines from the graphite electrode; gas concentrations are
referenced to the NMC BET area (left y-axis) and to the sum of graphite and
conductive carbon BET area (right x-axis). The OEMS data are smoothed,
baseline corrected, and converted into units of [μmol/m2
NMC]and[μmol/m2
C].
together with the corresponding evolution/consumption of CO2,H
2,
O2, CO, and C2H4in Figure 7b. From the beginning of the first charge,
a steep increase of the ethylene signal (dashed orange line) is observed,
which is caused by the reduction of EC in the course of the SEI
formation on the graphite electrode.12,60–62 Once the SEI is formed,
the reduction of EC stops, so that the ethylene concentration stays
constant at around 8 μmol/m2
C(s. right-hand y-axis), an amount equal
to ∼1.2 monolayers of the main EC reduction product lithium ethylene
dicarbonate (LEDC) on the graphite anode.12 Simultaneously with the
ethylene evolution, roughly 1.2 μmol/m2
Ccarbon monoxide (dashed
light-blue line) are evolved (after ca. 2.5 hours), which are typically
ascribed to a minor EC reduction pathway with the ring opening at
the carbonyl carbon atom.12,63 Subsequently, the CO signal shows
a stepwise increase, which will be discussed in the next paragraph.
Furthermore, hydrogen (dashed green line) starts to evolve from the
beginning of the measurement, due to the reduction of trace water in
the electrolyte.12,64 The H2signal initially evolves at a fast rate and
then gradually approaches a concentration of ∼12 μmol/m2
Cby the
end of the measurement. The reason why the H2evolution does not
stop after the first charge like the C2H4evolution is, we believe, caused
by the formation of protic species from electrolyte decomposition and
their subsequent reduction at the graphite anode yielding continuous
hydrogen evolution.12
Besides C2H4, CO, and H2one can also observe a linear increase
of the CO2concentration in the first four hours of the measurement
(up to ∼4.6 V cell potential) to 50 μmol/m2
NMC. This increase can
be assigned to the oxidation of carbonate impurities on the NMC
particles, which are typically around 0.1 %wt.12,18 The total CO2signal
of 50 μmol/m2
NMC corresponds to ∼50 μmol/m2
NMC ·0.26 m2/g ·
16.69 mgNMC =217 nmol CO2or to 217 nmol of carbonate (in the
case of Li2CO3, it was shown, that one mole of Li2CO3releases
one mole of CO2upon electrochemical oxidation65). If referenced to
Li2CO3(73.89 g/mol), which is customarily done, this would amount
to 16 μgLi
2CO3equal to 0.10 %wt.
A1368 Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017)
Figure 9. (a) Cell voltage vs. time of a NMC811-Graphite cell over four
charge/discharge cycles at 0.2 C and 25◦C between 2.6 and 4.4 V, in a cell
containing 400 μLof1.5MLiPF
6in ethylene carbonate (EC), glassfiber
separators and 16.40 mg NMC811. (b) Evolution of CO2(dark blue), H2
(green), C2H4(orange), CO (blue), and O2(black, 10-fold magnified) as a
function of time. Solid lines indicate the gases stemming from the NMC
electrode and dashed lines from the graphite electrode; gas concentrations are
referenced to the NMC BET area (left y-axis) and to the sum of graphite and
conductive carbon BET area (right x-axis). The OEMS data are smoothed,
baseline corrected, and converted into units of [μmol/m2
NMC]and[μmol/m2
C].
After four hours, the cell voltage reaches ∼4.6 V and oxygen starts
to evolve. The onset potential fits very well to the one found in Figure
6(note that in the Figures 7–9the NMC–graphite full-cell potential is
reported, whereas in Figure 6the potential is measured vs. a metallic
lithium counter electrode, the potential of which is ∼0.1 V below the
potential of lithiated graphite). Simultaneously to O2,theCOandCO
2
signals increase until the cell switches from the CV-phase at 4.8 V to
the discharge, from which point on the CO and CO2concentrations
stay constant until the cell voltage again increases above ∼4.6 V in
the following cycles, where O2, CO, and CO2evolve again, leading to
a stepwise increase of these signals. The fact that after the first cycle
no CO2is evolved below ∼4.6 V in any of the subsequent cycles
confirms our prior hypothesis that the CO2evolution below 4.6 V
is due to the oxidation of carbonate impurities in the first cycle (s.
discussion of Figure 5). By subtracting the amount of CO evolved
during the SEI formation (∼1.2 μmol/m2
Cor ∼25 μmol/m2NMC)and
the amount of CO2related to carbonate impurity oxidation in the first
cycle (∼50 μmol/m2
NMC), the total amount of CO and CO2evolved
exclusively due to processes at high voltage after the four cycles are
∼80 μmol/m2
NMC and ∼180 μmol/m2
NMC, respectively.
While the step-like profile of the oxygen signal is similar to that
of the CO and CO2signal, showing a rapid rise every time the po-
tential goes above ∼4.6 V, it does exhibit a superimposed potential-
independent continuous decrease. This consumption of oxygen in the
cell is most likely caused by a slow but steady reduction of the evolved
oxygen at the graphite anode, which would be consistent with the ob-
served decreasing consumption rate over time, as a more protective
SEI is being formed (note that an analogous consumption of CO2is
observed in the second and, to a much lesser degree in the third cycle,
which appears smaller in magnitude than the O2consumption only
due to the fact that the oxygen signal is magnified by a factor of ten).
Thus, in order to estimate the total amount of evolved oxygen over the
four cycles, one can sum up the steep increases of the oxygen signal in
each cycle, which gives a total oxygen evolution of ∼9μmol/m2
NMC.
Note that all values which are summed up over the four cycles are
corrected for the decreasing concentrations due to gas consumption
(for CO2and O2) on the graphite anode by summing up the increases
rather than considering the total concentrations measured at the end
of the experiment. Additionally, it is quite apparent that the amount
of evolved oxygen decreases from cycle to cycle, which would be
consistent with our assumption that the oxygen is released mainly
from the near-surface regions of the NMC particles and that its re-
lease becomes slower as the oxygen depleted surface layer increases
in thickness.36 This hypothesis is also supported by the total amount
of released oxygen, which will be discussed in further detail in the
Discussion section.
The results of the analogous experiment with an NMC622-graphite
cell cycling at 0.2 C between 2.6 V and 4.8 V are shown in Figure
8. For all gases, a very similar trend as for the NMC111-graphite
cell is observed. The total amounts of evolved gases during SEI-
formation are ∼8μmol/m2
Cof ethylene and ∼1.2 μmol/m2
Cof CO.
Additionally, ∼10 μmol/m2
Cof hydrogen is evolved over the course
of the experiment. Prior to the onset of oxygen evolution at ∼4.54
V (vertical dotted line), the oxidation of carbonate impurities re-
sults in ∼19 μmol/m2
NMC CO2. This corresponds to ∼19 μmol/m2
NMC·
0.35 m2/g ·20.23 mgNMC =135 nmol CO2or to 135 nmol of car-
bonate. Again, if referenced to Li2CO3, this would amount to 10 μg
Li2CO3equal to 0.05 %wt. As was observed for NMC111, no CO2
is evolved below ∼4.5 V in any of the subsequent cycles, confirming
our prior hypothesis that the CO2evolution below 4.5 V is due to the
oxidation of carbonate impurities in the first cycle (s. discussion of
Figure 5).
By subtracting the amount of CO evolved during the SEI formation
(∼1.2 μmol/m2
Cor ∼20 μmol/m2NMC) and the amount of CO2related
to carbonate impurity oxidation in the first cycle (∼19 μmol/m2
NMC),
the total amount of CO and CO2evolved exclusively due to pro-
cesses at high voltage after the four cycles are ∼79 μmol/m2
NMC and
∼171 μmol/m2
NMC, respectively. The estimated amount of evolved
oxygen over the four charge/discharge cycles using the above de-
scribed approach is ∼6μmol/m2
NMC. The total amounts of gaseous
species are very similar for both the NMC111-graphite and NMC622-
graphite cells, illustrating once again a similarity also in the gassing
behavior of NMC111 and NMC622 as it was shown in Figure 5and
Figure 6.
Lastly, a similar experiment was performed with an NMC811-
graphite cell (see Figure 9), except that the upper cutoff voltage was
reduced to 4.4 V, as this voltage is sufficient to include the complete
peak stemming from the H2 →H3 transition (see Figure 4). Dur-
ing the SEI formation a total of ∼8μmol/m2
Cof ethylene and ∼1.5
μmol/m2
Cof CO are evolved. These amounts fit very well to the gas
amounts detected in the experiments with NMC111 and NMC622
(see Figure 7and Figure 8), which is expected since the gases at this
initial stage of the first cycle were shown to originate solely from the
graphite electrode,12 i.e., they are independent of the cathode material.
Over the four cycles of the measurement, ∼13 μmol/m2
Cof hydrogen
are formed, which also fits to the amounts measured in the NMC111
and NMC622 cells. The upper cutoff potential can have an influence
mainly on the hydrogen evolution as a result of the cross-talk between
cathode and anode.12,42,66 The underlying assumption of the follow-
ing analysis is that the cross-talk effect be either similar for all the
measurements or has only a minor effect on the overall gas evolution.
Prior to the onset of oxygen evolution at already ∼4.2 V (verti-
cal dotted line), the oxidation of carbonate impurities results in ∼80
μmol/m2
NMC CO2. This corresponds to ∼80 μmol/m2
NMC·0.18 m2/g ·
16.40 mgNMC =236 nmol CO2(or carbonate). Again, if referenced
to Li2CO3, this would amount to 17 μgLi
2CO3equal to 0.11 %wt.
By subtracting the amount of CO evolved during the SEI formation
(∼50 μmol/m2NMC) and the amount of CO2related to carbonate im-
purity oxidation in the first cycle (∼80 μmol/m2
NMC), the total amount
of CO and CO2evolved exclusively at >4.2 V (i.e., after the onset
of oxygen release) after the four cycles are ∼70 μmol/m2
NMC and
∼170 μmol/m2
NMC, respectively. The estimated oxygen release over
the four cycles is ∼8μmol/m2
NMC. Thus, even though the upper cutoff
Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017) A1369
Figure 10. (a) Cell voltage vs. time of a NMC622-graphite cell over the first
charge/discharge cycles at 0.2 C and 25◦C between 2.6 and 4.8 V, in a cell
containing 400 μLof1.5MLiPF
6in ethylene carbonate (EC), glassfiber
separators and 18.40 mg NMC622. The NMC622 electrode was prepared with
13C-labeled carbon instead of Super C65. (b) Evolution of CO2(dark blue),
H2(green), C2H4(orange), CO (blue), and O2(black, 10-fold magnified),
13CO2(gray, 10-fold magnified) and 13CO (bright blue, 10-fold magnified)
as a function of time. Solid lines indicate the gases stemming from the NMC
electrode and dashed lines from the graphite electrode; gas concentrations are
referenced to the NMC BET area (left y-axis) and to the sum of graphite and
conductive carbon BET area (right x-axis). The OEMS data are smoothed,
baseline corrected, and converted into units of [μmol/m2
NMC]and[μmol/m2
C].
potential in the NMC811-graphite cell is reduced by 0.4 V compared
to the NMC111 and NMC622 cells, the amounts of evolved CO, CO2,
and O2are very similar. This finding is remarkable, because if the
CO and CO2were a result of electrochemical electrolyte oxidation a
difference between the amounts measured at 4.4 V and 4.8 V would
be expected, especially since the electrolyte is commonly believed to
be stable against oxidation to CO and CO2at potentials as low as
4.4 V.59,67,68 One could explain the similar amounts of gas with 4.4 V
(NMC811) and 4.8 V (NMC111 and NMC 622) cutoff potentials, if
one were to assume that CO2and CO are actually the result of the oxy-
gen release from the NMC lattice. In consequence, this could mean
that the electrochemical oxidation of carbonate electrolytes would ac-
tually be negligible or at least very low on NMC surfaces at both 4.4
V and even 4.8 V, if there were no release of lattice oxygen, which in
turn would explain the complete absence of CO/CO2up to 4.8 V in
EC-only electrolyte on a carbon black electrode at 25◦C.59 This is also
supported by the fact that LNMO can be operated at a potential of
4.8 V with insignificant CO/CO2evolution due to the absence of
oxygen release (see Figure 5).
Having presented substantial evidence that the CO/CO2evolution
at high potentials is mostly caused by a chemical reaction of the
released lattice oxygen, the question remains whether the evolved
CO/CO2derive from its reaction with the electrolyte or with the con-
ductive carbon in the NMC electrode. Therefore, an NMC622 elec-
trode with 4.4 %wt 13C-labeled carbon as conductive additive was pre-
pared, replacing the Super C65 conductive carbon, such that a reaction
of released lattice oxygen with carbon would result in 13CO/13 CO2,
while its reaction with electrolyte would result in 12CO/12 CO2.The
NMC622-graphite cell with 13C conductive carbon was charged to
4.8 V and subsequently discharged to 2.6 V (see Figure 10a). The
capacity reached during the CC-phase was only 198 mAh/gNMC, i.e.,
∼17% lower than for the NMC622 electrode with Super C65 (see
Figure 8); this inferior electrode performance is likely caused by the
strongly agglomerated structure of the 13C-carbon, resulting in a poor
electronic accessibility of the active material particles in the cathode.
Nevertheless, also for this electrode, the release of oxygen can be
clearly seen. It is shifted to a higher potential of 4.75 V, compared
to the 4.54 V for NMC622 with Super C65 (s. Figure 8), which can
be rationalized by the fact that the material contains more lithium at
∼4.6 V due to the worse cathode performance, which in turn renders
it more stable at this voltage. Additionally, the cutoff potential is only
50 mV above the O2onset, which is the reason for the overall lower
oxygen evolution.
In total, by the end of the first cycle, ∼1.5 μmol/m2
NMC,
∼23 μmol/m2
NMC and ∼9.3 μmol/m2
NMC of O2,CO
2, and CO were
formed, respectively, whereby these values were again corrected by
the ∼13 μmol/m2
NMC CO2stemming from carbonate oxidation prior
to the onset of oxygen evolution and by the ∼1.1 μmol/m2
CCO orig-
inating from EC reduction on the graphite anode. In comparison, in
the first cycle of the NMC622-graphite cell with Super C65 (Figure 8)
∼2.9 μmol/m2
NMC,∼40 μmol/m2
NMC and ∼19 μmol/m2
NMC of O2,CO
2
and CO were evolved, respectively (again, corrected for contributions
from carbonate oxidation and EC reduction). It is interesting to note,
that not only the amount of oxygen is cut in half, but also the amounts
of CO2and CO are cut in half, which shows once more that these gases
are linked to the oxygen evolution. Finally, Figure 10 clearly shows
that neither the evolution of 13CO nor 13CO2was observed, prov-
ing that the carbon additive in the cathode is stable at potentials of
4.8 V and also stable against the released oxygen from the NMC lat-
tice. Therefore, the observed CO/CO2formation at high potentials can
be ascribed to the oxidation of EC (possibly also the binder) rather
than of the conductive carbon by released lattice oxygen.
Discussion
Correlation between oxygen release and surface structure of
NMC.—We first want to focus on the correlation between the H2
→H3 phase transition at ∼4.2 V for NMC811 and the high-voltage
feature at ∼4.6 V of NMC111 and NMC622 observed in the dq/dV
analysis (Figure 4) and the oxygen release detected for the different
NMCs by OEMS (Figures 7–9). For NMC111 it is known that upon
lithium extraction the c-parameter increases until roughly 2/3 of the
lithium is removed and it is ascribed to repulsive interactions of the
negatively charged oxygen layers upon the removal of the positive
lithium ions.7,11 Upon further removal of lithium, i.e., at higher states
of charge, a decreasing c-parameter is reported, which has been linked
to increasing covalency between the metal and the oxygen.7,69 Increas-
ing covalency in principle corresponds to a decrease of the oxygen
anion charge density, i.e., an oxidation of the lattice oxygen anions
(from 2- in the idealized ionic structure to a lower charge density
of the oxygen atom). This hypothesized oxidation of the oxygen an-
ions (recently shown by Tarascon’s group for the model compounds
Li2Ru1-ySnyO3and Li2IrO3)50,51 would also be consistent with the
release of oxygen from the NMC material. The H2 →H3 phase
transition was described in the literature for LiNiO2(LNO), where it
occurs at ∼4.2 V vs. Li/Li+46–48 and Li et al. showed by in-situ XRD
that the c-parameter of the LNO unit cell at low states of charges
gradually increases and drastically shrinks at the H2 →H3 phase
transition.47 At roughly the same voltage, this phase transition also
occurs for NMC811, as described by Noh et al. and Woo et al.,18,49
and the associated volume contraction was hypothesized to lead to
capacity fading.18,49 We believe that in analogy to the interpretation
in the case of NMC111 the shrinkage of the c-parameter for NMC811
at the H2 →H3 phase transition can be also a result of a decreas-
ing repulsion between the oxygen layers, caused by the oxidation of
the oxygen anions, which finally may result in O2release. As was
reported by Strehle et al. on Li-rich NMC (Li1+x(Ni,Mn,Co)1-xO2),
we believe that due to the limited diffusion length of oxygen anions
in the bulk NMC particles at 25◦C, the oxygen release is limited to
the surface-near region yielding a disordered spinel or rock-salt type
layer while the bulk structure stays intact.36
A1370 Journal of The Electrochemical Society,164 (7) A1361-A1377 (2017)
By a detailed investigation of the dq/dV plot shown in Figure 4,
one can observe that the peak assigned to the H2 →H3 phase tran-
sition (NMC811) as well as the high-voltage feature (NMC111 and
NMC622) are reversible. If these features are at least partially re-
lated to oxygen redox, the reversibility of the peak also indicates a
reversibility of the oxygen redox upon relithiation. This reversibility
is not contradicting the experimentally observed irreversible oxygen
loss, since the dq/dV analysis reflects mostly processes in the bulk,
where no oxygen loss can occur due to the limited bulk diffusivity of
oxygen anions in the layered oxide particle. We believe that the oxy-
gen release would likely occur throughout the entire particle, if the
oxygen anion diffusion were fast enough and/or if the NMC particles
were small enough. In other words, the NMC structure is thermody-
namically unstable at high degrees of delithiation, and only retains its
oxygen and its layered bulk structure due to the kinetically hindered
oxygen diffusion. This hypothesis is supported by the literature, where
it is reported that layered NMC21,25,27,70,71 and NCA22,25,26,70 structures
in the charged state (low lithium content) are not stable at high tem-
peratures (>170◦C) and decompose under release of oxygen, forming
a disordered spinel or rock-salt structure, which are thermodynami-
cally more stable than the layered structure at low lithium contents.
In these reports, the oxygen release is a bulk phenomenon due to the
significantly faster oxygen anion diffusion at elevated temperatures.
Consequently, a complete transformation of the layered structure into
the spinel or rock-salt structure is observed.
The limitation of spinel or rock-salt structures to the surface-near
regions was already reported before for various layered oxides.9,28–31
In particular, Muto et al. found for NCA that the rock-salt forma-
tion on the surface can be up to 100 nm thick after 500 cycles at
80◦C.30 Jung et al. investigated NMC532 in the voltage range between
3–4.8 V and found a spinel layer thickness of 12–15 nm and a thickness
of the rock-salt phase of 2–3 nm after 50 cycles at room temperature.9
Abraham et al. observed a 35–45 nm thick rock-salt structure on
LiNi0.8Co0.2 O2after calendaric aging of a charged electrode at 60◦C
for 8 weeks.28 They also stated that the oxygen release was expected
to occur from the surface-near region of the material, as their XAS
and EELS data showed both that the Ni:O and Co:O ratios were twice
as high on the surface compared to the bulk and that the Ni oxidation
states on the surface matched NiO whereas in the bulk it matched
that of Ni in a layered structure.28,29 Even though a release of oxygen
could not be shown in these reports, it is implicitly required because
of the lower oxygen to metal ratios in the spinel and rock-salt phases
compared to the layered structure: MO2(layered) →M3O4(spinel)
→MO (rock-salt) (i.e., metal/oxygen =1:2 →3:4 →1:1).
In Figures 7–9, the amount of released oxygen is largest in the first
cycle and decreases in the subsequent cycles. This fits to the hypothesis
that the oxygen is released only from surface-near regions and is
therefore fastest in the first cycle, and lower in subsequent cycles, since
then it has to diffuse through the already formed disordered spinel or
rock-salt layer. In summary, a clear correlation can be made between
the structural rearrangement of the NMC particle surface and the
release of oxygen. Additionally, the spinel or rock-salt surface layer
is very likely the cause of the increase in the polarization (represented
by a decrease in the charge-averaged mean discharge voltage of the
cathode, ¯
Vcathode
discharge) observed during cycling in the Figures 1–3.
Connection between released O2and evolution of CO and CO2.—
The total amount of oxygen released during the four cycles (Figures
7–9) is similar for all three NMCs, ranging from 6–9 μmol/m2
NMC (see
Tab l e II) and for the chosen upper cutoff potentials there is no apparent
correlation with the Ni or Co contents in the NMCs. Furthermore,
Tab l e II summarizes the measured amounts of CO2and CO within
the four cycles shown in Figures 7–9. They are corrected for the CO2
derived from carbonate oxidation and for the CO originating from
EC reduction, such that only gassing processes at high-voltage are
regarded. As was already discussed, a closer examination of the 2nd,
3rd,and4
th cycle in Figures 7–9reveals that CO/CO2only evolve once
the evolution of O2is observed, confirming that CO/CO2produced at
low potentials in the 1st cycle is indeed due to SEI formation (CO) and
Table II. Total amounts of oxygen, carbon monoxide, and carbon
dioxide evolved at high potentials over the first four cycles in the
cells shown in the Figures 7–9(the amounts of CO2stemming from
oxidation of carbonate impurities as well as the CO originating
from EC reduction, both in the first cycle, were subtracted).
NMC111 NMC622 NMC811
O2[μmol/m2
NMC]9 6 8
CO [μmol/m2
NMC]80 79 70
CO2[μmol/m2
NMC] 180 171 170
carbonate impurity oxidation (CO2). This raises the question, whether
CO and CO2derive from the chemical reaction of the released lattice
oxygen with the electrolyte. A significant reaction of the evolved
oxygen with conductive carbon can be excluded, since it was shown
in Figure 10 that no 13CO and 13 CO2was evolved when 13C labeled
carbon was used as conductive additive in the NMC electrode instead
of conventional carbon (Super C65). Another interesting observation
is that in the case of NMC811-graphite cells, O2, CO, and CO2evolve
already at ∼4.2 V. At this potential, no gas evolution is observed for
the analogous cells with NMC111 (onset of O2evolution at ∼4.57 V)
or NMC622 (onset of O2evolution at ∼4.54 V), so that it is too low
to ascribe the evolved gases to the electrochemical oxidation of the
electrolyte, which strongly supports our hypothesis that the evolution
of O2, CO, and CO2are of the same origin.
The purely electrochemical oxidation of EC-only electrolyte on a
carbon electrode, i.e., in the absence of any possible catalytic effect
by transition metal surfaces, was studied in a recent report by Metzger
et al. by applying a linear sweep voltammetry procedure from OCV
up to 5.5 V with a scan rate of 0.2 mV/s.59 There, the onset of CO2
and CO evolution was at ∼4.8 V vs. Li/Li+, where the sum of the CO
and CO2-evolution rate was determined to be 0.3 μmol/(m2
C·h).59 For
comparison, in the NMC111-graphite cell (Figure 7), the total amount
of CO and CO2produced between 4.6 V and the end of the first
cycle is 59 μmol/m2
NMC and was detected within 1.5 h, corresponding
to an average evolution rate of ∼39 μmol/(m2
NMC·h); if referenced
to the total surface area of conductive carbon and NMC in the cell
(0.052 m2Cand 0.0043 m2
NMC), this equates to ∼3.0 μmol/(m2
NMC+C·h).
Comparing both values and excluding any catalytic effect of the active
material for the above-described reasons, it becomes clear that the
purely electrochemical oxidation of EC can only account for at best
∼10 % of the evolved CO and CO2.
This estimate shows that the electrochemical electrolyte oxidation
occurs to a certain extent at high potentials, consistent with previous
reports in the literature, which show that the voltage of the NMC
slowly drops during storage via a self-discharge caused by electro-
chemical electrolyte oxidation.72–75 However, once the potential is
above the threshold voltage for the release of lattice oxygen, the ma-
jority of CO and CO2generated in cells containing NMC stems from
chemical electrolyte oxidation. A detailed discussion of the chemical
and electrochemical pathways and their ratios on the total electrolyte
oxidation will be presented below.
While the absence of oxygen evolution for the high voltage spinel
makes sense, considering that the spinel phase is the stable phase
which forms upon oxygen release of the layered material, it is inter-
esting that no CO and CO2are evolved with LNMO up to a potential
of 5 V (see Figure 5), on a surface for which one would not expect a
substantially different catalytic effect (if there is any) for the electro-
chemical oxidation of electrolyte than for NMC surfaces. This implies
that the electrolyte should be very stable (i.e. negligible or very minor
electrochemical electrolyte oxidation) at the potentials used for the
NMC-graphite cells, further supporting our hypoth