You are on page 1of 15

An in-situ SEM study of failure modes in sulphide-based

all-solid-state batteries throughout cycling has been As featured in:


performed by a group of researchers from the Laboratory
of Reactivity and Chemistry of Solids, Université de
Picardie Jules Verne in France (LRCS, CNRS UMR 7314).

Study of failure modes in two sulphide-based solid


electrolyte all-solid-state batteries via in situ SEM

The failure modes of two sulphide based all-solid-state


batteries have been investigated in real time throughout
cycling by in-situ SEM. We were able to follow and classify
the observed degradation processes (e.g., dendrite)
into three interconnected failure modes: i) electrical, ii)
electrochemical, and iii) mechanical failures.
See Mathieu Morcrette,
Carine Davoisne et al.,
J. Mater. Chem. A, 2022, 10, 17142.

rsc.li/materials-a
Registered charity number: 207890
Journal of
Materials Chemistry A
View Article Online
PAPER View Journal | View Issue

Study of failure modes in two sulphide-based solid


electrolyte all-solid-state batteries via in situ SEM†
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

Cite this: J. Mater. Chem. A, 2022, 10,


17142
Neelam Ghanshyam Yadav,ab Nicolas Folastre,ab Mickael Bolmont,ab Arash Jamali,c
Mathieu Morcrette *ab and Carine Davoisne *ab

Lithium metal was left unused for a long time in conventional liquid-based batteries, with a fear of catching
fire for safety reasons. However, with advancements in solid electrolytes, these were considered the holy
grail in the development of Li metal-based all-solid-state batteries (ASSBs) alongside NMC which is
considered the next-generation cathode material. Unfortunately, Li metal faces tremendous challenges
when brought into contact with many currently considered promising solid electrolytes (SEs). The
associated challenges are electrical, electro-(chemical) and mechanical instabilities. Surprisingly, if one
studies the literature thoroughly, it seems that batteries based on SEs show much faster dendrite
formation than liquid electrolytes. For this particular study, thanks to in situ and operando SEM, we
studied these sulfide-based SEs, viz. b-Li3PS4 (LPS) and Li6PS5Cl (LPSCl). We highlight the key differences
in failure modes of two sulfide-based SEs, by keeping the anode (Li metal) and cathode (NMC111)
constant. For both the SEs, initially, electro-chemo-mechanical stress is induced, due to volume
expansion of active materials and plating during cycling. However, in the case of LPS, this leads to
electrical failure causing a short circuit, whereas in the case of LPSCl it leads to further mechanical
damage-causing delamination of the cathode. Thus, the porous structure of SEs heavily determines the
mechanical behaviour of the ASSB, the distribution of the induced electro-chemo-mechanical stress and
in turn, the mechanism of ASSB failure. The other striking observations are: (1) cracks travel ahead of the
dendrites (2) cracks grow and propagate at a much faster rate in the case of b-LPS than in LPSCl (3)
higher surface roughness of LPSCl gives rise to many different morphologies of Li deposition/dendrites
(4) the cathode electrolyte interface (CEI) forms and grows at a much faster rate in the case of LPS than
in LPSCl. We demonstrate that differences in cycling performance and mode of failure are due to the
Received 9th March 2022
Accepted 18th July 2022
differences in SEs. We also highlight that the key limitation in implementing Li metal as an anode in
ASSBs is the dendrite formation and mechanical instability. These findings again emphasize the
DOI: 10.1039/d2ta01889f
importance of coating active materials, the introduction of a buffer layer between the SE/Li metal
rsc.li/materials-a interface, coating of SE, or the need for using composite/hybrid or bilayer solid electrolytes at least.

density, thanks to its huge theoretical capacity. However, it


Introduction could not be employed with liquid-based electrolytes, due to
Currently commercial liquid electrolyte-based Li-ion batteries safety reasons. Replacing liquid electrolytes with solid electro-
show excellent performance, thanks to their stable or less lytes could allow the employment of the Li metal anode. Thus,
degrading liquid electrolytes with an anode/cathode interface. all-solid-state batteries (ASSBs) are considered to be next-
However, with increasing global CO2 emissions, it is imperative generation batteries or the future alternative to the current
today to switch from fossil fuels to electric vehicles. This battery technology. They can offer both safety and high energy
requires batteries with high energy and power densities. Usage density. They hold the power to revolutionize the electric
of Li metal as an anode can solve the issues of high energy transportation market. To reach such an achievement, the key
component is the solid electrolyte (SE) and different classes can
a
be found such as polymers, oxides, sulphides, uorides, phos-
Laboratoire de Réactivité et Chimie des Solides (LRCS), CNRS UMR 7314, Université
phides, etc. Sulphide solid electrolytes (see ESI-1† for an over-
de Picardie Jules Verne, 33 Rue Saint Leu, 80039 Amiens Cedex, France. E-mail:
mathieu.morcrette@u-picardie.fr; carine.davoisne@u-picardie.fr view of properties) such as b-Li3PS4 (LPS) and Li6PS5Cl (LPSCl)
b
Réseau sur le Stockage Electrochimique de l'Energie (RS2E), FR CNRS 3459, France are among the few promising SEs heavily explored in the liter-
c
Plateforme de microscopie électronique, Université de Picardie Jules Verne, 33 Rue ature. They can be easily processed simply by cold pressing,
Saint Leu, 80039 Amiens Cedex, France thanks to their ductile nature (young modulus  20 GPa).1 Also,
† Electronic supplementary information (ESI) available. See they offer high ionic conductivity at room temperature (103 to
https://doi.org/10.1039/d2ta01889f

17142 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A

102 S cm1)2,3, remarkably close to the values of liquid elec- at the interphase of Li4P2S6 allowed the battery to cycle stably
trolytes etc. The electrochemical stability window for most of the even at higher current densities without a short circuit. This
known sulphides lies in the range of 1.7–2.3 V.4–6 Although study highlights that interfacial behaviour can be very much
sulphides offer high ionic conductivity values, a lower operating different even if the two SEs belong to the same category of
potential window is still among its big limitations. These are electrolytes. This is because, even if the nature of the formed
important prerequisites for the development of ASSBs. The interphases is similar, the amount and the rate of such degra-
solid electrolyte plays a twofold role in ASSBs, by acting both as dation might be different affecting the total ionic conductivity
a solid electrolyte and a solid separator. Despite such promising at the interphase.
values that are even higher than those for liquid electrolytes The physical/mechanical instability arises rst due to poor
(LP30  10 mS cm1), ASSBs have not been able to achieve solid–solid contact. Even if one establishes a good contact
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

reliable electrochemical performance when compared to during cell fabrication with the application of stack pressure,
commercial batteries. This is due to ASSBs challenges which due to volume expansion in the active materials upon cycling,24
mainly lie in the interfaces.7 Different research groups around this physical contact is subjected to change. Also, the Li elec-
the world have investigated the interfacial challenges8 and they trochemical deposition is known to generate stresses as high as
can be categorized into different types such as chemical or 135 MPa,25 and this further adds to the huge loss of contact,
electrochemical instability,9 and physical or mechanical insta- especially at the anode interface. The phenomena such as Li
bility. All these types of instabilities/challenges impact the key deposition/plating26,27 and Li dissolution/stripping28,29 are
parameter, which is the ion transport at the interfaces. studied by in situ Auger electron microscopy and spectroscopy
The chemical or electrochemical instability arises from solid techniques,30 in situ transmission electron microscopy (TEM),27
ionic conductors and especially in the case of sulphide-based and operando X-ray tomography.31 Plating during the charging
SEs their instability against Li metal10 or with transition metal process is responsible for dendrite30 formation whereas the
oxides. Such SEs oen degrade at higher voltages, forming stripping process is known to form voids.32 Dendrites are
decomposed products known as “interphases” at the anode/ formed due to non-uniform plating. Nagao et al.33 showed via in
cathode interface. LPS and LPSCl when in contact with Li metal situ SEM of the bulk type solid state battery with Li2S–P2S5, that
are known to form Li3P, Li2S and LiX.11 Similarly, transition uniform lithium deposition (plating) and dissolution (strip-
metal oxides, such as LCO or NMC etc. when in contact with ping) are the key to suppression of dendrites and occurrence of
thiophosphate-based SE, oen degrade the SE.12,13 Chemical unfavourable cracks. In situ Auger electron spectroscopy/
characterisation performed on the cathode composite aer microscopy (in situ AES/AEM) showed that Li deposition on the
cycling showed the formation of S and Li2P2S6 (ref. 14 and 15) Li metal anode was strongly dependent upon the applied force
from LGPS/LCO and S, Li2Sn, P2Sx, phosphate-based to the sample.30 Several factors are known to affect this inho-
compounds, and LiCl at the interface of Li6PS5Cl and LCO and mogeneity of plating, such as the local surface topography, non-
NMC111 and LNMO.16 uniform contact between SE/Li,34 applied stack pressure,28
Such chemical or electrochemical instability leads to huge operating current density,35 etc. to name a few. Voids are formed
interfacial resistance as highlighted by electrochemical when the Li ions are removed/stripped at a much faster rate
impedance spectroscopy (EIS)17–19 and X-ray photoelectron than the rate of Li self-diffusion that can replenish back the Li-
spectroscopy (XPS).11,18,20 These formed decomposed products/ ions at this interface. Operando X-ray tomography-based studies
interphases can either allow or block the ion transfer across this on solid state batteries have conrmed the formation of voids at
interface depending on the nature of this formed interphase. the anode interface.31 Such voids reduce the contact area31
The interfaces can thus be broadly categorized into 3 types: (i) creating local current densities that drive dendrite formation
ion transfer interphase considered as non-reactive and ther- during the next plating cycle,28 dendrite penetration and nally
modynamically stable, (ii) electronically conductive interphase cell death via a short circuit. Also, less studied Li surface
which is thermodynamically unstable and results in constant impurities27 and other surface aws are some of the intrinsic
electrolyte decomposition and (iii) insulating interphase also issues associated with the usage of Li metal anodes. A study by
considered as unstable and whose growth depends on its ionic X-ray tomography of Li–LPSCl–Li showed that cracks are formed
conduction properties.21 in the SE, during the process of Li plating and when the Li
Sulphide-based SEs such as LPS and LPSCl fall into the 3rd plated structures relieve their stress around the edge.36 This
category of the interface type described above. Li2S specically study also pointed out the fact that the absence of short circuits
is low in ionic and electronic conductivities and hence forms when even huge cracks are formed, shows that cracks are
a highly resistive SEI at the anode interface that could lead to formed rst and only later does lithium ll up the crack.
the formation of decomposed products that might be insulating The volume expansion in active materials is another leading
in nature resulting in poor kinetics.22 Suyama et al.23 showed reason for physical/mechanical instability. Such volume
that the ionic conductivity of the formed interphase is one of expansion is known to develop stress/strain. This strain leads to
the leading factors in determining the fate of failure. They the cracking of active materials, causing loss of contact at the
studied two sulde based SEs, namely Li2S–P2S5 and Li4P2S6. AM/SE interface. Such damage is known to be the leading
Though similar compounds, Li2S and Li3P were formed at both reason for capacity fade.19
the SE interfaces in contact with the Li metal anode, the latter In summary, interfaces with solid electrolytes are at the core
SE had higher ionic conductivity. The higher ionic conductivity of the limitations of ASSBs and the understanding of the

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17143
View Article Online

Journal of Materials Chemistry A Paper

mechanisms involved there is the key to the success of all solid


state batteries. Chemical, morphological and microstructural
investigations of solid–solid interfaces without causing any
damage are challenging. Though many characterization tech-
niques exist and are used (EIS, XPS, X-ray tomography, TEM,
AFM, neutrons, and X-ray diffraction), some of them are
versatile enough to allow chemical and morphological analyses
at the same time on “bulk” batteries without very specic
preparation such as SEM coupled with EDX. For example, TEM
based in situ/operando requires the preparation of lamellar
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

based37 or nanobatteries which can show completely different


behaviour than bulk-type batteries. Also, beam energy has been
known to induce damage in structures such as dendrites which
gave birth to the idea of using cryo-TEM for observing such
structures. XPS will mainly provide chemical insight with few
imaging possibilities. In situ SEM has been used to investigate
Li metal-based polymer solid state batteries.38 Thus in this
work, in situ SEM in combination with EDX was used to study Fig. 1 Battery assembly steps (a) self-standing pellet with SE and
sulde-based ASSBs. We studied the failure modes of two a cathode composite. (b) Mechanical breaking of the pellet inside a die
sulphide-based solid electrolytes, namely b-Li3PS4 (LPS) and pelletizer to obtain three pieces of battery. (c) The mounted battery in
Li6PS5Cl (LPSCl) via live morphological and chemical investi- a homemade in situ cell (IS_cell).
gation. The electrochemical, chemical and morphological
evolution will be discussed and the categorisation of the failure
modes is proposed. cathode composite is obtained as shown in Fig. 1(a). A thin
cathode layer with an approximate thickness of 64 mm was used
purposely, to study battery failure based on morphological
Experimental section changes and eliminate many possible transport issues related
Battery assembly/fabrication to the thick cathode. Since we needed a at section of the
Materials. Two commercially available sulphide-based solid battery for better imaging and visualization during in situ SEM
electrolytes (SEs), viz. b-Li3PS4 and Li6PS5Cl were used for this experiments, we broke this pellet inside the die pelletizer to
study. LPS with SE particles in the range of 0.01–7 mm was obtain three small pieces of battery as shown in Fig. 1(b). The
purchased from NEI. Corp, while LPSCl with SE particle size AM loading is evaluated from the surface area of the pellet and
ranging from 50 to 150 mm was purchased from Ampcera. Both the weight fraction of the pellet with respect to the entire pellet.
SEs, b-Li3PS4 and Li6PS5Cl were studied separately. NMC111 Finally, 0.75 mm thick Li is punched into 1/3rd of 8 mm in
from Targray was used as an active material (AM). Two different diameter and is hand-pressed on the other side of the SE/
kinds of electronic carbon additives, viz., C45 from Timcal and cathode composite. The prepared battery is then mounted on
VGCF from Showa-Denko respectively were used. Li (0.75 mm a homemade in situ cell (IS cell), which is placed between two
thick) was purchased from Alfa Aesar. stainless steel current collectors as shown in Fig. 1(c). The
Cathode composites. The cathode composites were prepared initial mechanical integrity and the contact between different
by hand milling in a mortar and pestle. NMC111 along with layers in the battery are maintained via the screw on the top of
carbon additives was rst mechanically milled and later either the cell.
LPS or LPSCl was added. The following weight fractions of active The process of battery assembly and cell mounting was kept
materials, solid electrolytes, and carbon additives were used: constant for the study of both the solid electrolytes. The entire
AM (66.5%) + SE (28.5%) + C45 (2.5%) + VGCF (2.5%). Their cell assembly was carried out in a glove box (Jacomex) under an
respective volume fractions are as follows: AM (30.74%) + SE argon atmosphere with H2O and O2 at 1 and 0.6 ppm
(32.59%) + C45 (33.94%) + VGCF (2.71%). respectively.
Li surface cleaning process. The as obtained Li surface was
cleaned using a razor blade. Later this foil was pressed in
between two polypropylene sheets, to obtain a rather uniform Li Electrochemical measurements
surface. Ionic conductivity (EIS) and electronic conductivity (DC
Battery assembly. A 10 mm diameter die pelletizer with three polarization). EIS and DC polarization measurements were
pieces as shown in Fig. 1(b) was used to fabricate an ASSB. For carried out for the two solid electrolytes to verify their ionic and
the battery assembly, rst, SE powder (80 mg) is added to the electronic conductivity values, respectively shown in Fig. ESI-2.†
10 mm pelletizer and densied by the cold press for approxi- For EIS and DC measurements, pellets with a diameter of 10
mately 1 min at 1 ton. Next, the cathode composite powder (10 mm with 80 mg of SE, fabricated at 3 tons of pressure were
mg) is added and the two layers are pressed together at 3 tons used. The EIS measurements were carried out with an MTZ
for approx. 3 minutes. A self-standing pellet with SE and the impedance analyser provided by Biologic, with a frequency in

17144 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A

the range of 30 MHz to 1 Hz and a classical ion blocking setup battery, post-mortem SEM analyses were further performed
with a 10 mV potential perturbation. The SE pellet was placed aer the in situ analysis.
between gold-coated Cu plates during the measurements. The
pressure during EIS measurements is controlled by a spring in
an MTZ impedance cell provided by Biologic. A circular ring Results and discussion
controls the compression in the spring. The spring is In situ electrochemical analysis
compressed until the last circular motion is possible in the cell
and thus the force per unit area (pressure) in each case remains To verify the electrochemistry before carrying out experiments
constant. The pressure is enough to keep the electrodes in with the homemade IS-cell inside the SEM, electrochemical
contact with the pellet during the measurement. For DC testing outside the SEM was carried out as shown in Fig. ESI-3.†
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

measurements, a Bio-logic SP-200 battery cycler potentiostat The electrochemistry obtained with a classical ASSB cell with
with a sensitivity of current up to nano-amps (nA) was used. The uni-axial pressure like in a pmma (polymethyl methacrylate)
DC measurement set-up allowed a constant stack to be applied, matrix with two pistons/plungers on each side is provided in
and thus 9 Nm of torque was applied (similar setup as shown in Fig. ESI-3(a)† and was used as a reference to verify the obtained
ESI-3(a)†). The 9 Nm is the reference torque that is used during electrochemistry with the in situ cell shown in Fig. ESI-3(c).† For
battery cycling with the Li metal anode and hence the same both the cells, LPSCl–SE, with NMC111 as the cathode and Li
condition for electronic conductivity measurement was fol- metal as the anode was used. The cells were 1st cycled at C/50,
lowed. DC polarization experiments were also carried out under corresponding to a current density of 59 mA cm2. The electro-
ion blocking conditions with SE placed between two stainless chemistry with the homemade in situ cell (ESI-3(c-1)†) looked
steel pistons in the potential range from 0.1 V to 1 V with 0.1 V promising and was comparable to that of the pmma cell, con-
steps each. At each potential step, 3 hours of rest were allocated rming the reliability of the electrochemistry with the newly
to allow the stabilization of the current. developed cell.
Fig. 2(a) depicts the electrochemistry obtained during the in
situ SEM experiments for LPS at a current density of 36 mA cm2
Galvanostatic cycling with potential limitation (GCPL)
(C/25). There's a sudden potential drop during the 1st charge
The battery cycling was carried out in constant current mode at 3.75 V, indicating the possibility of a short circuit. We also
(GCPL) between 2.3 V and 4.3 V, the potential range used for the prepared batteries based on LPS–SE and LPSCl–SE with our
study of both the SEs. A high accuracy Biologic potentiostat reference ASSB cells as shown in Fig. ESI-3(a-1) and (a-2)†
SP200 was used for cycling. respectively. Though the cell does not fail at 1st charge, high
polarization and low capacity values are evident, depicting poor
In situ/operando scanning electron microscopy setup (in situ electrochemistry compared to LPSCl (Fig. ESI-3(a-2)†). Better
SEM) and uniform control of pressure by the two pistons could
A scanning electron microscope (SEM) FEI quanta-200 F was
used for all the in situ experiments. The prepared IS-cell is
positioned in a homemade airtight box that allows battery
transfer without air exposure to and from an argon glove box
and the SEM chamber and also allows its cycling inside the
SEM. The SEM chamber has been modied for the cable inlet
and outlet and to allow the opening and closing of the transfer
box. All the cycling and analysis processes were performed in
high vacuum mode. Morphological analysis was carried out by
secondary electron imaging (SE). The backscattered electron
(BSE) imaging providing images linked to the atomic number of
the elements was carried out to gain a visual understanding of
the differences in the chemical composition especially at buried
interfaces where EDX isn't accessible. The chemical composi-
tion and distribution were obtained using an X-ray energy
dispersive spectroscope INCA OXFORD (Bruker).
The following steps were performed to monitor the specic
area throughout the cycling: at 0 min the positions of different
areas of interest (electrodes, electrode–electrolyte interfaces,
etc.) were marked in the battery. These positions were followed
throughout the cycling and tracked for morphological and
chemical changes.
To carry out automatic imaging at a specied time interval
and especially during night cycling, we used a small “routine” Fig. 2 Electrochemistry inside the in situ cell with (a) Li3PS4 (LPS) and
code. To investigate further, and carry out analysis inside the (b) Li6PS5Cl (LPSCl).

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17145
View Article Online

Journal of Materials Chemistry A Paper

explain the improved electrochemistry obtained inside the performance in the electrochemistry inside the SEM compared
reference cell when compared to our IS-cell. Pressure plays a key to those cycled with full pellets with LPSCl outside the SEM in
role in lithium plating and dendrite formation and hence a lack the same in situ cell as shown in Fig. ESI-3.†
of good pressure control inside the SEM chamber could be
a possible reason for faster dendrite formation. This problem
aids our experiment and helps to visualize battery degradations In situ SEM: imaging and chemical analysis
in the allotted SEM time. To investigate the potential drop and Solid electrolyte. Fig. 3(a) depicts the backscattered electron
the possibility of a short circuit, in situ SEM imaging was per- (BSE) images of LPS–SE evolution during cycling. One can see
formed before any electrochemical analysis, i.e. at 0 min, the presence of pre-existing cracks and some dark spots in
intermediate imaging and one aer the short circuit during the pristine LPS. These pre-existing cracks seem to form an inter-
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

1st charge. As for the solid-state system, there's no complete connected network, and thus a dened path exists through
short circuit, so we let the system evolve aer the short circuit to which cracks can grow. Our theoretical calculations, based on
increase the probability of observing dendrites, if any during the LPS thickness and surface area approximate a 72% com-
the in situ experiments. pacity for LPS which agrees with the pristine LPS image,
Fig. 2(b) depicts the electrochemistry obtained during the in revealing the poor compacity. The presence of dark spots is
situ SEM experiments for LPSCl. We began our experiment at a common observation in the literature for LPS; however, they
a C rate of C/15, corresponding to a current density of 130 mA have been attributed to voids.39 We do not think of this as void,
cm2. The choice of the C rate was made for practical reasons since the secondary images reveal these dark spots to have
(SEM availability). However, cycling at C/15 showed some a solid form and shape as shown in Fig. ESI-5(a).† The EDX of
noise in the cycling data as evident from the 1st two cycles of the such dark spots in the LPS presents the complete absence of
cycling curve. We thought that such noises were due to a high sulphur, and phosphorus but the presence of oxygen (refer to
rate/growth of dendrites at higher C rates so we lowered the C the EDX map in Fig. ESI-5(a)).† Since we are not able to detect Li
rate to C/49 corresponding to a current density value of 40 with our EDX analysis, we cannot strike out the fact that there
mA cm2. From the 5th discharge onwards cycling was carried could also be Li metal inside the SE. It could also be a possible
out, outside the SEM chamber (SEM unavailability) under an impurity associated with oxygen such as Li2O. Another peculiar
airtight atmosphere. The electrochemical curve from the 5th feature of pristine LPS is that it possesses a layered morphology
cycle until the 15th charge is provided in Fig. ESI-4.† One as shown in its cross-section image in Fig. ESI-5(b).† This
observes a constant capacity fade with increasing polarization. feature of LPS also makes it more prone to crack formation,
A similar imaging strategy was followed for LPSCl, i.e. imaging since this layered morphology denes cleavage planes. During
at 0 min and end of charge (4.3 V) & discharge (2.3 V) steps and LPS rst charge, the analyses show the apparition of a big and
nal imaging was carried out at the end of the 15th charge. No relatively sharp crack of about 11 mm. Thus, LPS displayed in
imaging at intermediate potential was carried out unless stated. this region in Fig. 3(a) shows crack formation and growth, as
In the case of LPS, the battery (possibly) short circuits even at wide as 13 mm within the 1st charge interval. In addition, the
a lower C rate such as C/25, while for LPSCl though cycling at C/ pre-existing cracks widen and become more visible as the
15 results in increased cell polarization and lower capacity charging progresses. Electro-chemo-mechanical stress is
values, no such potential drop or short circuit is visible. These created during battery charging. This stress could be due to
observations from an electrochemical point of view show that plating (Li deposition) or the volume expansion of NMC  3%40
LPSCl performs better compared to LPS. or of the lithium anode41 that exerts stress on the SE or the
From our theoretical understanding, the possible reasons for expansion of SE42 due to its decomposition. This stress is then
differences in electrochemistry could be: exerted on LPS, that is on pre-existing cracks, interconnected
(a) Ionic conductivity of LPS (105 S cm1) is two orders of networks of cracks etc. causing faster growth of cracks. Thus,
magnitude less than that of LPSCl (103 S cm1) as shown in the intrinsic nature of LPS with its layered morphology (Fig. ESI-
Fig. ESI-2.† Hence the ion transport in the bulk of SE is 5(b)†) is prone to generating sharp, huge and fast propagating
different. Intuitively, also the effective ionic conductivity in cracks.
their respective cathode composites will also be lower. Thus, the Fig. 3(b) depicts the BSE images of LPSCl evolution during
ion transport across the interface will not be the same in both cycling. The pristine material surface seems more compact
cases. This adds an extra kinetic limitation for LPS compared to from the SEM images, in good agreement with our theoretical
LPSCl. The other reason for differences in ion transport across calculations of compacity  89%. However, the surface of LPSCl
the interface can come from a more porous LPS–SE and its seems more uneven/rough compared to that of LPS. This
cathode composite compared to LPSCl, which will be discussed roughness could be due to large particles of LPSCl (50–150 mm)
later. This highly porous nature creates hurdles in achieving compared to those of LPS (0.01–7 mm). During the cycling, an
sufficient contacts, and thus even with lower operating pressure approx. 12 mm uneven/blunt crack is visible by the end of the 5th
values, the effects get emphasized. This could explain the poor charge which grows to 33 mm by the end of the 15th charge.
electrochemical performance of LPS compared to LPSCl. Unfortunately, we do not have a 1st charge SEM image of LPSCl
(b) Also, for the in situ SEM experiments, the pellet is to compare the crack growth with LPS. However, the crack width
mechanically broken into 3 pieces, leaving some sort of at the 5th charge of LPSCl (12 mm) is comparable to the crack
mechanical shock in the battery which could explain the poor width by the end of the 1st charge of LPS (13 mm), thus,

17146 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A


Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

Fig. 3 Backscattered electron (BSE) images of solid electrolyte evolution during cycling of (a) b-Li3PS4 and (b) Li6PS5Cl. These images show the
crack formation and evolution.

validating the point that the crack grows (widens) at a much plausible explanations for faster short circuits with LPS than
faster rate in LPS than in LPSCl. The gures ESI-6 and ESI-7,† with LPSCl.
show the growth of cracks in LPSCl from 0 min up until the 5th (3) First cracks, then dendrites. The intuition about cracks in
charge at the anode and cathode interface respectively. These SEs is that they could be induced by dendrite penetration and
BSE images conrm that LPSCl indeed shows slow crack growth they grow with their growing propagation. In situ X-ray computed
and the new cracks are formed at or near the pre-defect area. tomography44 coupled with X-ray diffraction36 studies with a Li–
The areas in the LPSCl that are compact seem to remain LPSCl–Li symmetry cell have revealed that cracks travel ahead of
untouched irrespective of the interface. The following are the dendrites which very much aligns with our in situ SEM observa-
key observations: tions. Cracks seem to grow and propagate ahead of creped Li or
(1) Compacities and the surfaces of the two SEs. The difference Li dendrites. Also, every crack is expanding simultaneously,
in terms of compacity could come from the differences in rather than individual dendrites propagating through the SE.
particle size of the two SEs. LPS is ner (0.01–7 mm), while LPSCl From our observations, we believe it's the porous structure of the
is composed of a mix of different particle sizes (50–150 mm). SE that determines this fate of cracks being induced rst and
These particle size differences also lead to a rough surface in then dendrites. The generation of electro-chemo mechanical
LPSCl and a smoother surface in LPS. Thus, when the two SEs stress during battery cycling is now a known fact. The way this
are pressed at the same pressure, LPSCl compacts more as ner generated stress gets distributed is determined by the porous
particles ll in the voids le by large ones. This is not possible nature of SE. More porous SE (LPS in this case) will better
in the case of LPS. distribute this stress, than the compact one (LPSCl in this case).
(2) Nature of the cracks and their growth. The cracks in LPS This could also explain the faster expansion of pre-existing cracks
seem sharper/more dened, while those in LPSCl are blunt/ in LPS. The expanded cracks in turn will serve as a path for
uneven. Also, the crack growth is much faster in the former dendrites and nally the battery short circuits as evident with the
case, than in the latter case. The faster crack growth in LPS potential drop at 3.75 V in the case of LPS. Thus, also according
could also be explained based on higher porosities. There's to our observations, in compliance with previous observations
a high possibility that during Li plating, Li depositions could it’s cracks rst, and then dendrites follow the cracks.
take place inside these porosities at the anode interface, and Anode interface (Li/SE interface). Fig. 4(a) depicts the
exert mechanical stress on SE, causing elongation of neigh- evolution of the BSE images of the LPS/Li interface during the
bouring pre-existing cracks. Such local stress-induced inside rst charge. Good contact between LPS–Li is visible at 0 min. By
the surface aws of SE are reported to cause transgranular the end of the 1st charge, this interface experiences an approx-
cracks.43 LPS is a glass-ceramic-based SE that shows fractures imately 9 mm contact loss. In addition, the pre-existing cracks
associated with typical glassy materials.1 Fractures associated present on the LPS surface also seem to sharpen/widen and are
with glassy materials are classied as brittle fractures. Brittle more clearly visible.
fractures give rise to sharp cracks that are known to expand and Similarly, for LPSCl, an initial good contact at 0 min shown
grow much faster (speeds—up to 2100 m s1) than ductile in Fig. 4(b), experiences approx. 36 mm contact loss by the end of
cracks, seen in LPSCl. Ductile cracks grow slower and give blunt the 5th charge. Unfortunately, we couldn't carry out imaging at
cracks. The cracks that propagate in a brittle glassy material will the 1st charge during LPSCl cycling, and hence no direct
continue to grow once initiated and can serve as one of the comparison with the LPS 1st charge could be made. However,

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17147
View Article Online

Journal of Materials Chemistry A Paper


Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

Fig. 4 Anode interface: (a) and (b) backscattered electron (BSE) images at the lithium interface for LPS and LPSCl solid electrolyte respectively
showing the loss of contact. (c) Li dendrite morphologies observed in LPS (secondary electrons SE) and (d) LPSCl (black outlined, backscattered
electrons BSE).

approximately 8 mm contact loss at the same position aer the reverse is true for the cathode. This generated stress has been
2nd cycle, as shown in Fig. ESI-5(c),† shows that contact loss is known to strain the NMC particles,24 another leading reason for
still lower when compared with LPS. The presence of certain the capacity fade.
structures that create spacing between LPSCl–Li is visible by the The other striking difference between the two SEs is the
end of the 5th charge. These structures perhaps still keep LPSCl number of and the nature of morphologies (structures) of
and Li in contact even with the complete loss seen. Also, there is deposited Li at the anode interface. Fig. 4(c) shows the
a possibility that deeper in the battery, contact between the SE morphology of the deposited Li at the anode interface in the
and anode is still present allowing the cycling to continue. The case of LPS. Two kinds of morphologies are visible. A thick bre-
contact loss between SE/Li could be attributed to a combination like morphology labelled as 1 and a spherical morphology
of processes occurring in the battery. labelled as 2. Fig. 4(d), shows the observed morphologies at the
(1) Plating (during charging). Plating will induce a loss of LPSCl/Li interface. At least 3 kinds of morphologies, labelled as
contact at the interface as Li ions get deposited at the SE/Li bre-like (0.6 mm), globular (2.6 mm) and thick cylindrical
interface. This deposition puts stress on the SE/Li interface structures (4.6 mm), were seen during the in situ analysis.
causing contact loss. This could be one of the reasons for Mind you, these are only the surface observations and we do not
contact loss in both SEs. However, in the case of LPSCl, as the strike out the possibility of more such structures beneath the
battery discharges (stripping), the Li ions move back to the surface in contact with SE. As can be seen in Fig. 4(d), these
cathode, thus, the possibility of recovering this contact loss. But formed morphologies are buried deep down the Li/SE interface,
since the contact loss remains permanent, this highlights the and hence no EDX chemical analysis was possible. However, the
irreversibility32 of the process. With every cycle, a residual backscattered electron (BSE) images reveal that the formed
contact loss is le which keeps adding to the contact loss of the structures are not purely Li metallic structures. The BSE signal
previous cycle and this grows which could also explain the is proportional to the atomic number of the species. Thus,
capacity fade with each cycle and huge irreversibility in the a brighter area's signal possibly comes from high atomic
capacity values of LPSCl. Also, some areas experience higher number (Z) elements in the system meaning probably the
contact losses than others indicating the non-uniform Li presence of S or P and we assume that this could correspond to
deposition/plating at this interface. degradation products such as Li2S which is widely reported at
(2) Volume expansion/volume stress induced during cycling. The the anode interface45 while slightly grey areas could be Li
stress exerted from the volume expansion of active materials depositions/dendrites either still in the metallic state or covered
during charge in systems such as NMC–sulphide SE–Li–In can by a layer of decomposed SE, Li2S, Li3P or Li2O.
reach as high as 1–1.25 MPa.24 In our case, the axial screw on the As already highlighted, the lithium plating and stripping
top of the current collector cannot compensate for such pres- processes are non-uniform. Thus, different regions especially in
sure and the loss of contact becomes inevitable. While the SE/Li the case of the LPSCl/Li interface, experience local depositions
interface experiences loss of contact, a contact gain between Li/ giving rise to different morphologies of these deposited Li.
CC (current collector) is possible as the anode is forced towards Local depositions occur, due to surface inhomogeneities at the
the CC during charging and away during discharging. The Li/SE interface, such as defects, impurities or irregularities on

17148 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A

the Li surface, surface roughness or any surface aws etc. They a possible cathode electrolyte interphase (CEI) is seen. The EDX
are considered hotspots for Li depositions. They alter the analysis on this region, marked with a red cross (c) as shown in
topography and possibly affect the morphologies of the depos- Fig. ESI-8(a),† presents an S/P atomic ratio of approx. 3 as
ited Li. Considering that similar Li metal foil was used in both opposed to the expected value of 4 obtained from the LPS-ref.
the experiments and a similar cleaning process was used, we The obtained values have been tabulated (ESI†). The position of
expect that the Li surface impurities and rigorousities/surface this secondary NMC particle in the entire cathode composite of
roughness on both surfaces will remain constant. Thus, differ- LPS is shown in ESI-9(a).† Similarly, Fig. 5(b) shows the cathode
ences at the SE/Li interface come from differences in the interface evolution of LPSCl. A similar kind of surface layer on
surfaces of the two SEs. SE particle size affects the surface NMC in contact with SE is visible by the end of the 15th charge.
roughness. As explained earlier, LPS has a smoother surface This layer has an approximate thickness of about 170 nm. The
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

than LPSCl. Higher surface roughness and higher topographical EDX analysis of this surface layer, marked with the blue cross
variations give a more inhomogeneous deposition, leading to (c) and corresponding atomic ratios of elements tabulated
void formation, resulting in more observed dendrite morphol- below, as shown in Fig. ESI-8(b),† presents an S/P atomic ratio of
ogies which have already been highlighted by the Masahiro 3.3 as opposed to the expected value of 5 obtained from refer-
Tatsumisago group.20 This could explain why we observe so ence LPSCl. The lower atomic ratio of S/P (3 in the case of LPS,
many different morphologies with LPSCl than with LPS. as opposed to 4 and 3.3 in the case of LPSCl, as opposed to 5),
Cathode interface. Fig. 5 depicts the modications observed reveals a higher “P”, compared to S thus, highlighting the
at the cathode interface during the in situ cycling for both the possibility of forming a “P” rich compound at this interface. The
SEs. Fig. 5(a) shows the 0 min secondary electron (SE) image of position/location of this secondary NMC particle with the
a secondary NMC particle surrounded by LPS and the carbon formed surface layer in the bulk of the cathode composite of
additives in the cathode. By the end of the 1st charge, an LPSCl is shown in ESI-7(b).† We also found a similar kind of
approximately 22 nm thick surface layer/lm on NMC, surface lm/layer on the side of cathode composite that was

Fig. 5 Secondary electron (SE) images of the cathode composite showing the cathode interface in (a) LPS and (b) LPSCl (c) backscattered
electrons (BSE) images show the delamination of the cathode from LPSCl with increasing cycle.

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17149
View Article Online

Journal of Materials Chemistry A Paper

close to the current collector side, during the post-mortem of an voltages, forming a layer of decomposed products, known as an
ex situ cycled cell made with LPS–SE as shown in Fig. ESI-10(d) “interphase”. The VGCF/thiophosphate interface however is
and (e).† These observations conrm the formation of such known to form decomposed products that are redox-active.12
a surface lm/CEI to be a bulk phenomenon and not just These redox-active decomposed products are known to improve
surface observations. A study on NMC622/LPSCl13 by time-of- the initial capacity. However, increasing cycles lead to capacity
ight secondary ion mass spectrometry (ToF-SIMS) coupled fade, which is the case with our LPSCl. The formation of a space
with XPS, studied the nature of this interface with cycling. They charge layer, at NMC/sulphide SE, which occurs due to differ-
found decomposed products at this interface that are mostly ences in chemical potentials has also been reported but seems
composed of phosphates and sulphate like species, whereas to have negligible impact.47 The formation of poorly conducting
transition metal compounds such as TMS, TMP, and TMCl decomposition products (interphase) and poor or small contact
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

can be neglected. With NMC being the available oxygen source, areas is known to be more detrimental and increase interfacial
they found that the formation of Li2SO4 or Li3PO4 at this resistance. Thus, in this case, we would relate the formation of
interface is plausible. They also showed that the amount of this insulating CEI to being the leading cause of capacity fade in
Li3PO4 formed was much higher than that of Li2SO4. This is in the case of LPSCl. A stable interphase (SEI-anode interphase/
very good agreement with our conrmed atomic ratios of S/P. CEI-cathode interphase) is the one that is formed during the
Also, since Li2SO4 and Li3PO4 are both insulating in nature, this initial cycles and does not grow with the increasing cycle.
formed CEI will indeed be insulating. We also observed a higher However, in the case of unstable interphase, this keeps thick-
reactivity and charging effect of this formed CEI with an elec- ening with increasing cycles,48 which is the case with both SEs
tron beam, highlighting and conrming the possibility of the studied here. This is a big problem for battery cycling.
insulating nature of this formed interphase. Similarly, the S/Cl Fig. 5(c) shows the cathode-LPSCl interface at lower magni-
ratio is lower (3.3), than the expected value (5), highlighting cation. The cathode composite seems less compact at 0 min.
lower “S” and the possibility of the formation of compounds We expect less compacity in the cathode composite since
like LiCl. A study on interfacial degradation of Li6PS5Cl against materials of different crystallinities and ductilities are pressed
LiCoO2 (LCO) via in situ electrochemical Raman microscopy together at one dened pressure. As NMC is a highly crystalline
conrmed the formation of P2Sx (x $ 5) and LiCl46 during material, a higher pressure would be required to form
charging. a compact composite. Also going higher in pressure means
The cathode composite also comprises other interfaces such more mechanical damage and hence an optimum of 3 tons was
as C/SE and C/NMC, apart from NMC/SE. The C/NMC interface selected for the fabrication of the cathode composite. The areas
is a stable one, similar to the one found in conventional liquid marked as A1, A2 and A3, expand from 0.1 mm, 1.85 mm, and
electrolyte-based batteries. However, the NMC/SE as discussed 0.24 mm at the 0 min to 1.72 mm, 21.94 mm and 18.88 mm,
already and the C/SE18 are known to be detrimental to battery respectively by the end of the 15th charge. Also, complete
performance. They lead to the degradation of SE at higher delamination of the cathode (the electrode itself) takes place by

Fig. 6 Post mortem of a LPS-based battery: (x) a piece of the LPS battery used for post mortem analysis. (a) Backscattered electron (BSE) image
shows the formation of voids on the LPS surface in contact with the Li anode and (a0 ) the associated magnified image. (b) A secondary electron
(SE) image of the LPS surface shows the presence of dendrites inside the solid electrolyte and (b0 ) associated magnified BSE image.

17150 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A

the end of the 15th charge. In the present scenario, since LPSCl localized stripping thus forming such voids. A dark black area
acts as a better barrier for dendrite penetration due to its better with a thickness of 6 mm (Fig. 6(b0 )), is visible inside the crack
compacity, compared to LPS, the generated electro-chemo- present in the SE (Fig. 6(b)) and considering the high difference
mechanical stress during cycling is exerted upon the cathode in atomic number with LPS (white compound) the black phase
causing delamination of the cathode from the SE layer is probably lithium. Since the mechanical strength of the two
(mechanical failure) whereas, in the case of LPS, due to the high SEs, (20 GPa)1 is almost similar, the mechanical strength
porosity of SE, this stress is transferred throughout the SE, offered to block the dendrites should practically be the same.
causing crack growth (mechanical failure) and in turn causing Also, the electronic conductivities of the two SEs are almost
a dendritic short circuit (electrical failure). similar, as shown in Fig. ESI-2.† The electronic conductivity of
SE being the reason for dendrite penetration in LPS can also be
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

Post mortem analyses of in situ cycled batteries by SEM eradicated.43 The other ways that dendrites can thus propagate
are via grain boundaries, pre-existing or newly formed cracks,
To further our understanding obtained from the surface
voids/pores etc. For LPS as already stated, the poor compacity
observations via in situ cycling and to validate/conrm our
and thus high porosity could majorly serve the path for dendrite
observations, post mortem investigation of the in situ cycled
propagation. Thus, the post mortem conrms that the dendritic
batteries at different interfaces was carried out by SEM. Fig. 6
short circuit is indeed the cause of the sudden potential drop in
shows the post mortem images recorded on the LPS in contact
the case of LPS.
with lithium. Fig. 6(x) shows the general view of the investigated
Fig. 7 shows the images obtained from postmortem inves-
piece on which a part of the lithium remains (upper part of the
tigation of the LPSCl in situ cycled battery. A piece of Li surface
piece) in contact with the SE. In Fig. 6(a), the presence of
which was in contact with LPSCl was used for post mortem
a phase on top of the LPS surface which has a black contrast on
investigation as shown in Fig. 7(y). This represents the general
the BSE images (see also Fig. 6(a0 )) indicating a low atomic
view of the analysed piece. On this backscattered image (Z-
number compound, is visible. Its morphology (Fig. 6(a0 ) and (b))
dependence contrast) three colours (white, grey and black) are
looks like “lace” with the presence of round holes. Considering
visible indicating the presence of at least 3 different phases. The
the Z—information and the observed morphology, this phase
black color represents the Li foil. By investigating further
corresponds to deposited lithium on the LPS surface and the
(Fig. 7(a) and (a0 )), the grey area formed on top of the Li, seems
holes to voids which are a common observation29 induced
like it’s composed of a lot of small grains. Their analyses by EDX
during plating/stripping. However, the voids seen on the LPS
reveal the presence of a high sulphur content (S) with an atomic
surface are questionable since the LPS battery fails during the
ratio of S/P  14 (LPSCl atomic ratio S/P ¼ 5). The deviation
1st charge and no stripping (discharge) has been carried out.
from the expected ratio of LPSCl indicates the formation of
One of the possible ways to explain this void formation can be
decomposition products rich in sulfur, most probably Li2S or
found in Fig. 2(a). The continuous potential increase and drop
Li2Sn. The formation of the interphase between Li metal/
that occurs aer the battery short circuit could give rise to

Fig. 7 Post mortem of the LPS-based battery. (y) A piece of Li surface in contact with LPSCl used for postmortem analysis. (a) The backscattered
secondary electron (BSE) image shows the formation of reduced SE on the Li surface and (a0 ) associated magnified image. (b) BSE image of the Li
surface with Li plated structures and (b0 ) the associated magnified image of the plated Li.

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17151
View Article Online

Journal of Materials Chemistry A Paper

sulphide SE is now commonly known in the literature. This Postmortem of an ex situ cycled battery with LPS by SEM
occurs due to the intrinsic chemical instability of sulphide SEs
To further conrm our observations made from in situ and
with the Li metal anode. This interphase comprises Li2S, Li3P
post mortem of the in situ cycled battery, we also carried out
and LiX in the case of Li6PS5Cl. But in the case of Li6PS5Cl, this
postmortem of an ex situ cycled cell outside the SEM with our
interface is known to be stabilized aer a few hours of
ASSB reference cell (pmma-matrix). ESI-1(a)† shows the 1st
contact.6,49 The ionic and electronic conductivity of the formed
charge electrochemical curve of LPS obtained with the pmma
interphase at the Li/SE interface has been known to be an
reference cell, cycled outside the SEM. The observations such
important factor in improving the short-circuiting tolerance of as sharp cracks (ESI-10(b)†), the formation of anode inter-
all-solid-state Li–metal batteries. Suyama et al. showed that by phases (ESI-10(c)†), Li plated structures (ESI-10(f)†) and CEI/
increasing the Li3P content in the interphase, the cell can cycle
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

surface layers (ESI-10(d) and (f)†) are all coherent from the
longer without short-circuiting.23 So, could there be differences
observations made during the in situ experiment. This indeed
in total ionic and electronic conductivities of formed inter-
conrms the reliability of the surface observations made from
phases between the two sulphide SE? Or differences in the total
in situ SEM experiments. We believe that open surfaces can
amount of formed Li3P between b-Li3PS4 and Li6PS5Cl? This
indeed give fast reactions since the surface remains open and
could be another possible explanation for the different short-
exposed; however from our post-mortem analysis of the in situ
circuiting behaviours between the two SEs. In other words, this cell/ex situ cell we do not believe this behaviour to be different
could explain why it's easier to have a dendrite short circuit in b- from the bulk.
Li3PS4, than in LPSCl. This hypothesis however needs to be
To summarize our discussion, Fig. 8 provides a schematic
veried further in this case.
summary of various failure modes, observed in the two SEs via
Further investigation of the lithium surface is presented in
in situ SEM. We believe that these failure modes are the pieces
Fig. 7(b) and (b0 ). In Fig. 7(b) also three phases are present: (i)
of the same puzzle and hence are linked together, as has been
the black one at the bottom corresponding to Li foil (ii) the grey
shown in recent studies.50 They could be categorized into either
one discussed previously, corresponding to degradation prod-
of the following 3 interlinked broad modes of failure (mecha-
ucts, and (iii) the white one whose composition is on par with nism of failure):
the solid electrolyte. The morphology of lithium (Fig. 7(b0 )) (1) Electrical failure comprises dendrite short circuits or
shows a lot of reliefs and is composed of lament-like struc-
mechanically induced short circuits or an increase in the
tures. It is very different from the original lithium surface and
impedance value.
more consistent with plated structures. Li plated structures are
(2) Mechanical failure could be cracks in SE, loss of contact
formed, with depositions of dead Li-ions during the ion trans-
at the interface, the delamination phenomenon or the global
port thus, highlighting the capacity fade and extremely poor
volume expansion/contraction of the cell during cycling, etc.
capacity by the end of the 15th charge as shown in Fig. ESI-4.†
(3) Electro(chemical) failure is a joint term to describe
chemical and electrochemical failures together. It comprises

Fig. 8 Schematic summarizing all the observed failure modes observed in the LPS vs. LPSCl solid electrolytes.

17152 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A

failures associated with the intrinsic instability of two chemical analysis on two sulphide based solid electrolytes namely b-
compounds, e.g. Li metal in contact with thiophosphate SEs and Li3PS4 and Li6PS5Cl to fully comprehend and categorize the
also describes the degradation associated with the application different modications observed.
of electrochemical potentials, e.g. the formation of the CEI etc. We separate the evolutions as parts of three main inter-
Such evolutions have been observed separately in previous connected failure modes:
studies but they were all observed during our experiments  (1) Electrical failure with the formation of lithium
showing the complexity of such systems. X-ray CT coupled with dendrites propagating across the SEs as in the case of LPS
SEM studies on LPS highlighted that the main mode of failure leading to short-circuit.
associated with LPS kind of SE is a chemo-mechanical failure.42  (2) Electro-chemical failure with the formation of degra-
LPS expands during its chemical decomposition with Li metal dation products, a SEI and CEI, in both cases.
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

upon cycling, leading to the formation of small cracks. Li metal  (3) Mechanical failure with the apparition of cracks and
then penetrated through these small cracks. In our case, we their expansion in both solid electrolytes as well as the delam-
believe the intrinsic nature of b-Li3PS4 lies at the core of its ination of the cathode composite as in the case of LPSCl.
failure. The poor compacity of the LPS–SE separator (72%), even All these failure modes are interlinked and should be studied
poorer compacity of the cathode composite (51.75%), the as a whole to fully grasp the origin of the performance limita-
interconnected network of porosities, pre-existing cracks, and tion. Among them, we believe that mechanical failure is at the
layered morphology of LPS that denes the fracture planes are core of the capacity losses since these are irreversible in nature.
at the core. The electro-chemo-mechanical stress, generated We also highlighted that the key factor determining the differ-
during the 1st charge (Li deposition), is exerted on the poorly ences in the failure between the two SEs lies in their respective
compact LPS–SE. Since LPS is poorly compact, generated stress intrinsic natures.
is distributed throughout the SE, causing the simultaneous Throughout this work and in our discussions, we emphasize
growth of pre-existing cracks. There's a high probability that Li the role of surface roughness, size of SE particles etc. However,
depositions could occur inside the porosities of LPS. This this doesn't seem to be a direct comparison since the two above
creates higher stress in SE leading to faster crack growth. Since mentioned SEs are different. To perform a direct comparison
the fracture planes seem to be dened due to layered and validate the role of SE particle size and its impact on
morphology, cracks grow faster and sharper. Later on, these dendrite morphology, the formation of thick structures and
cracks serve as a path for dendrites. Also, faster crack propa- their origin, compacity/porosity, dendrite and crack propaga-
gation occurs due to its glassy nature. In short, LPS clearly tion etc. by keeping the SE constant will be the focus of our
shows electrochemically induced (plating) mechanical failure future work.
(cracking in SE), collectively electro-chemo-mechanical failure
leading to short circuit aka electrical failure (dendrites).
In LPSCl, upon 1st charge, due to electrical plating, Author contributions
mechanical stress is generated at the anode interface. Since this
SE is more compact, the generated stress cannot spread in the Conceptualization, data curation, formal analysis, investiga-
SE but gets exerted upon the electrodes. Similarly, volume tion, methodology, visualization and the writing of the original
expansion from NMC keeps building up. This generated dra were carried out by Neelam Ghanshyam Yadav. Nicolas
mechanical stress keeps on adding up until there's complete Folastre, Mickael Bolmont, and Arash Jamali aided in the
delamination of the SE/cathode interface and huge contact loss development of the in situ cell used during the experiment.
at the SE/anode interface. Dendrites formed at the anode Mathieu Morcrette and Carine Davoisne, the corresponding
interface do not penetrate inside SE, and thus we believe that authors were responsible for project administration, supervi-
LPSCl acts as a better barrier compared to LPS, in preventing sion, validation of the research output and reviewing the written
dendrite propagation. This ability to act as a better barrier manuscript. The above descriptions are accurate and have been
possibly comes from the good compacity of SE, due to large agreed upon by all authors.
particles and also a good particle size distribution (PSD). Also,
the nature of fracture with LPSCl is ductile, and hence the
cracks propagate at a slower rate compared to LPS. The higher Conflicts of interest
surface roughness due to the large particle size of SE could be
one plausible explanation for the many different dendrite There are no conicts to declare.
morphologies observed. Thus, the main mode of failure could
be a chemo-mechanical failure for LPSCl.
Acknowledgements
Conclusions
The authors would like to thank the French ministry (MESRI)
An understanding of the failure mechanisms in all-solid-state and the French National Research Agency (STORE-EX Labex
batteries using promising solid electrolytes is crucial for their Project ANR-10-LABX-76-01) for nancial support. We would
development. Hence we performed a thorough chemical and also like to thank the “cellule énergie” of CNRS for the funding
morphological investigation using in situ and operando SEM to develop the in situ cell used during this study.

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17153
View Article Online

Journal of Materials Chemistry A Paper

23 M. Suyama, S. Yubuchi, M. Deguchi, A. Sakuda,


References
M. Tatsumisago and A. Hayashi, J. Electrochem. Soc., 2021,
1 A. Sakuda, A. Hayashi and M. Tatsumisago, Sci. Rep., 2013, 3, 168, 060542.
2–6. 24 W. Zhang, D. Schröder, T. Arlt, I. Manke, R. Koerver,
2 H. Deiseroth, S. Kong, H. Eckert, J. Vannahme, C. Reiner, R. Pinedo, D. A. Weber, J. Sann, W. G. Zeier and J. Janek, J.
T. Zaiß and M. Schlosser, Angew. Chem., 2008, 47, 755–758. Mater. Chem. A, 2017, 5, 9929–9936.
3 Q. Zhang, D. Cao, Y. Ma, A. Natan, P. Aurora and H. Zhu, Adv. 25 L. Zhang, T. Yang, C. Du, Q. Liu, Y. Tang, J. Zhao, B. Wang,
Mater., 2019, 31, 1–42. T. Chen, Y. Sun, P. Jia, H. Li, L. Geng, J. Chen, H. Ye,
4 T. Hakari, M. Nagao, A. Hayashi and M. Tatsumisago, J. Z. Wang, Y. Li, H. Sun, X. Li, Q. Dai, Y. Tang, Q. Peng,
Power Sources, 2015, 293, 721–725. T. Shen, S. Zhang, T. Zhu and J. Huang, Nat. Nanotechnol.,
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

5 Y. Zhu, X. He and Y. Mo, ACS Appl. Mater. Interfaces, 2015, 7, 2020, 15, 94–98.
23685–23693. 26 K. J. Harry, D. T. Hallinan, D. Y. Parkinson, A. A. MacDowell
6 L. Zhou, N. Minafra, W. G. Zeier and L. F. Nazar, Acc. Chem. and N. P. Balsara, Nat. Mater., 2014, 13, 69.
Res., 2021, 54, 2717–2728. 27 K. B. Hatzell, X. C. Chen, C. L. Cobb, N. P. Dasgupta,
7 L. Xu, S. Tang, Y. Cheng, K. Wang, J. Liang, C. Liu, Y.-C. Cao, M. B. Dixit, L. E. Marbella, M. T. McDowell,
F. Wei and L. Mai, Joule, 2018, 2, 1991–2015. P. P. Mukherjee, A. Verma, V. Viswanathan, A. S. Westover
8 A. Banerjee, H. Tang, X. Wang, J.-H. Cheng, H. Nguyen, and W. G. Zeier, ACS Energy Lett., 2020, 5, 922–934.
M. Zhang, D. H. S. Tan, T. A. Wynn, E. A. Wu and 28 J. Kasemchainan, S. Zekoll, D. S. Jolly, Z. Ning, G. O. Hartley,
J.-M. Doux, ACS Appl. Mater. Interfaces, 2019, 11, 43138– J. Marrow and P. G. Bruce, Nat. Mater., 2019, 18, 1105–1111.
43145. 29 H. Liu, X. B. Cheng, R. Xu, X. Q. Zhang, C. Yan, J. Q. Huang
9 S. Wenzel, T. Leichtweiss, D. Krüger, J. Sann and J. Janek, and Q. Zhang, Adv. Energy Mater., 2019, 9, 1–7.
Solid State Ionics, 2015, 278, 98. 30 S. H. Kim, K. H. Kim, H. S. Choi, D. Im, S. Heo and
10 W. D. Richards, L. J. Miara, Y. Wang, J. C. Kim and G. Ceder, H. S. Choi, J. Mater. Chem. A, 2019, 7, 13650–13657.
Chem. Mater., 2016, 28, 266–273. 31 J. A. Lewis, F. J. Q. Cortes, Y. Liu, J. C. Miers, A. Verma,
11 S. Wenzel, S. J. Sedlmaier, C. Dietrich, W. G. Zeier and B. S. Vishnugopi, J. Tippens, D. Prakash, T. S. Marchese,
J. Janek, Solid State Ionics, 2018, 318, 102–112. S. Y. Han, C. Lee, P. P. Shetty, H. W. Lee, P. Shevchenko,
12 F. Walther, S. Randau, Y. Schneider, J. Sann, M. Rohnke, F. De Carlo, C. Saldana, P. P. Mukherjee and
F. H. Richter, W. G. Zeier and J. Janek, Chem. Mater., 2020, M. T. McDowell, Nat. Mater., 2021, 20, 503–510.
32, 6123–6136. 32 A. J. Sanchez, E. Kazyak, Y. Chen, J. Lasso and
13 F. Walther, R. Koerver, T. Fuchs, S. Ohno, J. Sann, N. P. Dasgupta, J. Mater. Chem. A, 2021, 9, 21013–21023.
M. Rohnke, W. G. Zeier and J. Janek, Chem. Mater., 2019, 33 M. Nagao, A. Hayashi, M. Tatsumisago, T. Kanetsuku,
31, 3745–3755. T. Tsuda and S. Kuwabata, Phys. Chem. Chem. Phys., 2013,
14 W. Zhang, D. A. Weber, H. Weigand, T. Arlt, I. Manke, 15, 18600.
D. Schröder, R. Koerver, T. Leichtweiss, P. Hartmann, 34 M. Du, Y. Sun, B. Liu, B. Chen, K. Liao, R. Ran, R. Cai,
W. G. Zeier, W. G. Zeier and J. Janek, ACS Appl. Mater. W. Zhou and Z. Shao, Adv. Funct. Mater., 2021, 2101556.
Interfaces, 2017, 9, 17835–17845. 35 K. Dong, Y. Xu, J. Tan, M. Osenberg, F. Sun, Z. Kochovski,
15 W. Zhang, T. Leichtweiß, S. P. Culver, R. Koerver, D. Das, D. T. Pham, S. Mei, A. Hilger, E. Ryan, Y. Lu, J. Banhart
D. A. Weber, W. G. Zeier and J. Janek, ACS Appl. Mater. and I. Manke, ACS Energy Lett., 2021, 6, 1719–1728.
Interfaces, 2017, 9, 35888–35896. 36 Z. Ning, D. S. Jolly, G. Li, R. De Meyere, S. D. Pu, Y. Chen,
16 J. Auvergniot, A. Cassel, J. B. J.-B. J. B. Ledeuil, V. Viallet, J. Kasemchainan, J. Ihli, C. Gong, B. Liu, D. L. R. Melvin,
V. Seznec and R. Dedryvère, Chem. Mater., 2017, 29, 3883. A. Bonnin, O. Magdysyuk, P. Adamson, G. O. Hartley,
17 B. Chen, J. Ju, J. Ma, J. Zhang, R. Xiao, G. Cui and L. Chen, C. W. Monroe, T. J. Marrow and P. G. Bruce, Nat. Mater.,
Phys. Chem. Chem. Phys., 2017, 19, 31436–31442. 2021, 20, 1121–1129.
18 K. Yoon, J. Kim, W. M. Seong, M. H. Lee and K. Kang, Sci. 37 H. Sun, Q. Liu, J. Chen, Y. Li, H. Ye, J. Zhao, L. Geng, Q. Dai,
Rep., 2018, 8, 8066. T. Yang, H. Li, Z. Wang, L. Zhang, Y. Tang and J. Huang, ACS
19 R. Koerver, I. Aygün, T. Leichtweiß, C. Dietrich, W. Zhang, Nano, 2021, 15, 19070–19079.
J. O. Binder, P. Hartmann, W. G. Zeier and J. Janek, Chem. 38 M. Golozar, A. Paolella, H. Demers, S. Bessette, M. Lagacé,
Mater., 2017, 29, 5574–5582. P. Bouchard, A. Guer, R. Gauvin and K. Zaghib, Commun.
20 A. Kato, H. Kowada, M. Deguchi, C. Hotehama, A. Hayashi Chem., 2019, 2, 1–9.
and M. Tatsumisago, Solid State Ionics, 2018, 322, 1–4. 39 M. B. Dixit, N. Singh, J. P. Horwath, P. Shevchenko, M. Jones,
21 S. Wenzel, T. Leichtweiss, D. Krüger, J. Sann and J. Janek, E. A. Stach, T. S. Arthur and K. Hatzell, Matter, 2020, 3, 2138–
Solid State Ionics, 2015, 278, 98–105. 2159.
22 T.-T. Zuo, R. Rueß, R. Pan, F. Walther, M. Rohnke, S. Hori, 40 R. Koerver, W. Zhang, L. De Biasi, S. Schweidler,
R. Kanno, D. Schröder and J. Janek, Nat. Commun., 2021, A. O. Kondrakov, S. Kolling, T. Brezesinski, P. Hartmann
12, 6669. and G. Zeier, Energy Environ. Sci., 2018, 11, 2142–2158.
41 H. Ye, Y. Zhang, Y. Yin, F. Cao and Y. Guo, ACS Cent. Sci.,
2020, 6, 661–671.

17154 | J. Mater. Chem. A, 2022, 10, 17142–17155 This journal is © The Royal Society of Chemistry 2022
View Article Online

Paper Journal of Materials Chemistry A

42 M. Otoyama, M. Suyama, C. Hotehama, H. Kowada, 46 Y. Zhou, C. Doerrer, J. Kasemchainan, P. G. Bruce, M. Pasta


Y. Takeda, K. Ito, A. Sakuda, M. Tatsumisago and and L. J. Hardwick, Batteries Supercaps, 2020, 3, 647–652.
A. Hayashi, ACS Appl. Mater. Interfaces, 2021, 13, 5000–5007. 47 N. J. J. De Klerk and M. Wagemaker, ACS Appl. Energy Mater.,
43 F. Han, A. S. Westover, J. Yue, X. Fan, F. Wang, M. Chi, 2018, 1, 5609–5618.
D. N. Leonard, N. J. Dudney, H. Wang and C. Wang, Nat. 48 K. Nie, Y. Hong, J. Qiu, Q. Li, X. Yu, H. Li and L. Chen, Front.
Energy, 2019, 4, 187. Chem., 2018, 6, 1–19.
44 S. Hao, S. R. Daemi, T. M. M. Heenan, W. Du, C. Tan, 49 L. Ye and X. Li, Nature, 2021, 593, 218–222.
M. Storm, C. Rau, D. J. L. Brett and P. R. Shearing, Nano 50 T. Yang, H. Li, Z. Wang, L. Zhang, Y. Tang, J. Huang, H. Sun,
Energy, 2021, 82, 105744. Q. Liu, J. Chen, Y. Li, H. Ye, J. Zhao, L. Geng, Q. Dai, T. Yang,
45 P. P. Paul, B. Chen, S. A. Langevin, E. J. Dufek, J. Nelson and H. Li, Z. Wang, L. Zhang, Y. Tang and J. Huang, ACS Nano,
Published on 19 July 2022. Downloaded by Universite de Picardie on 11/6/2023 10:37:53 AM.

J. S. Ko, Energy Storage Mater., 2022, 45, 969–1001. 2021, 19070–19079.

This journal is © The Royal Society of Chemistry 2022 J. Mater. Chem. A, 2022, 10, 17142–17155 | 17155

You might also like