Understanding the Capacity Loss in LNMO-LTO Lithium-Ion ...

doi.org/10.26434/chemrxiv.5944228.v1

Understanding the Capacity Loss in LNMO-LTO Lithium-Ion Cells atAmbient and Elevated TemperaturesBurak Aktekin, Matthew J. Lacey, Tim Nordh, Reza Younesi, Carl Tengstedt, Wolfgang Zipprich, DanielBrandell, Kristina Edström

Submitted date: 02/03/2018 • Posted date: 05/03/2018Licence: CC BY-NC-ND 4.0Citation information: Aktekin, Burak; Lacey, Matthew J.; Nordh, Tim; Younesi, Reza; Tengstedt, Carl;Zipprich, Wolfgang; et al. (2018): Understanding the Capacity Loss in LNMO-LTO Lithium-Ion Cells atAmbient and Elevated Temperatures. ChemRxiv. Preprint.

The high voltage spinel LiNi0.5Mn1.5O4 (LNMO) is an attractive positive electrode due to its operating voltagearound 4.7 V (vs Li/Li+) and high power capability. However, problems including electrolyte decomposition athigh voltage and transition metal dissolution, especially at elevated temperatures, have limited its potentialuse in practical full cells. In this paper, a fundamental study for LiNi0.5Mn1.5O4 || Li4Ti5O12 (LTO) full cells hasbeen performed to understand the effect of different capacity fading mechanisms contributing to overall cellfailure. Electrochemical characterization of cells in different configurations (regular full cells, back-to-backpseudo-full cells and 3-electrode full cells) combined with an intermittent current interruption technique havebeen performed. Capacity fade in the full cell configuration was mainly due to progressively limited lithiation ofelectrodes caused by a more severe degree of parasitic reactions at the LTO electrode, while thecontributions from active mass loss from LNMO or increases in internal cell resistance were minor.Comparison of cell formats constructed with and without the possibility of cross-talk indicate that the parasiticreactions on LTO occur because of the transfer of reaction products from the LNMO side. The efficiency ofLTO is more sensitive to temperature causing a dramatic increase in the fading rate at 55 °C. Theseobservations show how important the electrode interactions (cross-talk) can be for the overall cell behaviour.Additionally, internal resistance measurements showed that the positive electrode was mainly responsible forthe increase of resistance over cycling, especially at 55 °C. Surface characterization showed that LNMOsurface layers were relatively thin when compared to the SEI on LTO. The SEI on LTO does not contributesignificantly to overall cell resistance even though these films are relatively thick. XANES measurementsshowed that the Mn and Ni observed on the anode were not in metallic state; the presence of elemental metalsin the SEI is therefore not implicated in the observed fading mechanism through a simple reduction process ofmigrated metal cations.

File list (2)

download fileview on ChemRxivManuscript.pdf (2.43 MiB)

download fileview on ChemRxivSupporting Information.pdf (1.24 MiB)

1

Understanding the capacity loss in LNMO-LTO lithium-ion cells

at ambient and elevated temperatures

Burak Aktekin1, Matthew J. Lacey1, Tim Nordh1, Reza Younesi1, Carl Tengstedt2,

Wolfgang Zipprich3, Daniel Brandell1, Kristina Edström1

1Department of Chemistry – Ångström Laboratory, Uppsala University, Box 538, SE-75121, Uppsala, Sweden

2Scania CV AB, SE-151 87, Södertälje, Sweden 3Volkswagen AG, D-38436, Wolfsburg, Germany

Abstract

The high voltage spinel LiNi0.5Mn1.5O4 (LNMO) is an attractive positive electrode due to its operating

voltage around 4.7 V (vs. Li/Li+) and high power capability. However, problems including electrolyte

decomposition at high voltage and transition metal dissolution, especially at elevated temperatures, have

limited its potential use in practical full cells. In this paper, a fundamental study for LiNi0.5Mn1.5O4 ||

Li4Ti5O12 (LTO) full cells have been performed to understand the effect of different capacity fading

mechanisms contributing to overall cell failure. Electrochemical characterization of cells in different

configurations (regular full cells, back-to-back pseudo-full cells and 3-electrode full cells) combined

with an intermittent current interruption technique have been performed. Capacity fade in the full cell

configuration was mainly due to progressively limited lithiation of electrodes caused by a more severe

degree of parasitic reactions at the LTO electrode, while the contributions from active mass loss from

LNMO or increases in internal cell resistance were minor. Comparison of cell formats constructed with

and without the possibility of cross-talk indicate that the parasitic reactions on LTO occur because of

the transfer of reaction products from the LNMO side. The efficiency of LTO is more sensitive to

temperature causing a dramatic increase in the fading rate at 55 °C. These observations show how

important the electrode interactions (cross-talk) can be for the overall cell behaviour. Additionally,

internal resistance measurements showed that the positive electrode was mainly responsible for the

increase of resistance over cycling, especially at 55 °C. Surface characterization showed that LNMO

surface layers were relatively thin when compared to the SEI on LTO. The SEI on LTO does not

contribute significantly to overall internal resistance even though these films are relatively thick.

XANES measurements showed that the Mn and Ni observed on the anode were not in metallic state; the

presence of elemental metals in the SEI is therefore not implicated in the observed fading mechanism

through a simple reduction process of migrated metal cations.

Keywords: High voltage spinel, cross-talk, electrode interactions, LNMO-LTO, Mn dissolution

Introduction

LiNi0.5Mn1.5O4 (LNMO) is a promising spinel-type positive electrode (cathode) material for lithium-ion

batteries (LiBs), with a theoretical capacity of 147 mAh/g and an operating voltage around 4.7 V (vs.

Li/Li+). This high voltage plateau, due to the active Ni2+/Ni4+ redox couple, renders a high theoretical

energy density of 690 Wh/kg for the active material.1 Additionally, the intrinsic properties of the spinel

structure allow excellent rate capability, and combined with the comparatively high energy density,

these two main advantages make the material an ideal candidate for high power applications (e.g.,

electric vehicles).2 However, the anodic instability of conventional LiB electrolytes (namely LiPF6 in

organic carbonates) at high voltages cause interfacial side reactions which are also accompanied by

transition metal dissolution. This, in turn, causes rapid capacity fading in full cells, especially at elevated

temperatures.3,4

2

It is important to understand how undesired reactions originating from the cathode side affect the aging

of different cell components and how these contribute to the overall cell life. If classified into three main

groups, the different contributions to capacity fading can be due to ‘active mass loss’, ‘internal

resistance increase’ of the cell and ‘loss of cyclable lithium’.4

Active mass loss and internal resistance increase can result from the degradation of the active material

or other electrode components such as the binder, conductive additive and current collector. Similarly,

oxidation of the electrolyte at high potentials and interactions between the positive and negative

electrodes might influence the cell life-time significantly.5

In optimized LNMO-based composite electrodes, the electrode integrity remains relatively intact, as the

volume expansion of LNMO during cycling is moderate (around 7%)6 and binders such as PVdF7 have

acceptable anodic stability. At high potentials, even though conductive additives such as carbon black

have been shown to be reactive,8,9 and the Al current collector has been shown to dissolve to a minor

extent,10 it is unlikely that these processes damage the electrical network considerably during the limited

cell testing period of most published studies in literature. Therefore, the degradation of inactive electrode

components is not expected to have a major role in active mass loss or internal resistance increase during

this period. On the other hand, transition metal dissolution, which is more severe at elevated

temperatures,11 can cause active mass loss from the cathode. However, the contribution of metal

dissolution to capacity fading directly as a result of active cathode mass loss has been reported to be less

than 5%, as confirmed by the recovery up to 95% of the capacity in a LNMO half cell constructed with

an aged electrode from a LNMO-graphite full cell.12 This verifies that ‘the active mass loss’ which might

be due to metal dissolution (or electrode degradation) has only a minor contribution to the observed

rapid fading of LNMO based full cells during their limited testing time.

As previously mentioned, electrolyte oxidation and transition metal dissolution are often considered as

the main problems for the LNMO material, where both are expected to contribute to capacity fading via

‘internal resistance increase’ and ‘loss of cyclable lithium’. Capacity fading caused by internal resistance

increase can also be described as power fading, since it affects the cell capacity due to kinetic limitations

for a specific test or usage condition. For instance, the generation and precipitation of organic

compounds, LiF, C-F and P-Fx species on LNMO might result in sluggish Li+

intercalation/deintercalation kinetics13 and therefore might increase the electrode impedance.14

Apart from this, interactions between the electrodes (“cross-talk”) are known to occur in LNMO based

full cells. Migration of oxidation products from the positive electrode can result in the formation of

surface films on the negative electrode,15 and migration of dissolved metal cations can cause damage to

an existing Solid Electrolyte Interphase (SEI) on the negative electrode, resulting in further thickening

of the SEI.11 Further contributions may include the evolution of gas products such as CO2 after

electrolyte oxidation16 that can lead to increased cell resistance17; or clogging of electrode and separator

pores by oxidation products.4,18 In the study reported by Kim et al.,12 electrodes were harvested from a

LNMO-graphite full cell which had been cycled until significant capacity loss had occurred, and those

electrodes were then further tested in freshly assembled half cells. It was shown that both half cells were

largely able to recover their initial capacities. This indicates that the capacity loss in this optimized full

cell was dominated by a third mechanism, which can be expressed as ‘loss of cyclable lithium’.

Loss of cyclable lithium is an issue related to capacity balance between anode and cathode, and it is

obviously not a problem in half cells in which the lithium electrode provides an effectively unlimited

capacity. Such a capacity fading mechanism can occur via electrolyte reduction on the negative electrode

during SEI formation19 (also known as negative electrode self-discharge). On the other hand, electrolyte

oxidation at high potential would instead cause positive electrode self-discharge and increase the

charging time. The result of this is excess charge stored by the negative electrode, which therefore

perturbs capacity balance.19 This process, also, results in loss of cyclable lithium. Moreover, one

explanation for the capacity loss in LNMO-graphite full cells has been made by negative electrode self-

discharge as a result of SEI damage caused by migrating metal ions from the positive electrode.11

Interestingly, when LNMO was coupled to Li4Ti5O12 (LTO), which is known for its good

electrochemical stability in standard lithium-ion battery electrolytes (electrode potential around 1.5 V

3

vs. Li/Li+) and thereby limited SEI layer formation, the capacity fading was again shown to be due to

negative electrode self-discharge.15,20 This showed that some cross-talk between the electrodes – not

necessarily limited to dissolved metal ions – indeed exists and has a significant role in determining cell

lifetime, since electrolyte reduction on LTO could not cause such negative electrode self-discharge

alone.

The well-known example of dissolved metal ion (e.g. Mn) interactions between electrodes has been

studied extensively21 but the exact mechanisms leading to the cyclable lithium loss is not yet clear.

Similar ambiguities also exist regarding the oxidation state of transition metals deposited on the negative

electrode.22 Recently, some studies have reported a different kind of cross-talk: the interaction of gas

products evolved in systems such as LNMO-graphite23 and NMC-graphite.24,25,26 The main gas product

of electrolyte oxidation was identified as CO2 at high cathode voltages16 while gases such as C2H4, CO

and H2 might evolve on the negative electrode at potentials below 1 V (vs. Li/Li+).24 Xiong et al. showed

CO2 evolution when de-lithiated (charged) NMC was simply stored in the electrolyte at 60 °C, while no

considerable gas was present after storage in charged full cells (with a graphite electrode), indicating

CO2 consumption by the negative electrode.26 The reduction of CO2 to species such as lithium oxalate

or lithium carbonate would cause cyclable lithium loss – and this loss would be reversible if soluble

oxalate can migrate back to the positive electrode and be re-oxidized (i.e., a redox shuttle mechanism).27

In addition to gas-consuming reactions, gas-generating reactions occurring via cross-talk are also

possible. Metzger et al. for example observed persistent H2 evolution at a graphite electrode caused by

the reduction of protic species formed continuously by electrolyte oxidation at a NMC cathode.24 The

correlation between different types of electrode interactions and the overall fading of cell capacity

indicates that these are not independent processes, and more research is necessary to identify their exact

roles in full cells containing high voltage electrode materials.

In this paper, we aim to understand how different components of the cell contribute to capacity fading

in LNMO-LTO full cells cycled both at room temperature and elevated temperature (55 °C). As seen in

the literature, the fading rate for this type of cell at elevated temperature is much faster (e.g. at 55 °C).

In some reports, this performance difference is more pronounced; e.g., quite stable cycling for >1000

cycles has been reported at room temperature28 while fast fading was observed at 55 °C. We investigate

the reasons behind these observations by following independent internal resistances and electrode

potentials with respect to LNMO and LTO in three-electrode cells which have been cycled

galvanostatically with short-duration intermittent current interruptions.29 Pseudo-full cells (i.e., ‘back-

to-back’ cells) as suggested by Li et al.20 have also been tested to observe cycling behaviour when no

cross-talk is possible. Regular full cells and half cells have been tested for reference as well as for the

ex situ characterization of cycled electrodes. We believe that the findings from this combination of

different cell configurations are highly valuable for providing insights into the ultimate failure

mechanisms of ‘high voltage positive electrodes’, since LTO has a high electrode potential and does not

contribute to extensive SEI formation (as opposed to for example graphite, which consumes cyclable

lithium during initial film formation and film repair mechanisms following SEI damage from dissolved

metal ions).

Experimental

Preparation of electrodes & pouch cells. LNMO slurry was prepared via 2 hours ball milling of

commercial LNMO powders (90 wt%) with carbon black (Imerys, C65, 5 wt%) and a poly(vinylidene

difluoride)-based binder (5 wt%, PVdF-HFP, Kynar Flex 2801) in N-methyl-2-pyrrolidone (NMP,

VWR Chemicals). The binder was pre-dissolved in NMP in 4:96 wt% ratio. The slurry was casted onto

Al foil (20 μm thickness) and subsequently calendered to a final coating film thickness of approximately

50 μm. The average porosity was 40-50% and the active mass loading was around 10.6 mg/cm2 (or 1.56

mAh/cm2). Commercial LTO electrodes were provided by Leclanché and had a capacity of

approximately 1.7 mAh/cm2. Therefore, the capacity of full cells were always limited by the LNMO

electrode. A schematic of a standard pouch test cell is shown in Figure 1a. 2 cm diameter electrodes

were punched from LNMO and LTO sheets and 2 layers of Celgard 2500 separators with 4 × 4 cm2 area

4

were used. This separator is a monolayer polypropylene with 25 μm thickness and 64 nm average pore

size. The separators were dried overnight under vacuum at 70 °C prior to use. 1 M LiPF6 (Ferro Corp.)

dissolved in EC and DEC (BASF) with 1:1 volume ratio was used as electrolyte. The electrolyte volume

was fixed at 120 μl. The connecting tabs were shaped with 2 cm diameter circular ends inside the cell

to ensure homogeneous pressure distribution and good matching of electrode position. Al foil was used

to contact LNMO and Cu for LTO electrodes. Nickel strips were used to contact lithium metal

electrodes.

Figure 1. Schematic view of pouch cell designs used for cell testing in this study.

In three-electrode cells (Figure 1b), concentric lithium rings were punched with an inner diameter of 2.2

cm and outer diameter of 2.6 cm and carefully placed between LNMO and LTO, facing separators on

both sides. In half cells, the lithium counter electrode diameter was 2.6 cm. The cells were vacuum

sealed in an Ar-filled glove box with H2O < 5 ppm and O2 < 1 ppm. Pouch cell size was 10 × 5 cm and

during cycling, half of the pouch (5 × 5 cm) was compacted using flat plates to ensure better electrical

contact. The extra 5 × 5 cm area was for accumulation of evolving gases in order to reduce the risk of

cell leakage or bubble formation in pores. So-called “back-to-back cells” (Figure 1c) were constructed

by connecting the working electrodes of separate LTO and LNMO half cells as the negative and positive

terminal respectively, and shorting their respective lithium electrode terminals. The lithium electrode

terminals were connected to the reference electrode terminal on the potentiostat to allow for separate

measurements of the LTO and LNMO electrodes. A modified back to back cell was also constructed by

using only one pouch cell. This time, 4 sheets of separator were used and a large piece of nickel metal

foil covered with 2.6 cm diameter Li discs on both sides was placed between the LNMO and LTO

electrodes so that the lithium metal faced two separators on both sides. There was no direct electrolyte

contact between the two sides of this foil (Li+Ni+Li), however, evolving gases from one electrode would

not be prevented from coming into contact with the other.

5

Electrochemical characterization. All cells were galvanostatically cycled between cell voltages of 1.5

V and 3.5 V (i.e., vs. LTO) at C/5 rate according to the calculated theoretical capacity of LNMO. In all

measurements, the cells were kept under OCV condition for 10 hours at the testing temperature prior to

cycling. A Novonix HPC (High Precision Charger) system was used for testing of regular (two

electrode) cells at 55 °C while a Digatron BTS-600 was used for room temperature (24 °C – 25 °C)

testing. Testing of three electrode and back-to-back cells was made via galvanostatic cycling combined

with an intermittent current interruption (ICI) resistance determination method29,30 using a Bio-Logic

VMP2 instrument. Cells were tested at room temperature, 40 °C and 55 °C. The current was interrupted

after every 150 seconds for 1 second. The calculation of a time-independent cell internal resistance was

made by regression of E vs. t1/2 to t=0 for the interruption period, and calculated according to Ohm’s law

(i.e., R = ∆E/∆I). Data analysis was performed using the R programming language and environment.

Simple simulations of capacity fade due to coulombic efficiency imbalance were also written in R; the

code is supplied in the Supporting Information (SI).

Surface characterization. Analysis of pristine LNMO and LTO electrodes were performed prior to any

electrolyte contact, and on electrodes after cycling at C/5 rate for 50 cycles both at room temperature

and at 55 °C. Electrode morphologies were imaged via a Zeiss 1550 scanning electron microscope

(SEM). Cycled pouch cells were opened in an Ar-filled glove box. LNMO and LTO electrodes were

rinsed with DMC five times using 4-5 droplets each time to remove salt residues. Electrodes were placed

on carbon tapes after drying off the DMC. Samples were transferred to the SEM in airtight glass vials

and exposed to air for 15-20 seconds before being transferred to the SEM chamber. The accelerating

voltage was 3 kV and the working distance was 5 mm during analysis. All images were scaled to the

same magnification using the image analysis software ImageJ.

Surface characterization was made via X-ray photoelectron spectroscopy (XPS). The sample preparation

was the same as for SEM except an airtight transfer system was used for sample to avoid any exposure

to air. All samples were analyzed using a Phi-5500 instrument with monochromatized Al-Kα radiation

(1486.6 eV). Additionally, LTO electrodes were analyzed via HAXPES for depth-profiling of surface

films. To ensure reliability of results from the different analyses, multiple samples were cut from each

electrode and vacuum sealed into coffee bags (two times) using protective boxes. Measurements were

performed at the Helmholtz Zentrum Berlin, BESSY KMC-1 beamline, High Kinetic Energy

Photoelectron Spectrometer (HIKE) end station using a Scienta R4000 electron analyzer. Sealed

samples were opened in an Ar-filled glove box, mounted on a sample holder with copper adhesive tape

and transferred to the analysis chamber without any air exposure. Photon excitation energies were 2005

and 6015 eV. XANES measurements were performed at the same beamline using a Bruker XFlash 4010

fluorescence detector.

Data calibration was made by linear shifting of the hydrocarbon peak to 285 eV for the LNMO data.

The XPS data for the LTO electrodes were calibrated using the titanium (Ti 2p) peak. The titanium peak

was set to be 459 eV as measured previously for LTO with carboxymethyl cellulose (CMC) binder.31

For the 1486.6 eV measurement of the sample cycled at 55 °C, no Ti 2p peak was visible, therefore the

main peak in the Mn 2p spectra was adjusted to align with the corresponding peak observed with

HAXPES data (2005 eV) as shown in supplementary information (SI). This shift was applied to all other

spectra in that sample. All normalization was done by normalizing the area of the spectra to 1. For

measurements with no peaks, the noise level was adjusted to match instrument noise level for other

measurements. For XANES, energy calibration was made using gold foil mounted on the sample holder.

The monochromator was adjusted to the edge energy and Au 4f spectra were measured to determine the

energy shift. CasaXPS and Igor were used for the analysis of XPS and HAXPES data. Analysis and

normalization of XANES data were made using Athena software.

6

Theory

The effect of electrolyte oxidation and reduction. As mentioned in the introduction section, half cells

with optimized LNMO electrodes and a Li counter-reference electrode with a non-limiting capacity are

able to operate without significant capacity loss, while capacity loss is apparent in full cells where both

electrodes have a limiting lithium inventory: this is a clear indication of cyclable lithium loss during

cycling.

It is necessary at this point to clarify certain terms. The term ‘lithium inventory’ refers to the amount of

lithium hosted in an electrode at any given time; in the absence of side reactions, the sum of the lithium

inventory of both electrodes is constant. On the other hand, the term ‘cyclable lithium’ refers to the

usable (or electrochemically accessible) lithium content of the cell, and is distinct from the sum of the

lithium inventory of the constituent electrodes. A loss in cyclable lithium is observed both at room

temperature and elevated temperatures [20] in LNMO-based full cells. It is important to understand the

reasons that lead to such a loss to meaningfully interpret the response of cells to, for example, different

cycling conditions, and electrolytes [32].

We can consider the hypothetical self-discharge behavior of a partially charged full cell as an example

to illustrate two different mechanisms which result in cyclable lithium loss. In this demonstration (Figure

2), it is assumed that solvent molecules (denoted as ‘S’) decompose either on the positive electrode

(electrolyte oxidation) or on the negative electrode (electrolyte reduction). The capacity of the complete

cell in this scenario is limited by the capacity of the positive electrode (“cathode-limited”) and initially

at 50% state of charge (SOC) (Figure 2b).

In the event of electrolyte oxidation (Figure 2c), decomposition of the solvent molecule (S S+ + e-)

at the positive electrode necessitates simultaneous intercalation of a lithium ion from the electrolyte to

maintain charge neutrality. This increases the lithium inventory of the positive electrode, with no

corresponding change in the lithium inventory of the negative electrode. Conversely, in the case of

electrolyte reduction (Figure 2d), decomposition of the solvent molecule (S + e- S-) at the negative

electrode is accordingly accompanied by lithium ion de-intercalation from the negative electrode. In this

case, the lithium inventory of the negative electrode would decrease while the positive electrode remains

unchanged.

7

Figure 2. The schematic view of a cathode-limited full cell after assembly (a) and after charging to 50%

SOC. The distribution of lithium inventory after calendar aging would be similar to (c) if only electrolyte

oxidation had occurred and similar to (d) for electrolyte reduction.

In each of these cases, as depicted in Figure 2c-d, an imbalance in lithium inventory is created. In the

absence of either of the parasitic processes described above, a complete discharge of the cell would

return the electrodes to their initial state, with the positive electrode fully lithiated and the anode fully

de-lithiated (Figure 2a). In the case of electrolyte oxidation: on full discharge, some lithium inventory

would remain in the negative electrode once the positive electrode reaches its fully lithiated state. If this

remaining lithium inventory is larger than the excess capacity of the negative electrode relative to the

positive electrode, the negative electrode will become the limiting electrode in the system when the cell

is charged: the negative electrode will become fully lithiated before the positive electrode can be fully

delithiated. There is therefore a loss of cyclable lithium: the result of an imbalance of lithium inventory

between the electrodes which prevents total delithiation of either electrode during cycling. The sum of

the lithium inventory of both electrodes increases in this case.

In the event of electrolyte reduction (Figure 2d), on full discharge the negative electrode would be fully

de-lithiated, preventing the full lithiation of the positive electrode. In this case, the loss in cyclable

lithium arises from a similar imbalance in which both electrodes are prevented from being totally

lithiated. In this case, the sum of the lithium inventory of the electrodes is decreased.

The effect of positive and negative electrode efficiency on capacity fading. Of course, parasitic

reactions on the electrode surfaces occur not only at open circuit conditions but also during charging

and discharging of cells. We can consider three extreme cases with simple simulations of arbitrary full

cells in which different ‘coulombic efficiencies’ are assigned to each electrode. We can assume a larger

degree of parasitic reactions during charging compared to discharging, on both electrodes, due to

overpotential: for the purpose of this exercise we assume that parasitic reactions take place only during

charge, and that the discharge is 100% efficient for both electrodes. Simulations were made for LNMO-

8

limited full cells (i.e., excess capacity in LTO) for comparison with the cells tested experimentally in

this work (an example of the code is given in the Supporting Information).

In the simulations, we assigned initial capacities such that the ratio of the electrode capacities are

comparable with the experiments presented in this work. A capacity of 5 (“mAh”) was chosen for the

negative (LTO) electrode and 4 (“mAh”) for the positive (LNMO). For simplicity, it is assumed that the

electrode impedance does not affect the cycling performance and no active mass loss occurs.

First, we observe how the cyclable lithium loss would be manifested in full cells if parasitic reactions

are limited to the LNMO electrode (see Figure 2c and Figure 3a-b). Consider a case where LNMO has

an efficiency of 0.95 (i.e., 95%) and LTO an efficiency of 1. That is, for every 1 mole of charge passed

through the cell, the lithium inventory in LNMO changes by 0.95 moles, and 0.05 moles of charge is

passed through a parasitic reaction at the positive electrode. In Fig. 3b, the black line shows the

instantaneous amount of charge stored (Li+ ions, i.e., lithium inventory) in LNMO (blue for LTO). LTO

has zero initial lithium inventory, LNMO is, correspondingly, initially fully lithiated.

As clearly seen in Figure 3b, a parasitic process at the LNMO electrode effectively extends the charging

time and results in the intercalation of a greater amount of lithium into LTO (~4.2 mAh) than is initially

stored in LNMO (4 mAh); the excess Li+ is drawn from the electrolyte. However, this “excess” lithium

in LTO cannot later intercalate back into LNMO during discharging since the capacity of LNMO is still

limited to 4 mAh. Therefore, nearly 0.2 mAh of lithium inventory remains ‘trapped’ in LTO after one

complete cycle. It is seen in Figure 3a that, in this scenario, the overall cell capacity is stable over the

first 5 cycles but thereafter fades at an almost constant rate. The stable early cycling results from the

excess capacity in anode – allowing complete delithiation of the positive electrode, but for only a limited

time. It is clear from Figure 3b that the “active cycling window” shifts to the higher lithiation states of

both electrodes during cycling. From the 5th cycle onwards, the lithium inventory in the negative

electrode reaches the full 5 mAh capacity at the end of charging step. Complete delithiation of LNMO

is therefore no longer possible during the subsequent cycles. Capacity fading is observed from this point

because the parasitic process becomes progressively limiting, and the cycling window of both electrodes

becomes progressively more limited to the higher lithiation states.

Therefore, ultimately, LNMO and LTO would both be ‘trapped’ in their fully lithiated states and the

amount of cyclable lithium will become zero. In this type of failure mechanism, the cyclable lithium

loss need not be attributed to any chemical inactivation of lithium in the system (as compared to, for

example, formation of inert lithium-containing compounds) but rather an accumulation of excess lithium

in the negative electrode. In this scenario, the coulombic efficiency of the cell is equal to the coulombic

efficiency of the positive electrode. We can describe this fading mechanism as “limited delithiation”, as

the lithium inventory which can be extracted from either electrode during the respective delithiation

process progressively decreases.

9

Figure 3. Simulation of capacity fading with different ‘efficiencies’ of negative and positive electrodes

(a, c, d). Corresponding charge amounts in both electrodes are shown below the each graph (b, d, f). In

these graphs, 4 mAh refers to fully lithiated LNMO and 5 mAh to fully lithiated LTO.

In Figure 3(c, d), a different scenario is presented where a parasitic process occurs only on the negative

electrode (corresponding to an LTO efficiency of 0.95; an unrealistic scenario for LTO electrodes in a

standard electrolyte, but still a useful comparison). In this case, the positive electrode can be completely

delithiated during charging, but since some of the charge passed is consumed in side reactions on the

negative side, a smaller amount of lithium is intercalated into the negative electrode (cf. self-discharge

of the negative electrode, as illustrated in Figure 2d). Ultimately, both electrodes would be trapped in

their fully delithiated states where the amount of cyclable lithium again approaches zero. Here, the

overall coulombic efficiency of the cell is determined by the coulombic efficiency of the negative

electrode. We can correspondingly describe this fading mechanism as “limited lithiation”, as the lithium

inventory which can be inserted into each electrode during the respective lithiation process progressively

decreases.

In Figure 3(e, f), we consider a scenario where parasitic processes occur on both electrodes to the same

degree (i.e., they both have 0.95 efficiency). Here, the excess charge passed by the positive electrode

during charging is now balanced by an equivalent amount of excess charge passed at the negative

electrode. If the efficiencies of both electrodes are equal, no imbalance in lithium inventory is created

between the electrodes, and no capacity fade occurs as a result. Even though stable cycling is observed,

it should be noted that this scenario still represents the worst case for the real system. Assuming the side

reactions are irreversible, then electrolyte degradation (and the corresponding impedance increase)

would eventually degrade the performance of the cell.

Although the above scenarios represent extreme cases, parasitic reactions do occur on both electrodes

in practice. We can therefore expect behavior corresponding to one of the cases depicted here,

determined by the more inefficient electrode. This is true not just for the LNMO-LTO system considered

here but for any comparable battery system.

10

It has been observed in the literature that the cyclable lithium loss in LNMO-based full cells is due to

the “limited lithiation” mechanism previously discussed (i.e., the scenario depicted in Figure 3c-d). For

example, Kim et al. [12] demonstrated this mechanism in LNMO-graphite cells by further cycling of

half cells made with LNMO and graphite electrodes harvested from exhausted full cells. Following the

scenarios presented above, the negative electrode must therefore be less efficient than the positive

electrode. Where a passivated negative electrode such as graphite is used, this inefficiency implies that

the passivation layer (SEI) should therefore somehow be damaged by species generated at the positive

electrode, or that the SEI is permeable to these species. Moreover, Li et al. [20] studied LNMO-LTO

pseudo- (back-to-back) full cells alongside conventional full cells to show that cross-talk exists, as LTO

in isolation cannot be responsible for the high negative electrode inefficiency: products of electrolyte

oxidation at the positive electrode can be reactive towards the negative electrode.

These scenarios described above serve as a foundation for the interpretation of our experiments here,

where different cycling conditions also have been investigated.

Results and Discussion

Electrochemical Cycling Performance of 2-electrode full cells. LNMO-LTO full cells cycled at C/5

rate and at room temperature showed quite stable cycling while the fading rate was much faster when

similar cells were cycled at 55 °C at the same C-rate (Figure 4). After 50 cycles, another cycle at C/20

rate was also tested for both cells and showed only a small improvement in cell capacity, indicating that

the change in cell resistance had only a limited effect on the observed capacity. For the 55 °C cell, the

changes in voltage curves between the first and 50th cycle were especially apparent for regions of high

LNMO lithiation, verifying cyclable lithium loss through the “limited lithiation” mechanism. The

coulombic efficiency for room temperature cycling reached around 0.997 at the 20th cycle while it was

around 0.98 for the 55 °C cell. When the two last cycles are considered (RT), the coulombic efficiency

decreased from 0.998 to 0.993 when the C-rate changed from C/5 to C/20. The corresponding decrease

at 55 °C was from 0.98 to 0.93. Both observations show that the degree of side reactions is time

dependent, rather than dependent on cycle number.

11

Figure 4. (a) Galvanostatic cycling of LNMO-LTO full cells at room temperature and 55 °C. Smaller

symbols refer to coulombic efficiency, and larger symbols refer to charge (hollow) and discharge (filled)

capacity. (b) Selected voltage curves from these cells.

Although the difference in cycling performance for the two temperatures is large, it would be even more

dramatic if evolving gases disrupted the electrolyte-electrode surface (e.g. pore blocking, etc.) or caused

leakage of the cell. Such a rapid failure would limit the cycling duration over which a meaningful

interpretation of results and ex-situ characterization of electrodes can be made, and we therefore

intentionally added extra space in the pouch cell that could delay any effect of gas evolution (see

experimental section) which according to our experience is a significant problem at elevated

temperature. The performance of a cell prepared in an identical manner to the standard cells used in this

work, but without extra buffer space in the pouch cell, is shown in Figure 5. The rapid cell failure seen

in Figure 5 shows that the fast capacity fading at elevated temperature reported by a number of studies

might be partially due to experimental conditions such as cell design. It also shows the importance of

cell inspection (e.g. for leakage) after certain durations during cell testing.

The effect of waiting time (at OCV) in the fully charged state is also shown in Figure 5. The waiting

time between cycles is gradually increased from zero to 48 hours to observe the effect of parasitic

reactions under static conditions. As the waiting time increases, the discharge capacity is expected to

decrease significantly due to self-discharge. If the difference between charge and discharge capacities

are compared for two subsequent cycles (with different OCV time, e.g. 25th and 26th cycle), it is seen

that self-discharge rate varies between 0.25-0.35 mAh g-1 h-1. A sudden change in charge capacity can

be observed when the waiting time is 48 hours. This indicates that the cell reaches a break-down point

where other contributions apart from cyclable lithium loss start to have a significant role. One such

example is likely to be increased internal cell resistance. The observations in Figure 5 demonstrates that

12

rapid cell failure can eventually happen in any cell with the same chemistry, although the cell lifetime

until this failure is highly dependent on experimental conditions including cell design. Therefore, it can

change significantly between different studies, which can explain some of the controversies in literature.

Our focus here remains in the region before the onset of such a rapid failure, and the contribution of

different fading mechanisms are therefore investigated using back-to-back cells and 3-electrode cells in

the following sections.

Figure 5. Galvanostatic cycling results of LNMO-LTO full cells at 55 °C. Red symbols refer to ordinary

cells used in this study, but paused in charged state with varying durations (as showed on top of the

graph). Black symbols refer to another cell prepared similarly but without leaving extra space in the cell

for evolved gases (without OCV in charged state). Large symbols are capacity while small symbols are

Coulombic efficiency.

Galvanostatic Cycling in Back-to-back Cells. This type of full cell has been constructed by connecting

one LNMO half cell to another LTO half cell by their negative poles (lithium ends) as previously

proposed in reference.20 In such a pseudo-full cell, no chemical interactions are possible between the

electrodes and it is chosen to see the mechanisms leading to cyclable lithium loss as well as the rate of

fading when any cross-talk between electrodes can be excluded. As a modification of the test conducted

by the reference paper,20 the lithium terminals of the half cells were here used as a reference electrode.

Cycling of these cells was conducted using an intermittent current interruption (ICI) technique to follow

individual cell resistances in both half cells throughout the cycling. The cycling behaviour at room

temperature, 40 and 55 °C are shown in Figure 6. The inset in the figure shows the expected evolution

of capacity according to our previously described simulations for two different coulombic efficiencies

of the LNMO, and assuming that LTO is 100% efficient.

13

Figure 6. The capacity fading in back-to-back cells cycled at different temperatures.

It is seen in Figure 6 that the capacity fading rate is quite fast compared to ordinary full cells. In the

inset, simulation of capacity fading is also shown. In this simulation, LTO efficiency is assumed to be 1

while LNMO efficiency is either assumed to be 0.997 or 0.98 (values obtained experimentally in Figure

4). As the inset shows, the experimental fading rates in back-to-back cells are comparable to the

simulation. The second important observation from these cells is the initial stable period for room

temperature cycling. A similar short period is also seen at 40 °C, while it is entirely missing at 55 °C.

As in the scenario depicted in Figure 3(a-b), the extra capacity of LTO (10-15% excess) is able to

compensate the electrode imbalance caused by LNMO self-discharge during cycling – but only for a

limited time. It is seen that this time is only limited to the formation cycles at 55 °C. It should be noted

that a constant coulombic efficiency is assumed in the simulations, while the real cells showed lower

coulombic efficiency during the initial cycles (Figure 4a).

The change in individual electrode voltages and internal resistances are shown in Figure 7. The change

of median resistance with cycling is also shown in Figure S1. It is clearly seen from the LNMO voltage

profiles that the active cycling window of LNMO shifts from the high voltage end, meaning that the

cycling is becoming more and more limited to the higher lithiation state (i.e. lower voltages). Individual

LTO voltages show that the voltage drop at the beginning of charging disappears, while a charge end-

point appears later, showing that the active cycling window of LTO also shifts to higher lithiation states.

This constitutes the “limited delithiation” mechanism discussed previously. The shift of LNMO voltages

explains the slightly better cycling of the 55 °C sample as compared to the simulation (Figure 6), since

the cycling of LNMO is limited to the low voltage parts which reduces the degree of side reactions.

After cycling and with the back-to-back cell in a fully discharged state, the constituent half cells were

separated and cycled further under the same conditions. The results are given in the Supporting

Information, Figure S3. Both electrodes showed reversible capacities close to their initial capacities;

furthermore the LTO electrode was found to be partially charged, with the LNMO electrode in its fully

discharged (lithiated) state. This further confirms the limited delithiation mechanism.

The measured resistances for each electrode in this cell construction include considerable contributions

from the lithium electrodes, so caution must be exercised in their interpretation. The contribution of the

lithium electrode is primarily manifested as relatively high resistance in the initial cycles and an increase

in resistance during the lithium stripping process; the latter is most clearly seen in the LTO compartment

in Figure 7. Both electrode resistances show comparable and minor increases during cycling at room

temperature. Based on this limited increase, as well as the observations of individual voltages showing

no considerable increase in overpotential, it can be assumed that resistance increase does not have any

major effect on the capacity fading observed. At elevated temperature, lower resistance is observed for

14

both electrodes in the formation cycle due to improved kinetics at higher temperature. However, a

gradual increase of resistance for LNMO is observed while the resistance at the LTO electrode remains

almost constant. It can therefore be concluded that the resistance increase mainly comes from the LNMO

part and it is more pronounced at high temperature, but this resistance increase seems not to be large

enough to dominate the cycling behaviour in the back-to-back cells.

Figure 7. Individual voltage curves together with resistance measurements throughout the cycles of

LNMO-LTO back-to-back cells at room temperature (RT), 40 and 55 °C for (a) the first cycle, (b) the

10th cycle, (c) the 30th cycle and (d) the 50th cycle.

Galvanostatic Cycling in Three-electrode Cells. In these cells, circular lithium rings were placed as

reference electrodes concentric to the circular LNMO and LTO electrodes. As seen in Figure 8, when

electrode interactions are allowed, the capacity retention is much more stable compared to the back-to-

back cells, especially at room temperature. As demonstrated in the back-to-back cells, there is a

significantly higher degree of side reactions on LNMO compared to LTO. In a regular cell, these side

reactions on LNMO should still be present. However, if fading is not observed, this implies that

migration of reaction products to LTO causes additional side reactions which are not observed when

LTO is only in contact with the standard electrolyte. If the degree of additional side reactions on LTO

is comparable to LNMO, this would give an overall result similar to the third scenario described in

Figure 3e-f. Additionally, when the relative increase of fading rate with increasing temperature is

15

compared between the cell formats, it is much higher in regular full cells. This shows that the

temperature dependence of the individual coulombic efficiencies are different for LNMO and LTO

electrodes, the cycling stability at 55 °C would otherwise be comparable to RT cycling. This can explain

why the fading behaviour does not follow an Arrhenius type trend in regular full cells.

Figure 8. Capacity fading in three-electrode LNMO-LTO full cells at room temperature, at 40 and 55

°C.

More insight into the fading mechanisms can be gained from individual electrode potentials and internal

resistance measurements (see Figure 9). As seen from the voltage curves, the cycling window is

progressively shifting to the lower LNMO lithiation states as the cycling proceeds. This represents the

“limited lithiation” mechanism discussed previously, and in contrast to the mechanism observed in the

back-to-back cells. This implies that the LTO electrode in this case is more inefficient (i.e., has a lower

coulombic efficiency) than the LNMO electrode. At room temperature, the capacity loss is relatively

small, and the shift of LNMO voltage is not easily seen; however, it is clearly seen that the voltage drop

at the beginning of charging is persistent for LTO. This shows that LTO is always returning to its initial

delithiated state at the end of discharge and that the coulombic efficiency should be dominated by the

higher degree of inefficiency on the LTO side (similar to scenario in Figure 3d). The shift of the LNMO

cycling window is very clear at 55 °C. Similar to the back-to-back cells, the three-electrode cell was

disassembled in the discharged state with the LNMO and LTO electrodes reassembled into new half

cells (Supporting Information, Figure S3). In this case, the LTO electrode was found to be in its fully

discharged state, while the LNMO was partially charged, confirming again the limited lithiation

mechanism. The LTO electrode showed a reversible capacity comparable to its initial capacity; the

LNMO however showed a reduced capacity due to a high impedance, but which is attributable to the

disassembly process.

At all three temperatures, the loss of cyclable lithium occurs through the “limited lithiation” mechanism.

If the effect of temperature is considered, it is seen that the efficiency of LTO should be more dependent

on temperature in comparison to LNMO. However, the opposite behaviour was observed in back-to-

back cells (Figure 7). When no cross-talk is allowed between the electrodes, increased temperature

caused more severe reactions on LNMO compared to LTO. These results show that the reactions on

LTO which are initiated by cross-talk products are indeed more sensitive to temperature.

It is observed from resistance measurements that the resistance of LTO does not change significantly on

cycling, although there is a more pronounced increase at 55 °C. For LNMO, there is a gradual increase

16

in resistance over cycling which is again more significant at 55 °C. In all cases, the increased resistance

does not become high enough to significantly reduce capacity, as evidenced by the clear end-points in

the voltage profiles. Compared with the back-to-back cells, the initial resistances are smaller but later

reach higher values for the 55 °C sample (see Figure S1 and S2). The initially lower values are due to

the absence of contributions from the lithium electrodes in the back-to-back cells. However, since

capacity fading is slower in regular full cells than in back-to-back cells, and cycling shifts to higher

voltage windows, the cells are cycled at more harsh conditions overall, and combined with the effect of

electrode interactions it is hence possible to observe higher resistances. Even though the resistance

contribution to the capacity is minor, it is likely that it would start to play a major role in later stages

near the cell break-down point.

Figure 9. Individual voltage curves together with resistance measurements throughout the cycles of

LNMO-LTO 3-electrode cells at room temperature (RT), 40 and 55 °C for (a) the first cycle, (b) the

10th cycle, (c) the 30th cycle and (d) the 50th cycle.

As mentioned in the introduction, gas formation associated with electrolyte oxidation and interaction of

gas species between two electrodes has been shown to exist in other full cell chemistries.23,24,25,26 For

17

instance, CO2 evolved after electrolyte oxidation can migrate and subsequently be reduced on the

negative electrode.27 With an aim to gain further insight into the role of gas interaction in the cross-talk

mechanisms for this system, a modified “back-to-back” cell using a bipolar electrode and contained

within a single pouch has been tested (see experimental part for details). In this experiment, gas

interactions are possible between the electrodes but interaction of species in the electrolyte is prevented.

This type of cell showed faster fading as compared to the regular back-to-back cell. The cycling results

of this cell is shown in Figure 10. As the voltage curves show, the cyclable lithium loss originates from

the “limited delithiation” mechanism as in the regular back-to-back cells. This indicates that the cross-

talk due to interaction of gas species is unlikely to significantly contribute to “limited lithiation”

mechanism observed in regular full cells.

Therefore, the capacity fade due to the “limited lithiation” mechanism is most likely primarily due to

the interaction of dissolved species in the electrolyte with LTO, rather than gaseous species (e.g. CO2;

nevertheless, the possibility of cross-talk due to the much slower process of gas evolution and

subsequent re-dissolution cannot be completely ruled out). Such species can be soluble electrolyte

decomposition products, for example, the oxidation of EC can generate radical cations on positive

electrode.33 These reactive radical cations can participate in further side reactions in different parts of

the cell, or be reduced on the negative electrode. Similarly, dissolved transition metal cations (Mn, Ni)

can travel to the negative electrode and be reduced there.19 The reduction of such species on LTO could

contribute to the observed fading mechanism. As stated in the introduction, fast fading of full cells with

graphite electrode is ascribed to SEI damage caused by the incorporation of transition metal ions.12

Interaction of Mn ions can induce pores, cracks and increase the electronic conductivity of the graphite

SEI.34 Mn in the SEI can coordinate with EC inside the SEI and facilitate its reduction35 or decrease the

reduction barriers via an electrocatalytic mechanism.36 Therefore, migrated species could cause

additional reduction of electrolyte rather than being reduced itself. One of the reasons for studying this

full cell system was to eliminate this second possibility as LTO operates at a significantly higher

potential than graphite and does not require a comparable SEI for its operation. However, it should still

be noted that the nature of the interfacial reactions on LTO are still not well understood and the second

possibility should not be neglected (i.e., possibility of catalytic effects). In any of these cases, at higher

temperatures, either new reaction routes which consume more electrons at LTO become favorable, or

the kinetics related to existing cross-talk reactions are accelerated. In order to get further insight into

these mechanisms, we performed ex situ characterization of both electrodes obtained from regular full

cells (two-electrode cells).

18

Figure 10. Cycling performance of ‘single pouch’ back-to-back cells allowing only gas interactions

(a) and the selected voltage curves from the same cell (b). Large symbols are capacity and small

symbols are Coulombic efficiency.

Surface characterization of LNMO electrodes after 50 cycles.

The morphology of the LNMO electrodes after cycling (50 cycles, C/5 rate) at room temperature and at

55 °C have been observed via SEM (Figure 11). Both conductive carbon and active LNMO particles

showed no significant observable differences after cycling at either temperature. This indicates that there

is no considerable damage to nor thick surface film formation on the LNMO particles. The same

electrodes were also characterized via XPS to investigate the surface film thickness and composition.

19

Figure 11. SEM images of pristine LNMO electrodes (a), after cycling at room temperature (b) and at

55 °C (c).

C 1s, O 1s and F 1s spectra of pristine and cycled LNMO electrodes obtained with the in-house XPS

instrument (Al-Kα radiation 1486.6 eV) are shown in Figure 12a-c. The main differences are observed

in the C 1s and O 1s spectra. For the pristine electrode, a large contribution appears from carbon black

which is assigned to the peak at 284.5 eV (all spectra were calibrated with respect to hydrocarbon peak;

0.5 eV difference was fixed between the hydrocarbon and carbon black peaks). The green peaks in

Figure 12 are assigned to the binder which is a co-polymer of PVdF and HFP (Kynar Flex 2801 binder,

88 mol% PVdF and 12 mol% HFP). In C 1s spectrum, C=O/O-C-O and C-O related adsorbed species

could be observed before cycling. In the O 1s spectrum, a metal oxide peak (529.8 eV) is observed

together with C=O and C-O related adsorbed species. The F 1s spectrum shows one large peak around

688 eV corresponding to the PVdF-HFP binder.

After cycling, the binder-related peaks shift to lower energies, as previously reported.13,37 The main

differences in both the C 1s and O 1s spectra are due to evolution of surface species which are related

to organic surface film components as a result of electrolyte oxidation. For both temperatures, no

significant differences can be observed when comparing the relative intensities of the surface species,

except for a slight increase in the peaks for C=O related species for the 55 °C sample. Comparing the

relative intensity of the peaks corresponding to the surface species to the peaks for the bulk in the C 1s

spectrum reveals an increase in the relative amount of surface species with respect to the binder and C-

C peaks. No similar observation can be made in the O 1s spectrum in, for example, the relative ratio of

surface species to the metal oxide (i.e. substrate) peak. This indicates that a comparatively larger

deposition of surface species occurs on carbon black and binder-rich regions than on the active material

when the temperature is 55 °C. In addition, slightly more LiF or M-F (e.g. MnF2, NiF2) can be seen at

55 °C. If general trends in atomic concentrations are compared (see Figure 12d), a significant decrease

in F content is observed in accordance with intensity loss from the binder-related peaks. The decrease

in concentration of C is comparatively smaller. Here, surface films increase C concentration while at the

same time these films decrease the contribution from carbon black. The latter effect seems to dominate

the overall trend in C concentration. For O 1s, the contribution from surface films dominates over the

intensity loss from the bulk metal oxide supporting a slightly more preferential film formation on carbon

black-binder regions. The spectra for P 2p, Mn 2p and Ni 2p are shown in Figure S4.

20

It is clear, and in agreement with the SEM observations, that there is no significantly thick surface film

formed on LNMO. A thin layer is indeed observed, but its thickness does not seem to increase by cycling

at elevated temperature as seen in the O 1s spectra. Since it is obvious that this layer does not stop

electrolyte decomposition and also does not change in thickness, side-reaction products are not

continuously deposited on active particle surface and instead released into the electrolyte or in a broad

sense into the cell (e.g. gases). In the internal resistance measurements (Figure 9), it was seen that the

major contributions to the overall resistance originated from the LNMO electrode. When the cells were

opened for sample preparation, we observed deposited species on the separator facing the LNMO

electrode which were partly adhered to the electrode outer surface. It is thus possible that highly reactive

species formed after electrolyte oxidation react further with the polypropylene separator, creating

deposits which partially block the pores of the separator.

Figure 12. C 1s, O 1s and F 1s XPS spectra of pristine and cycled LNMO electrodes.

Surface characterization of LTO electrodes after 50 cycles.

As shown in Figure 13, and in contrast to the LNMO electrodes, the morphology of the LTO electrodes

changes clearly on cycling. Surface films seem to be formed both on the LTO active particles and on

the conductive carbon additives. Measurement of 100 carbon black particles gives an average particle

size of 54 nm (±11 nm) in the pristine electrode (Figure 13a), which increases to 70 nm (±15 nm) after

room temperature cycling and 76 nm (±13 nm) after 55 °C cycling. According to the SEM images of

the LTO electrodes cycled at 55 °C, apparent surface layer formation can be observed since previous

sharp edges and voids appear to be filled with previously unobserved deposits. According to the

resistance measurements (Figure 9), relatively thicker layers could be expected for LNMO electrodes,

but instead only an extremely thin layer exists on the surface according to SEM and XPS analysis. On

the other hand, the LTO electrode did not contribute much to the overall internal resistance, but thicker

SEI layer formation is indeed observed by SEM. We performed XPS analysis using both in-house XPS

and high energy synchrotron XPS in order to gain further insight into the surface products deposited on

the LTO electrodes.

21

Figure 13. SEM images of pristine LTO electrodes (a), after cycling at room temperature (b) and at 55

°C (c).

The survey spectra (see Figure 14) of the pristine sample display characteristic peaks of carbon (~285

eV), oxygen (~531 eV), and titanium (~459 eV). This is expected for an electrode comprising water-

soluble binders and LTO. The most notable differences between the pristine and the cycled samples are

the disappearance of the titanium peak, the emergence of the fluorine peak (~680 eV) and an increasing

O/C ratio for the cycled samples. This is in accordance with previous results.38,31 The most notable

difference is the increased F concentration in the 55 °C compared to the RT sample.

Figure 14. Survey spectra of pristine and cycled LTO samples measured using different photon

energies.

22

The C 1s spectra of the different samples are displayed in Figure 15 (note that the cycled samples were

measured using three different photon energies). In the pristine electrode, the conductive carbon is

identified as the peak at 284 eV in the pristine sample. General C-H bonds are found at 285 eV. Weak

contributions to intensity are observed in the region where carboxylates (289.3 eV), O-C-O / C=O

related species (288 eV) and ethers (286.7 eV) might be present.39 Since the exact ratio of binders used,

occurrence of particle surface treatment or presence of additives are unknown, a detailed assignment

cannot be made.

The general trend in the C 1s spectra is that from the pristine to the cycled samples there is a significant

increase in the C-O region, correlated to a decrease in C-C peak (which can be considered as a bulk

peak). A peak at around 289 eV is also observed, corresponding to carboxylate groups. When comparing

the different photon energies, for the room temperature (RT) sample large contributions are still visible

at 284 eV, but is significantly decreasing with lower excitation energies. This indicates that the SEI is

thick enough to barely allow the in-house XPS to probe through it. For the 55 °C sample, it is observed

that no significant signal at 284 eV is observed at any photon energy, indicating an even thicker SEI at

elevated temperatures. The relative intensity of organic species decreases as the photon energy increases

(i.e. increasing probing depth) for both temperatures. A similar trend is also seen when the RT sample

is compared with the 55 °C sample indicating that the main difference between them is due to film

thickness rather than the film composition.

Figure 15. C 1s and O 1s spectra of the pristine, room temperature, and 55 °C LTO sample at different

energies.

In the O 1s spectra, reference lines for C-O and C=O related species with binding energies of 533.5 eV

and 532.2 eV were placed on the graph, respectively, while titanium oxide could be observed at ~530.5

eV.38,40 The different oxygen peaks in the organic compounds usually have a large overlap and are hard

to separate, however, a lower binding energy can be expected for C-O contribution from ethers (532.7

eV) compared to C-O in carboxylates (*O-C=O, 533.6 eV).39 The most intense peak for the RT sample

(at 532.7 eV) is in accordance with ethers, which is also observed in the C 1s spectrum. At 55 °C, the

main peak becomes broader, indicating compositional changes in addition to the increase of film

23

thickness. This can be due to contributions from salt decomposition products (e.g. LixPFyOz) at 534.6

eV as well as C-O and C=O contributions from carboxylates while non-homogeneous deposition of

species (e.g. on active material vs. carbon black) might also play a role in this broadening. One

interesting feature in the O 1s spectra is that the oxide peak, being clearly visible at 6015 eV excitation

energy, hardly has any contribution at lower energies. The probing depth of the O 1s spectra can be

estimated to be around 9, 14 and 40 nm for excitation energies of 1486.6, 2005 and 6015 eV,

respectively.41 By comparison of the integrals of the metal oxide peak to the organic oxygen peaks, the

SEI thickness can be assumed to be between 9-14 nm for the RT sample since the metal oxide peak

becomes slightly visible when measured by photon energy of 2005 eV (blue curve). However, the SEI

thickness is seen to be above 14 nm for the 55 °C sample (note that there is a number of assumptions

made in estimations of these numbers and see SI for further discussion).

As can be seen in Figure 16, the pristine sample contains neither phosphorous nor fluorine. Both

phosphorous and fluorine are present in the hexafluorophosphate salt. The double peak feature in the P

2p spectra indicates that there are both P-O bonds and P-F bonds present. This suggests that salt

decomposition products are incorporated into the SEI while some undecomposed salt may also be

present. The F 1s spectra also shows signs of salt decomposition products, since the peak width indicates

that several different components are present. One important difference is observed only for the RT

sample in the region where contributions can come from LiF (or MnF2, NiF2). These species seem to be

deeper in the SEI as they are only seen at high excitation energies. Absence of related peaks at 55 °C

can be due to increased solubility of LiF or M-F species at this temperature, which could lead to partial

dissolution of some SEI components. The spectra for Mn 2p, Ni 2p and Ti 2p are shown in Figure S5

and S6.

Figure 16. P 2p and F 1s spectra of the pristine, room temperature (RT), and 55 °C LTO sample

measured using different photon energies.

In summary, it can be concluded that the cycled samples of LTO show signs of a relatively thick SEI

comprising ethers, carboxylates, phosphates and F-containing species. Ni and Mn are obviously present

24

in the SEI on the LTO anode (see SI). It is interesting to note that even though a considerably thick

surface layer is formed on the LTO surface, the impedance rise from LTO is still insignificant (Figure

9). Electrochemical testing of these cells showed that there should be steady side reactions occurring on

the LTO side - meaning electrolyte oxidation products are able to reach the LTO surface through these

layers. It is therefore not surprising that these layers do not cause considerable ionic resistance. These

results also show the importance of individual internal cell resistance measurements combined with the

XPS surface characterization, since it is common to correlate thick surface layers to cell resistance even

when this is not supported by any electrochemical testing of cells.

XANES. The deposition of dissolved cathode transition metal ions on the anode surface is a well-known

example of ‘cross-talk’ in full cells (e.g. for systems with LMO, LNMO or NMC positive electrodes).

The reduction of these migrating ions to their metallic state has been reported,11 however, more recent

papers claim non-metallic oxidation states for Mn and Ni.22,42 In our previous study, Mn and Ni peaks

were observed after only one cycle43 on LTO, and this has also been seen in this current work for both

room temperature and high temperature cells. However, it was not possible to determine the oxidation

state of the metals via XPS. Therefore, in order to determine these oxidation states, XANES

measurements have been performed on cycled LTO electrodes as well as reference samples for different

oxidation states of Mn and Ni. The results are shown in Figure 15 and indicate the Mn oxidation state

to be close to +III and the Ni state to be close to +II. This shows migrated metal ions are not reduced to

their metallic state, in agreement with the recent literature.44 Although metal ion deposition on the anode

is not a direct cause for capacity loss via reduction to metallic state, it is still associated with the cross-

talk and possibly with side reactions on LTO. It should be noted that temperature does not appear to

have any effect on the oxidation state of the metal ions in the surface layer.

Figure 17. XANES measurements of LTO samples cycled at room temperature and 55 °C.

Conclusions

In this study, a comprehensive analysis of capacity fading occurring in LNMO-LTO cells has been

performed using different cell configurations and galvanostatic cycling combined with an intermittent

25

current interruption technique. One of the primary motivations in this study was also to understand the

mechanisms causing the observed fast fading at elevated temperature.

It is observed that severe parasitic reactions are occurring on the LNMO surface in both half cells and

full cells. These reactions, in isolation, would be expected to cause fading of cells via a loss of cyclable

lithium originating from an increasingly “limited delithiation” of both electrodes in the full cell.

Electrode interactions (cross-talk) are in fact observed to result in a greater degree of parasitic reactions

(i.e. a lower coulombic efficiency) at the LTO electrode relative to the LNMO electrode in the full cell.

This lower efficiency of the negative electrode results instead in an increasingly “limited lithiation” of

the electrode materials. Accelerated failure at elevated temperature results from significantly lower

coulombic efficiency of LTO. Evolution of gases can also cause rapid cell failure, but can be mitigated

with optimization of cell design.

Stable capacity retention at room temperature may be misleading as an indication of cycle life in

practice, as a significant degree of side reactions may occur even at room temperature, resulting in

progressive decomposition of the electrolyte. Coulombic efficiency must therefore be taken into

account, as it may be significantly below 100% even if there is no apparent fade in capacity. A typical

lab-scale test cell such as those investigated in this work may cycle with good reversibility for more than

three months, or 1000 cycles (see Figures S2 and S7) before exhibiting sudden capacity fade. This is

partly because the electrolyte is in significant excess compared to realistic cells, and this variable should

be considered when investigating long-term or high temperature cycling behaviour.

According to the capacity fading mechanisms shown in this study, some strategies to increase the

efficiency of the negative electrode (e.g., separator decorated with lithiated LTO or a larger particle size

of LTO) may create a situation in which the efficiencies of both electrodes are equivalent and result in

deceptively ‘stable’ cycling at certain temperatures. Based on our results, it might be expected that

system with equivalent electrode efficiencies at high temperature may show a lower efficiency for

LNMO at room temperature, which would be expected to result in the “limited delithiation” mechanism.

Therefore, it is important to check the cycling stability at different temperatures and preferably over

long durations to verify that such strategies do indeed improve the cell stability. However, based on the

present results, we suggest that future strategies preferably target improving the efficiency of the LNMO

electrode, as this is ultimately the major source of inefficiency at LTO.

Surface film formation after electrolyte oxidation and transition metal dissolution from the cathode are

widely considered to cause failure in LNMO-based cells. It has been demonstrated here that the observed

capacity fading is mainly due to perturbation of capacity balance as a result of a progressively “limited

lithiation” of the electrodes. Resistance measurements and voltage curves obtained from three-electrode

measurements confirm limited contributions to capacity loss from internal resistance (or overpotential

increase) and active mass loss due to metal dissolution. The resistance increase in the system is observed

to mainly originate from the LNMO side being more pronounced at 55 °C. The surface film thickness

on LNMO did not change when temperature was raised to 55 °C. This indicates that the contribution to

resistance from LNMO surface films were minor. A reaction of oxidation products with the separator

might cause this, or pore clogging due to gases evolved. Another possibility is the increase of contact

resistance between LNMO and the current collector.45 Even though considerable oxygen evolution from

the structure is not expected for LNMO, evolution of surface reconstruction layers (rock-salt phase)

might also contribute to such a resistance, as observed in layered oxide electrodes.46,47 In contrast to the

LNMO, surface films formed on the LTO electrode are found to be relatively thick but show

comparatively little resistance. Those films also consisted of Mn and Ni migrated from LNMO and they

are shown via XANES to be in a non-metallic state.

Acknowledgement

We acknowledge Helmholtz Zentrum Berlin for synchrotron radiation beam time for HAXPES and

XANES measurements. Dr. Maria Hahlin and Dr. Roberto Felix Duarte are acknowledged for their help

26

during HAXPES and XANES measurements, Dr. William Brant for his help for data processing in

Athena software. Leclanché are also gratefully acknowledged for providing the LTO electrodes.

References

(1) Julien, C. M.; Mauger, A. Review of 5-V Electrodes for Li-Ion Batteries: Status and Trends.

Ionics (Kiel). 2013, 19 (7), 951–988.

(2) Kraytsberg, A.; Ein-Eli, Y. Higher, Stronger, Better… A Review of 5 Volt Cathode Materials

for Advanced Lithium-Ion Batteries. Adv. Energy Mater. 2012, 2 (8), 922–939.

(3) Kim, J. H.; Pieczonka, N. P. W.; Yang, L. Challenges and Approaches for High-Voltage Spinel

Lithium-Ion Batteries. ChemPhysChem 2014, 15, 1940–1954.

(4) Li, W.; Song, B.; Manthiram, A. High-Voltage Positive Electrode Materials for Lithium-Ion

Batteries. Chem. Soc. Rev. 2017, 46, 3006–3059.

(5) Vetter, J.; Novák, P.; Wagner, M. R.; Veit, C.; Möller, K. C.; Besenhard, J. O.; Winter, M.;

Wohlfahrt-Mehrens, M.; Vogler, C.; Hammouche, A. Ageing Mechanisms in Lithium-Ion

Batteries. J. Power Sources 2005, 147 (1–2), 269–281.

(6) Brutti, S.; Greco, G.; Reale, P.; Panero, S. Insights about the Irreversible Capacity of

LiNi0.5Mn1.5O4 Cathode Materials in Lithium Batteries. Electrochim. Acta 2013, 106, 483–

493.

(7) Fransson, L.; Eriksson, T.; Edström, K.; Gustafsson, T.; Thomas, J. . Influence of Carbon

Black and Binder on Li-Ion Batteries. J. Power Sources 2001, 101 (1), 1–9.

(8) Younesi, R.; Christiansen, A. S.; Scipioni, R.; Ngo, D.-T.; Simonsen, S. B.; Edstrom, K.;

Hjelm, J.; Norby, P. Analysis of the Interphase on Carbon Black Formed in High Voltage

Batteries. J. Electrochem. Soc. 2015, 162 (7), A1289–A1296.

(9) Metzger, M.; Marino, C.; Sicklinger, J.; Haering, D.; Gasteiger, H. A. Anodic Oxidation of

Conductive Carbon and Ethylene Carbonate in High-Voltage Li-Ion Batteries Quantified by

On-Line Electrochemical Mass Spectrometry. J. Electrochem. Soc. 2015, 162 (7), A1123–

A1134.

(10) Streipert, B.; Röser, S.; Kasnatscheew, J.; Janßen, P.; Cao, X.; Wagner, R.; Cekic-Laskovic, I.;

Winter, M. Influence of LiPF 6 on the Aluminum Current Collector Dissolution in High

Voltage Lithium Ion Batteries after Long-Term Charge/Discharge Experiments. J.

Electrochem. Soc. 2017, 164 (7), A1474–A1479.

(11) Pieczonka, N. P. W.; Liu, Z.; Lu, P.; Olson, K. L.; Moote, J.; Powell, B. R.; Kim, J.

Understanding Transition-Metal Dissolution Behavior in LiNi0.5Mn1.5O4 High-Voltage

Spinel for Lithium Ion Batteries. J. Phys. Chem. C 2013, 117, 15947−15957 2013.

(12) Kim, J. H.; Pieczonka, N. P. W.; Li, Z.; Wu, Y.; Harris, S.; Powell, B. R. Understanding the

Capacity Fading Mechanism in LiNi0.5Mn 1.5O4/graphite Li-Ion Batteries. Electrochim. Acta

2013, 90, 556–562.

(13) Lu, D.; Xu, M.; Zhou, L.; Garsuch, A.; Lucht, B. L. Failure Mechanism of

Graphite/LiNi0.5Mn1.5O4 Cells at High Voltage and Elevated Temperature. J. Electrochem.

Soc. 2013, 160 (5), A3138–A3143.

(14) Aurbach, D.; Markovsky, B.; Talyossef, Y.; Salitra, G.; Kim, H.-J.; Choi, S. Studies of Cycling

Behavior, Ageing, and Interfacial Reactions of LiNi0.5Mn1.5O4 and Carbon Electrodes for

Lithium-Ion 5-V Cells. J. Power Sources 2006, 162 (2), 780–789.

27

(15) Dedryvere, R.; Foix, D.; Franger, S.; Patoux, S.; Daniel, L.; Gonbeau, D. Electrode/Electrolyte

Interface Reactivity in High-Voltage Spinel LiMn 1.6 Ni 0.4 O 4 /Li 4 Ti 5 O 12 Lithium-Ion

Battery. J. Phys. Chem. C 2010, 114 (24), 10999–11008.

(16) Michalak, B.; Berkes, B. B.; Sommer, H.; Bergfeldt, T.; Brezesinski, T.; Janek, J. Gas

Evolution in LiNi0.5Mn1.5O4/Graphite Cells Studied In Operando by a Combination of

Differential Electrochemical Mass Spectrometry, Neutron Imaging, and Pressure

Measurements. Anal. Chem. 2016, 88 (5), 2877–2883.

(17) Goers, D.; Holzapfel, M.; Scheifele, W.; Lehmann, E.; Vontobel, P.; Novák, P. In Situ Neutron

Radiography of Lithium-Ion Batteries: The Gas Evolution on Graphite Electrodes during the

Charging. J. Power Sources 2004, 130 (1–2), 221–226.

(18) Kostecki, R.; Norin, L.; Song, X.; McLarnon, F. Diagnostic Studies of Polyolefin Separators in

High-Power Li-Ion Cells. J. Electrochem. Soc. 2004, 151 (4), A522.

(19) Arora, P. Capacity Fade Mechanisms and Side Reactions in Lithium-Ion Batteries. J.

Electrochem. Soc. 1998, 145 (10), 3647.

(20) Li, S. R.; Chen, C. H.; Xia, X.; Dahn, J. R. The Impact of Electrolyte Oxidation Products in

LiNi0.5Mn1.5O4/Li4Ti5O12 Cells. J. Electrochem. Soc. 2013, 160 (9), A1524–A1528.

(21) Bhandari, A.; Bhattacharya, J. Review — Manganese Dissolution from Spinel Cathode : Few

Unanswered Questions. J. Electrochem. Soc. 2017, 164 (2), A106–A127.

(22) Wandt, J.; Freiberg, A.; Thomas, R.; Gorlin, Y.; Siebel, A.; Jung, R.; Gasteiger, H. A.; Tromp,

M. Transition Metal Dissolution and Deposition in Li-Ion Batteries Investigated by Operando

X-Ray Absorption Spectroscopy. J. Mater. Chem. A 2016, 4 (47), 18300–18305.

(23) Michalak, B.; Sommer, H.; Mannes, D.; Kaestner, A.; Brezesinski, T.; Janek, J. Gas Evolution

in Operating Lithium-Ion Batteries Studied In Situ by Neutron Imaging. Sci. Rep. 2015, 5,

15627.

(24) Metzger, M.; Strehle, B.; Solchenbach, S.; Gasteiger, H. A. Origin of H 2 Evolution in LIBs: H

2 O Reduction vs. Electrolyte Oxidation. J. Electrochem. Soc. 2016, 163 (5), A798–A809.

(25) Xiong, D. J.; Ellis, L. D.; Petibon, R.; Hynes, T.; Liu, Q. Q.; Dahn, J. R. Studies of Gas

Generation, Gas Consumption and Impedance Growth in Li-Ion Cells with Carbonate or

Fluorinated Electrolytes Using the Pouch Bag Method. J. Electrochem. Soc. 2017, 164 (2),

A340–A347.

(26) Xiong, D. J.; Petibon, R.; Nie, M.; Ma, L.; Xia, J.; Dahn, J. R. Interactions between Positive

and Negative Electrodes in Li-Ion Cells Operated at High Temperature and High Voltage. J.

Electrochem. Soc. 2016, 163 (3), A546–A551.

(27) Sloop, S. E.; Kerr, J. B.; Kinoshita, K. The Role of Li-Ion Battery Electrolyte Reactivity in

Performance Decline and Self-Discharge. J. Power Sources 2003, 119–121, 330–337.

(28) Wu, H. M.; Belharouak, I.; Deng, H.; Abouimrane, A.; Sun, Y.-K.; Amine, K. Development of

LiNi0.5Mn1.5O4/Li4Ti5O12 System with Long Cycle Life. J. Electrochem. Soc. 2009, 156,

A1047–A1050.

(29) Lacey, M. J. Influence of the Electrolyte on the Internal Resistance of Lithium−Sulfur Batteries

Studied with an Intermittent Current Interruption Method. ChemElectroChem 2017, 4 (8),

1997–2004.

(30) Lacey, M. J.; Edström, K.; Brandell, D. Visualising the Problems with Balancing Lithium–

sulfur Batteries by “mapping” Internal Resistance. Chem. Commun. 2015, 51 (92), 16502–

16505.

28

(31) Nordh, T.; Jeschull, F.; Younesi, R.; Koçak, T.; Tengstedt, C.; Edström, K.; Brandell, D.

Different Shades of Li 4 Ti 5 O 12 Composites: The Impact of the Binder on Interface Layer

Formation. ChemElectroChem 2017, 4 (10), 2683–2692.

(32) Li, S. R.; Sinha, N. N.; Chen, C. H.; Xu, K.; Dahn, J. R. A Consideration of Electrolyte

Additives for LiNi0.5Mn1.5O4/Li4Ti5O12 Li-Ion Cells. J. Electrochem. Soc. 2013, 160 (11),

A2014–A2020.

(33) Xing, L.; Li, W.; Wang, C.; Gu, F.; Xu, M.; Tan, C.; Yi, J. Theoretical Investigations on

Oxidative Stability of Solvents and Oxidative Decomposition Mechanism of Ethylene

Carbonate for Lithium Ion Battery Use. J. Phys. Chem. B 2009, 113 (52), 16596–16602.

(34) Delacourt, C.; Kwong, A.; Liu, X.; Qiao, R.; Yang, W. L.; Lu, P.; Harris, S. J.; Srinivasan, V.

Effect of Manganese Contamination on the Solid-Electrolyte-Interphase Properties in Li-Ion

Batteries. J. Electrochem. Soc. 2013, 160 (8), A1099–A1107.

(35) Vissers, D. R.; Chen, Z.; Shao, Y.; Engelhard, M.; Das, U.; Redfern, P.; Curtiss, L. A.; Pan, B.;

Liu, J.; Amine, K. Role of Manganese Deposition on Graphite in the Capacity Fading of

Lithium Ion Batteries. ACS Appl. Mater. Interfaces 2016, 8 (22), 14244–14251.

(36) Gilbert, J. A.; Shkrob, I. A.; Abraham, D. P. Transition Metal Dissolution, Ion Migration,

Electrocatalytic Reduction and Capacity Loss in Lithium-Ion Full Cells. J. Electrochem. Soc.

2017, 164 (2), A389–A399.

(37) Lindgren, F.; Rehnlund, D.; Källquist, I.; Nyholm, L.; Edström, K.; Hahlin, M.; Maibach, J.

Breaking Down a Complex System: Interpreting PES Peak Positions for Cycled Li-Ion Battery

Electrodes. J. Phys. Chem. C 2017, acs.jpcc.7b08923.

(38) Nordh, T.; Younesi, R.; Brandell, D.; Edström, K. Depth Profiling the Solid Electrolyte

Interphase on Lithium Titanate (Li 4 Ti 5 O 12 ) Using Synchrotron-Based Photoelectron

Spectroscopy. J. Power Sources 2015, 294 (0), 173–179.

(39) Beamson, G.; Briggs, D. High Resolution XPS of Organic Polymers: The Scienta ESCA300

Database. J. Chem. Educ. 1993, 70 (1), A25.

(40) Younesi, R.; Hahlin, M.; Björefors, F.; Johansson, P.; Edström, K. Li–O 2 Battery Degradation

by Lithium Peroxide (Li 2 O 2 ): A Model Study. Chem. Mater. 2013, 25 (1), 77–84.

(41) Painter, L. R.; Arakawa, E. T.; Williams, M. W.; Ashley, J. C. Optical Properties of

Polyethylene: Measurement and Applications. Radiat. Res. 1980, 83 (1), 1–18.

(42) Nordh, T.; Younesi, R.; Hahlin, M.; Duarte, R. F.; Tengstedt, C.; Brandell, D.; Edström, K.

Manganese in the SEI Layer of Li4Ti5O12 Studied by Combined NEXAFS and HAXPES

Techniques. J. Phys. Chem. C 2016, 120 (6), 3206–3213.

(43) Aktekin, B.; Younesi, R.; Zipprich, W.; Tengstedt, C.; Brandell, D.; Edström, K. The Effect of

the Fluoroethylene Carbonate Additive in LiNi 0.5 Mn 1.5 O 4 - Li 4 Ti 5 O 12 Lithium-Ion Cells.

J. Electrochem. Soc. 2017, 164 (4), A942–A948.

(44) Jarry, A.; Gottis, S.; Yu, Y. S.; Roque-Rosell, J.; Kim, C.; Cabana, J.; Kerr, J.; Kostecki, R.

The Formation Mechanism of Fluorescent Metal Complexes at the LixNi0.5Mn1.5O4-

δ/Carbonate Ester Electrolyte Interface. J. Am. Chem. Soc. 2015, 137 (10), 3533–3539.

(45) Landesfeind, J.; Pritzl, D.; Gasteiger, H. A. An Analysis Protocol for Three-Electrode Li-Ion

Battery Impedance Spectra: Part I. Analysis of a High-Voltage Positive Electrode. J.

Electrochem. Soc. 2017, 164 (7), A1773–A1783.

(46) Muto, S.; Sasano, Y.; Tatsumi, K.; Sasaki, T.; Horibuchi, K.; Takeuchi, Y.; Ukyo, Y. Capacity-

Fading Mechanisms of LiNiO2-Based Lithium-Ion Batteries. J. Electrochem. Soc. 2009, 156

29

(5), A371.

(47) Streich, D.; Erk, C.; Guéguen, A.; Müller, P.; Chesneau, F. F.; Berg, E. J. Operando

Monitoring of Early Ni-Mediated Surface Reconstruction in Layered Lithiated Ni-Co-Mn

Oxides. J. Phys. Chem. C 2017, 121 (25), 13481–13486.

download fileview on ChemRxivManuscript.pdf (2.43 MiB)

Understanding the capacity loss in LNMO-LTO lithium-ion cells

at ambient and elevated temperatures

Burak Aktekin1, Matthew J. Lacey1, Tim Nordh1, Reza Younesi1, Carl Tengstedt2,

Wolfgang Zipprich3, Daniel Brandell1, Kristina Edström1

1Department of Chemistry – Ångström Laboratory, Uppsala University, Box 538, SE-75121, Uppsala, Sweden

2Scania CV AB, SE-151 87, Södertälje, Sweden 3Volkswagen AG, D-38436, Wolfsburg, Germany

Supplementary Information

An example of the code used to numerically simulate the limited lithiation and delithiation

mechanisms, written in the R programming language is given below.

# set values maxQ_p = 4 # positive electrode capacity maxQ_n = 5 # negative electrode capacity Eff_p = 0.98 # positive efficiency Eff_n = 1 # negative efficiency dQ = 0.005 # time increment # starting values Q_p = maxQ_p # charge stored in +ve Q_n = 0 # charge stored in -ve DQ = 0 # total charged passed (decreases on 'discharge') t = 0 # time out <- data.frame() # create blank table to store values cyc.n <- 1 # cycle number # loop over 10 cycles while(cyc.n < 10) { # 'charge' step while(Q_p - dQ > 0 && Q_n + dQ < maxQ_n) { t <- t + dQ DQ <- DQ + dQ # charge stored is incremented according to efficiency Q_p <- Q_p - (dQ * Eff_p) Q_n <- Q_n + (dQ * Eff_n) state = "C" # identifies charging step # collect data for each time increment summary <- data.frame(t, cyc.n, state, DQ, Q_p, Q_n) # append to 'out' table out <- rbind(out, summary) } # 'discharge step' while(Q_p + dQ < maxQ_p && Q_n - dQ > 0) { t <- t + dQ DQ <- DQ - dQ # discharge is assumed 100% efficient Q_p <- Q_p + dQ Q_n <- Q_n - dQ state = "D" summary <- data.frame(t, cyc.n, state, DQ, Q_p, Q_n) out <- rbind(out, summary) } cyc.n = cyc.n + 1 } # write output to text file write.table(proc, file = "results.txt", quote = FALSE, row.names = FALSE, sep = "\t")

Figure S1. The median resistance change with respect to LNMO (a) and LTO (b) electrodes in LNMO-LTO back-

to-back cells as obtained from intermittent current interruption (ICI) technique. Empty symbols refer to median

resistance during charging while filled symbols refer to median resistance during discharging.

Figure S2. The median resistance change with respect to LNMO (a) and LTO (b) electrodes in LNMO-LTO 3-

electrode cells as obtained from intermittent current interruption (ICI) technique. Empty symbols refer to median

resistance during charging while filled symbols refer to median resistance during discharging.

Figure S3. Potential profiles of LNMO and LTO half cells at C/5 and 55 °C from the cycled and fully

discharged a) back to back cells in Figure 7 and b) three-electrode cells in Figure 9. In a), the half cells

are simply the separated component half cells of the back-to-back setup, cycled without further

modification; in b), the electrodes from a three-electrode cell were removed and assembled into new

half cells with Li metal counter-reference electrodes.

Electrochemical behaviour of half cells after cycling in back-to-back and three-electrode cell

formats: In a), the fully discharged back-to-back cell presented in Figure 7 was separated into its half

cell components which were individually cycled under the same conditions. The potential profiles

show that in the discharged back-to-back cell, LNMO was in its fully discharged state while LTO was

lithiated to a significant degree (approximately 3/4 charged). Both electrodes had close to their initial

reversible capacities. This confirms the limited delithiation mechanism for the back-to-back cell.

In b), the fully discharged three-electrode cell presented in Figure 9 was disassembled, and the LNMO

and LTO electrodes assembled into fresh half cells with Li counter-reference electrodes. The potential

profiles show that LTO was initially in its fully lithiated (discharged) state, and for LNMO, was in a

partially charged state (the lower plateau is not seen in the first charge in Fig. S3 b)). The LTO

electrode showed close to its initial reversible capacity. However, for LNMO, the freshly assembled

half cell showed a considerably higher impedance compared to the original three-electrode cell. We

attribute this to the disassembly of the cell: the separator was stuck firmly to the LNMO and could not

be removed without damaging the electrode; the electrode was therefore moved to the new cell with

the old separator, and fresh electrolyte added. We cannot account for the possibility of precipitation of

salt or other species in the separator during this process, and it is possible that the highly oxidising

nature of the partially charged LNMO causes reactions during the re-assembly process which increase

the impedance of the new half cell. This was attempted a number of times with the same observation.

Nonetheless, the results here are sufficient to confirm the limited lithiation mechanism for the three-

electrode cell.

Figure S4. The XPS spectra (1486.6 eV photon energy) of LNMO electrodes for P 2p, Mn 2p and Ni 2p after 50

cycles. The data is shown without any normalization.

Thickness of surface films on LTO. The probing depth of XPS is determined by the kinetic energy of

photoelectrons emitted. The kinetic energy determines the inelastic mean free path (IMFP, λ). In XPS

measurements, 95 % of the total signal obtained originates from a depth of 3λ or less; 3λ is therefore

said to be the probing depth. Since the probing depth varies with energy, the thickness of the SEI can be

estimated from analyzing the layer using different energies. By approximating the SEI density to that of

polyethylene [1], the probing depth of the O 1s spectra at 6015 eV can be estimated as 40 nm. For 2005

eV and 1486.6 eV, the probing depth would be 13-14 nm and 9-10 nm, respectively. For the room

temperature sample, no oxide is observed at 1486.6 eV whereas a small signal is observed at 2005 eV,

indicating that the SEI thickness could be between 9 nm and 14 nm. With a similar approach, the

consideration of C 1s spectra would estimate the SEI thickness between 11 nm and 16 nm for the room

temperature sample, if using the -C-C- peak at 284 eV as indicative of the bulk. For the sample cycled

at elevated temperatures, the SEI appears thicker. When observing the Ti 2p spectra for cycled samples

(the only spectra that did not yield any signal at 1486.6 eV), the SEI thickness estimation is only 10-14

nm. These values should be handled with care and not be considered as real thickness values since many

simplifications are made in making these estimates (e.g. complex electrode morphology, heterogeneous

film nature and also deposition on different electrode components, etc. are neglected).

Figure S5. Mn 2p and Ni 2p XPS spectra of LTO electrodes measured at different photon energies.

Figure S6. Ti 2p XPS spectra of LTO electrodes measured at different photon energies.

Figure S7. Galvanostatic cycling results of 2-electrode regular LNMO-LTO full cell (LNMO limited) cycled at

1C charge-discharge rate and at room temperature. After 1600 cycles (around 3 months of cycling duration), it is

observed that the fading rate is significantly accelerated (a). The voltage curves at different rates show the presence

of significant overpotential and its effect on capacity (b). This overpotential (during charging) is also expected to

accelerate cell fading further because of more severe side reactions.

Reference

[1] L. R. Painter, E. T. Arakawa, M. W. Williams, and J. C. Ashley, “Optical Properties of Polyethylene:

Measurement and Applications,” Radiat. Res., vol. 83, no. 1, pp. 1–18, 1980.

download fileview on ChemRxivSupporting Information.pdf (1.24 MiB)