Volumetric expansion of Lithium-Sulfur cell during operation – Fundamental insight into applicable characteristics

Abstract

During the operation of a Lithium-Sulfur (Li-S) cell, structural changes take place within both positive and negative electrodes. During discharge, the sulfur cathode expands as solid products (mainly Li 2S or Li2S/Li2S2) are precipitated on its surface, whereas metallic Li anode contracts due to Li oxidation/stripping. The opposite processes occur during charge, where Li anode tends to expand due to lithium plating and solid precipitates from the cathode side are removed, causing its thickness to decrease. Most research literature describe these processes as they occur within single electrode cell constructions. Since a large format Li-S pouch cell is composed of multiple layers of electrodes stacked together, and antagonistic effects (i.e. expansion and shrinkage) occur simultaneously during both charge and discharge, it is important to investigate the volumetric changes of a complete cell. Herein, we report for the first time the thickness variation of a Li-S pouch cell prototype. In these studies we used a laser gauge for monitoring the cell thickness variation under operation. The effects of different voltage windows as well as discharge regimes are explored. It was found that the thickness evolution of a complete pouch cell is mostly governed by Li anodes volume changes, which mask the response of the sulfur cathodes. Interesting findings on cell swelling when cycled at slow currents and full voltage windows are presented. A correlation between capacity retention and cell thickness variation is demonstrated, which could be potentially incorporated into Battery Management System (BMS) design for Li-S batteries.

Keywords:

Lithium-Sulfur batteryPouch cellVolume expansionThickness changeLaser gauge

1. Introduction

The low weight, low cost and high specific energy of Lithium-Sulfur (Li-S) batteries make this technology one of the most promising energy storage system for the future. Predicted to exceed the energy density of secondary Li-ion batteries by five times [1, 2], they have been extensively researched in academia and industry over past years [3-5]. The lithiation of sulfur during discharge, unlike in a Li-ion cell, involves multiple state changes, where intermediate species (lithium polysulfides, Li2Sx) of different chain lengths and solubility in the electrolyte are formed. This transition is governed by electrochemical and chemical interactions occurring in parallel, making the working mechanism very complex and abstruse despite of decades of research [6-9]. Li-S cells face several major issues which impede their commercialization [10, 11]. That results in the practical values of specific energy being much lower than theoretical values (2500 Wh kg -1). To date a tremendous amount of research has been conducted, tackling existing problems from many different angles. A majority of this work is devoted to the development of new materials to improve cell performance [12-14]. Experimental studies are often supported by theoretical work, and vice versa [15-17]. Recently, more literature concerning working mechanism has appeared, which also

1

involve the application of sophisticated in situ and operando techniques to better understand the functioning of Li-S cells under operation [18-20]. This aspect is undoubtedly critical as better fundamental knowledge is necessary for achieving further improvements. Most of this work however, has been conducted at a small scale, in a laboratory set up (coin cell, small pouches). Although it provides crucial fundamental information, the drawback of such small scale research is that some phenomenon cannot be observed at easily measurable levels as it would be in a larger cell (i.e. thermal response, volumetric cell expansion, etc.). Also, results obtained from a cell with a bespoke design (often used for in situ and/or operando studies) can be misleading due to non-representative conditions in the cell. This research can be overly simplistic, which can lead to issues when the technology is commercialized. During the past few years the effect of cell parameters, such as electrolyte amount and type, areal capacity of cathode and many more, have been researched, which when well optimised, could lead to more competitive specific energy values [21-24]. Therefore, to accelerate the process of Li-S commercialisation, research should now focus on more applied aspect, as rightly pointed by some researchers [25, 26]. Going a step further, in a real application, a Li-S battery pack will be composed of multiple single Li-S cells connected in series and/or parallel to meet the power and voltage requirement of desired application, such as electric engine, stationary storage, spacecraft, etc. These cells will cycle at different operating conditions and environments. The measurement of engineering variables such as thickness and temperature become vital not only for fundamental understanding but for use in BMS (Battery Management System) as well. When designing and/or modelling the battery pack, an important parameter to consider is dynamic expansion of the individual cell while under operation. Volumetric expansion of commercial Li-ion cells has been widely studied, with a variety of different techniques, such as Digital Image Correlation (DIC) [27, 28], high-precision displacement sensors [29], in situ strain gauge [30], mechanical instrumentation with laser [31] and synchrotron-based computed tomography (CT) [32], to name a few. To the authors best knowledge, similar studies on a large-scale Li-S cells have not yet been performed. Several examples can be found in Li-S related literature, where a fundamental explanation is given concerning volumetric expansion of individual electrodes, including experimental [33, 34] and modelling work [35]. A lot of attention has been paid to cathode structural changes, beginning at the microscopic level [35] and extending to a complete electrode [34, 36-38]. Nevertheless, since the volume change of a complete cell is a combination of the simultaneous response of both electrodes and gas formation, it is therefore important to evaluate complete cell behaviour.

Herein, we investigate for the first time, the volumetric changes of a pre-commercial, large-scale Li-S pouch cell (21 Ah and 11 Ah versions) through monitoring the variations of a thickness of the cell during cycling at different operating conditions, which are close to those of real world applications. The effects of discharge C-rates and voltage window are taken into consideration. The obtained information is important, not only from a battery pack design perspective, but also from the potential to shed light on the behaviour of pre-commercial, large-scale Li-S cells.

2. Theory

2.1. Volumetric changes of individual electrodes in Li-S cell

During the discharge of a Li-S cell, reduction of elemental solid S 8 occurs on the cathode, which leads to the formation of soluble intermediate species, lithium polysulfides Li2Sx (8 ≥ x > 2). The latter undergoes further reduction and solid precipitates, i.e. short chain polysulfides (Li2S2, Li2S) are formed on the cathode surface [38, 39]. Majority of the precipitated insulating product is assumed

2

to be Li2S, since the formation of Li2S2 is still questionable and has been rarely confirmed by experimental techniques [19, 40, 41]. Due to significant differences in densities between end-reaction solid product, i.e. sulfur (2.03 g cm-3) and Li2S (1.67 g cm-3), and considering that each mole of sulfur can be transformed into eight moles of Li2S, almost 80% of extra volume in the cathode is necessary to accommodate the precipitated Li2S [9-11]. Earlier experimental studies demonstrated that a cathode at the of discharge was 22% thicker as its pristine form [34]. Recent work of Fan et al.[42], supported by a model of Ren et al. [15] demonstrate that the morphology of Li2S precipitates is rate-dependent, where slower discharge currents lead to lower nuclei density and formation of larger crystallites, whereas at higher discharge rates more conformal and fine structure of Li2S is created. Although not yet fully explored, the differences in the morphology of Li2S precipitates may also affect the final volume change of the cathode. During charging, the opposite process takes place wherein solid Li2S is oxidized into soluble polysulfides, therefore, the total volume of the cathode at the end of charge is expected to be lower than at the end of discharge. However, elemental sulfur is expected to be formed on the the cathode surface at a potential higher than 2.4 V [43], as a result of oxidation of soluble polysulfides species. That may cause the thickness of the cathode to expand. Experimental results have showed that the thickness of the cathode at the end of charge increased about 4% from the initial state [34]. Due to the insulating nature of Li2S species, their oxidation is difficult (this is seen as a characteristic overpotential kick at the beginning of the charge voltage profile [10, 44]) and it may appear that some residual Li2S precipitates can stay on the cathode surface even at the end of charge. In long term cycling this may lead to the building up of a layer of insulating precipitates on the cathode surface, causing the internal resistance of the cell [39] and overall electrode thickness to increase.

The lithium metal anode is a key component for next generation high specific energy cells. The mechanism of Li plating is different for different operating conditions and current densities [45]. Two main types of plating have been identified [46]. Root-growing mossy deposits of lithium, whose fresh surface when exposed to the electrolyte leads to electrolyte consumption and formation of SEI (Solid Electrolyte Interphase), which in turn results in increased internal resistance, low efficiency and short cycle life. Dendritic growth or finger-like deposition risks the penetration of the porous separator and causing the cell to short-circuit. In Li-S cells, unlike in Li-ion counterparts with graphite anode where intercalation of Li+ occurs into the lattice of graphite, pure metallic Li is used and all the deposition/precipitation processes take place on the lithium anode surface. During cell discharge, lithium from the anode oxidizes causing contraction of the electrode [47]. The amount of Li being depleted depends on the amount of charge being passed between the electrodes. During charge the opposite process takes place where Li ions (Li+) are moving towards the metallic Li anode. They are subsequently plated on the anode surface causing electrode expansion [47, 48]. Moreover, the undesired but unique shuttle effect which occurs at the end of charge of a Li-S cell, widely described in the literature [49, 50], on one hand leads to reduction of soluble polysulfides species onto the surface of anode in the form of insulating and insoluble Li2S layer [11], on the other hand prevents the dendritic growth of Li. Shuttle phenomenon, in brief, is a parasictic reaction that occurs at the anode surface. Lower order polysulfides are oxidised to mid-to-high order polysulfides during charge at the sulfur electrode. Due to an excess of higher order polysulfides at the sulfur electrode surface compared to the lithium electrode’s one, a concentration gradient is set up. Therefore, the higher order polysulfides diffuse towards the lithium electrode and are reduced to lower order polysulfides, which then begin the journey back to the sulfur electrode side. The presence of soluble polyslufides does not allow for Li dendrites to develop, as Li pins are consumed in the reactions with soluble polysulfides species [51]. The presence of shuttle is noticeable in the voltage curve at potentials close to 2.45 V, where at the end of charge, the voltage profile starts to progressively flat line rather

3

than increase, as complete oxidation of sulfur cannot be achieved (Figure S1). As a result of this parasitic reaction, the amount of current supplied for charge is usually greater than that accessible during discharge, causing a reduction in Coulombic efficiency. Moreover, the shuttle reaction generates heat which may cause enhanced degradation of the cell and curtails the life of the cell.

An aspect for consideration is the presence of stress and strains in this system. As we show later on in this study, it is difficult to deconvolute the thickness changes of each individual component of the cell easily. The presence of multiple layers stacked together in a Li-S cell along with multiple phase changes that occur, as species dissolve and precipitate, call for careful consideration of the strain responses of the system and as such are beyond the scope of this particular study.

2.2. Gas formation

To the best of authors knowledge, very little work has been done on the characterisation of the gaseous products within Li-S cells. Jozwiuk et al. [52] have studied the gassing process of a bespoke Li-S cell under operation, through the measurement of cell pressure. The electrolyte used was ether-based and the effect of LiNO3 as a shuttle inhibitor was investigated. In general, formation of gas in a Li-S cell with ether-based electrolyte is caused by decomposition of a solvent molecules due to reaction with fresh surface of lithium anode. Formation of gas is reduced, but not eliminated, when LiNO3 additive is used. The increase in the pressure is detected during charge, and it was attributed by the authors to the gasous product formation.

2.3.Voltage region selection

In a Li-S system different reactions and phenomena take place at different voltage levels during charge an discharge. Some of them may be detrimental to the life time of a cell, such as shuttle effect being present at the end of charge at the voltages above 2.4 V. Severe passivation of the cathode surface by insulating Li2S occurrs at the very end of discharge, which may cause the premature end of discharge visible as a rapid voltage drop. These phenomena, among others, can be minimized by controlling the voltage limits in which Li-S cells are cycled [44, 53]. In this work, to maximize the capacity of tested cells, the voltage window of 1.5 V - 2.45 V was applied. The reason behind this limited charge cut-off voltage of 2.45V (instead of the limits typically found in literature of 2.8 V/3.0 V) is two-fold. First, the voltage window was restricted mainly to prevent shuttle phenomena, which occurs primarily at voltages close to 2.45 V in the studied OXIS cells, and which have a detrimental effect on electrochemical performance, as previously explained. Secondly, in a real world application, the battery is never taken up to the maximum voltage limits, unlike in the fundamental research studies. The aim of the real world application, and therefore this study, is to maximise the cycle life of the battery as much as possible. Therefore, our tests described in further section contain results obtained when charging the cell to a maximum of 2.45V. The cell was also examined under charge taken up to 3.0 V (Figure S2). In terms of capacity and electrochemical behaviour, there is not much to gain when operating in this range, but the risk of enhancing shuttle phenomenon greatly increases.

To further maximize the lifetime and improve coulombic efficiency, a strategy to avoid cycling the cell in extreme voltage ranges was applied, and will be further referred to 80% DoD cycling (see section 2.4).

2.4.Terminology

4

In this manuscript two terms will be used to describe the nature of cell swelling: solid and gaseous. Although in this work a quantitative distinction between the contribution of both types of swelling was not possible, the cells were inspected visually which gave a fairly good indication of the main cause of swelling. Gaseous swelling is distinguished as causing deformation non-uniformly in the z-direction, in a more ‘pillow-like’ shape, as schematically illustrated in Figure S3. When touching the pouch, presence of a gas can be easily recognised. On the other hand, solid swelling is attributed to the swelling caused by volumetric changes of the electrodes. The cell, when touched, is hard and a more uniform expansion in the z-axes is seen. Sometimes a cavity-like shape can be seen, where the cell is slightly thicker on the edges and concave in the middle. This behaviour could possibly be explained by more preferential deposition of Li on the edges, as seen in Li-ion pouch cells [48]. Thickness measurement applied in this work only collected the data on a single point, in the middle of the cell, therefore the effect of thickening the edges was not included in the laser measurement.

In this work the Li-S cells underwent different testing protocols and following terms will be used throughout the manuscript:

• 100% DoD – cycling the cell in a full voltage window, i.e. 1.5 V – 2.45 V. This represents the optimal voltage range for OXIS Long Life cells which maximises the capcity.

• 80% DoD – cycling the cell in a limited voltage window, where discharge terminates just above 1.7 V and charge does not exceed 2.4 V. This strategy maximizes the lifetime of the Li-S cell used in this work.

• Standard cycling conditions – cycling the cell in the 1.5 V – 2.45 V voltage window (100% DoD) at C-rates of 0.1 C (C/10) and 0.2 C (C/5) for charge and discharge, respectively.

• Preconditioning – initial 10 cycles performed at standard cycling conditions (see point above)

3. Experimental setup

3.1. Pouch cells description

These studies were performed on Long Life Li-S pouch cells (OXIS Energy Ltd) [4], composed of multiple layers of anodes, porous polymeric separators and cathodes, all trimmed to the desired size and stacked together alternately in the pouch material (aluminium film laminated with polymer). The anode is a metallic lithium foil, while the cathode is a proprietary mixture of sulfur, carbon and binder, coated double-sided on an aluminium foil serving as a current collector. The thickness range of individual components is as follow: 10 – 30 µm (separator), 30 – 100 µm (double sided cathode), 40 – 110 µm (Li metal)*. All layers of cathodes as well as anodes are welded together to the current collector tabs, i.e. Aluminum and Nickel, for positive and negative electrodes, respectively. The electrolyte is a sulfone-based solution* and is absorbed within the cathode’s and separator’s pores and it provides transport of Li+ between the electrodes. The amount of electrolyte* used in this cell was optimized in order to maximize the cycle life (through sufficient excess of electrolyte) without causing a drastic decrease of the specific energy density.

Two types of cells were used in this work, both having the same components and x-y dimensions, with the only difference being in the number of layers stacked together. As a result, cells of different thicknesses and nominal capacities were obtained (see Figure 1 and Table 1).

* Details of the exact cathode and electrolyte composition, type of separator, precise amount of electrolyte with respect to the mass of sulfur, as well as number of layers stacked in the pouch are proprietary information of OXIS Energy. However, the results and discussion provided in further part of the manuscript should be sufficient to assess the hypothesis we proposed, even though we are not able to provide explicit proprietary cell information to the reader.

5

Single layer pouch cells composed of only one anode/separator/cathode layer were also built and cycled prior to post mortem analysis. The x-y dimensions of the electrodes were identical to the one used in larger cells described above.

It is important to note that the stack of the layers is vacuum sealed in a pre-formed pouch. The pouch, made of laminated aluminium film, is shaped during the pre-forming process in order to enclose the cell tightly within the pouch on sealing the cell. The depth * of the pre-formed pouch is a result of a cell design optimisation process. Due to the nature of the chemical and electrochemical processes taking place within the cell, most of the thickness growth occurs in the z-direction, perpendicular to the layers. As there is minimal slack volume between the pouch and the layers, the volume expansion in the z-direction directly correlates to thickness changes within the layers. We believe that for most cases the volume changes are within the elastic limits of the pouch i.e. there is no permanent deformation of the pouch.

3.2.Thickness measurement

The thickness of each fresh cell before cycling was initially measured by a micrometer (Kroeplin IP67; resolution 10 µm), in the center of the x-y footprint surface (marked with yellow dot in Figure 1). The thickness during cycling was monitored with a laser thickness gauge (Keyence; model IL-065) in the set up presented in Figure 2. A special cycling rig was designed, so that the head of the laser was at a fixed distance from the bottom plate, on which the cell was placed. The laser measured the thickness at a single point, being approximately in the middle of the surface of the cell. As described previously, the assumption is made that any volumetric changes primarily occur in the z-directions, and therefore should be detectable through thickness changes in this direction. This set up, however, did not account for the potential deformation on the edges. The tabs were screwed in place under the Cu blocks, which then where plugged into the battery cycler. The laser settings were calibrated prior to the cell thickness measurement on a known thickness object. The laser was also verified against the potential issue for a measurement, i.e. laser drift (Supplementary Info, section S4 and Figure S5). After placing the cell under the laser beam, the reading of the cell thickness was compared with the one obtained by a micrometer. A good agreement between both values was achieved. The laser reading was then reset to zero, so that only variations in thickness were captured during cycling. A 21 Ah cells were cycled in a Bitrode battery cycler and the thickness/laser data were collected using a Graphtec GL240 data logger. A 11 Ah cells were cycled in a Maccor 4000 battery cycler where thickness laser gauge was connected to a Maccor auxiliary voltage board, which enabled us to record the thickness data along with other electrochemical data.

3.3. Electrochemical testing

Cycling tests of the 21 Ah cells were performed on a Bitrode battery cycler, while a Maccor series 4000 battery cycler was used for testing the 11 Ah versions. The cells were tested in constant current mode. After being assembled, the cells were discharged at 0.2 C rate to 1.5 V and left for storage over the period of several days† before further tests were performed. The cells were then preconditioned (see section 2.4), unless otherwise stated.

The whole set up (Figure 2) was kept in a thermally controlled chamber at 30 oC. Thermocouple was used for the cell temperature control. It was attached with Kapton tape to the cell surface and

† The rest period after initial discharge was applied only to 21 Ah cells due to logistical issues. The internal OXIS Energy studies on effect of storage on cell performances showed that such short resting storage period has no effect on further testing results.

6

covered with a thermally insulating tape to minimize the influence of the surrounding environment. An important fact to note is a thermal expansion. When the temperature of the oven was immediately changed, an instantaneous change in the thickness was noticed. This is attributed to thermal expansion of the cell. Therefore, care has been taken to limit the temperature change the cell was exposed to, in order to ensure that all thickness changes seen are due to changes in the internal electrodes due to chemical and electrochemical reactions taking place, rather than a mechanical response to a thermal stimulus.

Several different tests were performed on both formats of cells, at different voltage windows and currents conditions, to fully explore the effect of each factor on cell thickness variations. Table 2 summarizes the details of cycling procedure applied for each cell.

3.4. Gaseous cell swelling

In order to study swelling due to gas formation, a freshly built 11 Ah cell was left for 72 h hours in a fully charged state. We postulate that this allows residual water trapped in the cell components/electrolyte to react with the fresh lithium anode. The gas released, potentially due to a reaction of Li with water, causes visible cell swelling. Further work is needed to confirm the exact gas composition. After 72 h of storage, the cell was found to be ~3 mm thicker than the initial value (see Table 1) and deformation of the pouch clearly indicated the presence of gas inside the cell (Figure S3a). The cell was then cycled under the laser beam and thickness variation during cycling was recorded.

3.5. Post mortem characterisation of a single layer pouch cells

Single layer cells were used for post mortem analysis, where after being cycled in a desired voltage and C-rate regime, the cells were taken apart and thickness of the electrodes was measured (with the micrometer; precision ±10 µm). The obtained values were then compared with the values of the pristine electrodes. The morphology of cycled electrodes was investigated using Scanning Electron Microscopy (JEOL JCM-6000). This work was done to study the dependence of the morphology and electrodes thickness on the discharge current regime as well as the state of charge/discharge. Four single layer cells were cycled for a certain number of cycles (see Table 3), until the discharge capacity faded significantly. The charge regime was maintained at 0.1 C and only the discharge current was altered. Two cells were discharged at 0.5 C, while two others at 0.1 C. In each pair of cells, one of them was stopped at the end of charge (2.45 V), while another one was stopped in a fully discharged state (1.5 V).

4. Results and discussion

4.1. Cell performance under standard cycling conditions

The capacity retention and voltage profile of a 11 Ah cell cycled during the initial 10 cycles under standard cycling conditions (see Table 2, Cell#1) is shown in Figure 3. The cell after being built was immediately discharged with 2.2 A current (equal to 0.2 C rate) and left for storage over the period of several days before further cycling (including operando thickness monitoring) was done. The initial discharge capacity (marked with a red star in Figure 3a) is higher than the value obtained during further cycles, which is commonly seen in liquid electrolyte Li-S system [54]. At the 10th cycle discharge capacity was equal to 11 Ah. Charge capacity, on the other hand, stabilized from the 5 th

cycle at the constant value of 12.1 Ah, as the charge process terminated by the time cut-off limit before the voltage limit was triggered (Figure 3c). This is normal behaviour of a Li-S cell exhibiting

7

shuttling behaviour, as previously explained in section 2.1, where the voltage curve progressively flattenes at the potential close to 2.4 V [10, 55]. The thickness of a freshly built cell (measured with a micrometer) was 7.75 mm and it contracted to 7.3 mm after the initial discharge. This shrinkage is primarily due to the lithium metal anode deplating/oxidising. Lithium metal, in reaction with sulfur, gets converted to lithium sulfide (Li2S) which is accompanied, in theory, by 80% of volumetric expansion of the cathode. Therefore, even though the thickness of the cathode is expected to increase due to deposition of solid Li2S, the response of the cathode seems to be masked by the more significant anode thickness change.

The thickness variation recorded during cell preconditioning is shown in Figure 4. During cycling, the cell reaches the maximum thickness value at the end of charge, and the minimum value at the end of discharge. This behaviour is repeatable at each cycle, however, the cell thickness is monotonously decreasing, being thinner by approximately 0.2 mm as compared to the beginning of cycling (red dashed arrow in Figure 4a). At the end of charge, the thickness of the cathode is expected to be the lowest, since Li2S deposits are oxidized, leaving the carbon-binder structure uncovered. Elemental sulfur is deposited towards the end of charge, which may cause slight cathode expansion. However, its appearance is usually observed at a potential closer to 2.5 V [43], which is not the case in our cycling procedure, as charging process was stopped at the potentials ≤ 2.45 V. At the same time Li +

ions are being reduced and plated onto the lithium anode, which leads to more noticeable thickness changes and therefore, the overall cell response is governed by Li deposition at the end of charge. At the end of discharge the cell is the thinnest, most likely due to lithium deplating as explained before.

In these studies we focused only on one type of an OXIS cell and we characterised it according to a protocol which is closer to a real world application. The cell described in the manuscript is therefore only one cell configuration and one test case among the variety of other possibilities. The effect of different cell components on the global response of the cell thickness was studied only in a limited manner and results are shown in section S6 (Supplementary Info). It was found that despite significant differences in the cell parameters, the general trend of the thickness behaviour of both cells is quite comparable, with minor differences. Further work should be carried out in order to explore in a more thorough fashion the effect of each individual parameter on the Li-S pouch cell thickness response.

These experiments do not allow us to quantitatively identify the contributions of the solid and gaseous swelling. Only a visual inspection suggested that these thickness variations were caused by solid swelling (i.e. volumetric expansion of the electrodes; see section 2.4 and Figure S3.b) in larger degree. Thickness variation of a swollen cell due to gasous swelling was also examined and the results clearly show the difference in thickness changes (section S8 and Figure S8 in Supplementary Info).

The plot of thickness variation overlaid with the voltage profile (Figure 4c) helps us to identify the hysteresis between the thickness variation during charge and discharge, especially along the lower discharge plateau. This strongly indicates that the mechanisms hidden behind charge and discharge processes are not linearly reversible and they cause different cell behaviour during charge/discharge. This observation is in fairly good agreement with previous operando studies, where hysteresis between the measured values (such as peak area of Li2S reflection measured by XRD [18]) during charge and discharge was clearly visible. It is worth mentioning that the hysteresis between the charge and discharge voltage profile is due to the difference in concentrations as well as type of polysulfides species (soluble, insoluble) present at a particular state of discharge and charge of the cell. However, the voltage hysteresis cannot be translated into the hysteresis seen in the

8

measurement of other variables, for example, thickness. During charging, the thickness variation is more linear, whereas during discharge a visible bump is present, starting from the beginning of the lower discharge plateau. It is commonly known that deposition of solid precipitates (mainly Li2S, existence of Li2S2 is still unclear [19, 40, 41]) commence exactly at the transition point between upper and lower discharge plateau [10, 43]. Therefore, this bump may indicate the formation of solid precipitates on the cathode surface during discharge, leading to cathode volumetric expansion. This effect however, is masked by simultaneous deplating/thinning of lithium anode.

An interesting feature can be also observed during the initial 60 h of cycling, where the value of the maximum thickness variation goes through oscillations, with some significant peaks at the end of charge reaching an absolute value of ~0.5 mm (Figure 4a). We denote this phenomena as a ‘stabilisation’ period. During initial cycling stabilisation of Li anode surface occurs, potentially including continuous growth of a more stable SEI layer. Since the surface of lithium is fairly fresh at the beginning of cycling, it is more likely to react with the electrolyte or with residual water in the cell components/electrolyte, if there is any, leading to gas formation. As we showed previously (Figure S8 in Supplementary Info), this gas formation process is reversible. After the ‘stabilisation’ period, a reversible thickness contraction/expansion of the cell is seen in the range of approximately 0.35 mm within a single cycle, which is 4-5% of the total cell thickness (7.3 mm).

Almost identical features were observed on the 21 Ah version of the cell (see Table 2; Cell#2), where more layers of anode/separator/cathode are stacked together. Similarly, a ‘stabilisation’ region during the initial 90 h of cycling was observed, after which reversible expansion and contraction of the cell occured during charge and discharge, respectively (Figure S9). The magnitude of the reversible expansion within a single cycle was measured to be 0.55 mm, which is again 4-5 % of the total cell thickness (13.0 mm). This value is in a very good agreement with the data obtained for the 11 Ah version of the cell. This results prove that when scaling up the cell by increasing the number of layers, thickness variation can be predicted, which from a battery pack design point of view is important. This also proves our initial assertion that the volume expansion can be primarily expressed as a thickness increase in a direction perpendicular to the plane of the stacked layers for a given combination of materials for the cell components, namely the cathodes, anodes and electrolyte.

4.2. Cell thickness evolution in a limited voltage range

A 11 Ah cell after being preconditioned for 10 cycles, was then switched to a limited voltage range test, where both charge and discharge cut-offs were limited (see Table 2, Cell#3). By doing so, capacity values decreased to about 80% of the value obtained at the full range. Discharge capacity dropped from 11 Ah (on the 10th cycle) to 8.95 Ah (on the 11th cycle), as shown in Figure 5. As the charge was stopped at lower potential (below 2.40 V), oxidation of medium order polysulfides to higher ones was not completed. As a matter of fact, the upper discharge plateau was curtailed and the associated capacity was lost. Coulombic efficiency improved significantly, since charging was stopped before the shuttle process started. On the other hand, modifying the discharge cut-off of the voltage did not have a significant effect on the value of capacity, since the capacity delivered between 1.75 V and 1.5 V is almost negligible.

The thickness evolution (Figure 6) clearly shows that when changing both voltage cut-offs simultaneously, the charge limit affects the cell expansion more significantly, while the change in the discharge cut-off voltage does not cause an immediate change to the cell contraction. The reversible thickness change within a single cycle was reduced to 77% of the value in the full voltage range, i.e. from 0.354 mm during 10th cycle to 0.271 mm during 11th cycle. Reduction of the voltage window

9

leads to a reduction in capacity by 81%. This value agrees quite well with the thickness variation change. During further cycles at the limited voltage range, the end-of-charge-thickness stays at fairly stable level (blue line in Figure 6a), while end-of-discharge-thickness is progressively shifting towards higher values (red line in Figure 6a). As a result, after 600 h of cycling the thickness variation within a single cycle decreased to 0.164 mm, whereas delivered capacity stayed unchanged. The explanation for this behaviour is still unclear, nevertheless, it is visible that cycling the cell in a limited voltage window results in less of a cell thickness variation. In a real application, it is more likely that the cell will not operate in the extreme voltage regions for both, charge and discharge. From a battery pack design perspective, it is therefore beneficial to deal with smaller thickness variation, as the space between the cells can be reduced, which in turn can improve the volumetric energy density of the battery pack.

4.3. Cell behaviour at different C-rates

Figure 7 presents the results of the 21 Ah cell submitted to a C-rate screening protocol (Table 2; Cell#4). The charge rate was kept at a constant value of 0.1 C irrespective of the discharge current, to eliminate the uncertainty of the way lithium was plated during charge, depending on the current density. Only for higher discharge currents the voltage cut-off was reduced to 1.2 V, to account for eventual overpotential and premature triggering of the voltage limits. Due to the higher ohmic drop, the overpotential at the juncture between the high and low plateau can sometimes dip below 1.5 V necessitating further drop in the cut-off voltage. The general trend observed for capacity retention behaviour (Figure 7a) is that at slower discharge rates, discharge capacity is initially higher but then fades rapidly within the next few cycles. At the same time, the charge capacity stabilises at the shuttle value of 23.1 Ah, as the charge process terminates by a time limit. The time limit of 11 h is calculated at the 0.1 C rate with an addition of an hour as a safety limit. Higher discharge currents lead to slightly (although not much) lower discharge capacity values, but more stable capacity retention is observed. Charge capacities are of lower values since charging terminates with the voltage cut-off, suggesting that shuttle phenomenon was not present. It is interesting to see that although the charge process was performed at a constant rate of 0.1 C, it was significantly affected by different discharge currents in the preceding cycle. As already mentioned, for cycles with discharge rates ≥ 0.5 C, no sign of shuttle phenomenon is seen, while lower discharge rates (≤ 0.2 C) led to the occurrence of the shuttle effect as the voltage of the cell was unable to reach the charge cut-off. This could be related with the fact that the concentration gradient between the electrodes necessary for shuttle mechanism to develop is time dependent. Therefore, the longer the charge process lasts, more likely that the shuttle will occur. Fast discharge rates result in different composition of polysulfides remaining in the electrolyte, as the discharge is not complete. The higher the discharge rate, the lower the capacity extracted, and thus the more the proportion of higher order polysulfides remained within the electrolyte. The proceeding charge therefore will not be as long, simply because only part of the capacity needs to be replaced. This shortening of the charge results in a reduction of time for the concentration gradient to develop. This is in opposite to the slow discharge rates, where during long discharge processes more of a concentration gradient can be formed, as conversion of soluble species to solid ones is more complete. As an example, discharge of 2 C which lasted 24 min (blue curve in Figure 7b) resulted in 17 Ah of capacity, and following charge terminated after 9 h of charging, delivering 19 Ah of capacity. On the other hand, 0.05 C discharge cycle (black curve in Figure 7b) lasting >18 h and delivering 19.1 Ah of capacity, resulted in the ‘shuttle’ charge terminating after 11 h, thus 23.1 Ah of capacity.

10

It can also be seen that high discharge rates resulted in better electrochemical utilisation of the higher discharge plateau, whereas the cells discharged at slower rates show a slightly reduced upper plateau (dashed cirle in Figure 7b). A possible explanation could be related to the presence/absence of shuttle phenomenon in the preceeding charge. In brief, the occurrence of shuttle phenomena can be clearly distinguished by the stable charge capacity (Figure 7a) due to the fact that the charge process is terminated by time cut-off rather than voltage cut-off (Figure 7c). This is because, as explained earlier, the voltage flat-lines around 2.45 V and is unable to reach the voltage cut-off. Therefore, during such a ‘shuttling charge’, some part of active material gets lost due to irreversible deposition on the anode/lithium surface. As a result, the capacity corresponding to upper discharge plateau obtained in a subsequent discharge is consequently reduced, as the amount of species to be reduced during discharge is lower. In the cycles with no ‘shuttling charge’, the following discharge voltage profile has a complete upper discharge plateau. In addition to that, the cycles with a reduced upper voltage plateau (0.1C, 0.05C and 0.2C) occurred at the end of testing (more than 35 cycles; Figure 7a), where the cell was potentially degraded and therefore shuttle phenomenon was more likely to occur.

A significant difference in the magnitude of the overpotential kick just at the beginning of charge can also be seen (zoomed in image in Figure 7c) depending on the preceding discharge rate. This overpotential is related with the initial step of Li2S oxidation, which forms on the surface of the cathode during the discharge. The higher the kick, the more difficult it is to oxidize the insulating precipitates. It is clearly visible that higher discharge current resulted in lower overpotential at the beginning of following charge. In contrast, slow discharge rates lead to a very high charge overpotential kick. This difference in oxidizing Li2S cannot only be attributed to the amount of Li2S deposited, since the length of the 2nd discharge plateau during which Li2S was formed is fairly similar for all cases, especially for the rates ≤ 0.5 C (Figure 7b). It could potentially be related to the different morphologies of Li2S crystallites, since they demonstrate a rate-dependent morphology, as recently explained by Fan et al. [42]. Higher C-rates lead to formation of high nuclei density thus a continuous morphology composed of smaller crystallites, which are more likely to be easier oxidized in a charging process. Lower discharge C-rates on the other hand favourite formation of fewer but larger crystals, being more difficult to dissolve back, thus resulting in a significant kick at the beginning of charge.

In order to have deeper insight into the cell thickness variation upon cycling at different C-rates, another 21 Ah cell was cycled in a C-rate screening protocol (Table 2; Cell#5), in the laser thickness gauge setup. Voltage profile and cell thickness variation are shown in Figure 8a. The initial ‘stabilisation’ period (approximately 100 h; previously shown in Figure 4a) is not shown in this graph. Similarly to already discussed cases, the cell expands during charge and contracts during discharge. Discharge capacity together with the end-of-discharge-thickness value as a function of cycle number are shown in Figure 8b. It is interesting to observe such a symmetrical correlation between these two values. 1 C discharge cycles resulted in a very stable and reversible thickness variation, in line with a stable discharge capacity. When changing the discharge rate to 0.1 C, progressive increase of the thickness from cycle to cycle is visible. At the same time, discharge capacity is fading, as previously seen (Figure 7a). When switching the discharge current to the standard 0.2 C rate, the thickness starts to decrease while capacity is slowly recovering/increasing. At the very end, when reducing the discharge current to a very slow rate of 0.05 C, discharge capacity fade is observed again, accompanied by a rapid and significant increase of the cell thickness. After barely 7 cycles at 0.1 C/0.05 C (charge/discharge), the thickness of the cell recorded at the end of charge was 1.9 mm thicker as its initial value, which stands for ~15% of the thickness increase.

11

4.4. Continuous slow discharge (0.1 C)

Further test was performed to study the effect of a prolonged slow discharge regime on the cell capacity and thickness (see Table 2, Cell#6). For that, a 21 Ah cell was cycled continuously at a slower current regime of 0.1 C/0.1 C‡ for charge and discharge, respectively. The cell was left on test until the capacity dropped below 10% of the initial value and the thickness increase reached the limits of the laser detection (5 mm). The voltage profile (Figure 9a) shows that the hysteresis between charge and discharge cycle starts to increase after approximately 400 h of cycling, indicating that internal resistance in the cell progressively builds up. After 615 h of cycling (which corresponds to 38 cycles) the cell voltage profile deteriorated due to a high degree of polarisation and the capacity drastically dropped. At the same time, the thickness of the cell increased, reaching the limit of a laser reading of 5 mm, which corresponds to a 40% thickness increase. At the end of cycling ( i.e. after 730 h), the cell was found to be significantly swollen (Figure S10). The thickness measured with calliper was found to be 21.6 mm, which corresponds to 68% increase (8.1 mm) of the initial thickness. It is believed that this swelling occurred due to volumetric changes of the electrodes (so called ‘solid swelling’, section 2.4), rather than gas formation.

Three main regions can be distinguished in the capacity retention behaviour (Figure 9b). Until cycle 22, only discharge capacity is fading while charge capacity stayed at the stable value limited by time cut-off, indicating the presence of shuttling. The thickness of the cell within that period increased by 1 mm (approx. 7%). Starting from cycle 23, internal resistance in the cell increased sufficiently to cause polarisation in both charge and discharge curves. As a result of that, during charge the voltage cut-off of 2.45 V is reached much before the time limit. Until cycle 35, discharge capacity is continuously fading in a similar manner as during previous cycles, however, charge capacity starts to fade drastically. The explanation of this behaviour is still unclear. It could be potentially explained by a vicious cycle of undercharging leading to lower discharge capacity. At the cycle 35 the increase of the thickness at the end of charge was recorded to be 5 mm (further operando thickness values are missing due to laser limitations). Starting from cycle 35, both capacities faded in an identical manner, dropping to 2 Ah after 50 cycles. A symmetric correlation between end-of-discharge thickness increase and discharge capacity fade can be observed. It can be also clearly noticed that the most rapid thickness increase occurred when discharge capacity started to fade rapidly, i.e. after 35th

cycle.

An in-depth look into the cell thickness changes at different state of health when cycling at 0.1 C/0.1 C charge/discharge protocol is presented in Figure 10. At the initial stage of cycling (after 35 h; 3 rd

cycle) a fairly reversible thickness change of 0.6 mm within a single cycle occurred. The characteristic bump along discharge plateau in the thickness profile is clearly visible. After 340 h of cycling ( i.e. 18th

cycle) cell expansion during charge (0.8 mm) is higher than the cell contraction during discharge (0.65 mm). Although the same amount of charge has passed during the 3 rd and 18th charge, the cell thickness increase was not the same. It is believed that the main process which causes cell expansion at the end of charge is Li plating. However, slow discharge current may lead to more severe passivation of cathode and irreversible accumulation of Li2S which is not fully removed from the cathode’s surface during oxidation/charge process. As a result, at the end of charge two processes with antagonistic effects on the final cell thickness coincide, i.e. anode expansion due to Li plating in more mossy form and incomplete cathode contraction because of residual Li2S. The shape of the

‡ The current used in this test was 2.23 A instead of 2.1 A used for testing the cells: #1, #4 and #5 (Table 2). Therefore, the effective C-rate was 0.106 C (i.e. C/9.4), which is a 6% of the difference. It is believed that this difference did not have the effect on the obtained results.

12

thickness profile also changes as compared with the initial state of health. The bump in the thickness profile is no longer visible along the lower discharge plateau, and the charge-thickness profile has a slightly different shape. It may suggest that, although difficult to quantify, the mechanism of precipitation and dissolution of lithium and Li2S has been modified. Finally, the thickness of a very aged cell (after almost 600 h of cycling) increased more than 4 mm. The thickness increase during single charge is now as high as 1 mm, and contraction during discharge stayed at a similar value as before, at the level of 0.6 mm. Such a significant increase during charge may be related with more Li mossy-like growth and other precipitates being formed on the anode, such as deposition of active material in the form of Li2S caused by shuttle phenomenon. An interesting feature is also observed during discharge, where a significant kick of magnitude of 0.2 mm is seen just at the beginning of the lower discharge plateau (Figure 10c; marked with a black star). Its presence cannot be explained as an experimental artefact, as it appears exactly at the same point during several cycles towards the end of the cell life. This kick can potentially be related with Li 2S deposition, however, in the aged cell the mechanism of its precipitation may change, since coverage of the cathode surface is different as compared with the fresh electrode. Moreover, as slow discharge currents lead to formation of larger crystallites of Li2S [15, 42], its formation in the very aged cell may explain the occurrence of this kick. The thickness measurement was done on a single point on the cell surface. Therefore, it is difficult to state either this rapid thickness increase was a local phenomenon detected by a laser or the entire cell expanded in the z-direction. Nevertheless, these results can give an indication of the change in the mechanism in the Li2S precipitation as a result of cell aging.

4.5. Post mortem characterisation - single layer cells

To better understand the cause of a significant cell thickness increase when cycling continuously at low discharge rates of 0.1 C, two single layer cells were cycled in similar conditions and post mortem analysis of the electrodes were performed. For comparison, two other cells were discharged with a current density being 5 times higher (0.5 C rate). A summary of cells performance and electrode thickness increase (in %) as compared to the pristine values are presented in Table 3. Although the cells cycled at different conditions were not stopped at the comparable state of health ( i.e. similar number of cycles or delivered discharge capacity), these results can still be helpful to draw conclusions. The cell cycled at slow discharge regime (0.1 C), displayed similar behaviour to that previously described in section 4.4, where ‘shuttling’/stable charge capacity during initial 22 cycles was present, followed by a rapid fade, whereas discharge capacity from the beginning of cycling was progressively decreasing (Figure S11).

The thicknesses of the electrodes were measured with a micrometer, whose precision (±10 µm) allowed only for an estimated value. Nevertheless, it can be clearly seen (Table 3) that while the thickness of cathode increased by approx. 10% for both cases studied, it is the lithium anode which expanded significantly. As expected, the thickness of the lithium anode from a fully charged cell in both cases was higher than from a discharged cell, since deposition of Li occurs at the end of charging process. Moreover, a significant difference in the anode thickness is seen between two different discharge rates examined. The anode from a cell cycled in a slow discharge rate (0.1 C) expanded by 100% of the original thickness, while faster discharge rate (0.5 C) led to reduced anode expansion.

More details can be seen in the SEM pictures (Figure 11) of the morphology of both electrodes after cycling. Only electrodes from fully discharged cells are shown, since at this state precipitates of Li 2S are expected to be found on the cathode side, whereas the lithium anode should be the thinnest due to depletion/oxidation of Li. It can be clearly seen that the morphology of the cathodes

13

discharged at different currents is significantly different. Fast discharge rate led to more uniform deposits (Figure 11b) while slow discharge current favoured formation of larger crystallites of Li2S (Figure 11a). These results are in a very good agreement with the work presented by Fan et al. [42]. However, even if different morphologies of Li2S crystallites caused differences in the electrode thickness, it is more likely to occur at the microscopic level, and the changes are too small to be detected with the micrometer (±10 µm of precision). Significant differences in the surface morphologies and the thickness are clearly visible for Li anodes (Figure 11c-f). The anode which was cycled at slow rates (0.1 C / 0.1 C charge/discharge) has accumulated residuals of the deposits, even in the fully discharged state. The same electrode but at the end of charge (fully lithiated) looked similar, with more deposits on it. These deposits could be attributed to the mossy lithium and precipitates of short chain polysulfides (Li2S, potentially Li2S2) which reduced on the anode surface, during severe shuttle phenomenon occurring during charge. As previously explained, stable charge capacity indicates the presence of shuttle as the cell potential could not reach the voltage cut-off at 2.45 V. As a consequence, the charge process terminated with the time limit; the shuttle effect was present during the last ~2 h of charging (over the total charging time of 11 h). The anode from the cell cycled at faster discharge current displays more cavity like pits on the surface of the pristine lithium without additional deposits. Cycling data (not presented in this manuscript) showed that this cell did not experience the shuttle effect, which may explain the lack of short chain polysulfides deposits.

It is important to keep in mind that both cells did not deliver similar capacities, therefore the amount of Li deplated and precipitated is not equal. Nevertheless, the morphological differences between anodes cycled at different currents are significant and clearly indicates that slow discharge currents lead to significant growth of anode thickness, most likely due to the accumulated precipitates during shuttle phenomenon present during the initial tens of cycles.

4.6. Irreversible cell expansion at slow discharge rates – proposed mechanism

The results presented in the section 4.5 showed that prolonged cycling of a Li-S cell at slow discharge currents and in a voltage range of 1.5 V - 2.45 V led to destructive changes in the cell: rapid capacity fade accompanied by a significant increase in cell thickness. A schematic illustration of the cell thickness behaviour when cycling at a slow discharge rate (≤ 0.1 C) is shown in Figure 12. The assumptions are made for a single layer format, where one layer of each component is stacked together. A real cell is composed of multiple layers therefore the effect of the thickness variation is magnified. The initial thicknesses of the fresh cell and Li anode are marked with the arrows as the references. At the early stage of cycling (Figure 12b,c), thickness variation is as typically observed for other C-rates previously described, where the cell is the thinnest at the end of discharge and the thickest at the end of charge due to Li stripping and plating, respectively. The thickness variation of a complete cell is mostly governed by the volumetric changes of the Li anode, which partially masks the response of the cathode. In a very aged cell (Figure 12d,e), it is still the response of the Li anode which govern the thickness change of the complete cell. The shuttle mechanism together with deposition of mossy Li causes significant increase of the Li anode thickness at the end of charge. Lithium is plating on already existing mossy Li on the surface (Figure 12f), which cause even more significant thickness increase. At the end of discharge, despite of Li depletion, the surface of Li anode does not clear out and mossy deposits are still present. It is more likely that only the bulk of pristine Li gets depleted, whereas the mossy deposits on the surface stayed unchanged (Figure 12g). This results are in good agreement with the work of Cheng et al.[56], who demonstrated that Li metal anode contributes significantly to a pouch cell failure.

14

As discussed previously, low discharge currents cause the formation of larger crystallites of Li2S on the surface of the cathode. The latter are more difficult to oxidize in subsequent charge than the uniform layer of small Li2S crystallites formed at high discharge rates. As a consequence, irreversible cumulation of Li2S over cycling can occur, causing the thickness of the cathode to increase. Nevertheless, the changes occurring on the cathode side are more refined and maight affect the electrochemical performances of the cell rather than contribute visibly to the complete cell thickness changes. As mentioned already, the thickness changes of the cathode are masked by the response of the Li anode.

5. Conclusions

This work investigated, for the first time, the volumetric expansion of a pre-commercial Li-S prototype cell under different operating conditions. A thickness laser gauge was used to monitor the thickness of the cell while cycling. It was found that the thickness evolution was mostly due to volumetric expansion of the electrodes, rather than gas formation. Moreover, the response of the global cell is mostly driven by the behaviour of the metallic lithium anode, which masks the volumetric changes of the cathode. Lithium plating during charge causes cell expansion, with the cell thickness reaching the maximum value at the very end of charge. During discharge, despite the fact that volumetric expansion of the cathode is known to occur due to formation of insulating layer of Li2S, the cell contracts as a result of lithium stripping. Different discharge currents lead to different morphologies of Li2S precipitates on the cathode side. Neverthless, the resulting thickness changes on the cathode side are rather subtle and are hard to deconvolute from the complete cell response. Changes of the cycling conditions significantly affect the cell thickness evolution. Limiting the voltage cut-off limits resulted in less of a thickness variation and more stable capacity retention. This is due to avoiding shuttle mechanism at the end of charge and less of lithium plating, as the capacity of the cell is limited. Modification of discharge C-rate led to an interesting observation, where faster currents caused more stable thickness change, whereas slower C-rates caused progressive thickness increase. Continuous cycling at very slow rates caused rapid capacity fade and significant thickness increase. Post mortem analysis helped us to understand the reason for such behaviour and attribute it to large lithium expansion caused more likely by mossy-like growth of Li and precipitation of polysulfides active material during shuttle. This results highlight the importance of developing a Lithium protection layer to inhibit the shuttle phenomenon from occuring as well as help to control the mossy growth of plated Li. The correlation between the thickness and the voltage profile, and more importantly, the life time of the cell depending on the C-rate is quite remarkable. These results are therefore important for proving that the thickness can be an important marker for cell aging and State of Health (SoH) determintation. This can ultimately be considered as a useful parameter to be incorporated into Battery Management Systems (BMS) developed for these types of cells. The ultimate goal of our studies is to highlight that the volumetric expansion of a Li-S pouch cell is an important phenomena which can be linked to cell degradation and is significant in a larger format pouch cell. These preliminary results should encourage other researchers to explore the area of volume changes in a Li-S pouch cell in a more thorough fashion, where the effect of different parameters (such as wider operating voltage window, sulfur to electrolyte ratio, electrode balance, etc.) into cell volume evolution is studied.

Acknowledgements

15

This work was supported by Innovate UK (in UK) under the Revolutionary Electric Vehicle Battery (REVB) project (EP/L505298/1).

References

[1] P.G. Bruce, S.A. Freunberger, L.J. Hardwick, J.M. Tarascon, Nature materials, 11 (2011) 19-29.[2] X. Ji, L.F. Nazar, Journal of Materials Chemistry, 20 (2010) 9821.[3] L. Chen, L.L. Shaw, Journal of Power Sources, 267 (2014) 770-783.[4] www.oxisenergy.com (last visited on 2017/03/01)[5] www.sionpower.com (last visited on 2017/03/01)[6] M. Wild, L. O'Neill, T. Zhang, R. Purkayastha, G. Minton, M. Marinescu, G.J. Offer, Energy Environ. Sci., 8 (2015) 3477-3494.[7] R. Xu, I. Belharouak, X. Zhang, R. Chamoun, C. Yu, Y. Ren, A. Nie, R. Shahbazian-Yassar, J. Lu, J.C. Li, K. Amine, ACS applied materials & interfaces, 6 (2014) 21938-21945.[8] C. Barchasz, F. Molton, C. Duboc, J.C. Lepretre, S. Patoux, F. Alloin, Analytical chemistry, 84 (2012) 3973-3980.[9] R. Xu, J. Lu, K. Amine, Advanced Energy Materials, 5 (2015) 1500408.[10] S.S. Zhang, Journal of Power Sources, 231 (2013) 153-162.[11] M.K. Song, E.J. Cairns, Y. Zhang, Nanoscale, 5 (2013) 2186-2204.[12] L.N. S. Evers, Accounts of Chemical Research, 46 (2012) 1135-1143.[13] L. Zhu, W. Zhu, X.-B. Cheng, J.-Q. Huang, H.-J. Peng, S.-H. Yang, Q. Zhang, Carbon, 75 (2014) 161-168.[14] S.S. Zhang, D.T. Tran, Journal of Materials Chemistry A, 2 (2014) 7383.[15] Y.X. Ren, T.S. Zhao, M. Liu, P. Tan, Y.K. Zeng, Journal of Power Sources, 336 (2016) 115-125.[16] A.F. Hofmann, D.N. Fronczek, W.G. Bessler, Journal of Power Sources, 259 (2014) 300-310.[17] T. Zhang, M. Marinescu, S. Walus, G.J. Offer, Electrochimica Acta, 219 (2016) 502-508.[18] S. Waluś, C. Barchasz, R. Bouchet, J.-C. Leprêtre, J.-F. Colin, J.-F. Martin, E. Elkaïm, C. Baehtz, F. Alloin, Advanced Energy Materials, 5 (2015) 1500165.[19] S.Y. Lang, Y. Shi, Y.G. Guo, D. Wang, R. Wen, L.J. Wan, Angewandte Chemie, 55 (2016) 15835-15839.[20] M.U. Patel, R. Demir-Cakan, M. Morcrette, J.M. Tarascon, M. Gaberscek, R. Dominko, ChemSusChem, 6 (2013) 1177-1181.[21] C. Li, H. Zhang, L. Otaegui, G. Singh, M. Armand, L.M. Rodriguez-Martinez, Journal of Power Sources, 326 (2016) 1-5.[22] S.-H. Kang, X. Zhao, J. Manuel, H.-J. Ahn, K.-W. Kim, K.-K. Cho, J.-H. Ahn, Physica Status Solidi (a), 211 (2014) 1895-1899.[23] S. Urbonaite, P. Novák, Journal of Power Sources, 249 (2014) 497-502.[24] M. Hagen, P. Fanz, J. Tübke, Journal of Power Sources, 264 (2014) 30-34.[25] S. Urbonaite, T. Poux, P. Novák, Advanced Energy Materials, 5 (2015) 1500118.[26] M. Hagen, D. Hanselmann, K. Ahlbrecht, R. Maça, D. Gerber, J. Tübke, Advanced Energy Materials, 5 (2015) 1401986.[27] J. Luo, C.Y. Dai, Z. Wang, K. Liu, W.G. Mao, D.N. Fang, X. Chen, Measurement, 94 (2016) 759-770.[28] P.K. Leung, C. Moreno, I. Masters, S. Hazra, B. Conde, M.R. Mohamed, R.J. Dashwood, R. Bhagat, Journal of Power Sources, 271 (2014) 82-86.[29] K.-Y. Oh, J.B. Siegel, L. Secondo, S.U. Kim, N.A. Samad, J. Qin, D. Anderson, K. Garikipati, A. Knobloch, B.I. Epureanu, C.W. Monroe, A. Stefanopoulou, Journal of Power Sources, 267 (2014) 197-202.[30] X. Wang, Y. Sone, G. Segami, H. Naito, C. Yamada, K. Kibe, Journal of The Electrochemical Society, 154 (2007) A14.[31] E. Gibellini, C. Lanciotti, R. Giovanardi, M. Bononi, G. Davolio, A. Marchetti, C. Fontanesi, Journal of The Electrochemical Society, 163 (2016) A2304-A2311.[32] T. Bond, J. Zhou, J. Cutler, Journal of The Electrochemical Society, 164 (2016) A6158-A6162.[33] M. Lécuyer, J. Gaubicher, M. Deschamps, B. Lestriez, T. Brousse, D. Guyomard, Journal of Power Sources, 241 (2013) 249-254.[34] X. He, J. Ren, L. Wang, W. Pu, C. Jiang, C. Wan, Journal of Power Sources, 190 (2009) 154-156.[35] P. Barai, A. Mistry, P.P. Mukherjee, Extreme Mechanics Letters, 9 (2016) 359-370.[36] L. Zielke, C. Barchasz, S. Walus, F. Alloin, J.C. Lepretre, A. Spettl, V. Schmidt, A. Hilger, I. Manke, J. Banhart, R. Zengerle, S. Thiele, Scientific reports, 5 (2015) 10921.

16

[37] R. Elazari, G. Salitra, Y. Talyosef, J. Grinblat, C. Scordilis-Kelley, A. Xiao, J. Affinito, D. Aurbach, Journal of The Electrochemical Society, 157 (2010) A1131.[38] S.-E. Cheon, K.-S. Ko, J.-H. Cho, S.-W. Kim, E.-Y. Chin, H.-T. Kim, Journal of The Electrochemical Society, 150 (2003) A796.[39] R. Demir-Cakan, Journal of Power Sources, 282 (2015) 437-443.[40] Z. Feng, C. Kim, A. Vijh, M. Armand, K.H. Bevan, K. Zaghib, Journal of Power Sources, 272 (2014) 518-521.[41] A. Paolella, W. Zhu, H. Marceau, C.-s. Kim, Z. Feng, D. Liu, C. Gagnon, J. Trottier, G. Abdelbast, P. Hovington, A. Vijh, G.P. Demopoulos, M. Armand, K. Zaghib, Journal of Power Sources, 325 (2016) 641-645.[42] F.Y. Fan, W.C. Carter, Y.M. Chiang, Advanced materials, 27 (2015) 5203-5209.[43] S. Walus, C. Barchasz, J.F. Colin, J.F. Martin, E. Elkaim, J.C. Lepretre, F. Alloin, Chemical Communications, 49 (2013) 7899-7901.[44] J. Zheng, M. Gu, C. Wang, P. Zuo, P.K. Koech, J.G. Zhang, J. Liu, J. Xiao, Journal of the Electrochemical Society, 160 (2013) A1992-A1996.[45] L. Gireaud, S. Grugeon, S. Laruelle, B. Yrieix, J.M. Tarascon, Electrochemistry Communications, 8 (2006) 1639-1649.[46] P. Bai, J. Li, F.R. Brushett, M.Z. Bazant, Energy Environ. Sci., 9 (2016) 3221-3229.[47] R. Cao, W. Xu, D. Lv, J. Xiao, J.-G. Zhang, Advanced Energy Materials, 5 (2015) 1402273.[48] B. Bitzer, A. Gruhle, Journal of Power Sources, 262 (2014) 297-302.[49] Y.V. Mikhaylik, J.R. Akridge, Journal of The Electrochemical Society, 151 (2004) A1969.[50] M.R. Busche, P. Adelhelm, H. Sommer, H. Schneider, K. Leitner, J. Janek, Journal of Power Sources, 259 (2014) 289-299.[51] W. Li, H. Yao, K. Yan, G. Zheng, Z. Liang, Y.M. Chiang, Y. Cui, Nature communications, 6 (2015) 7436.[52] A. Jozwiuk, B.B. Berkes, T. Weiß, H. Sommer, J. Janek, T. Brezesinski, Energy Environ. Sci., 9 (2016) 2603-2608.[53] Y.S. Su, Y. Fu, T. Cochell, A. Manthiram, Nature communications, 4 (2013) 2985.[54] C. Barchasz, J.-C. Leprêtre, F. Alloin, S. Patoux, Journal of Power Sources, 199 (2012) 322-330.[55] Y. Diao, K. Xie, S. Xiong, X. Hong, Journal of Power Sources, 235 (2013) 181-186.[56] C.Y. X.-B. Cheng, J.-Q. Huang, P. Li, L. Zhu, L. Zhao, Y. Zhang, W. Zhu, S.-T. Yang, Q. Zhang, Energy Storage Materials, 6 (2017) 18-25.

17

Graphical abstract

Figure 1. A 11 Ah OXIS Energy Li-S cell and schematic representation of a cross-sectional view of the cell components.

Figure 2. Thickness measurement set up with Keyence laser gauge.

Figure 3. Capacity retention (a) and voltage profile (b) of a 11 Ah cell cycled at standard cycling conditions: 0.1 C charge to 2.45 V and 0.2 C discharge to 1.5 V. Zoomed in image on the end of charge voltage curves (c).

18

Figure 4. Cell thickness and voltage variation (a) of a 11 Ah cell cycled during 10 preconditioning cycles (see Table 2, Cell #1), with zoomed in image on the 9th cycle (b,c).

Figure 5. Capacity retention (a) of a 11 Ah cell when switched from a full voltage range to a limited voltage window. Comparison of voltage profiles (b) obtained during the transition cycles.

19

Figure 6. Thickness and voltage variation as a function of time (a) of a 11 Ah cell presented in Error: Reference source not found, with the zoomed in image on the thickness/voltage difference when changing the voltage cut-off limits (b).

Figure 7. Capacity retention (a) of a 21 Ah cell submitted to a C-rate screening protocol (Table 2, Cell#4), with capacity fade rate (%) for each discharge C-rate marked. Representative discharge voltage profiles at different C-rates (b). The voltage curves were aligned/shifted to the voltage dip at the beginning of the lower discharge plateau, for easier comparison of the length of this plateau. Subsequent charge profiles obtained at 0.1 C (c), with the zoomed in image on the overpotential kick at the beginning of charge.

20

Figure 8. Thickness and voltage profiles (a) of a 21 Ah cell cycled at varying discharge currents and constant 0.1 C charge (Table 2; Cell#5). Blue and red spheres mark cell thickness at the end of charge and discharge, respectively. Discharge capacity and end-of-discharge-thickness as a function of cycle number (b).

21

(a)

(b) (c)

(a)

Figure 9. Voltage and thickness profiles (a) of a 21 Ah cell cycled in a slow current regime of 0.106 C for both charge and discharge cycles. End of charge and discharge thickness values together with capacity retention as a function of cycle number (b).

Figure 10. Zoomed in images of the thickness change and voltage profile of a Cell#6 (from Figure 9) at different states of health, which corresponds to cycles: 3 (a), 18 (b), 36 (c).

22

(a)

(b)

Figure 11. SEM pictures of cathodes (a,b) and anodes (c,d) from a fully discharged single layer cells, cycled in different current regimes, as indicated on the figures: 0.1 C and 0.5 C. Cross section images of Li anodes (e,f).

Figure 12. Schematic illustration of a Li-S cell thickness increase when cycled at slow rates (≤ 0.1 C), at different state of health, with respect to the thickness change of the individual cell components: anode (blue arrow), cathode, as well as complete cell (black arrow). Separator was not included in this schematic for clarity. Zoomed in image on Li plating (f) and stripping (g) occurring in aged cells.

23