Review—Post-Mortem Analysis of Aged Lithium-Ion Batteries: Disassembly Methodology and Physico-Chemical Analysis Techniques

Improvement of life-time is an important issue in the development of Li-ion batteries. Aging mechanisms limiting the life-time can efﬁciently be characterized by physico-chemical analysis of aged cells with a variety of complementary methods. This study reviewsthestate-of-the-artliteratureonPost-MortemanalysisofLi-ioncells,includingdisassemblymethodologyaswellasphysico-chemicalcharacterizationmethodsforbatterymaterials.AdetailedschemeforPost-Mortemanalysisisdeducedfromliterature,includingpre-inspection,conditionsandsafeenvironmentfordisassemblyofcells,aswellasseparationandpost-processingofcomponents.Specialattentionispaidtothecharacterizationofagedmaterialsincludinganodes,cathodes,separators,andelectrolyte.Morespeciﬁcally,microscopy,chemicalmethodssensitivetoelectrodesurfacesortoelectrodebulk,andelectrolyteanalysisare reviewed in detail. The techniques are complemented by electrochemical measurements using reconstruction methods for electrodes built into half and full cells with reference electrode. The changes happening to the materials during aging as well as abilities of the reviewed analysis methods to observe them are critically discussed. determined their LiF concentrations. By with the capacity fade of the cycled cell, the authors were able to propose several reaction paths for the different aging mechanisms of the examined electrodes. proposed thermal decomposition mecha- nisms due to autocatalysis and protic impurities for several carbonate solvents frequently used in state-of-the-art electrolytes like 220,221 Their study was a combination of NMR methods ( 1 H, F, 31 COSY and GC-MS and SEC and was done on a model system, which did not contain material collected an aged Li-ion cell. in addition to protic impurities

Li-ion batteries are currently used in everyday objects such as smart-phones, power tools and tablet computers as well as in the growing fields of light electric vehicles (LEVs), unmanned aerial vehicles (UAVs), battery electric vehicles (BEVs), hybrid electric vehicles (HEVs), and plug-in hybrid electric vehicles (PHEVs). [1][2][3][4] Furthermore, the rise of renewable energy sources like wind and solar power, which are only intermittently available, demands reliable and highly flexible stationary energy storage solutions, which provide high capacities and predictable life-times. 2,5 Aging of Li-ion batteries is a general problem for manufacturers as they have to guarantee long-term reliability of their products. For state-of-the-art cells, degradation effects on the material level lead to capacity fade and resistance increase on the cell level.  The aging state of a battery is often characterized by the state-of-health (SOH) in % according to 3,16,22,[29][30][31] SO H(t) = discharge capacity (t) discharge capacity (t = 0) [1] where t represents the aging time. In general, one has to differentiate between cycling 7,16,18,21,[23][24][25]32 and calendar aging. 7,19,[21][22][23][24]27 Since commercial Li-ion cells can be subject to calendar aging in the time between manufacturing and delivery, it is good practice to measure the discharge capacity at t = 0 for each cell that undergoes an aging test. Since the discharge capacity depends mainly on temperature, depth-of-discharge (DOD), and discharge current, the SOH is usually monitored by regular check-ups with defined parameter sets, 7,16,21,23,24 which can vary depending on the application. Typically, a temperature of 25 • C, 16,22,24 DOD of 100%, 16,21 and discharge rates of 1C 7,16,21,22,24 or lower 23 are used in check-ups. The performance decrease on the cell level is mainly based on chemical degradation reactions on the material and on the electrode level (see Figure 9). 3,9,[15][16][17]25,28,[33][34][35][36][37][38][39][40][41] In this sense, a thorough understanding of the degradation mechanisms happening inside the cells is crucial to improve their life-time.
In order to conclude on aging mechanisms, it is mandatory to disassemble the cells and analyze the relevant cell components. For a deep z E-mail: waldmann@zsw-bw.de understanding of a battery's aging process, a homogenized procedure including cell opening, disassembly, sample processing, and analysis is important to avoid damage, contamination, and modification of the cell components and to yield interpretable data.
However, as displayed on the housing of almost every commercial Li-ion cell, disassembly is not recommended by the manufacturers. This is due to safety hazards, e.g. the possibility of creating shortcircuits during cell opening which can lead to thermal runaway of a cell. Furthermore, there are severe health issues that arise from the chemical compounds and the risk of damaging the samples by inappropriate treatment. 30 However, if certain protocols are followed, the disassembly of Li-ion cells is safe and gives reliable results on the composition of the built-in materials and changes during aging.
In 2011, Williard et al. presented a methodology for analysis of failed Li-ion batteries, e.g. after thermal runaway. 30 However, to best of our knowledge, no standard method is available for the disassembly and analysis of aged Li-ion cells, although until now, much research has been conducted involving disassembly of aged batteries without failure. 12,16,17,25,26,28,32,[42][43][44][45][46][47] In this paper, we review the state-of-the-art methods for the disassembly of aged Li-ion cells as well as the physico-chemical methods for analysis of materials from disassembled cells. For each method, the revealed aging mechanisms and the observable changes on the material level occurring during aging are discussed. Special attention is drawn to the question which changes can be observed with specific analysis methods. Finally, we conclude on combinations of methods to get a full understanding of aging processes. Figure 1. Before disassembly of cells, nondestructive characterization methods are useful to gain first insights into aging mechanisms. Additionally to capacity tests (see Equation 1), incremental capacity analysis (ICA) 48,49 and Electrochemical impedance spectroscopy (EIS) are powerful techniques to obtain information on aging mechanisms. 37,50-52 ICA is based on dQ/dV vs. V plots and therefore transforms voltage plateaus and inflections points in voltage curves into dQ/dV peaks. 48 The changes in the dQ/dV peaks (peak intensities and peak shifts) can be monitored during aging and allow conclusions on loss of active material/loss of electrical contact, changes in cell chemistry, underdischarge, undercharge, 48 and stripping of Li plating. 53 Li stripping was also determined by differential voltage analysis (DVA) in dV/dQ vs. Q plots. 53 EIS is another non-destructive method to characterize aged cells. 25,34,37,39,[50][51][52] During aging, the cell impedance is typically increasing, leading to slower kinetics which are partly the cause for capacity fade. 34,37 The reason for the impedance increase are physicochemical processes inside the cells, such as increase of resistive layers. 25,39,50 Klett et al. found significant differences in the Bode plot of calendar and cycling aged cells. 39 The main reason for this difference was found to be a more pronounced film on the anode surface for cycle aging. 39 However, cell impedance is influenced by many factors, requiring modelling. 50 A more straightforward and fast method to obtain very basic information on impedance changes of cells are measurements at only one frequency, typically 1 kHz. 28 Such measurements allowed finding direct correlations between impedance rise during aging and the increase of Mn, P, and Li on graphite anodes by Post-Mortem analysis. 28 Although non-invasive electrochemical methods are powerful tools to get information on aging mechanisms, a direct observation of chem-ical changes is only possible by Post-Mortem analysis. Furthermore, localized aging phenomena, representing only small fractions of the electrodes, are often not visible in electrochemical measurements, since they are averaging over the entire electrodes of a cell.

Pre-inspection and non-destructive methods before opening of Li-ion cells.-An overview of the individual steps in Post-Mortem analysis is given in
After electrochemical characterization, visual inspection, pictorial documentation, and weighing are the next reasonable steps in the analysis of aged Li-ion cells. It can indicate external deformation or leakage, which may influence the aging behavior or lead to cell failure. Furthermore, these steps can give first hints on the best position for cell opening. While for standard cell designs -such as 18650 or 26650 cells -the cutting positions are similar in most cases (∼1 mm near the positive or negative connector), it might be necessary to conduct additional tests for other geometries like prismatic and pouch cells.
Non-destructive methods that reveal the interior of batteries are X-ray analysis, 30,54-56 X-ray computed tomography (CT), 9,42,[54][55][56][57][58][59][60][61][62][63][64][65] and neutron tomography. 66,67 Since X-ray analysis yields 2D transmission images (Figure 2a), depending on the cell design, it can be necessary to conduct X-ray transmission measurements in several viewing angles. 55 In contrast, CT data are acquired by rotating a cell in small angular steps, while X-ray images are recorded for every angle. From this dataset, a 3D model of the cell is determined by a mathematical algorithm that allows calculating axial and frontal 2D sections at defined positions (see Figure 2b and Figure 3a). Thus, CT is an expensive method and usually requires longer measurement time if compared to X-ray transmission measurements. Furthermore, CT results in higher amounts of data and higher workload for interpreting this data. However, CT is able to reveal many details on the battery's internals, such as deformations in the interior of cells after aging, 9,32,42 stress, 64 failure, [54][55][56]58 or abuse tests. 57,59,60 In the case of internal deformations, CT is very useful to image their shape without application of a mechanical force that could alter the cell 32,42,64 (see Figure 3a). Both, X-ray analysis and CT imaging are appropriate to determine cutting positions for cell opening as shown in Figure 2.
Techniques involving neutrons are also suitable to obtain information on the macroscopic design inside Li-ion cells 66 and can even deliver chemical information 66,68 in a non-destructive way. However, due to the very high effort of this method, neutron tomography is not practical to determine cutting positions for cell disassembly. Also it has to be kept in mind that the sample may be radioactive after treatment with neutrons. is desirable, since it lowers the energy content of the cell. In case of an unwanted creation of a short circuit, deep discharge will decrease the risk of thermal runaway.
On the other hand, the cell voltage must not leave the normal operating window in order to avoid undesired material changes which are not caused by aging. Therefore, most authors discharge the cells to the end-of-discharge voltage prior to disassembly corresponding to SOC = 0%. 12,16,17,28,30,34,46,69,70,72 A defined SOC is also important for comparability of the results of different cells, e.g. aged and fresh cells of the same type. The exact discharge procedure prior to disassembly is unfortunately not provided by most authors. Kobayashi et al. mentioned that the open circuit voltage (OCV) of aged cells discharged to 2.5 V at C/20 was larger than the OCV of fresh cells due to the increase of internal cell resistance. 12 Hence, the authors held all cells at 3.0 V for over 10 h before disassembly, leading to an OCV of 3.0 V ± 0.01 V. 12 A similar discharge method was used by Takahara et al. 26 Kumaresan et al. discharged pouch cells in two stages, first with C/33 and after a rest period of 30 min with C/83 to ensure a complete discharge. 73 Disassembly of cells at higher SOCs has been carried out for Tcells, 70 whose capacity is very low (∼0.2 mAh) and therefore the risk is small compared to commercial cells (few Ah). Burns et al. recently opened commercial 0.22 Ah pouch cells at ∼50% SOC and found Li plating after cycling with high currents. 45 The same authors opened also 3.4 Ah 18650-type cells after discharging to 0 V for safety reasons 45 due to their higher capacity. Hence, Li plating was not directly visible anymore (but clear differences in color and texture of negative electrode), although it was expected from Coulometry measurements. 45 This discrepancy was attributed to the deep discharge to 0 V. 45 Since some components of Li-ion cells react with O 2 and H 2 O, a glove box filled with highly pure Ar atmosphere containing H 2 O and O 2 only in the lower ppm range has to be utilized. 30,32,34,[45][46][47][69][70][71][73][74][75][76][77][78] Especially Li x C 6 , metallic Li, and LiPF 6 show reactivity with the components of air. LiPF 6 reacts with water to form HF gas, 30,36,79 which can cause significant health problems without appropriate protection gear 30 as well as corrosion of cathode materials. 36 We note that usage of N 2 as inert gas is not suitable, due to its reactivity with metallic Li to form Li 3 N. 80 In their paper from 2002, Aurbach et al. used an Ar-filled glove box with an O 2 content ranging from 5 to 10 ppm and a H 2 O content from 2 to 5 ppm. 34  In some cases, it is less important to protect samples from air. 30 Examples are XRD or ICP-OES measurements of washed cathode materials. The authors recommended a fume hood with a draw of 60-100 feet per minute as a minimum requirement for the disassembly of small commercial cells after cycling under non-abusive conditions. 30 Consequently, Amanieu et al. opened 18650 cells inside an Ar-filled glove box for safety reasons, however, after removing the electrolyte with DMC, the LiMn 2 O 4 samples were dried in a steady air stream of a fume hood overnight, since the samples were not air sensitive. 74 We note that the safety during disassembly of cells on air depends also on the humidity. Cell opening in humid air is also critical and might lead to critical conditions resulting in lab fires.
In any case, aged electrodes that are used in order to obtain reassembled cells (see Electrochemical analysis of reassembled electrodes section) should be kept inside a glove box 12,34,73,81 before they are inside a sealed cell. Kostecki et al. conducted cell opening and washing in an Ar-filled glove box and stored the electrode samples in an airtight cell in the glove box before further investigations. 69 We note that electrode samples that are in contact with electrolyte deteriorate even being sealed airtight, thus we recommend using the electrodes for further electrochemical tests on the day of disassembly.
Hightower et al. used a special protection by covering Li x C 6 samples with an inert liquid (Fluorinert FC-43) inside an Ar-filled glove box before transfer through air into the vacuum chamber of a TEM device where the inert liquid evaporated during evacuation of the chamber. 82 We conclude at this point that disassembly of Li-ion cells should be performed in a chemically inert environment, such as an Ar-filled glove-box. Even if cells are discharged to the end-of-discharge voltage, disassembly of aged Li-ion cells still has to be done with great caution. The procedure and therefore the costs of a cell disassembly critically depend on the risks for the operator and the sensitivity of the materials to air and moisture.

Li-ion cell opening procedure and separation of components.-
External short circuits can occur by unintentional contact of the external tabs, e.g. with conductive tools, by a metal flake during cutting, or by contact with the metallic surface of the glove box. Depending on the specific cell design, the cell case can be connected either to the positive or negative tab. This can be easily determined using a voltmeter prior to disassembly.
Furthermore, during cell opening, care must be taken to prevent internal short circuits of the cell, 30,32,54,74 as well as of the samples. 30 Internal short circuits are most probable during cutting of the cell housing, either by penetration or deformation of the electrode stack / jelly roll or by mechanical pressure. Therefore, an ideal cutting position has to be determined for each cell type by applying nondestructive techniques, as shown above in the Pre-inspection and non-destructive methods before opening of Li-ion cells section. Furthermore, it is advantageous to use non-conductive tools, e.g. made from ceramics or with non-conductive coatings.
Aurbach et al. presented a special device for opening of 18650 cells that can be operated inside a glove box. 34 In this device, a cylindrical cell is turned by a remote-controlled motor while the cap of the cell case is cut by a carbide-tipped saw. 34 As illustrated in Figure 4a, a Dremel tool could also be used for opening cells. Once the cap of the cell is removed (Figure 4b) the tabs connected to the casing have to be cut. Then, the bottom of the cell can be cut and finally the casing is sawn along the cylinder axis and the jelly roll is unrolled (Figure 4c).  Figures 4d-4f, since the pouch foil can simply be cut by ceramic scissors 30 or a knife. In the case of prismatic cells, it was suggested to make a shallow cut by a cutting tool on one side of the cell prior to peeling the remaining casing off by using isolated pliers. 30 In any case, the cell opening must be carried out very carefully and excessive force on the jelly roll or electrode stack must be avoided.
Formation of metal dust or swarfs depends on the cutting method. Dust may get into the cell and contaminate the materials, 30 whereas swarfs can get several mm long and can create short circuits leading to an unwanted discharge of the cell and generation of heat. Furthermore, it has to be considered that local heating is also produced during cutting, which may cause changes of the cell materials or even result in safety issues.
In most cases, the cell components will be separated from each other in order to analyze them separately (see Figures 4c, 4f). For aged anodes, it can occur that active material is sticking to the separator, 32 leading to problems in separation of components. This might be solved by dipping anode and separator into DMC. In contrast, for aged cathodes this is often less problematic.
Typical configurations of cells are wound jelly rolls in cylindrical cells, flat wound jelly rolls in prismatic and pouch cells, as well as stacked electrodes/separators, z-folded separators, or combinations of stacking and winding in pouch and prismatic cells. Unfortunately, most authors do not comment on this step of cell disassembly. We note that special care has to be taken to avoid cross-contamination by contact between anode and cathode. If electrolyte is present, direct contact of anode and cathode creates a short circuit resulting in consequences as discussed above.
Right from the start of opening, electrolyte drops can be recovered if contained in sufficient excess. 83 Otherwise, electrolyte should be sampled by immersing the jelly roll immediately after case removal in CH 2 Cl 2 84 or the separated wetted components in acetonitrile. 85 This last method allows extracting electrolyte as well as compounds produced by its decomposition upon aging at each electrode. As many solvents of electrolyte are highly volatile, fast recovery of electrolyte is advised to keep the composition unchanged.

Post-processing of samples from disassembled Li-ion cells.-
After separation of cell components, most experimenters wash these components with typical electrolyte solvents, such as DMC, 12 69 whereas only some authors did not perform washing of their samples. 34,45,53,65,78,92 This is possible, when only visual inspection 45,53 and/or electrochemical testing are required. 46,65 Unwashed electrodes are likely to contain residual crystallized LiPF 6 or non-volatile solvents that can hardly be distinguished from elements in the SEI or intercalated Li. Furthermore, a washing step is also useful to reduce corrosion of the samples since LiPF 6 reacts with H 2 O and O 2 and as already mentioned to protect sensitive analytical equipment if samples are exposed to air. Somerville et al. showed that washing is not necessary to remove EC and other typical carbonates when the samples are brought into vacuum (∼10 −4 kPa), e.g. in vacuum-based devices such as XPS or SEM. 89 Unfortunately, most authors do not comment on the washing procedure (time, temperature, volume, type of solvent) 26 65 Williard et al. commented that washing may lead to the absence of particular SEI components. 30 Abraham et al. demonstrated that DMC rinsing should remove insulating species deposited at the graphite surface after aging. 93 Recently, Somerville et al. investigated this topic in detail for graphite anodes with films formed by different amounts of VC additive. 89 Depending on the amount of VC in the electrolyte and therefore on the composition of the film, it was also found that the SEI can be altered at least partly by washing with DMC. 89 For one particular case, LiPF 6 and LiF were completely removed and LiP x F y species were reduced after 1 min. 89 According to their study, the washing duration and/or washing or no washing should be tested for each cell chemistry. 89 From our experience, two washing steps in the order of 1 min to 2 min with pure solvent are required to remove traces of Li salt from samples. Furthermore, it is important to perform the washing steps always in the same way to get comparable results.
Some techniques such as ICP-OES analysis 16,28 use active material scraped off from the electrodes. Such mechanical treatment will not change the chemical composition and is therefore not problematic. XRD is possible with both, electrodes and scraped off powder-like material, however, it has to be considered that preferred particle orientations in the electrodes which are not present in scrapped off material can lead to differences in the peak intensities. 34 If cracks in the active material are investigated, it is possible to prepare cross-sections of a whole Li-ion cell (see Figure 3b). In this case, the casing of the cell is not removed. Instead, a cut with a non-conductive saw blade is performed through the whole cell. The position of the cut might be determined by CT scanning beforehand (see Figure 3a). After cutting a cell, the electrolyte is removed, followed by a stabilization via epoxy and a metallographic polishing step. 30  Compared to CT imaging, cell cross-sections are more costly in terms of labor and lead to destruction of a cell. However, cell cross-sections can provide a significantly higher resolution for specific parts of the cell (compare Figures 3a and 3b) as well as the possibility to perform measurements with other powerful methods such as focused ion beam (FIB) cutting 42,74 and observation by optical microscopy, 18,19,30,42,74,94,95 SEM, 30,32,42,74,[95][96][97] or EDX. 95,97 Crosssections of complete cells provide the thickness of electrodes in working condition (of the respective charge level), i.e. with a similar pressure as in the closed cell. We note that this is not the case for crosssections of single electrodes, 18,19,74,96,97 which might have expanded after separation of the cell components.

Physico-Chemical Analysis of Aged Materials after Disassembly of Li-ion Cells
In this section, methods for physico-chemical characterization of battery materials are reviewed. Evidence on aging mechanisms obtained by the respective methods is discussed in order to give an overview of the possibilities for observing specific degradation mechanisms. Figure 5 shows a scheme of the main cell components and corresponding available physico-chemical analysis methods to characterize them. Samples can originate from anode, cathode, separator, current collector, or electrolyte, however, only a cathode is shown as an example in Figure 5 to simplify matters. It can be seen from Figure 5 that different parts of a solid sample can be distinguished: electrode surface, bulk, cross-sections, and different analysis methods can be assigned to them, respectively.
The reasons for surface sensitivity of analysis methods are related to the physical nature of the involved types of radiation or particles. A simplified overview of physical/chemical principles (irradiation by and / or detection of electrons e − , electromagnetic radiation / photons hv, neutral particles, and ions) is shown in Figure 6. These are briefly explained for each method in the sections below. For more elaborate details on the excitation and detection mechanisms of the respective analysis methods, we refer to text books. [98][99][100][101][102] Surface sensitivity is created either by reflection of radiation/particles on the sample surface (e.g. when an electrode is examined by optical microscopy) or due to short mean free paths of particles inside solid samples (e.g. methods involving e − or ions). In contrast, other methods are not surface sensitive and include information from the electrode bulk. In this case, material from the sample has to be scraped off, e.g. in ICP-OES analysis, or the radiation which is detected is not hindered by the sample (e.g. X-rays in case of XRD).
Typical Post-Mortem analysis methods for cell components and the aging mechanisms they revealed are discussed separately in the following section. However, due to the sensitivity of the methods to the different parts of the samples mentioned above, an overview on the capabilities for each method is given in the Combination of methods for full characterization of aging mechanisms section.
Microscopy.-Optical microscopy.-Optical microscopy is based on the reflectance of visible light from a sample surface (Figure 6a). In general, the resolution of optical microscopes is limited by Abbe's diffraction limit, corresponding to a range of 0.2 μm. 99 This enables resolving particles in the μm range with low effort compared to electron microscopy techniques. 18,19,30,42,74,95 Therefore, it is possible to detect aging effects such as changes in electrode thickness 18,19 or depositions on electrode surfaces which are in the μm size range. 18,78 Due to the limited resolution of optical microscopy, the detection of particle cracks or of very thin films is difficult or might not be observable. However, optical microscopy is a very effective method to get an overview of the surface of a sample.
Brand et al. observed scorching of a separator via optical microscopy after shaking tests of 18650 cells. 64 Some groups studied Li deposition and dendrite formation during the charging process in situ by optical microscopy. 96,[103][104][105][106][107][108][109] Furthermore, color changes in graphite 96 and rutile 110 electrodes were monitored in situ by optical microscopes.
In the case of new electrode materials under development, optical microscopy has also proven to be helpful. Pharr et al. were able to use optical microscopy techniques to determine the fracture energy of lithiated Si thin-film electrodes as a function of Li concentration. 111 Li and Fedkiw were successful in studying the effect of silica nanoparticles added to gel electrolytes to prevent Al current collector corrosion. 112 These are only a few exclusive examples to demonstrate the use of optical microscopy in Post-Mortem analyses of Li-ion cells, however, they exemplify the wide range of possibilities of this characterization technique being underestimated quite often.
Scanning electron microscopy.-Scanning electron microscopy (SEM) provides improved resolution compared to optical microscopy due to the smaller de Broglie wave lengths of electrons compared to visible light. The resolution of SEM is mainly limited by spherical aberration of the electron lenses. 99 Furthermore, the fact that electrons are used instead of photons (Figure 6b) has to be considered in the interpretation of SEM images. The image contrast depends strongly on the selected detector, which collects either backscattered or secondary electrons. 99 Furthermore, SEM observations are restricted to vacuum, leading to evaporation of volatile components, such as carbonate solvents.
Due to the higher resolution of SEM, the observed areas can be much smaller compared to optical microscopy. Therefore, much care must be taken to record data which is representative for the whole sample. This is usually done by first recording overview images and then zooming into different parts of the sample. Different cell components are commonly observed by SEM as it offers fundamental information on the microstructure, which can be related to degradation mechanisms.
Furthermore, SEM is limited to observations of the sample's surface. To gain information on the bulk and / or the chemical composition, SEM is usually complemented by other methods. For example, SEM is often combined with EDX analysis in order to determine the chemical composition and/or combined with cross-sectioning techniques such as metallographic preparation, 30  cutting, 39,42,74,[114][115][116][117][118][119] or ion milling. 120 Furthermore, removal of thin slices by FIB and subsequent SEM imaging, allows creating videos 119 and construction of 3D models of electrodes [116][117][118]121 (FIB/SEM tomography). Such 3D models of electrodes were useful in multi-scale calculations where the microstructure of electrodes was considered. 118,121 On the negative side, graphite is the most common anode material and in combination with other techniques, SEM has brought significant results to identify the degradation mechanisms occurring on the surface of this material.
The growth of the solid electrolyte interface (SEI) on the surface of graphite particles during aging were observed by SEM 16,18,40 (see upper part of Figure 7). The growth of the SEI during aging is related to electrolyte decomposition and is the reason for Li loss and therefore of a capacity drop. 16,18,28,33,122 Another aging mechanism is deposition of metallic Li on graphite anodes. Honbo et al. studied Li deposition on graphite via SEM and revealed dendritic and granular morphologies on pristine and grinded carbon, respectively. 123 Zier et al. showed that it is possible to enhance the material contrast of Li deposition on graphite electrodes by reaction with OsO 4 . 119 The study of other anode materials such as Li 4 Ti 5 O 12 by SEM has so far not given precise information on the degradation mechanisms. 124 On the cathode side, often no changes are visible by SEM imaging between pristine and aged cathodes 16,40,125 (see lower part of Figure  7). When a visible surface film was reported after prolonged cycling on top of LiCoO 2 , it has not been possible to relate it with a clear degradation mechanism. 34 On the other hand, mechanical stress [126][127][128] due to volume changes during cycling leads to cracks in the particles which are observable by SEM. 42,74,75,129,130 Additionally to aging mechanisms involving the electrode materials, degradation of other cell components such as corrosion of Al current collectors [131][132][133] and pore-closure 14,134,135 or melting 64 of separators are observable by SEM.
Transmission electron microscopy.-Compared to SEM, transmission electron microscopy (TEM) usually employs higher acceleration voltages for the electrons, allowing transmission through materials ( Figure 6c) and a higher resolution down to the atomic scale. 119,136,137 Therefore, TEM unveils sample characteristics in terms of particle morphology, crystallinity, stress or even magnetic domains. However, due to the higher energy, beam damage has to be taken into account for battery materials. 138 It has to be mentioned that TEM measurements are limited to localized areas of the sample and therefore it is difficult to survey a large sample accurately.
As for all microscopic methods, sample preparation and sample monitoring during data acquisition are crucial for TEM, e.g. FIB cutting has proven to be useful. Furthermore, the size (the thinner the better) and cleanliness of sample are of great importance. The higher effort in sample preparation makes TEM a more time consuming method compared to SEM. Several overviews in terms of experimental possibilities and comparison with other microscopic methods can be found in textbooks. 101,102 Structural changes in particle morphology resulting from calendar and cycling aging were investigated by Watanabe et al. for LiNi 0.8 Co 0.15 Al 0.05 O 2 cathode material. 139,140 TEM analysis has also given valuable insight regarding binder dependency on cell performance, 141 SEI formation on cathodes 142 and evaluation of new electrode materials. 110,[143][144][145][146] In this overview, only a few possibilities of TEM in Post-Mortem analysis have been mentioned, however, usage of this method is shifting from a Post-Mortem characterization method to an in situ and operando technique. 147 This trend is becoming much more obvious as instruments are providing low-Z-element-sensitive analytical instrumentation, sample environmental control, and high-speed and sensitive direct electron detectors are becoming more and more available. 148 Chemical analysis methods sensitive to electrode surfaces.-Energy dispersive X-ray spectroscopy.-Energy dispersive X-ray spectroscopy (EDX) is often combined with SEM devices. The EDX principle is based on irradiation of a sample with electrons and detection of the generated characteristic X-ray photons ( Figure 6d). Therefore, EDX allows gaining information on the chemical composition of the sample. Additionally, if the sample surface is scanned by the electron beam, maps of the chemical composition on the surface can be created by superposition with SEM images (EDX mapping). 39,95,97,119,149,150 However, EDX presents a major drawback since it is not able to detect Li. Therefore, complementary methods are necessary for its detection and quantification. 16,17,72,150 EDX analysis performed during Post-Mortem studies enables to verify the composition of the active materials and to detect the presence of additional phases. E.g., EDX has enabled to detect the redeposition of dissolved Mn on top of graphite electrode after dissolution from NMC/LiMn 2 O 4 blend cathodes. 16,28 Similarly, Klett et al. observed Fe on anodes after dissolution from LiFePO 4 cathodes. 17 It is also possible to measure the presence of F and P on anodes due to the degradation of the electrolyte. 16,17,28,151 For the analysis of such elements the sample preparation and electrode washing is very important as described in the Methods for cell opening section. However, one may pay attention that in some cases the presence of such elements can actually be part of the active material. 152 Krämer et al. modified anodes with Li deposition using isopropanol. 150 EDX mapping allowed to detect O and C suggesting the formation of Li 2 CO 3 , however, the authors had to perform further measurements using FTIR and XRD for verification. 150 The modification of the Li deposition with isopropanol allowed evaluat-ing the area on the anode surface covered with Li 2 CO 3 from EDX mapping. 150 Maleki et al. investigated the effects of deep discharge below the end-of-discharge voltage for commercial LiCoO 2 / graphite cells. 153 The authors found that discharging to 0 V can lead to dissolution of Cu from the negative current collector, which was accordingly detected on both anode and cathode by EDX. 153 Furthermore, EDX was used in combination with SEM to detect contaminations after failure of cells. 54 X-ray photoelectron spectroscopy.-X-ray photoelectron spectroscopy (XPS) is based on the photoelectric effect. 98,154 Atoms in the sample are ionized by X-rays and the kinetic energies of the emitted photoelectrons are measured (Figure 6e). 98 As the kinetic energy of an emitted photoelectron is characteristic for its originating element, XPS allows analysis and determination of all elements (except H and He), their oxidation states and -to a certain degree -their chemical environment. 98 XPS is surface sensitive due to the small mean free path of the emitted electrons in solid bodies (few nm). 98 Therefore, XPS is able to characterize chemical changes on particle surfaces making it valuable in terms of Post-Mortem analysis.
On the lab-scale, typically Al K-alpha X-ray sources are used. Furthermore, also synchrotron radiation can be employed to perform experiments with in the hard X-ray region (HAXPES), 17,155,156 however, this results in a much higher effort.
However, due to the high energy content of X-rays, the possibility of irradiative sample damage has to be taken into account. Especially SEI components might alter their chemical nature. Therefore, XPS data interpretation requires a high level of knowledge of the system under investigation. Furthermore, it has to be mentioned that XPS measurements are strongly localized, making it necessary to probe a larger sample at different areas to get an overview. XPS can be combined with ion sputtering to obtain depth profiles. The uneven surface of an electrode makes this exercise particularly difficult and special care is necessary for data analysis. However, in combination with sputtering, XPS is not able to measure a depth profile through the whole electrode sample. More specifically, XPS is limited to the first nanometers of the surface. Therefore, only the SEI layer can be observed and often the signal from the active material remains hidden.
A useful overview of possible catalytic reactions happening at the electrolyte graphite interface and their observation by XPS was recently given by Ross. 157 A general review about SEI analysis including XPS has been provided by Verma et al. 158 Commercial LiFePO 4 / graphite cells investigated by Klett et al. showed uneven aging of electrodes for cycled cells, whereas the electrodes of calendar aged cells were uniform 17 which is related to temperature 77,159-161 and pressure gradients evolving during cycling.
Lu et al. discussed aging focusing on LiCoO 2 / graphite cells. 162 The authors performed XPS surface and depth profiling analyses and observed an increase of the SEI thickness in aged cells. 162 A recent report by Ehrenberg's group deals with the formation of the SEI in commercial pouch cells. 163 Via XPS, the authors were able to identify the constituents of the outer and inner layers of the SEI, however, it was not possible to figure out whether there are any differences in SEI characteristics for different formation procedures. 163 Several groups gathered information about the SEI compositions on anode materials other than graphite, like SiO 164 or Sn. 47 Several authors reported on transition metals which were dissolved from the cathode, migrated through the electrolyte and precipitated or were included in the SEI layer of the aged anode. 165,166 Such behavior was also reported using complementary methods. 16,18,26,28 The relationship between different aging conditions and the chemical composition of the SEI was of interest for Zheng et al. 167 The authors studied the degradation of commercial LiFePO 4 / graphite cells during calendar aging over 10 months at different temperatures and SOCs. 167 For temperatures raised to 55 • C and high storage SOCs, they observed a significant increase in bulk resistance and chargetransfer resistance as well as capacity loss. 167 Post-Mortem XPS analysis confirmed that newly formed layers of Li 2 CO 3 and LiF on the anode surface were responsible for the changes in cell behavior. 167 In the development of new electrode materials, Post-Mortem XPS can help to identify unknown products of side-reactions taking place on surfaces. As pointed out by Feng et al., the main problem of Li-S cells is the accumulation of S species on the electrode surfaces as detected by XPS and the resulting capacity fade. 168 XPS is a versatile tool to obtain information on the chemical composition of surface species being formed during aging. In contrast to EDX, NMR, XRD, and IR spectroscopy, nearly all elements can be semi-quantitatively detected by XPS. Furthermore, XPS is one of the few methods to study SEI forming reaction and decomposition products.
Fourier transform infrared spectroscopy.-Fourier transform infrared spectroscopy (FTIR) is based on the interaction of a sample with infrared radiation. Figure 6h shows this interaction in reflectance mode. We note that transmission mode is also possible for FTIR, which is a bulk method and therefore not discussed here. High resolution data are collected simultaneously in the selected spectral range. By applying a Fourier transformation to the signal, the actual spectrum is created.
Early FTIR investigations of materials in Li-ion cells were conducted in Aurbach's group 169,170 and by Yoshida et al. 171 and focused on understanding the chemical characteristics of the SEI on Li and graphite-based anodes. These studies allowed identifying important reflections bands of the SEI as the asymmetric carbonyl stretching at 1650 cm −1 characteristic for (ROCO 2 Li) 2 and 1450 and 870 cm −1 characteristic for Li 2 CO 3 .
FTIR studies on samples from Post-Mortem analyses have also been conducted aiming to tackle differences when using electrolyte additives. [172][173][174] In these cases both, anodes and cathodes were studied. Likewise, FTIR results are used to compare SEI characteristics when changing the Li-based salt. 175 Many other FTIR studies have been carried out to follow aging effects. 34 176 The electrodes were rinsed and dried before the experiments and very similar spectra were found for both cases, with bands at 864 cm −1 , 1008 cm −1 and 1240 cm −1 assigned to Li x PF y and Li x PF y O z , which derived from the LiPF 6 thermal decomposition. 176 Norberg et al. investigated LiNi 0.5 Mn 1.5 O 4 -based cathodes cycled with 1 M LiPF 6 in EC/DEC mixture. 177 After cycling, FTIR tests unveiled characteristic bands of alkyl carbonates along with bands at 1310 cm −1 and 1110 cm −1 assigned to C-O and C-C stretching modes in ketones, which evidence of electrolyte decomposition on the cathode surface. 177 However, the identification of the specific decomposition compounds was not possible with this technique.
For FTIR experiments, we note that the electrode sample preparation protocol is crucial, since electrolyte traces must be removed to avoid undesirable reflections. Alternatively to measuring reflectance of electrodes, it is also possible to scrape off the active material and build KBr pellets. 177,178 The sample transport from the glove box to the FTIR device as well as the analyses should be done under an inert atmosphere, because (ROCO 2 Li) 2 on the electrode surfaces could react with H 2 O to form Li 2 CO 3 . 179 Finally, it is important to remark that FTIR does not allow compound quantification. Therefore, the FTIR result interpretations are often complementary to Post-Mortem analysis by other methods such as electrochemical testing, XPS, and SEM/EDX. Secondary ion mass spectroscopy.-Secondary Ion Mass Spectroscopy (SIMS) allows the characterization of elemental and molecular composition of the surface of a material. Molecular fragments, clusters, as well as positive and negative ions are teared off the surface by using a primary (pulsed) ion beam (Figure 6i). In the case of static SIMS or TOF-SIMS (Time of Flight SIMS), the secondary ions coming from the sample are collected and analyzed in a "time of flight" mass analyzer: mass separation of the ionized fragments is based on the time necessary to reach the detector. Though quantification is difficult to apply, SIMS and in particular TOF-SIMS are very sensitive techniques (down to some ppm). Moreover, due to sophisticated electronics, it is possible to focus the primary ion beam and achieve spectroscopic imaging of surfaces.
Though being a surface sensitive technique, TOF-SIMS is thoroughly used for the study of the surface of bulk materials. This is done by sputtering the sample with a beam of Cs + or Ar + ions allowing acquisition of mass concentration depth profiles. This makes TOF-SIMS a powerful tool for the characterization of thin layers, such as those used in micro-battery systems. 180,181 Surface spectroscopy can help determining the nature of electrochemical passivation layers or coatings on collectors and electrode materials.
The use of TOF-SIMS can also help in studying the aging of current collectors or electrode materials during. 182 The SEI can also be studied by using TOF-SIMS. It has been proven over the years to be a powerful complementary approach to XPS, allowing a better understanding of the SEI chemical structure. First SIMS-based studies in relation with Li-ion battery materials arose in the 2000's. Peled et al. initiated the first attempts to study the SEI at the electrode surface by using TOF-SIMS on HOPG single crystals, as this material can be considered as a model electrode for graphite systems. 189,190 The authors provided evidence for the presence of polymers in the SEI and a dependence of the SEI chemistry on the nature of HOPG planes. Then, other groups considered TOF-SIMS spectroscopy to study the influence of additives or alternative electrolytes (e.g. ionic liquids) on the SEI chemical structure. [191][192][193][194] Though TOF-SIMS is still underused in the field of energy storage, and in particular for Li-ion battery applications, the amount of such studies has grown in the last years. The unique strengths of this powerful surface analysis spectroscopy method are sensitivity, ability to analyze isotopes, better lateral resolution compared to other surface analysis spectroscopy such as XPS.
Glow discharge optical emission spectroscopy depth profiling.-Glow discharge optical emission spectroscopy (GD-OES) depth profiling provides elemental analysis of samples by sputtering and detection of emitted visible light from the teared off particles which are excited in the plasma of a Grimm lamp 195 (Figure 6g) and detected by a Rowland circle spectrometer. 100 GD-OES is well established for the quality control of surface treatments and steel coatings using direct current (DC) potential. Due to its ease of use and high sensitivity, further developments were carried out to make the technique applicable to non-conductive material by means of radio frequency (RF) potential, which made it possible to extend GD-OES application to thin film and non-conductive coating analysis. [196][197][198] More recently, electrodes of Li-ion batteries were subject to GD-OES investigations. 26,28,43,44,72,199,200 The depth profile data is achieved through removing the sample atoms in a layer-by-layer manner using plasma through sputtering. 201 The setup is as follows: 195 the sample is placed in front of an anode and plays the role of a cathode. The anode of the GD-OES device is a hollow cylinder which will be filled with low pressure Ar gas (∼10 −4 hPa). Gas ionization and plasma generation is achieved upon applying a potential difference (∼500-1000 V). Sputtered sample atoms diffuse into the plasma and get excited through further collisions. This results in emitting characteristic photons which will be recorded by the optical emission spectrometer. The cylindrical anode has a typical diameter of 2.5 mm or 4.0 mm, which is the size of the analysis spot. The design of the Grimm lamp makes GD-OES analysis independent of the sample matrix. 100 Unlike a number of other methods such as XPS and SIMS, GD-OES depth profiling is not limited to the surface vicinity of the sample but can analyze it from the electrode surface to the current collector. Therefore, GD-OES can yields information on both, electrode surfaces and on the electrode bulk. 26,28,43,44,72,199,200 Saito et al. observed the Li distribution in cathodes from highpower Li-ion cells. 199 The authors reported a Li gradient toward the surface for the discharged state and vice versa, which was attributed to the slow diffusion of Li in both the electrode and electrolyte. 199 Aldeficient regions in NCA were as well reported to have formed during cycling. 199 Takahara et al. have carried out extensive investigation on not only cathodes but also graphite-based anodes, focusing on SEI growth upon cycling aging. 26,43,44,200 The authors were able to perform calibrations based on the specific case of study and achieve quantified Li distribution throughout the graphite anode. They have as well reported faster and more sensitive depth profiling of graphitebased anodes using Ar gas with additional 1% of H 2 . 200 GD-OES has been applied to graphite electrodes from aged commercial 18650 cells, where a correlation with the electrochemical data was achieved. 28 The study differentiated between "surface" and "bulk" Li using GD-OES depth profiling data calibrated for Li based on ICP-OES results. 28 The study revealed that the surface Li content is correlated to the amount of capacity loss, implying the important role of side reactions on the graphite surface in cell degradation. 28 GD-OES has recently been used in Post-Mortem analysis to detect Li plating on graphite anodes, 72 which is difficult or not possible by other methods. Compared to anodes with SEI, the Li gradient and Li content is significantly increased in case of Li plating. 72 Furthermore, it was observed that most of the metallic Li is placed on the surface of the graphite anodes, 72 which is in agreement with simulations by Hein and Latz. 202 Due to comparably short measuring time, small sample size, high sensitivity, and its possibility to detect Li in depth profiles through the whole electrodes, the GD-OES method is a promising analytical tool to gain a better understanding of Li-ion battery aging mechanisms. However, to get a full picture of aging mechanisms, GD-OES must be combined with complementary methods.

Chemical analysis methods for electrode bulk analysis.-
Inductively coupled plasma optical emission spectrometry.-In Post-Mortem analyses, inductively coupled plasma optical emission spectrometry (ICP-OES) is used to determine the elemental composition of electrodes. 16,18,19,28,151 In ICP-OES, inductively coupled plasma is used to produce excited ions and atoms from the sample, which are emitting electromagnetic radiation in the visible range (Figure 6f). The wavelengths of this radiation are characteristic of a particular element. In this way, ICP-OES is able to give the ratio between elements present in a sample. An advantage of this method is that elements from the ppm range up to the major elements of a sample can be detected. However, one drawback is that ICP-OES does not give the complete sample composition, making the use of additional methods a requirement. ICP-OES is often compared to EDX (see above), however ICP-OES has the advantage of detecting Li. Samples are fully dissolved in an acid solution and then measured. This means that the bulk of a sample is subject to investigation and not only the surface. Furthermore, for ICP-OES sample areas in the cm 2 range are required. However, ICP-OES cannot provide depth profiles and the material has to be scraped-off from a few cm 2 of electrode samples limiting its capability to study local phenomena.
In Post-Mortem analysis, ICP-OES measurements are useful, to prove dissolution of transition metals from the cathode by detecting the migrated material on the anode. 16,18,19,28,151 This dissolution was shown to contribute to the aging mechanism of the anode 16,19,28,36,37 and is caused by HF. 36,37 Stiaszny et al. detected transition metal concentrations of fresh and aged anodes of LiMn 2 O 4 -NMC/graphite cells with ICP-OES. 19 This result was confirmed by the decrease in peak height of the NMC peak in the cyclic voltammetry. 19 The amount of Mn on graphite anodes dissolved from NMC/LiMn 2 O 4 blend cathodes was found to increase with temperature 16,28 203 The authors observed the lowest Mn dissolution from the pure olivine, whereas it was two orders of magnitude higher for the pure spinel. 203 For all blends, the authors found drastically reduced amounts of Mn dissolution by ICP-OES in the electrolye. 203 Growth of SEI thickness is another aging mechanism and was studied via ICP-OES limited to the elements Li, P, and Mn and supported by EDX. 16,28 The results are consistent with SEI growth via decomposition of LiPF 6 salt at the anode | electrolyte interface. 16,28,33 The consumption of cyclable Li on the anode measured by ICP-OES was found to directly correlate to capacity fade, 12,28 the decrease Li in the cathode and increase of cell resistance. 28 On the other hand, reaction of electrolyte on the anode surface is discussed to lead to drying of cells, leading to further capacity fade. 25,204 Nuclear magnetic resonance spectroscopy.-Nuclear magnetic resonance (NMR) spectroscopy is a powerful method being able to characterize materials and chemical compounds in solid state and diluted in solvents. 205,206 It does not only provide chemical and structural information but also information about transport properties and mobility of ions, electronic, magnetic, as well as thermodynamic and kinetic properties. [207][208][209][210][211] The sample is placed in a magnetic field, while it is excited with a radio frequency pulse (Figure 6j). The recorded free induction decay (FID) is processed via Fourier Transformation to yield the NMR spectrum. One or more NMR active nuclei (nuclear spin = 0) are required which serve as probe to detect changes in their chemical environment and their electronic properties. Both kind of samples, liquids and solids, can also be investigated in situ using specific measurement setups. 212,213 Several authors have provided overview articles concerning NMR spectroscopy in Li-ion cell development and Post-Mortem analysis. 214 NMR is a useful tool to facilitate the development of new anode materials based on findings gathered from Post-Mortem analyses. According to Delpuech et al., the high irreversible capacity loss of Sibased anodes mainly originates from the degradation of the carbonate solvents followed by the formation of non-lithiated carbon species in oligomeric or polymeric form. 215 Grey and co-workers showed that the capacity loss and self-discharge is directly linked to structural changes in Si anodes and can be avoided by the correct choice of binders. 216 Pérez-Vicente and colleagues studied Sn 4 P 3 as a possible new anode material. 217 While most NMR studies dealing with Li-ion cell materials are composed of solid state magic angle spinning (MAS) measurements, it is often neglected in the Li-ion cell community that it is also a powerful method to study liquid samples and solutions. Modern liquid electrolytes being mixtures of a vast number of organic and inorganic compounds provide a large variety of NMR active nuclei, such as 1 H, 13 C (in organic molecules), 7 Li, 31 P, 19 F (in LiPF 6 ), in the case of more recently developed conducting salts like LiTFSI or LiFSI, even 14 218 Their LiPF 6 based electrolytes were carbonate-based solvents (EC, EMC) containing fluorinated additives like 1-FEC, DT-FEC (Bis-(2,2,2-Trifluoroethyl carbonate), 2,2,2-trifluoroethyl methyl carbonate (TFEMC) and Triphenyl phosphate. 218 In case of the cathode, the authors could determine different amounts of irreversible Li concentrations for the same nominal electrochemical SOC after cell disassembly. 218 Furthermore, they could show that additive decomposition and deposition also occurs on the cathode. 218 From the anode point of view, varying amounts of LiF were detected as well as decomposition products of the fluorinated carbonates. Dupré and co-workers examined aged Li 4 Ti 5 O 12 and LiFePO 4 electrodes and determined their LiF concentrations. 219 By correlation with the capacity fade of the cycled cell, the authors were able to propose several reaction paths for the different aging mechanisms of the examined electrodes. 219 Lucht and co-workers proposed thermal decomposition mechanisms due to autocatalysis and protic impurities for several carbonate solvents frequently used in state-of-the-art electrolytes like DMC, EC and DEC. 220,221 Their study was a combination of NMR methods ( 1 H, 13 C, 19 F, 31 P, DEPT, COSY and HETCOR), GC-MS and SEC and was done on a model system, which did not contain material collected from an aged Li-ion cell. However, in addition to protic impurities like H 2 O or ethanol, the authors were able to identify DEC as main reason for thermal decomposition of LiPF 6 and observed a number of decomposition products which might also alter anode and cathode materials during the life of a Li-ion cell. 220 In a follow up study they investigated the interaction of several cathode materials with organic electrolytes at elevated temperatures. 221 Obviously Li 2 CO 3 present on the surface of cathode particles is able to inhibit thermal decomposition of organic electrolytes. 221 Furthermore, electrolyte decomposition products found in this study were similar to those observed on the cathodes removed from thermally abused Li-ion cells. 222 To summarize, model studies including NMR measurements in combination with other characterization techniques can be helpful to interpret data gained from Post-Mortem analyses of aged Li-ion cells.
Structural characterization.-In Post-Mortem analysis, the X-ray diffraction (XRD) technique is usually applied for the structural analysis of the active materials in the electrodes (Figure 6k). Like for all diffraction techniques, XRD is only applicable to materials whose atoms exhibit a certain degree of periodicity. XRD has been extensively used in many aging mechanism diagnosis as it provides relevant information on the structural changes that crystalline active materials can suffer during aging. 16,34,81,135,223 Additionally, XRD provides information about changes in the orientation of particles and film formation on the electrode | electrolyte interface. 34,135 The latter phenomenon can be seen in XRD by a decrease of peak intensity. 34,135 Similarly, Liu et al. showed that a broadening of the XRD peaks indicates the occurrence of graphite exfoliation. 223 Crystalline degradation products on anode surfaces are observed by additional peaks. 122 XRD also allows detecting chemical decomposition/dissolution reactions by a decrease of lattice volume of the cathode particles. Stiaszny et al. analyzed a commercial Li-ion battery with a mixed LiMn 2 O 4 /NMC cathode and graphite anode cycled at room temperature. 18 The authors measured a variation of the lattice parameters of aged NMC active material that was influenced by dissolution of transition metals into the electrolyte which is causing a decreased amount of Li in the cathode. 18,19 The temperature dependent change of aging mechanism (Li plating / SEI growth) in a commercial 18650 cell with graphite anode and LiMn 2 O 4 /NMC blend cathode was visible in XRD measurements by a change of the lattice constants a and c of the NMC. 16 The lattice constants were also correlated to the Li content measured by ICP-OES of aged anodes and cathodes. 16 After conservation of Li deposition by chemical reaction of its surface with isopropanol, powder X-ray diffraction (PXRD) was used to identify Li 2 CO 3 . 150 The results were in agreement with EDX and FTIR measurements. 150 XRD is a common method to conclude on lattice constants of crystalline active materials. However, since XRD is limited to measurements on the electrode bulk, it often has to be combined with other methods, e.g. surface sensitive method such as SEM or EDX.
Electrolyte analysis.-Electrolyte degradation occurs due to side reactions that provide insoluble, soluble and gaseous products. 37,224,225 Identification of such products is fundamental to trace back the side reactions responsible for battery aging. Thus, many studies have implemented techniques to analyze electrolytes and gases generated during battery aging. Next to liquid NMR spectroscopy (see above) chromatographic techniques have proven to be very successful in Post-Mortem electrolyte characterization. The basic principle of chromatography is the separation of the components in a mixture and subsequent detection. Separation is for example achieved by different retention time in the adsorbed state on the wall while travelling through a capillary (Figure 6l). Experiments with other techniques are often complementary, specifically for the analyses of insoluble products; those methods are described in other subsections of this manuscript. Most studies concerning chromatography techniques are focused on LiPF 6 -based carbonate (EC, PC, DMC, EMC, and DEC) solvent mixtures, since these are the more common electrolytes used in Li-ion batteries.
In the following, reviewing literature on the Post-Mortem analysis of electrolyte and gas is segregated into two main groups: investigations with lab scaled cells designed especially to collect a high amount of liquid and gas samples and investigations with commercial cells, generally containing only a slight excess of electrolyte which is difficult to recover. 171 were performed upon the first charge of LiCoO 2 /graphite cells. The authors implemented Liquid Chromatography coupled with Fourier infrared spectrometry (LC-FTIR) for analyzing the recovered electrolyte, which provided alkyl dicarbonates as the main soluble compound deriving from the electrolyte decomposition. 171 Gas Chromatography (GC) coupled with a thermal conductivity detector (GC-TCD) allowed the observation of H 2 , CO, CO 2 , CH 4 , C 2 H 4 , C 2 H 6 and C 3 H 8 upon the first charge. To gain insights in the formation mechanism of alkyl dicarbonates, Sasaki et al. carried out Post-Mortem GC-mass spectrometry (GC-MS) analyses of electrolyte recovered after cycling of Li / graphite half-cells. 226 The presence of alkyl dicarbonate was confirmed, and parallel chemical simulations signaled that Li alkoxides could trigger the alkyl dicarbonate formation. 226 Aiming on the understanding of the reductive decomposition of LiPF 6 -carbonate solvents, Laruelle's group [227][228][229] carried out Post-Mortem electrolyte and gas analyses on lab scaled cells. The authors used Electro Spray Ionization coupled with High Resolution Mass Spectrometry (ESI-HRMS) and GC-MS for analyzing the electrolytes, though a different set of capillary columns was implemented for GC-MS analyses of the stemmed gas. The parallel use of these techniques permits the detection of compounds on a wide mass range, thus a global electrolyte degradation mechanism was elucidated. 228,229 The authors found that most degradation compounds derived from the linear carbonate reduction, which provide the Li alkoxides that further trigger the electrolyte esterification, while the two-step reduction of EC was less important.

Electrolyte and gas analyses performed on lab scaled cells.-Early investigations
On the other hand, H 2 O traces in LiPF 6 lead to the formation of POF 3 , HF, and LiF. This mechanism is enhanced by temperature, consequently, electrolyte storage tests implementing chromatographic analyses are also found in the literature. [230][231][232][233][234] Terborg et al. 230 investigated the thermal aging and hydrolysis mechanisms of LiPF 6 by means of Ion Chromatography (IC) coupled with ESI-MS. In these studies, it is interesting to note that the implementation of IC allows the detection of HF in the electrolytes, thanks to the identification of F − . Kraft et al. 234 studied the thermal aging decomposition products of LP30 and LP50 electrolytes, by developing separation methods and comparing the reliability of three different IC columns. Moreover, the authors coupled IC-ESI-MS-MS for the identification of new organophosphate compounds. 234 Handel et al. investigated the thermal degradation of EC/DEC+LiPF 6 mixtures with deionized water contamination by applying GC-MS for liquid electrolyte analyses and Headspace-GC-MS for volatile compound analyses. 233 Complementary NMR and acid-titration were also performed. The authors concluded that the electrolyte aging had low speed as far as catalytic surfaces, ambient air and protic impurities are excluded. 233 Electrolyte and gas analyses performed on commercial cells.-Chromatographic techniques have also been applied to analyze electrolyte and gas taken from commercial Li-ion cells. 83,225,[235][236][237][238] The challenge is to apply the knowledge gained from analyses at laboratory scaled cells on the commercial ones to indicate the electrolyte degradation pathways followed upon a given aging protocol. Identification of undesirable reaction products can point out the future optimizations in commercial cells. Nevertheless, the sample taking method is crucial, since commercial cells usually do not have electrolyte excess neither gas pockets.
Kumai et al. designed a "gas release" vessel for harvesting gas samples from LiCoO 2 graphite cells after cycling, overcharge and over discharge tests. 235 CO 2 , CO, CH 4 , C 2 H 6 and C 3 H 8 were detected by performing GC-TCD analyses and GC coupled with a flame ionization detector (GC-FID). Terborg et al. recovered electrolyte from a commercial cell after 1400 cycles by washing the separators, anode and cathodes in PC. 237 Subsequently, the solutions were subject to GC-MS analyses. 237 More recently, the GC-FTIR-MS equipment was introduced 236 as a useful technique to analyze the gases from a swollen commercial cell. The gas recovery was performed in an Ar-filled glove box by piercing the pouch cell bag with an air-tight syringe. The authors identified CO, CO 2 , CH 4 and C 3 H 8 in the GC/FTIR Gram-Schmidt graph, whereas the GC/MS chromatogram allowed the detection of other less abundant volatile compounds. 236 Complementary GC-MS electrolyte analyses allowed the detection of alkyl dicarbonates and longer carbonate chains, as well as organophosphate compounds, which indicated that water traces present in the commercial cell play a role in the electrolyte decomposition. 236 Grützke et al. recovered electrolyte from 5 Ah NMC-based Liion cells that were field tested in HEVs. 83 The pressure valve of each cell was crushed and the electrolyte was collected. 83 GC-MS analyses yielded the electrolyte components, while GC-FID analyses permitted estimating the composition. 83 The authors found F − and PO 2 F 2 − species by means of IC-ESI-MS when opening the cells under inert atmosphere, whereas HPO 3 F − and H 2 PO 4 − were also detected when opening in humid environment. 83 In commercial cells, water traces seem to be unavoidable due to the hygroscopic character of LiPF 6 , then the electrochemical/chemical electrolyte degradation, which leads to solvent esterification is concomitant with the production of organophosphates since the decomposition products react with POF 3 . 228 In order to understand the influence of additives Post-Mortem analysis of degraded electrolytes and generated gas has to be carried out at lab scaled cells 239 as well as in larger prototypes. 225 Moreover, Post-Mortem chromatography techniques are starting to be applied for assessing new electrolyte formulations dedicated for high voltage applications, for which electrolyte oxidation and thermal decomposition are the main issues to solve.

Electrochemical analysis of reassembled electrodes.-Reconstruction into half cells.-Post-mortem electrochemical characteriza-
tions can be carried out in cells by reconstructing anodes or cathodes together with metallic Li as counter electrode. 12,[17][18][19]34,88 For double side coated electrodes, one coating has to be removed, e.g. by N-methylpyrrolidone 88 or by laser blanking. 87,90 Electrodes extracted from the fresh cells and those from aged cells are studied following the same protocols. The goals of these electrochemical tests are (i) accessing to the residual (or remaining) capacity of the electrodes (in mAh/cm 2 ) and (ii) measuring the reversible capacity (in mAh/cm 2 ), considering that the cells have necessarily been disassembled at the same SOC (often 0% of SOC).
For the negative electrode/Li cells, the first electrochemical test consists in a charge to extract Li from the anode (delithiation). Whereas for the positive electrode/Li cells, the first tests consists in a discharge to insert Li to the cathode (lithiation). The corresponding capacities are then the residual capacities of the electrodes. To obtain the reversible capacities, the negative electrode in the half cell is lithiated again, while the positive electrode is delithiated.
In electrodes extracted from fresh cells, the residual capacities permit the evaluation of the cell initial irreversibility due to the building of the SEI layer during the formation step. 46 The reversible capacity unveils the initial balancing. 12 In aged electrodes, the evolution of the residual capacities indicates the Li consumption in side reactions, and the evolution of reversible capacities serves to track cell unbalancing upon aging as well as to point out aging mechanism properly. 17,46 Figure 8 illustrates the principle of capacity determination in coin half cells. It is worth mentioning that care should be taken when using the indicated capacity determination method, if inhomogeneities are present in the electrode due to fabrication errors or aging. One way to overcome this problem is not to consider the average surface capacity (in mAh/cm 2 ) of the samples, but to consider the average weight of the samples harvested from "fresh" electrodes before building coin-cells to calculate and compare the mass capacity (in mAh/g).
Kobayashi et al. proposed a similar procedure for capacity determination of each electrode. 12 They studied a LiMn 2 O 4 / LiNi 0.8 Co 0.15 Al 0.05 O 2 blend cathode vs. a carbon graphite anode system. 12 The residual capacity of cathodes harvested from cells after cycling or calendaric aging (converted in the cathode SOC in the discharge state) increased if compared with the value obtained for the fresh cell. 12 The authors demonstrated that there is a relationship between the capacity retention of the studied cell and the SOC of the cathode in the discharge state (identically with the open-circuitvoltage of the half coin-cell that deduces the state of lithiation of the electrode). 12 That suggests that Li ions are not only irreversibly stored on the anode side in the initial SEI formation cycle but also continuously accumulated during cycling or calendaric aging. Similar results were found by other methods. 16,28 Aurbach et al. reassembled fresh and cycled electrodes from a 18650 cell with carbon anodes and LiCoO 2 cathodes into cells with Li counter and reference electrodes. 34 The authors performed cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) tests with these reassembled cells. 34 From the CV and EIS measurements they observed that the kinetics of the aged anodes get slower compared to the fresh anodes. 34 The reason is growth of a thick surface film during aging. 28,34,135,240 Such basic electrochemical tests in half-cell configuration offer an image of the real state of lithiation of each electrode of the full-cell at the discharge state. The remaining capacities can indirectly yield the irreversible accumulation of Li ions in the anode that can be confirmed also by chemical analysis of the anode by ICP-OES. The reversible capacities can evaluate the evolution of the insertion ability of the host structure of each electrode and the most significant factor of capacity fading can be determined. A key issue is consistency of the reassembled cells, therefore at least two cells should be built from the same electrodes.
Reconstruction into full cells with reference electrodes.-Additionally to reconstruction of either anode or cathode into half cells vs. Li,12,34,81 it is possible to build full cells using anode, cathode, and an additional reference electrode (RE). 75,87,241 The RE allows obtaining the potentials of both anode and cathode during charging and discharging. 87 We would like to note that measurements in half cells give a different result, since the interaction between anode and cathode is absent.
The stability in time of the RE potential is fundamental, and depends on the testing temperature and on the nature of the electrochemical couple chosen as RE ( It is well known that the position of the RE is particularly important. For example in aqueous systems the Luggin-Haber capillary is well established which is positioned (i) near the working electrode and (ii) between corking and counter electrode. 266 Recently, Hogg and Wohlfahrt-Mehrens performed measurements in 4-electrode full cells with two REs. 241 The authors found that the position of the RE between anode and cathode is also very important to measure anode potentials correctly in a Li-ion full cell. 241  81 The reassembled cells were built in a glove box using the separator from the disassembled aged Sony cell and 1 M LiPF 6 in EC:DMC = 1:1 as electrolyte. 81 Graphite anode and NCA cathode from commercial high-energy 18650-type cells were recently reassembled into 3-electrode full cells with an additional Li reference electrode. 87 Using this technique, it was possible to measure the anode potential vs. Li/Li + and therefore to determine the conditions for Li deposition. 87,90 Consequently, optimized charging procedures could be developed to prevent Li deposition in the commercial 18650 cells and improve the battery life significantly. 87 In addition to measurements of electrode potentials with the RE, it is also possible to perform impedance measurements of both anode and cathode recovered from fresh and aged cells simultaneously at various states of charge. 261 This type of measurement requires the optimization of the morphology of the RE and its placement inside the cell to obtain reliable impedance spectra.
The position of the RE is important to get reliable potential and impedance values. 248 Dees et al. simulated the electrolyte potential distribution inside the cell to find the best positioning of the RE inside pouch-cells. 267 Other articles 248,249,268 show the distortion or artefact of simulated impedance spectra of combined geometric and electrical asymmetries in the electrodes. These artefacts are often inductive loops in the low frequency region of one of the electrodes.
In any case it is very important to keep the electrodes strictly under controlled conditions. 75 Itou et al. reassembled electrodes from cycled cells together with fresh electrodes into new cells with Li reference electrode in order to measure resistance increase of the electrodes. 75 By this method, the authors found that the cathode was mainly responsible for the resistance increase during cycling at 60 • C. 75 In consistency, FIB/SEM and XAFS revealed cracking at grain boundary interfaces inside particles and local changes in the LiNi 0.8 Co 0.15 Al 0.05 O 2 cathode, respectively. 75 Noteworthy, another configuration for Post-Mortem impedance tests are symmetric coin cells with electrodes of the same polarity. This allows the evaluation of each electrode without the effect of the counter electrode or the RE. 269,270 The disadvantage of this configuration is that the impedance can be obtained only at one unique state of lithiation, corresponding to the SOC of the cell before disassembly.
An alternative solution could be the integration of a RE directly in the still functioning commercial cell without disassembly and reconstruction of the electrodes. This approach is complex because it will be necessary to assure proper cell re-sealing.

Combination of Methods for Full Characterization of Aging Mechanisms
As discussed in detail in the Physico-chemical analysis of aged materials after disassembly of Li-ion cells section, each physico-chemical analysis method has its specific advantages and drawbacks, allowing observing only specific aspects of an aging mechanism while not being able to characterize others. For instance surface sensitive methods cannot access the bulk properties of electrodes. In contrast, bulk sensitive methods mix the surface properties with those of the electrode bulk. Since the bulk is usually much larger than the surface, the influence of the surface on the measurement is often negligible. Depth profiling methods detect both, electrode surface and bulk, however, they do not observe morphological or structural changes.
The required capability of an analysis method depends strongly on the aging mechanism that is to be observed. Figure 9 shows a schematic overview of degradation mechanisms of electrodes and materials, which is often not distinguished in literature. Electrode degradation includes growth of films on top of electrodes (electrolyte decomposition 16,18,26,28,40,122 or Li deposition 25,45,72 ), clogging of pores of electrodes or the separator, 37 delamination of the separator, 42,95 cracks in the electrode coating 32,42,223 or deformation of electrodes or separator. 9,14,32 Material degradation involves particle cracks, 42,74,75,129,130 exfoliation, 37 changes at particle surfaces, 37,158 film formation on particles, 37,271 dissolution/migration of transition metals, 36,37,272 electrolyte degradation, 175,228,229 or closing of separator pores (e.g. by applied pressure). 134,273 The respective recommended analysis methods are indicated in Figure 9. Figure 10 shows an overview of the capabilities of the analysis methods discussed in detail in the Physico-chemical analysis of aged materials after disassembly of Li-ion cells section. Green, orange, and red colors indicate good, limited, and no capability for detection of a specific aging mechanism, respectively. It can clearly be seen from Figure 10. Overview of analysis methods and phenomena they are able to detect. Green, orange, and red colors indicate good, limited, and no capability for detection of a specific aging mechanism, respectively.
) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 207.241.231.108 Downloaded on 2020-02-10 to IP Figure 10 that the abilities of the different analysis methods is wide spread, but there is no method which is covering all aging mechanisms. Therefore, we give the clear recommendation to investigate samples with a variety of complementary analysis methods to get a full picture of aging mechanisms in Li-ion cells.

Conclusions
Disassembly of Li-ion batteries is mandatory to collect samples for determination of aging mechanisms and improvement of materials, including step by step improvement of state-of-the-art materials as well as the development of new material generations.
In the present paper, the state-of-the-art procedures for Post-Mortem analysis of aged Li-ion cells are reviewed. In particular, methods for disassembly of aged Li-ion cells as well as physico-chemical analysis of their components are reviewed extensively.
Chemically inert environments during cell opening are decisive to ensure reliable results with air sensitive samples and safe operation for the experimenter. For post-processing of samples it is recommended to wash the electrodes in a solvent which is already an ingredient of the electrolyte (e.g. DMC) to maintain the quality of the samples. However, at the moment it is not completely clear how washing affects SEI layers on electrodes. A highly experienced experimenter using appropriate equipment for cell opening is mandatory to get interpretable results in the analysis of the samples obtained from Li-ion cells.
The available physico-chemical analysis methods for Post-Mortem analysis of Li-ion batteries were reviewed and include microscopy, chemical methods which are sensitive to electrode surfaces and electrode bulk, as well as electrolyte analysis techniques and reconstruction of electrodes into half and full cells with reference electrode. For the latter case, there is a significant difference between reconstruction into half and full cells. Half cells with anodes or cathode vs. a Li counter electrode provide capacities of the single electrodes. In contrast, reconstruction of anodes and cathodes into 3-electrode cells with an additional reference electrode contain information on the interaction between anode and cathode. 3-electrode cells therefore allow gaining insights into electrode resistances and electrode potentials which are decisive for the main aging mechanisms (e.g. Li plating for negative anode potentials).
Each physico-chemical analysis method is able to observe only specific aspects of Li-ion degradation. Therefore, it is recommended to investigate samples with a number of complementary analysis methods to obtain a full picture of the aging mechanisms. By combination of the reviewed methods, it is possible to characterize all relevant parts of cells (anodes, cathodes, separators, and electrolytes) with regard to their micro structure, crystallographic structure, and chemical composition.
Only by detailed knowledge of the aging mechanisms, state-ofthe-art materials can be satisfactorily improved and new materials be developed to meet the challenges and demands of future battery usage applications in form of increased power and energy densities. Thus we consider the procedures reviewed in this paper appropriate for cell disassembly of future battery generations with increased power and energy densities, after slight modification of the disassembly method if necessary.