Surface analysis insight note: X‐ray photoelectron spectroscopy analysis of battery electrodes—Challenges with nickel–manganese–cobalt and Li examples using an Al Kα x‐ray source

X‐ray photoelectron spectroscopy (XPS) has become a highly important tool for the analysis of battery materials and components. However, both anecdotal and detailed analysis of selected parts of the literature indicate that many reports of XPS on battery electrodes have significant analysis or data flaws. In this paper, we highlight several of the common challenges that analysts face when using XPS for battery materials, pointing to recent literature that addresses many of the critical issues associated with sample preparation as well as data collection and analysis. A common error for battery materials (and other materials) involves ignoring peak overlaps and interferences. Specifically, when a “minor” peak associated with a component in relatively high concentration overlaps or contributes to the primary peak (or one recommended for quantitative analysis) from a different element in the material. Overlap issues apply to many battery electrodes composed of many elements with complex photoelectron peak structures, as well as those involving peaks with seemingly simpler spectral envelopes such as Li and F. Examples of issues associated with battery systems are highlighted by a discussion of challenges associated with XPS analysis of Li and nickel–manganese–cobalt (NMC) electrodes in battery systems. Lithium analysis has challenges associated with the preparation and an often‐unrecognized peak overlap with F. In our laboratory and in the literature, NMC electrodes are often examined and new XPS users do not always recognize interference of the Auger signal from FKLL (in or on the electrode) with Ni 2p photoelectron spectrum when generated with Al Kα X‐rays. The use of simulated spectra involving both F and NiO demonstrates the extent of F Auger contributions to the Ni 2p signal strength as a function of the F/Ni atom ratio in the material and suggests spectra information that can be used to identify how significant effects will be on the resultant spectra. Our analysis demonstrates that in many cases overlap issues are significant for real electrode materials.

several of the common challenges that analysts face when using XPS for battery materials, pointing to recent literature that addresses many of the critical issues associated with sample preparation as well as data collection and analysis. A common error for battery materials (and other materials) involves ignoring peak overlaps and interferences. Specifically, when a "minor" peak associated with a component in relatively high concentration overlaps or contributes to the primary peak (or one recommended for quantitative analysis) from a different element in the material.
Overlap issues apply to many battery electrodes composed of many elements with complex photoelectron peak structures, as well as those involving peaks with seemingly simpler spectral envelopes such as Li and F. Examples of issues associated with battery systems are highlighted by a discussion of challenges associated with XPS analysis of Li and nickel-manganese-cobalt (NMC) electrodes in battery systems.
Lithium analysis has challenges associated with the preparation and an oftenunrecognized peak overlap with F. In our laboratory and in the literature, NMC electrodes are often examined and new XPS users do not always recognize interference of the Auger signal from F KLL (in or on the electrode) with Ni 2p photoelectron spectrum when generated with Al Kα X-rays. The use of simulated spectra involving both F and NiO demonstrates the extent of F Auger contributions to the Ni 2p signal strength as a function of the F/Ni atom ratio in the material and suggests spectra information that can be used to identify how significant effects will be on the Notice: Manuscript Authored by Battelle Memorial Institute Under Contract Number DE-AC05-76RL01830 with the US Department of Energy. The US Government retains and the publisher, by accepting this article for publication, acknowledges that the US Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so for US Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan: (http://energy.gov/downloads/doe-public-access-plan) resultant spectra. Our analysis demonstrates that in many cases overlap issues are significant for real electrode materials.  [1][2][3][4][5] This paper is part of a planned ongoing series of short papers, which identify challenges analysts face when applying XPS to specific materials or systems and provide example solutions.
Here, we focus attention on one of the common problems, too often ignored in battery analysis and other applications, associated with peak overlap in the analysis of electrodes. The paper provides some general guidance related to the magnitude of the problem. Because of the significant application of XPS in battery research, we also identify other common challenges that need to be considered when using XPS for battery materials and point to relevant literature that further discusses the issues and provides possible solutions. Included herein are important and relatively common issues, with no claim that the list is complete.
XPS is a powerful surface and interface analysis method. Modern XPS instruments are remarkably reliable and capable of collecting highly precise and reproducible data. However, depending on the questions being addressed, the successful application of XPS is often not as simple or straightforward as many analysts assume. [6][7][8] Therefore, it is important for analysts to become aware of challenges and pitfalls that need to be considered in sample preparation, data collection, data analysis, and result reporting. Generic analysis challenges apply to many XPS applications and a collection of guides dealing with a variety of topics has been developed. 9 However, some specific analysis challenges must be recognized and addressed when applying XPS to battery materials. An assessment of XPS data in the literature found that more than 70% of the XPS data appearing in one energy and battery journal had significant flaws. 6 2 | COMMON CHALLENGES TO XPS ANALYSIS OF BATTERY MATERIALS AND SYSTEMS Important issues that need to be addressed for XPS analysis of battery materials include:

| Preparing samples for analysis
Many of the materials used in batteries are environmentally sensitive.
Specifically, these materials easily change chemical state in response to sample handling, environmental conditions, and changes in sample potential. Furthermore, samples taken from highly cycled cells may contain toxic chemicals, which requires additional considerations regarding exhaust gas from the vacuum pumps. Thus, care is required during sample preparation to avoid unintended changes and minimize environmental exposure for many sample types, 10 especially those associated with highly reactive battery materials such as Li. 11,12 At Pacific Northwest National Laboratory (PNNL) a glove box is connected to several of our XPS systems to minimize environmental exposure. 13 However, as observed by several research groups, 11,12 the atmosphere in a glove box is not adequate to keep highly reactive materials such as Li metal from reacting. Some experiments are accomplished using in situ or near-in situ measurements. 3 It is useful to validate sample preparation methods by looking at the impacts of changes with time and environmental exposure and sometimes examining the effects of alternate preparation approaches to determine if the results are consistent and sensible.
Lithium is a common battery component that is challenging to prepare for analysis. Characterization challenges of Li have been described by Otto et al 12 They examined the surfaces of Li and Li electrodes and found they were always covered by some type of passivating layer, often containing an outer Li -hydroxide and carbonate layer over an inner Li-oxide layer. The exact nature of these layers depends on sample treatment, use, storage, and nature of sample transport. Neither vacuum nor glove box environments are adequate to maintain a clean Li metal surface. Of particular interest are observations that Ar + -sputtering did not always result in a pure lithium metal surface, but a mixture of Li metal and oxide. Additionally, exposure of a passivated Li sample to an electron beam causes electrochemically driven mass flow of Li toward the surface of the sample resulting in a deposition of Li metal on top of the passivating layer (referred to as electroplating by the authors). They note that such surface layers may be useful for some studies, but this is not a "cleaned" Li surface in the conventional sense. Experiments by the current authors have confirmed this specific process only occurs when the sample is grounded.
Therefore, it is recommended that Li metal samples be mounted in a way that they are floating with respect to the spectrometer. For relatively clean Li metal samples grounded to the spectrometer, analysis can be performed without inadvertently changing the sample though electroplating if the charge neutralizer setup is not used during analysis; this can be useful for measuring the true binding energy of Li (0), as well as Li oxides and Li 2 CO 3 that usually appear in small quantities on nominally clean Li coin samples.
Although some researchers have gone to considerable work to produce clean Li metal surfaces for fundamental study, it is useful to note that commercial Li metal usually has a protective passive layer applied, and the rapidity of layer formation means that in most conditions Li metal in battery systems will not have a "clean" Li metal surface. In an effort to examine the fundamental interactions of clean Li metal with an electrolyte, a study at PNNL placed Li metal in an electrolyte in a glove box and used a clean knife to slice the metal. The resulting surface was analyzed by XPS, thereby ensuring the resulting surface that is analyzed is lithium metal reacted with electrolyte only.

| Planning and collecting XPS data
To achieve the desired performance, battery materials and systems are often relatively complex involving many elements. Consequently, it can be important to use reference spectra to determine the full range of photoelectron peaks that should be expected, identify likely peak interferences or overlaps, and select the most appropriate X-ray energy if there are options. For several decades some XPS users have had the option of using either Al Kα or Mg Kα sources. The switch from an Al Kα to a Mg Kα source would avoid the specific overlap which is discussed below. 1 Switching to a non-monochromated source from a monochromated Al Kα provides its own issues such as peak broadening, X-ray satellite, and ghost (breakthrough) peaks that should be considered. There are also a variety of higher energy sources in use today including Cr Kα, Ag Lα, and Ga Kα as well variable energies at synchrotrons. Each of these shifts the Auger electrons relative to photoelectron energies and enables analysis of a variety of photoelectron and Auger electron peaks. However, the peak overlap and interference issues discussed in this paper, particularly for complex samples, need to be checked for each source. Significantly higher energies offered by alternatives to Al or Mg may not always be appropriate if you are attempting to probe the very top surface and as such the sampling depth of each source should be considered. Currently, a monochromated Al Kα source is the standard offering from manufacturers for a base XPS instrument. While in the past inclusion of a secondary X-ray source was not always feasible, either due to cost or lack of available space on the instrument, dual monochromatic X-ray sources are now available from most manufacturers and can be a relatively inexpensive upgrade. Appropriate analysis planning often simplifies the required data analysis.
XPS data are usually, and should be, collected in both survey (low-energy resolution) and specific element (high-energy resolution) modes. Although some researchers choose only to collect high energy-resolution data on the peaks they assume to be of interest, most experienced users recognize the importance of collecting survey data before the higher resolution narrow window data. Survey spectra are useful to confirm that the sample has the expected composition and/or to identify contamination and damage. The information collected during a survey is often critical to setting up the narrow window measurements if they are required. A survey spectrum from an as-made electrode containing Ni, Mn, and Co oxides (NMC 811), Electrical potential or potential gradients often develop during XPS measurements on materials or samples that are not highly conducting and for heterogeneous samples the potential will have an inhomogeneous distribution. Although ideal electrode materials are conductive, the phases that form on electrodes during cycling are often non-unform in distribution and non-conductive. 15 The presence of such potential variations can shift and/or broaden the F I G U R E 1 Wide scan survey spectrum of a fresh NMC electrode obtained with an Al Kα source. In addition to the core Ni, Mn, Co, and O components, which contribute multiple photoelectrons and Auger peaks to this complex spectrum, note the presence of F, N, and C, which occur in the binder, surface contamination, graphite, and carbon black additives. The region in the box includes F and Mn Auger peaks and Ni 2p photoelectron peaks. photoelectron peaks and this general effect is identified as sample charging during XPS measurements. The non-uniform distribution of surface potential is identified as differential charging. Most XPS systems have methods to control charge buildup on samples during XPS. 16 The effectiveness of the neutralizer's capability can be assessed by determining the FWHM in (polyethylene terephthalate) PET in relation to the manufacturer's specifications. 16 Potential variations are integral to the operation of batteries and the development of potential in a sample during XPS might be expected to influence the measurements for some samples. Analysts need to look for photoelectron peak shifts and peak widths as a function of measurement time (and electrode cycling) and use/develop appropriate degrees of sample neutralization to maintain sample stability. It is useful to check for shifts and shape distortions in the C-C photoelectron peak in the C 1s narrow scan and also for shifts and changes in other photoelectron peaks. The charge neutralization system can also be used to deliberately vary the sample potential to extract the electrical properties of the material, which can vary depending on the neutralizer design. The extent of sample charging during the measurement will need to be assessed during data analysis. 2 Readers are referred to three papers that highlight challenges (and some solutions) associated with charging on complex materials such as the layers on battery electrodes, Oswald et al, 2 Woods and Teeter, 11 and Marchesini et al. 15 The study by Woods and Teeter 11 provides a useful example of a sample of both charging issues and a solution for battery materials. They examined the wide spread of Li 1s binding energies (BEs) in the literature making several important observations including the fact that surface charging during XPS was a major cause of these variations. They examined a range of materials associated with battery electrolytes and provided useful information about them. An important observation is that during charging the BE separation of major peaks (ΔBE) of specific phases did not change, which provides a way to both identify materials and observe or identify charging-related peak shifts during a measurement. They conclude that "Combining ΔBE values with elemental intensity ratios and valence-band spectra provides a means for accurately identifying phases using XPS analysis, even for spectra from difficult to analyze battery samples, where charging is a significant issue, and multiple overlapping peaks are com- including the complications of differential charging in heterogeneous materials (e.g., SEI layers) can cause for peak assignment/identification. Figure 1 and the discussion above, electrode materials are often complex and consist of many materials. Heidrich et al 5 highlight the additional complexity of uniformity and heterogeneity in electrode coatings. They emphasize the need for multiple analyses to get an appropriate understanding of the nature of SEI coatings and note the non-uniformity of the coating layers is a significant source of analysis variability. Their study also highlights some aspects of sample preparation, both opportunities and challenges. The current authors have found high value in collecting spectra from multiple analysis points on complex specimens. Not only is it possible to assess the compositional consistency or variation of a sample, but sometimes, the presence of differential charging can be observed for some of the spectra. The presence of charging on some portion of the composition of a complex specimen can be explored by overlaying multiple spectra.

As highlighted in
The presence of charging almost always appears as a broadening of the spectra. An example of such an effect is shown in Figure S1 from multiple spectra collected on Li coin cell batteries.

| Data analysis and reporting
The approach to the analysis of XPS data depends on the questions that need to be addressed. 8 At the simplest level, it can be used to identify the elements present on the surface of the sample, often looking for the presence of contamination. Other analyses may seek to quantify the composition, consider the chemical state of some of the elements, or determine the thickness of a surface coating (such as the SEI layer). At a minimum, most analyses require identification of the elements present, which requires identifying the elements responsible for the peaks observed in the collected spectrum.
The survey spectrum from the NMC electrode Figure 1, identifies the many photoelectron and Auger electron peaks observed for the primary elements (Ni, Mn, and Co) as well as other elements associated with the electrodes C, O, and F that are included as part of the electrode composition (as well as surface contamination). Transition metals have relatively complex spectra with multiple peaks. Analysis of such a spectrum, including the identification of peak overlaps, requires a systematic approach starting with the largest peaks, such as outlined in a tutorial on peak identification in a survey spectrum. 21 Of particular note for the NMC survey spectrum, which was collected with Al Kα X-rays, there is a significant overlap of Auger lines from F with the Ni 2p signal (the largest Ni photoelectron peaks most often used to quantify Ni and for determination of Ni chemistry). Relatively simple single compound spectra are shown in Figure 2 of components that contribute to the composition of the NMC electrodes to help with visualization of the source of the complications in the NMC spectrum. In addition to the F Auger overlaps with the Ni 2p, there are overlaps or interferences with Mn 2s and Co 2p peaks. If Mn is present in sufficient quantity the Mn 2p core levels would be better for quantification since they are close in energy to the Li 1s. In this work, only the peak interference of the F Auger is considered to illustrate one of several complexities associated with analyzing NMC XPS spectra.
Additionally, an analyst might consider the impact of Mn LMM on the Ni 2p region if there is enough Mn in the electrode. The extent of peak interference depends on the relative elemental concentrations, but it is not always possible to simply use the "normal" metal photoelectron peaks recommended for quantification when there is peak overlap.
The impact of F Auger lines on the intensity of Ni 2p in materials such as NMC electrodes will be described in more detail below. who identify issues related to peak overlap-the impact of F-and inappropriate application of peak fitting to determine Ni chemical state. They usefully note that the Ni 2p photoelectron peaks can be separated from the F Auger lines if a Mg Kα X-ray source is used, but even when the Mg source is used, appropriate peak fitting is often a problem. Their discussion is consistent with the study of XPS data quality which found that inappropriate and inconsistent fitting of peaks was the most common source of incorrect XPS data analysis. 6 To help address this problem, a guide to peak fitting was prepared 22 as part of the series of XPS guides mentioned earlier. Although there can be many issues associated with information obtained involving peak fitting, inappropriate peak shapes and widths, errors in addressing peak backgrounds, and incorrect identification of the peak structures are all issues that frequently appear in the literature.
Peak overlap issues can also occur when spectra and peak shapes are not nearly as complex as shown in Figure 1. Because many elements have peaks in the low binding energy range (see the list of the potential overlaps of other photoelectron peaks with the Li 1s peak in Appendix A.4 Figure A3-3). Of particular relevance to many battery systems are some more subtle overlaps that occur when both Li and F are present. Difficulties associated with accurate quantification of LiF, which consists of two common battery components, have been examined by Brundle et al. 23 The primary challenge is associated with the loss feature from the F 2s photoelectron peaks overlapping with the Li 1s peak, see for example Figure 5 of the Brundle et al paper 23 and Section 4 below (which discusses approaches that have been used to quantify Li in spite of the peak interference). Although a LiF crystal will likely have a somewhat different peak structure (including more intense plasmon satellite peaks) than Li and F in battery materials, some of the same overlap issues will likely be present.
Reporting of factors important to data collection and data analysis and critical to the assessment of the reported data has been found to be lacking in many papers. This is true in many areas, but the basis or assumptions and parameters involved in peak fitting are often not

| IMPACT OF F AUGER SIGNALS ON NI 2P PEAKS IN NMC ELECTRODE MATERIALS
At PNNL and as reported in the literature, NMC battery electrodes have been frequently analyzed using XPS 1 to verify electrode composition and to examine the thickness and composition of SEI layers. As shown in Figure 2, there is an overlap of Auger lines from F and the Ni 2p photoelectron peaks when data is collected using X-rays from an Al anode. This overlap has been ignored or not recognized in some F I G U R E 2 XPS survey spectra for NiO, 17 Co 3 O 4 nanoparticles (NPs), 18 MnO 2 , 19 and Nafion on Si 20 collected using Al Kα X-rays that illustrate Auger overlap from both Mn LMM and F KLL with Ni 2p region of an NMC electrode.
of the published literature. We have sometimes needed to call this issue to the attention of new XPS users.
Nickel photoelectron peaks are frequently used in two ways for the analysis of these electrode materials and the SEI layers that form.
First, the visibility of the Ni peaks indicates that the SEI layer is relatively thin and/or patchy. Second, peak structures have been used to identify Ni species in the SEI layer. If the XPS photoelectron peaks have been collected using Al Kα X-rays, the Auger electrons from any F in the sample contribute to both the area intensity and structural shape of the Ni 2p spectrum, frequently leading to totally erroneous conclusions as reported in the literature. Bondarchuk et al 1 have discussed several aspects of this problem. In addition, examination of the Ni LMM peak shape is often necessary to differentiate between Ni oxides and hydroxides, 24 but the presence of the Mn 2p doublet in the same region means that this is not viable when analyzing NMC batteries. In the following, we explore the impacts of the F Auger lines on the intensity and shape of the Ni 2p signals to demonstrate the degree to which the Auger signals can impact the Ni 2p information.
The portion of a survey spectrum around the Ni 2p photoelectron peaks for NiO collected with an Al Kα anode is shown in Figure 3 along with the F Auger peaks from Nafion, LiF, and a thick SEI layer formed on NMC after many cycles. The significant overlap of these peaks, with both Ni and F present, complicates the quantification of the Ni 2p peak intensity. To demonstrate the impact of the presence of both peaks on quantification, we have created a series of artificial spectra adding the survey spectrum of F containing Nafion and NiO (shown in Figure 4) in different ratios, measured the Ni 2p, Ni 3p, and F 1s peak areas from these simulated spectra and determined the F/Ni atomic ratios using the "standard" methods of quantification and compared to the "model" atom ratios of the simulated spectra. Furthermore, these spectra were first normalized so that the areas of the Ni 2p from the NiO and the F 1s from the Nafion were the same, which is important since these data were acquired on different instruments.
where I i is the intensity of a photoelectron peak in the sample from each element and S i is the photoelectron line-specific sensitivity factor. Using this approach to determine elemental ratios simplifies the calculations needed for the atomic percent ratios. For example, F/Ni using the F 1s and Ni 2p photoelectron peaks reduces to Using this approach, the F/Ni atomic ratios were determined from areas measured for the F 1s, Ni 2p, and Ni 3p peaks as described above as shown in Figure 5, where the "measured" F/Ni ratio is plotted against the ideal ratio (that is incorporated in the simulated spectra). It is readily observed that the F/Ni 2p ratio and the ideal atomic ratios increasingly deviate as the F concentration increases.
The ratio obtained using the Ni 3p peak intensity follows the appropriate linear trend but appears higher than ideal when using the sensitivity factor included in the standard instrument package. This could be adjusted by using an instrument-and sample-determined sensitivity factor. 25 As shown in Appendix A.1, the sensitivity factors for Ni 2p and Ni 3p differ by more than a factor of 5, contributing to some uncertainty in the analysis. Also included in Figure 5 is the corrected data from the Ni 2p peak with the contribution from the F Auger lines removed, which is possible for the synthetic spectra based on a known ratio of the Auger signal strength in the window of the Ni 2p data analysis to that of the F 1s signal intensity.
F I G U R E 3 Comparison of F Auger peaks in the Ni 2p region from Nafion, a thick SEI layer, and LiF. The overlap of the F and other Auger peaks with the Ni 2p spectrum complicates and can invalidate the use of Ni 2p for analysis of Ni in SEI layers of NMC electrodes when data are collected using Al Kα X-rays.
The shapes of the signals in the Ni 2p region are shown in Figure 6.
Somewhat remarkably, the shape of the signal in this analysis window does not seem to alter significantly although the relative signal strength from the F Auger peak increases from 0% to 66% of the signal strength for the F/Ni ratio of 28.4. As suggested by the nature of the Auger signals shown in Figure 3, the F Auger line shape for F-containing materials other than Nafion would likely impact the peak shapes to a greater extent due to the covalent vs ionic nature of the bond.

F I G U R E 4
Normalized survey spectra of NiO and Nafion obtained with an Al Kα x-ray source. These were the source spectra used to create synthetic spectra with different F/Ni atomic ratios. The initial spectra were normalized as described in Appendix A.1 so that when the sum spectra were analyzed using traditional peak area and sensitivity factor analysis, the atom ratios in Table 1 were obtained.
F I G U R E 5 F/Ni2p, F/Ni3p, and F/Ni2p atomic ratios from the measured peak areas in the synthetic/model spectra. The F Auger signal was removed from the Ni 2p region in the corrected F/Ni 2p ratio by subtraction of the Auger signal expected to appear in the analyzed Ni 2p region of the spectra based on the intensity of the F 1s signal.
The information in Figure 5 clearly demonstrates the nature of the complication, but it is appropriate for an analyst to ask three additional questions: • Is this a significant problem for measurements of actual electrodes?
• Is there a simple way for me to know when I have this problem?
• The Ni 2p peaks are used because of their significant amplitude. If the Ni 2p spectra correction worked for the model data, can the corrected data be used for quantitative analysis?

| Data from NMC electrodes
To consider the extent of Ni 2p and F Auger signal overlap for actual NMC electrodes for data collected using Al Kα X-rays, we examine spectra from a freshly made NMC electrode ( Figure 1) and two sets of NMC data from used/cycled electrodes (Figure 7). Based on an analysis using F 1s and Ni 3p, the fresh electrode had a F/Ni ratio of 1.6 and the two used electrodes had ratios of $133:1 and $79:1 (these ratios were consistent when determined using only the survey spectra as shown or from higher energy resolution elemental spectra collected as indicated in Appendix A.2). Thus, for the analysis of used electrodes, the use of the Ni 2p signal is highly compromised if using data generated by an Al Kα X-ray source due to signal attenuation.
These data suggest that high F/Ni ratios will be common for used

| F concentrations at which peak overlap impacts quantification
The model (or synthetic) electrode data contained in Figure 5 can be plotted with a different x-axis and some additional data to identify F I G U R E 7 XPS survey spectra of thin SEI layers formed on electrochemically processed NMC electrodes exposed to (a) FATF electrolyte and (b) FATE electrolytes (ingredients identified in Appendix A.2).

F I G U R E 6
Ni 2p synthetic spectra of varying at F/Ni ratios, which show seemingly subtle changes in the Ni 2p structure. However, the percent area contributions of F auger with respect area ratio of F/Ni grow to be quite significant varying from 0%, 34.8% (4.1), 42.5% (6.1), 50.2% (9.3), and 58.0% (15.0) to 65.7%, for the F/Ni ratio of 28.4.
when the issue of peak overlap becomes significant. Note that the sensitivity factors for Ni 2p and F 1s are significantly different (by about a factor of 5), with the Auger signal strength small relative to the F 1s signal. Thus, for small F/Ni ratios, the impact is not large. An indication of when the impacts are significant can be estimated based on data in Figure 8 for which the x-axis is the ratio of the F 1s to Ni 3p signal strengths (area ratio). A comparison of the ideal F/Ni atom ratio and the ratio inferred using the Ni 2p measured signal suggests that if the F 1s to Ni 3p ratio is around 6 or less, the impact is small.
Also plotted in Figure 8 is the Ni 3p to Ni 2p ratio (Â10). This is included because the ratios of low BE (higher kinetic energy) and high

| Can the Ni 2p peak intensity be corrected to give useful F/Ni values?
Because the Ni 2p photoelectron peak intensity is much stronger than the Ni 3p, it is natural to want to use the Ni 2p peaks if possible. Subtraction or removal of the F Auger signal from the Ni 2p peak strength was possible from the synthetic data because the peak shape of the Auger signal was known and constant. This cannot be generally assumed to be accurate for real materials. As indicated in Figure 2, only a portion of the F Auger line overlaps the portion of the energy range frequently used to determine the Ni 2p signal strength. For an F/Ni atomic ratio of 28, 66% of the signal in the Ni 2p integration region was from the F Auger intensity. As the Auger signal increases in intensity, any corrections would be increasingly risky. Even subtle changes in the F Auger peak structure, as demonstrated in Figure 3 and discussed below, would likely invalidate the subtraction of F Auger peak intensity from the Ni 2p region. Thus, the Ni 2p peak intensity cannot be reliably corrected to give useful F/Ni values.
Three F Auger signals are plotted in Figure 3 to illustrate variances in the F Auger peak structure. The spectrum identified as an SEI layer was collected from an electrode with an SEI layer formed from one electrolyte after multiple cycles, whereas the spectrum for LiF includes components of common battery electrolytes. Although there are some similarities in peak shape for the LiF and the SEI layer, the differences are sufficient that such intensity corrections are likely to be risky. In addition, there is no certainty that the F Auger peak shape would remain constant as the SEI layer formed and evolved during multiple cycles of a battery electrode.
Independent of concern about changes in F Auger peak shapes as the SEI layer formed, we measured the intensity of the F Auger signals within the Ni 2p window of peak integration, using the same energy endpoints and background we would apply for Ni analysis, and compared them to the F 1s signal intensity. The ratios of the F Auger (in the Ni 2p window) to the F 1s signal were 0.4 for Nafion, 0.2 for LiF, and 0.4 for the SEI layer. Although the peak structure looked similar for LiF and the SEI layer, the peak shape is not a reliable indication of the ratio of the signal strength for a portion of the signal in the Ni 2p window relative to the F 1s signal strength, which was the basis of the correction applied in Figure 5.

| QUANTIFYING Li IN THE PRESENCE OF F
As discussed in detail by Brundle and Crist, 26 it might be expected that the elements with simple spectra such as Li and F would be those most easily and accurately quantified. However, these elements have proven F I G U R E 8 Peak area ratios as a function of the calculated ratio of F1s/Ni3p peak area ratio. This figure identifies the conditions for which the F/Ni 2p ratio is significantly altered by the presence of the F Auger signal. The impact of the Auger signal also shows up as a decrease in the Ni 3p/Ni 2p ratio when the F 1s/Ni 3p area ratio increases above ≈6.
to be problematic with empirical and theoretical sensitivity factors having significant differences. With careful analysis, 23,26 it was found that F has significant photoelectron intensity appearing in satellite loss features at 15-30 eV higher than the photoelectron peak, for both F 1s and F 2s. This means that loss signals from a F 2s peak, as we know a common and significant element in battery systems, have a signal appearing within the Li 1s photoelectron peak structure. The nature of the overlap issue can be understood based on an examination of  Figure 9. This linear background is just above what is assumed to be the peak of the F 2s loss feature that appears adjacent to the Li 1s peak to avoid counting the intensity from the F satellites as Li intensity. From nine analysis points across three different Li samples, the average difference in the total measured area for Li 1s was 4.4%.
Further refinement of the end and start points for the linear background reduced the average difference to 3.1%. Considering the very low RSF for Li, and thus, the high error that is associated with quantifying Li 1s, this simple method is both an improvement on the many measurements that ignore the peak overlaps and may be adequate for the analyses needed in many cases. The consistency and relative accuracy of this approach was validated by more detailed fitting based on approaches described in the literature, as shown in Appendix A.3.

| CONCLUSIONS AND RECOMMENDATIONS FOR Al Kα X-RAYS AND NMC ELECTRODES
• Battery materials are complex, and there are many erroneous XPS analyses appearing in the literature.
• There are challenges related to sample preparation and handling, data collection (getting the proper data and minimizing any damage), and, of great importance, data analysis and reporting.
References in this paper point to other discussions of many of these problems and some solutions.
• XPS data on NMC battery electrodes is a common and often source of problems and incorrect data collection and analysis.
• Survey spectra are very important as a prelude to many types of XPS analysis. A tutorial on the interpretation of XPS peaks in survey spectra provides a highly useful approach for identifying the many major and minor peaks that appear in an XPS spectrum.
Survey spectra collected before and after detailed high energyresolution measurements are useful for assessing sample damage.
• This paper explores the nature and extent of the F Auger line and Ni 2p peak overlaps using synthetic spectra to suggest that a simple analysis of survey spectra can be used to indicate when the impacts are significant.
• Although the Ni 2p peak is commonly used to quantify Ni composition, the data from NMC electrodes exposed to electrolytes demonstrate that this peak is highly altered by F Auger lines and is often not useful for quantification or determination of chemical state when the data is collected using Al Kα X-rays. Although the signal strength is smaller, the Ni 3p has less peak interference and is often useful for quantitative analysis.
• There are a few challenges that are unique to analyzing Li metal electrodes. In addition to being highly reactive and thus able to rapidly change state even within the ultra-high vacuum environment of a typical spectrometer, analysts should take note of how they mount their samples to avoid "electroplating." Quantification values for Li 1s can be artificially inflated in the presence of F if satellite peaks associated with F 2s are ignored. A simplified approximate method for minimizing this problem was discussed.
F I G U R E 9 XPS spectra from LiF. The top spectrum extends from 0 eV BE to 120 eV, which includes the valance band, the F 2s and Li 1s photoelectron peaks, and their assorted loss lines. The bottom spectrum is from the F 1s region with the energy shifted such that the F 1s and F 2s peaks align for the purpose of demonstrating the position of the F loss lines relative to the main core line, that is, the relative BE difference will be the same for loss lines relative to F 2s and F 1s. The overlap enables loss features from F to be identified and shows that these features overlap and interfere with the Li 1s photoelectron peak. Brundle et al 23 fit the peaks labeled 2 and 3 to remove them from the Li 1s peak area. The dashed line in red shown in the top spectrum shows an approximate but simple approach for background placement to extract the intensity of this Li peak in the presence of F used at CISRO (also see inset of Figure A3-2). Figure adapted from Brundle and Crist. 26 And for further reading on the F 1s and F 2s loss lines, please consult this citation.

ACKNOWLEDGMENTS
We thank Prof. Alberto Herrera-Gomez for providing the NiO spectrum used in this study. The LiF spectrum from which the Li Auger signal is shown in Figure 3 is from the database contained in

SUPPORTING INFORMATION
Additional supporting information can be found online in the Supporting Information section at the end of this article. Peak areas and calculated atom ratios for the synthesized and NMC electrode spectra discussed in this short paper are summarized in Table 1. All spectra, while on different instruments, were collected using Al Kα X-rays. The NiO spectrum used in the simulation was provided by Alberto Herrera-Gomez with details published in a recent paper. 27 The Nafion spectrum was collected at PNNL on a Thermo-Fisher NEXSA system. These spectra were first normalized so that the areas of the Ni 2p from the NiO and the F 1s from the Nafion were of the same intensity. These intensities were then further normalized by the sensitivity factors for Ni2p and F1s respectively. These spectra were then added so that the Ni2p and F peak area ratios were those listed in Table 1. From these simulated or synthesized spectra, the areas from the Ni2p region, and the Ni 3p and F 1s photoelectron peaks were extracted as listed in the table. The quantification processes in the Thermo Fisher NEXSA system were used to quantify the electrode compositions, after paying attention to peak overlap issues.
We note that the software in the NEXSA system uses theoretical sensitivity factors based on cross-sections for producing photoelectrons.
In their reported analysis results they normalize the measured peak intensities by the relevant attenuation length to get the peak intensities used in Equations 1 and 2. Experimental sensitivity factors effectively include these attenuation lengths and are applied directly to the measured peak areas. The ratios of the sensitivity factors we used for the simulated spectra are equivalent to those in CasaXPS and reduced to the following effective sensitivity factor values (normalizing to S F = 1, with the attenuation length incorporated into the sensitivity factors) as follows: 1 for F 1s, 4.28 for Ni 2p, and 0.774 for Ni 3p.

A.2 | Details of NMC and Nafion measurements
The Nafion and NMC electrode measurements were collected on a Thermo-Fisher NEXSA system. The samples analyzed were mounted using double-sided carbon tape. Charge neutralization was used with no beam voltage, 150 mA emission, and a 40.0 kV extractor potential.
The narrow scan spectra were collected using a pass energy of 50 eV with a step size of 0.1 eV. For the Ag 3d 5/2 line, these conditions produced an FWHM of 0.84 eV ± 0.02 eV. The binding energy (BE) scale is calibrated using the Cu 2p 3/2 feature at 932.6 eV and Au 4f 7/2 at 84.0 eV. that B-Ag 3d 5/2 .
The samples for the spectra shown in Figure 7 were prepared for XPS by exposure to the electrolyte (and cycled 500 times).
These two samples were exposed to similar but slightly T A B L E 1 Peak areas with corresponding peak area ratios for the synthesized NMC spectra. The fit developed for F 1s was then copied to the spectrum of the F 2s and Li 1s region. The peak positions of the components associated with the satellite structure were fixed relative to the main peak to ensure that they remained the same distance when transferred to the new spectrum. An additional three components were added around F 2s to account for the overlapping satellite structure that originates from the valence band region, and a component was added to account for the intensity from Li 1s, as shown in Figure A3 This dashed line effectively represents the "true" background for Li 1s or as close as we can expect to get with this approach employing a fit.
Compared with the dotted red line in the same inset, which represents the simplified linear background for Li 1s as discussed in the main text, visually, it appears that there would be minimal difference in the peak area for Li 1s using the two approaches. To confirm that the simplified linear background approach was viable, the difference in peak area for Li 1s using the two approaches was calculated for  Figure A3-3, along with the Li 1s photoelectron peak from an SEI layer.
F I G U R E A 3 -1 Selected, representative fitted high-resolution F 1s spectrum from LiF. The components in red (dashed line) represent the intensity associated with the satellite structure that overlaps with the L 1s region when this satellite structure is shifted equidistance from F 2s. "*" denotes the red component where the area intensity was fixed as a multiple of the unmarked red component. The inset is of the same spectrum with different X-and Y-axis limits applied to facilitate observation of the fine structure of the satellite region.
F I G U R E A 3 -2 Selected, representative fitted high-resolution F 2s and L 1s spectrum from LiF. The fit developed for F 1s was transferred to this spectrum, with additional components added around F 2s to account for the satellite structure that originates from the valence band region, and a component (dotted line in yellow) for the intensity associated with Li 1s. "*" denotes the red component where the area intensity was fixed as a multiple of the unmarked red component. The inset is of the same spectrum with different X-and Y-axis limits applied to facilitate observation of the Li 1s peak background, where the dashed line in black represents the fit envelope of the fit for the satellite structure (or "true" background) and the dotted line in red represents the linear background that is also presented in Figure 9 of the main text.
F I G U R E A 3 -3 Li 1s region has potential overlaps with many other elements including a few common ones that may appear, such as Fe, in a battery system. The Li 1s peak from an SEI layer is shown along with typical binding energy ranges for some common elements to illustrate possible spectral overlap.