Probing the Mysterious Behavior of Tungsten as a Dopant Inside Pristine Cobalt‐Free Nickel‐Rich Cathode Materials

Nickel‐rich cathode materials with small amounts of tungsten (W) dopants have attracted extensive attention in recent years. However, the chemical state, crystalline form, compound chemistry, and location of W in these layered cathodes are still not well‐understood. In this study, these missing structural properties are determined through a combination of macro‐, to atomic‐sensitive characterization techniques and density functional theory (DFT). W‐doped LiNiO2 (LNO) particles, prepared with mechanofusion and coprecipitation methods, are used to probe changes in the structure and location of W‐species. The results indicate that W is mainly distributed on the surfaces and inside grain boundaries of the secondary particles, regardless of the doping method. Electron energy loss spectroscopy (EELS) mapping confirms the simultaneous presence of W, O, with and without Ni in the grain boundaries as well as W‐ and O‐rich regions on the very surface. The W‐rich areas inside the grain boundaries are found to be in two forms, crystalline and amorphous. This paper suggests the presence of kinetically stabilized‐Li4+xNi1‐xWO6 (x = 0, 0.1) with the possibility of LixWyOz phases in LNO which are consistent with the electron microscopy, X‐ray absorption and diffraction data. The multiple roles of W in this complex microstructure are discussed considering the W distribution.


Introduction
High energy density lithium-ion batteries (LIBs) are the best potential energy storage system for electric vehicles and to support a more sustainable and green future for the planet. [1][2][3][4] To fulfill commercial expectations, these batteries require higher capacity, cyclic lifetime, rate capability, and thermal stability. [5][6][7][8] Besides many attempts to enhance the main components in Li-ion batteries by discovering new materials and adjusting currently available materials, the cathode is still the main limiting factor for energy density and is the highest price component of a Li-ion cell. [1,9,10] Layered cathode materials are common in commercial applications. In these materials, higher levels of nickel lead to higher energy densities. [11,12] However, Ni enrichment brings several problems, such as decreased cycling life span, safety, and thermal stability, leading to quicker failure. [2,[13][14][15] A new generation of optimized engineered complex cathode structures have emerged, utilizing ultra-high Ni cathode materials, along with various methods like doping, core-shell structures, and coatings. These approaches can be categorized as either bulk or surface modification strategies, with the ultimate goal for all being to stabilize the active material and/or to increase its lifetime during charge-discharge cycling. [4,[16][17][18][19][20][21][22][23][24] Atomic doping is one of the most conventional methods to alter the cathode host crystal and reinforce its structure. [19] In this regard, tremendous work has been done to identify the best dopant candidates and some cations with a high-valence state have drawn significant attention. [13,[25][26][27] Using high oxidation state cations like Ti 4+ , Zr 4+ , Ta 5+ , Mo +6 , W +6 , etc., would stabilize the Ni-rich cathode structures, alleviate Jahn-Taller active elements (Ni +3 ), suppress the diffusion of Ni 2+ , protect the cathode from side reactions with the electrolyte, or prevent undesirable phase transformation. [17,[28][29][30] Lately, a family of W-doped Ni-rich cathode materials has demonstrated significantly improved LIB performance, as well as thermal and structural stabilities. [31][32][33] These recent studies have pointed out a few general effects of W on the behavior of LIBs and proposed some hypotheses as to what drives the observed enhancements. However, no comprehensive investigation has been conducted thus far to experimentally validate some of the hypotheses found in the literature. The high-valence charge and the complex electronic configuration of W in comparison with other common cation dopants, make the prediction of its role in the layered structure more complicated. [34] Therefore, understanding the W role would require identifying the location and distribution of W in bulk or surface, its effect on the structure of the layered material after synthesis, and the presence of possible W-phases inside the Ni-rich cathode structure.
In recently published papers, [34][35][36] the general aspects of performance enhancement and structural modification have been the focus of discussion. From our previous work, [34][35][36] a Nirich cathode material with only 1% W dopant showed the best cycling behavior and fracture resistance among the tested materials as compared to other doping levels. It is worth mentioning that based on our previous paper, the optimum heat treatment temperature to reach a better W-enriched structure and some level of heterogeneity regarding the W distribution in the host structure have been reported. [36] To better understand the role of W in battery performance, this article presents in-depth characterization results and therefore reveals the preferred location of W, its forms as well as new possible W-variants in the doped-LiNiO 2 (LNO) cathode materials. For this purpose, both X-raybased and electron-based characterization techniques, as well as modeling, have been used. These characterization techniques, alongside simulation and fitting of experimental data, provide sufficient information to predict and experimentally identify the location of W inside the cathode. Our results show that W is detected within the grain boundaries of the secondary particles and forming a thin layer, few nm thickness, on the edge of primary particles adjacent to these grain boundaries. Additionally, W is also found on the surface of secondary particles. As will be discussed in detail below, W-rich compounds which contain Ni with chemical composition Li 4+x Ni 1-x WO 6 (x = 0 and 0.1 which are kinetically stabilized compounds) along with the probability of Li x W y O z phases, with the preferred concentration on the top surface and inside the grain boundaries of LNO structure, play a protective layer role as well as stress absorber medium, respectively.

Synchrotron-based X-ray Diffraction (SXRD) and Pair Distribution Function (PDF)
The SXRD powder data for LNO doped with different amounts of W is illustrated in Figure 1a. The patterns quantification in Table 1 shows that both lattice volume and cation mixing (amount of Ni in the Li layer inside the LNO) are increased with increased W dopants. Both of these changes are consistent with the presence of more Ni 2+ , with a larger ionic radius than Ni 3+ in the transition metal (TM) sites, and in agreement with other published works. [31][32][33] As shown in Figure 1b, for higher amounts of W, additional peaks can be observed at 2θ angles between 19° to 35°, which is consistent with either a lower symmetry structure in additional phases or with superlattice reflections in LNO. Additionally, the presence of a broad hump in the scattering distribution ≈19° to 35° in doped materials suggests the existence of an amorphous phase. However, in the LNO_pristine sample, the detection of a weak hump might be caused by the capillary tube scattering.
According to Figure 1a, the major peaks can be indexed with high precision using the α-NaFeO2 (R3̅ m) structure. [37] For lower dopant amounts, namely LNO_1%W_mech, a clear separation between the (006), (012) and (018), (110) peaks, along with the higher intensity of the (003) peak, are consistent with a clear layered structure. The ratio of (003) (104) I I , identified as the R-value, has been used to track the amount of cation mixing inside the layered structure. Based on the literature, values below 1.2 are indicative of undesirable cation mixing due to more disordering or the formation of inactive cubic phase of LNO in the layered structures, which reduces electrochemical performance. [38][39][40] Our results show that the R-value decreased from LNO (≈1.37) to 1% W (≈1.34) and to 8% W (≈0.76), indicating higher levels of cation mixing in the higher W-doped LNO structures and for samples prepared by coprecipitation. Rietveld refinement of the patterns provides a much more accurate way of calculating the amount of Ni in Li layers sites, as presented in Table 1. The phase fractions of the Li 4 NiWO 6 (see further below), the lattice parameters of LNO and the atomic occupancy of Li and Ni in Li sites were refined, and the R-value was extracted directly from the raw XRD patterns. In general, the absence of the (003) peak is consistent with the Fm(-)3m phase rather than the layered structure. [15] In Figure 1a, the intensity of the (003) peak was significantly reduced in W-rich structures. Work in the literature suggested that this could be due to the presence of a rock salt/spinel phase based on electron diffraction patterns, caused by the presence of W in the structure of Ni-rich NMC (LiNi x Co y Mn 1-x-y O 2 ). [31,33] As shown below, however, the changes induced by the presence of W alter the local symmetry at the edge of particles where W is detected (see section 2.2). Also, the peak broadening noticeably increased with increasing the W content. This effect could either be caused by higher lattice strain induced by W, if present in the lattice, or by smaller particle size, as reported by other researchers. [33] Our previous work [36] showed that higher levels of W-doping in LNO materials generates smaller primary particles or, in other words, there are more grain boundaries inside their secondary particles.
According to Figure 1a, for the higher W content (e.g., 8% W) material, the (110) peak shifted to lower angles, which could be ascribed to there being a higher amount of larger radius ions in the TM planes in the W-rich samples. One possible explanation for this is the presence of more Ni 2+ (0.69 Å) with larger radii than Ni 3+ (0.56 Å) in doped materials. Some of the previous literature suggested that the W 6+ ions tend to occupy the TM sites and reduce some of the Ni +3 to Ni 2+ ions to maintain charge balance, which inevitably increases cation mixing due to the similar ionic radii of Ni 2+ and Li + (0.72 Å). In our density functional theory (DFT) work, we considered three different scenarios to maintain the charge balance when W +6 is in the LNO structure: 1) one Ni 3+ vacancy for each W +6 , 2) three atoms single valence reductions by formation of three Ni 2+ ions with and without their replacement in Li sites, and 3) formation of three Li + vacancies. However, our DFT calculations show that none of these configurations is energetically favorable. Additionally, the uniform presence of W in the LNO could not be entirely  by the Extended X-ray Absorption Fine Structure (EXAFS) data analysis, which will be discussed in section 2.3.
Other investigations indicated that the presence of XRD peaks in the range of 20° and 35° (2ϴ range based on λCuKα) originated from the ordering in the TM layers with Li 2 MnO 3type structure. For the Li 2 MnO 3-type structure, different layer stacking along the c-direction might be possible, leading to various space groups such as C2/m, C2/c, and P3 1 12. However, the C2/m system is more energetically favored (albeit by small margin) and results in a better fit for the Li 2 MnO 3-type structure. [41] Other work also reported the presence of a tiny broad peak at larger angles close to the (003) peak possibly arising from [√3a hex × √3a hex ] R30° superlattice ordering resulting from stacking faults alongside the c-direction. [42] On the other hand, these extra peaks might originate from minute contents of different W phases or variants present within the doped LNO particles as second phases. To validate this possibility, various probable W compounds were considered either based on the DFT calculations (predicting which phases are energetically and thermodynamically favored) or based on existing W compounds, accounting for the combination of elements (W, O, Li, Ni) and similar space groups (C2/m, C2/c, Cm,…) to the Li 2 MnO 3-type structure that were used above to fit the tiny peaks in XRD patterns.
Li 4 NiWO 6 and Li 4.1 Ni 0.9 WO 6, with C2/m and Cm space group symmetry, respectively, have been found to improve the fit quality of the XRD patterns beyond the bulk LNO phase. As demonstrated further below, this is also consistent with simulations to fit the X-ray absorption fine structure (XAFS) modulations from the W atoms' environments in the doped materials. To confirm these results with the XRD refinement, the LNO together with either one or both of these phases were studied.  Due to the similarity of XRD peak positions in these two W-variants, especially between 20° and 35° (based on the λ CuKα ) 2ϴ range, the fitting residuals were considered for different scenarios. First, the fitting was carried out considering LNO with one of the compounds (either Li 4 NiWO 6 or Li 4.1 Ni 0.9 WO 6 ) at a time, then considering LNO with both phases at the same time. While the fitting residues were relatively low and approximately similar in both cases, the concentrations yielded some unphysical outputs when fitting the peaks with both phases at the time. Based on this result and the similarity of the residues, the fitting of LNO with only one of the phases was considered. The results of fitting with LNO and Li 4 NiWO 6 are represented in Table 1. While there is a quantitative improvement of the residuals of fit, it is also clear that a perfect match to the full series of SXRD peaks is not obtained as illustrated in Figure S1 (Supporting Information) especially for lower W amounts. This suggests a combination of phases or of changes in the structure induced by multiple effects that are not accounted for by a simple Rietveld refinement.
Further insight into the complexity of the doped material and how the main lattice of the LNO phase is affected can be obtained from PDF analysis. PDF analysis provides an average distribution of pairs of interatomic distances. This information would make it possible to detect if interatomic spacings and order change. By inspecting the short and medium spacing ranges, we can deduce that the average structure of the doped compounds (with the concentration of W used in this work) shows no detectable changes in the PDF (such as clear extra peaks) as compared to the pristine material (see SI document for completeness). These results agree with SXRD analysis due to the low concentration of W-variants inside the LNO.

Scanning Transmission Electron Microscopy (STEM) and Electron Energy Loss Spectroscopy (EELS)
High-resolution-STEM images of the doped samples in their pristine state, from both the mechanofusion process and coprecipitation synthesis, are shown in Figure 2a-c. The contrast mechanism, in high-angle annular dark field (HAADF) mode of detection, is related to the atomic number and the thickness of regions that are scattering electrons and thus the brighter areas denote heavier elements or thicker regions of the sample. These high-resolution images, Figure 2a,b, mainly show the same apparent behavior at grain boundaries: two bright bands (mostly crystalline) separated by a dark region in between. Since W is the heaviest element inside these materials, based on the intensities in the images, the W-rich area would be located in those bright band regions. Because the dark areas could also imply a different thickness, a more detailed spectroscopic analysis with EELS is thus required. From inspection, however, two conclusions can be drawn: 1) there is W diffusion along grain boundaries between the primary particles, very deep in the structure of secondary particles (this is more evident in Figure 4a, and 2) W is located at grain boundaries but it is not clear if W is also in the structure, given that the concentration of W away from the boundaries is lower. In other words, these images demonstrate that W diffuses through the secondary particles and is mainly located inside the grain boundaries. Further analysis using Fourier Transform (FT) and crystal distance matching was performed to thoroughly understand the high-resolution atomic lattice images, as will be discussed in the following.
Detailed atomic resolution imaging of a primary particle (in the core part of the secondary particle, Figure 3), near a grain boundary and including a region further away, was carried out in order to probe in detail the changes to the structure induced by the presence of W. First, an area further from the grain boundary was selected, marked "I" in Figure 3a. The FT calculation providing a local numerical diffraction pattern and the intensity profiles from two different atomic planes confirm that the <010> zone axis of LNO for region "I" is presented in Figure 3b,c. Based on the line profile (Figure 3c), one can notice the high-intensity TM planes and troughs corresponding to the Li planes. However, within the Li planes where there should be very clear intensity minima since the Li does not contribute to the HAADF signal, there are additional intensity maxima (marked by orange arrows), locally consistent with cation mixing corresponding to either Ni or possible W atoms. Since the concentration of W is low and the intensity within the Li layers is quite uniform within these planes, this suggests that, in these regions, Ni would be present within the Li layers, consistent with results from structure refinement where some cation mixing is deduced. In the <010> zone axis, the distances between the atomic columns from the line profiles match well with the spacing between the Li and Ni sites, as shown in Figure 3d. In other areas, however, no cation mixing is detected, and the pristine layered structure is visible, which is shown in the purple dotted line box in Figure 3a. This also suggests that this effect is not an electron beam propagation artefact due to the channeling of the electron beam to the TM columns, since the thickness of the two regions would be very similar. Very clear from Figure 3a is also a very bright crystalline region (identified as region "II") at the edge of the grain, adjacent to the grain boundary. Based on the intensity of the atomic columns, it is expected that this is due to the presence of W. As will be explored further in the discussion of the EELS mapping, this W-rich area also contains O and Ni. The image of the numerical diffraction with FT from this bright area in Figure 3b does not show layered structural phase symmetry. Prior literature has shown that a spinel phase (cubic symmetry) which only originates from cation mixing [43,44] can appear after electrochemical cycling on the edge of primary particles. Here, our results illustrate that, although there is cation mixing between the Ni and Li atoms, the precise measurements with EELS (shown further below) reveal more information about the concentration of W in such non-layered phase areas and demonstrate the presence of some W phase variants on the very surface of the primary particles.
XRD refinement results (as well as the subsequent validation by EELS mapping and XAFS fitting) indicate a better fit to the pattern with the presence of Li, Ni, W, and O phase, which suggests that the crystalline bright area "II" could be either the stoichiometric compound Li 4 NiWO 6 or its solid solution Li 4.1 Ni 0.9 WO 6 . Further clarification is obtained from the highresolution HAADF image through the numerical FT diffraction from region "II" (Figure 3b), which is perfectly consistent with the <311> zone axis direction of either of these two W-compounds. Furthermore, from the line profiles and the calculated atomic distance between W and Ni atoms (Figure 3c,d), the plane spacings are in good agreement with the expectations from this zone axis. For this comparison, the spacing between the W and Ni atomic positions have been considered since they are the most visible sites from the HAADF intensities. However, as demonstrated in Figure 3d, these W variants show almost identical atomic distances in this zone axis. It is therefore impossible to differentiate between them even by detailed local analysis. The stoichiometric compound Li 4 NiWO 6 with C2/m symmetry has three shared sites (4g, 4h and 2d Wyckoff positions) for Ni and Li with Ni occupancies between 16% to 26% and a single site for W (2a Wyckoff position); however, in the solid solution of a general composition Li 4+x Ni 1-x WO 6 (in this specific case corresponding to a previously-identified compound with composition  The low magnification EELS maps from different doped materials, regardless of doping methods (Figure 4a,b) reveal that W is mainly concentrated on the very surface of the primary and secondary particles as well as at the grain boundaries in between the primary particles. Also, the W concentration inside the deeper level of the secondary particle, especially in the grain boundary, seems to increase with more W dopant. These results are consistent with the previously presented XRD peak broadening since smaller primary particles provide more accessible pathways for W to reach inside the secondary particles through their grain boundaries. This reduction of primary grain size along with the presence of W in the grain boundaries, improves the strength of this family of doped structure and increase their resistance toward cracking. [35] Having W-rich areas on the surface of the secondary particles can play a protective role in reducing the direct contact between the Ni-rich cathode materials and the electrolyte. A second role of W, given this microstructure, in addition to strengthening the grain boundaries, would be acting as a damping medium for the  cycling stress due to the presence of W-rich areas. The detailed experimental discussion regarding the protective role from side reactions during cycling, along with strengthening of the grain boundaries and damping cycling stress through the presence of W-variants, are out of this paper's scope and will be discussed in more depth in a follow-up work focused on the effect of W on the degradation mechanisms through electrochemical cycling. For in-depth characterization of the form and location of W inside the doped materials, higher magnification EELS mapping was implemented.
A high-magnification EELS map and the related Annular Dark-Field (ADF) image are presented in Figure 4c (as well as Figure S4a, Supporting Information), which illustrate that the W is noticeably present in regions within grain boundaries of the secondary particles with an amorphous structure. Additionally, the W signal is still apparent within a few nanometers of the surface of the primary particles within the crystalline atomic arrangement of the grain. Furthermore, there are marked differences between the O K-near-edge structure of bulk areas and those containing W, visible in the pre-edge peak feature. This pre-edge peak is very clearly detectable in the inner parts of the primary particles and is consistent with the results for pure LNO, which has a sharp and very clearly visible pre-edge peak. In Ni-rich materials, the general expectation for the O K-edge peak's shape is the appearance of the pre-edge peak arising from the transition between O 1s and the hybridized state of O 2p with Ni 3d states. In addition, the other features over 534 eV are due to the transition from O 1s to the hybridized state of O 2p with Ni 4sp states. Based on the literature, changes in the crystal structure through cation mixing could substantially affect the hybridization state between TM 3d and O 2p bands, and induce changes in the O pre-edge peak. [45] In our case, the alteration in the mentioned hybridization state seems to be induced by changes in the chemical environment as a result of the W presence. The pre-edge peak intensity decreases from the core of particles (i.e., away from the grain boundaries) to the grain boundary until the pre-edge finally disappeared inside the grain boundaries where the shape is significantly different, as shown in more detail in Figures S4b,S5b (Supporting Information). No spectroscopic evidence indicates an O K-edge consistent with WO 3 . Based on Figure S4a   of MLLS fitting results is illustrated in Figure S4d (Supporting Information). Further evidence obtained from the secondary particle, Figure S.6, also shows the presence of a ≈10 nm thick region on the top of the surface with W-and O-rich layers without any Ni element. Moreover, Energy-Dispersive X-ray Spectroscopy (EDXS) maps were also acquired and are consistent with the segregation as detected with EELS ( Figure S7, Supporting Information).
The high level of noise in EELS maps and the line profiles is due to the use of a very low electron beam current in order to minimize radiation damage. As illustrated in Figure 4d, the W, Ni, and O elements were detected near the primary particle perimeters, which encouraged the authors to perform a complementary analysis in addition to the EELS measurement to provide further insights on the W phase(s). XAFS analysis was utilized to fulfill this demand. The strengths of XAFS over EELS analysis are its better sensitivity to low concentrations and the ability to acquire the higher energy peaks of W. In EELS, only the tungsten M-edge, with its delayed and broad shape is accessible, but in XAFS, the W L 3 -edge is easily captured.

XAFS analysis
The normalized X-ray absorption near-edge structures (XANES) spectra of eight samples, shown in Figure S8 (Supporting Information), reveal a double peak feature in all the doped W materials regardless of the doping amount (1%, 2%, 4%, 8%) and the doping approach. The energy position of the W L 3 -edge in doped materials is nearly overlapping with the WO 3 reference, which was attributed to the expected oxidation state of tungsten (W +6 ). In other research, the W L 3 -edge peak in both WO 3 and WO 2 showed a single peak shape as well. [46] However, these double-peak shapes can result from either a distorted environment around the W or the coexistence of multiple W species inside the doped samples. To fit the XAFS data, two different hypotheses were considered as is discussed below.
As a first consideration, it was assumed that W is uniformly doped in the LNO structure, and these doped structures are described by considering different scenarios calculated by DFT. Among all doped structural assumptions, W occupies the Ni site with octahedral geometry. To balance the tungsten's extra charge and to maintain neutrality, several possibilities were considered, as discussed above, even though these were not energetically favorable [34] (i.e., 1) one Ni 3+ or 2) three Li + vacancies as well as 3) three reductions of Ni +3 to Ni +2 with and without exchanging position with 3 Li + ). In addition, an experimental comparison of the Radial Distribution Function (RDF) around Ni and W atoms, based on EXAFS data derived from the Ni K-edge from pure LNO and the W L 3 -edge from W doped LNO, was made to determine if these two distributions would clarify the same environment for the Ni and W species. In Figure 5a, Ni sites, Ni1 (on a Li site) and Ni2 (on a Ni site) and the experimental data from the LNO material are demonstrated. The RDF peaks from Feff modeling can be used to identify the major differences between two Ni local structural environments, notably after the first two shells (up to ≈3Å, identified as features "I" and "II"). The main difference between these two sites is the presence of features "III" and "V" which are exclusively specific for sites Ni1 and Ni2, respectively. The "V" feature is comparatively less practical than "III" since it is located at a higher R value and is influenced by the Debye-Waller effect in the experimental data. According to Figure 5a, the experimental data can be described very well by the model with Ni only at Ni2 site and is congruous with the expectation based on structure refinement from XRD data showing low cation mixing. Considering these fingerprint features, the comparison between the two experimental RDF patterns was carried out as depicted in Figure 5b. According to Figure 5b, both materials reveal the same location for the first peak with good first shell match. However, the significantly weaker peak intensity of the "II" peak in the doped material, with respect to the second neighboring shell of W, indicates that the W atoms' 2nd shell coordination is either incomplete or W is located at the surface site. Also, the missing "IV" and "V" peaks might be resulting from having large site disordering at the W occupied site, which is consistent with the "II" peak's low intensity. Consequently, the uniform distribution of W inside the Ni site of the LNO, cannot be experimentally justified.
Since the EELS maps demonstrate the presence of W within more localized areas and not simply uniform doping through LNO, a second configuration was investigated in which W-rich compounds being found inside the doped materials is assumed, and the impact on the consistency of the EELS and XANES was investigated. The compounds considered here are suggested based on DFT calculation or are previously published phases. Furthermore, four reference compounds (Li 2 WO 4 , Li 2 W 2 O 7 , Li 4 WO 5, and NiWO 4 ) were synthesized and examined through the same experimental methods, Figure S9 (Supporting Information), to compare data with the doped materials. Normalized XANES features were used for the linear combination fitting (LCF) analysis since the preliminary matching results from the 1st derivative XANES showed the same major answer as the normalized one. In the end, eight possible compounds were compared, with either different chemistry or crystal symmetry for the final fitting. All eight structures are shown in Figure S10 (Supporting Information). The main spectral features considered for identifying the candidate phases are a double-peak shape in the XANES, presenting almost the same intensity ratio between the first and the second peak as doped samples' experimental data, or compounds containing simultaneously W, O, and sometime Ni elements in which by combining their individual XANES features, the best fitting toward the experimental data could be achieved. During the final fitting step (fitting the range −20 to 70 eV with respect to the absorption edge), whenever the edge energy E 0 difference was larger than +/-10 eV, that compound was eliminated since such a large discrepancy in the energy calibration is unlikely.
Comparing the XANES features of the eight phases (247 total possible combinations), the stoichiometric compound Li 4 NiWO 6, and the solid solution Li 4.1 Ni 0.9 WO 6 , presented the most similarities of their W edge with the doped materials. Though they are not thermodynamically stable, these two compounds seem kinetically favorable. The final fitting results and their phase weightings between two finalized LCF answers are displayed in the Table S2 (Supporting Information).
The experimentally resolved XANES white-line has two main features: 1) a double peak fine structure, and 2) a signal intensity trend with the lower energy peak stronger than the higher energy one. According to the simulations in Figure S11 (Supporting Information), the double peak white-line structure and its intensity trend are reproduced by LCF fitting. However, the best fit does not perfectly overlap with the experimental data. The variance, specifically for the relative intensity feature, could be driven by: 1) the subtle structural differences in the first shell WO bond distances, bond angles, and the further beyond local structure environments between those actual W species in samples and the models which guided the XANES fitting, and/or 2) certain code-related defects in the XANES modeling. For these reasons, based just on XANES simulations, one cannot completely exclude the possibility of other coexisting W species besides these picked ones within the doped material structures. As discussed in the EELS sections, based on Figures S4a,S6e (Supporting Information), there is clearly a lack of Ni element inside some W-rich areas. Considering several possible Li x W y O z compounds in the XAFS fitting, none of them has been selected to better match the spectra. This may originate from incomplete matches between these models and the actual Li x Ni y O z compounds in the doped materials. The experimental data from the four mentioned reference compounds have been considered for LCF analysis to investigate this effect further. According to Figure S12 (Supporting Information), the best fittings in the presented doped samples show a calculated phase weight of >0.84 for the Li 4 WO 5 reference phase. However, the best match between the experimental and LCF results in LNO_1% W_Mech, considering only these four reference compounds, does not seem entirely enough to describe all the possible W species inside the rest of the doped samples. On the other hand, the Li 4 WO 5 reference compound was not a single phase based on our powder XRD data. Therefore, to consider this compound as a potential candidate through the final XANES modeling (based on the crystal structure files), phase identification from XRD analysis and then separate XANES modeling have been made to validate which of the identified phases would be a better representative of this reference compound. As can be seen in Figure S13 (Supporting Information), two Li 4 WO 5 phases, with P-1 and Fm-3m space groups, reveal double peak XANES features. However, none of the simulations of these two phases could completely reproduce the full experimental features: the correct intensity ratio of the first two peaks, and the features at higher energy of the Li 4 WO 5 reference compound. This mismatch could be due to the discrepancies between the XANES experiments and fitting for either of these two considered Li 4 WO 5 models. We can therefore conclude that a combination of W-rich phases (with and without Ni) would be required to explain the XANES features through fitting. Some possible factors contributing to these discrepancies are the unknown relative fraction of phases with Ni and without Ni, as well as the amorphous nature of the Ni-deficient phase due to the changes this would generate on the XANES.
The following reactions present the energetically favorable products regarding the presence of a deficient (1) or sufficient (2) amount of Li source in the environment.

Li NiWO Li
For the first reaction, the contact between the inner side of the LNO particle with excess Ni (1/32) in the Li layer, and the surface of the Li 4 NiWO 6 , is considered. In both reactions, the entropy of O 2 , which depends on pressure and temperature, was not taken into account for ∆E.
On the one hand, those W-rich regions between the primary grains might lack access to sufficient oxygen or Li sources to form more stable compounds. On the other hand, the heat treatment temperature here was above the melting point of the Li x W y O z compounds, and if they were produced at the beginning, they would melt and therefore be able to diffuse through the grain boundaries of the secondary particles and after that react with the LNO. Then the Ni/W ratio in those small gaps would be much different from the synthesis condition, making these Li 4+x Ni 1-x WO 6 compounds metastable. Other studies showed that the essential parameters for synthesizing Li 4 NiWO 6 were the amount of Ni/W ratio as well as the extra amount of precursor combination (Li 2 CO 3 , NiO, and WO 3 ) for eliminating the Li deficiency. [46] Furthermore, under high temperature and low O 2 partial pressure during synthesis, the ∆G = ∆E-T∆S could turn positive.
It is worth noting that the heterogeneous distribution of W and the presence of W-containing compounds, regardless of used synthesis methods in this research, reached the same results. However, further investigation on the cycled particles demonstrated that the high amount of porosities in coprecipitated particles which are visible in Figure 2c and Figure 4b, made these particles more vulnerable to electrolyte infiltration and faster capacity degradation. An unoptimized protocol causes this effect for our coprecipitated samples, which are discussed in more detail in our future work.

Conclusion
The role of W in the crystalline structure and its elemental distribution on cathode materials when used as a dopant in LNO was investigated using electron microscopy, X-ray scattering, and X-ray spectroscopy methods. SXRD analysis showed the appearance of minor peaks consistent with the additional presence of minor fractions of the stoichiometric compound Li 4 NiWO 6 or the solid solution form Li 4+x Ni 1-x WO 6 consistent with the Li 4.1 Ni 0.9 WO 6 phase in doped LNO. However, not all minor features can be fully explained by structural refinement of the powder patterns with these phases. Synchrotron-based PDF analysis revealed no significant structural differences induced by doping. Electron microscopy analysis with STEM imaging along with numerical diffraction of atomic-resolved images from nanometer scale areas and EELS, however, showed very clearly the presence of W in two main distinct types of regions: at grain boundaries and in their proximity on the surfaces of primary particles, as well as, on the surface of secondary particles. Within crystalline regions at the surface of primary particles, the W presence within a few ≈1-3 nm of the grain boundaries was deduced and showed a pattern consistent with either Li 4 NiWO 6 or Li 4.1 Ni 0.9 WO 6 phase structure. Detailed high-resolution EELS analysis mapping also provided more thorough information regarding the spatial distribution of W, demonstrating that this element was mostly in crystalline regions on the primary particles' perimeter (together with Ni) and within grain boundaries of the secondary particles where W-rich amorphous phases (sometime without Ni) were also found, thus indicating the presence of Li x W y O z compounds at the center of the boundaries. Additionally, these Li x W y O z compounds might also be present on the very surface of the secondary particles. Therefore, W was found to propagate during the synthesis along the grain boundaries, well within the core of the secondary particles, even if WO 3 had been initially placed on the surface of the particles. From the spatially resolved spectroscopic mapping of the O K-edge, the bonding environment of oxygen atoms at the grain boundaries was different and not consistent with WO 3 . Spectral fitting of the W L 3-edge XANES indicated that the spectral features could not be solely explained with the Li 4+x Ni 1-x WO 6 phases (x = 0 and 0.1) but must contain other compounds, such as Li x W y O z phases as evident from EELS. Due to different environmental conditions inside the grain boundaries and the surface of the primary particles, these metastable Li 4+x Ni 1-x WO 6 phases (x = 0 and 0.1) in their bulk form, might be stabilized and kinetically favored. These results demonstrated that W played a significant role through its presence in the grain boundaries rather than the minor structural modifications of the bulk phases. This is very important from a mechanical strengthening perspective since the most common fracture type in the Ni-rich cathode materials is intergranular cracking. The presence of W in grain boundaries increased the doped material's total resistance to cracking initiation and growth, making these series of cathodes suitable for use due to their longer life and higher capacity. This complex microstructure with W-rich areas at the surface of secondary particles and within grain boundaries, therefore, suggests that W might play multiple roles, protecting the surfaces from side reactions, strengthening grain boundaries, and providing a damping medium from stresses occurring during cycling. These effects will be discussed in more detail in a follow-up publication.

Experimental Section
Synthesis Procedures: In this work, two different synthesis methods, mechanofusion (a dry particle fusion method) and coprecipitation in a Continuously Stirred Tank Reactor (CSTR), were used to dope W within the cathode. These two different approaches were selected to fully understand the tungsten's behavior as a dopant, regardless of the synthesis method. Mechanofusion was conducted using a spinning speed of 2400 rpm for 60 min. During this process, a commercial Ni(OH) 2 with a primary particle size of ≈15 µm was coated with a nanosized (<100 nm) WO 3 powder from Sigma-Aldrich. 50 g of Ni(OH) 2 was loaded inside the mechanofusion bowl in addition to 1.263, 2.552, 5.210, and 10.873 g of WO 3 to reach 1, 2, 4, and 8 mol% of W in LNO, respectively. On the other hand, (Ni 1-x (OH) 2 ) 0.98 (NiWO 4 ) 0.01 , and (Ni 1-x (OH) 2 ) 0.96 (NiWO 4 ) 0.02 precursors having W/(Ni+W) molar ratios of 0.01 and 0.02 were synthesized using the coprecipitation method via the CSTR. The synthesis steps were similar to those in the study conducted by Van Bommel et al. [47] Two aqueous solutions were prepared containing 400 mL of 2.0 m NiSO 4 in addition to either 100 mL of 0.0808 or 0.1633 m NaWO 4 for the former and latter precursors, respectively. More details regarding the aqueous solution, reactor temperature, stirring speed, and other processing parameters have been explained in the authors' previous work. [36] In the end, all the precursors from the mechanofusion and coprecipitation methods were ground with LiOH·H 2 O (molar ratio of Li to (Ni + W) = 1.02/1). After this grinding step, the blended powders were placed under an oxygen flow at a preheated temperature of 480 °C. Next, these preheated batches were ground again to reach a uniformly homogenized mixture. For the heating steps under oxygen flow, the ground powder particles were heated at 480 and 800 °C for 2 and 20 h, respectively. Although different calcination temperatures were used in the previous studies, the present paper only focuses on the optimum temperature results at 800 °C. [36] Finally, pure LNO and W-doped LNO with 1, 2, 4 and 8% W were synthesized and used for other characterization techniques as reference materials. The higher W content samples (4% and 8%) were used as references to extend the structural investigation although there were no benefits to these concentrations from a performance point of view. For the sake of brevity and clarity, two labels will be used in this manuscript whereby "Mech" stands for mechanofusion and "Copr" indicates coprecipitation. For reference purpose, the pure LNO was subjected to the same heat treatment steps with the exception of the heating temperature in the second step which was set to 700 °C. SXRD and PDF: SXRD was conducted on the high-energy wiggler beamline of the Brockhouse X-ray Diffraction and Scattering sector at the Canadian Light Source (CLS) in Saskatoon, Canada, for meticulous crystallographic analysis. The selected energy was 30.3383 keV (λ = 0.4087 Å), and a nickel powder was utilized for detector calibration. All the powder samples were placed inside polyimide capillaries. All the angles in the XRD data were modified based on λCuKα for normalization with respect to the literature.
In addition, PDF analysis was performed on the same synchrotron beamline, in which an X-ray energy of 65 keV was employed to acquire a higher magnitude wave vector, Q (Q = 4πsinϴ/λ). [48] This setting reaches a reasonably high Q max (26 A −1 ) and results in sharp peaks in the PDF.
STEM and Analytical Electron Microscopy: STEM and EELS was employed for detailed spectroscopic analysis. A ThermoFisher Scientific (TFS) Helios G4 plasma-focused ion beam (PFIB) was used to create the ultra-thin samples required for these techniques. The samples were first coated with a very thin layer of carbon to reduce the charging effect in scanning electron microscopy (SEM) imaging mode, then two more layers of carbon and tungsten were added to the top surface in order to protect the surface structure of the particles from potential ion beam damage during the milling and lift-out procedures. The protective tungsten layer was applied due to its high resistance under ion milling. This step does not influence the amount of W inside the doped samples, as it was demonstrated by analyzing non-doped samples prepared with the same coating conditions, which resulted in no detection of W even in trace concentration in thin lamella. All the STEM images were acquired at 200 keV incident electron energy using a HAADF detector within an FEI Titan 80-300 equipped with aberration correctors for the probe and imaging lenses. The convergence semi-angle in STEM was 19.1 mrad. EELS data was acquired in STEM mode with a collection angle of 55 mrad, and a direct electron detector Gatan K2 Summit® was used to improve the quality of EELS signals, especially at low beam current (to reduce electron beam irradiation damage). To reduce the noise level on the EELS maps, PCA was utilized. In addition, more detailed EELS features fitting was carried out using MLLS to separate spectroscopic maps based on different peak shapes of the O K-edge. EDXS maps were obtained on a TFS Talos 200X fitted with a Super-EDS detector (results shown in Supporting Information).
XAFS: The XAFS data acquisition was carried out on the Hard X-ray Micro-Analysis (HXMA) beamline at the CLS. The HXMA photon source was a superconducting wiggler insertion device running at 1.9 T, and energy selection was performed based on the W L 3 -and Ni K-absorption edges with a Si (111) and (220) crystals monochromator, respectively. XAFS data were collected in both transmission mode and fluorescence modes using Oxford straight ion chamber detectors (100% filled with He) and 13 element Ge detector with Soller slits and Cu filter (several layers of Al foil), respectively. The same powder-form WO 3 model compound that was utilized for doping was set between the I 1 and I 2 ion chambers downstream of the sample for in-step energy calibration for each tungsten XAFS scan.
DFT Calculations: DFT was exploited to deduce crystallographic information for the different W-doped LNO structural scenarios and calculate the most thermodynamically stable W compound during synthesis. DFT was also employed to assess separate crystallographic information regarding other possible W-compounds with almost identical energy.
The Vienna Ab initio Simulation Package (VASP) was used to conduct these calculations with Projector Augmented Wave (PAW) pseudopotentials. [49][50][51][52] The exchange and correlation terms were defined by the Strongly Constrained and Appropriately Normed (SCAN) functional. [53] An energy cutoff of 520 eV was considered for the plane-wave basis. The Brillouin zone was sampled with <0.05 Å −1 k-point spacing. All the calculations were spin polarized and adopted ferromagnetic spin orderings.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.