Determining effects of doping lithium nickel oxide with tungsten using Compton scattering

With a rich history spanning over a century, nickel (Ni)-based cathodes have evolved from early alkaline rechargeable batteries to the prominent Ni-rich variant in lithium-ion batteries (LIBs).¹ Despite the initial dominance of LiCoO₂ in LIBs, Ni-based cathodes have come back in the mainstream battery industry due to their unique ability to provide high energy density and storage capacity at low cost.^2,3 The physical and chemical properties of LiNiO₂ (LNO) play a crucial role in driving the underlying redox reactions responsible for the superior performance of LNO as a battery material despite the challenges of structural and thermal stability.⁴ Improvements for large scale applications have been realized by applying surface coating techniques and adjustments to the composition via atomic substitutions, which promise an enhanced battery performance.⁵ 

Recent investigations^6–13 have focused on the effects of tungsten (W) doping on nickel-rich lithium-ion battery cathodes in LNO (WLNO) for improved energy density and overall battery performance. W doping acts to stabilize the crystal structure of the cathode, mitigate phase transitions, and improve longevity during charge/discharge cycles. W also enhances the electronic conductivity and thus facilitates electron transport and actively participates in oxygen redox reactions in nickel-rich cathodes to improve battery safety and stability.

Here, we employ high-energy x-ray Compton scattering experiments in combination with first-principles modeling to unravel fundamental quantum mechanical aspects of W doping on LNO. Through an analysis of the line shape of the Compton-scattered x rays, we adduce an atomic level understanding of how W influences the electronic structure of the LNO cathode. Our study contributes to the understanding of doping effects in LIBs for advancing battery performance and efficiency.

To calculate the electronic structure and the Bloch wave functions of WLNO and LNO, we employed density functional theory (DFT) as implemented in the Vienna Ab initio Simulation Package (VASP),^24,25 based on projector-augmented waves.²⁶ The Perdew, Burke, and Ernzerhof (PBE) version of the generalized gradient approximation (GGA) was employed to treat exchange–correlation effects.²⁷ A plane wave basis set with a kinetic energy cutoff of 470 eV was used. To account for strong electron correlation effects, a Hubbard parameter U = 4 eV was applied to Ni 3d electrons, consistent with prior studies.^28,29

We simulated W dopings of 12.5%, 3%, and 1% using 2 × 2 × 1, 2 × 4 × 2, and 4 × 4 × 3 supercells, respectively. For the pristine LNO, we employed the monoclinic crystal structure with an 8-atom primitive cell, characterized by a =4.84/Å, b = 2.81/Å, and c = 5.12/Å (description here, how many atoms, lattice constants, crystal structure). The 12.5% case is illustrated in Fig. 1. The k-space grids used in the relaxation and Compton profiles calculations had a density of at least 0.02 2π/Å. The density of states (DOS) was computed with the tetrahedron method and with finer k meshes: for 12.5% and 3% cases as well as for LNO, we used 8 × 8 × 8, and for 1% case, we used 3 × 5 × 4 grids. VASPKIT was employed for postprocessing the VASP results.³⁰ 

Samples were synthesized using the method described by Välikangas et al.,^3,5 which involved the following steps. Coprecipitation of Ni(OH)₂: Spherical Ni(OH)₂ precursor was first synthesized using alkali metal hydroxide (NaOH) coprecipitation in an inert gas (nitrogen) atmosphere.³² An inert gas was used to prevent the oxidation of the precursor. Precipitation was done in a continuous-flow stirred-tank reactor (CSTR) with a reactor volume of 3 l at 50 °C under vigorous agitation. Particle size growth during coprecipitation was followed by determining the particle size distribution of the slurry sampled from the reactor’s overflow tubing. The solution was heated to a precipitation temperature, and the pH was adjusted to the desired level with a NaOH solution. The feeding rates of nickel sulfate, Na₂WO₃, NaOH, and ammonia solutions were adjusted to maintain the desired residence time. After coprecipitation, the precursor slurry was filtered in a vacuum, and the precipitate was carefully washed with a sufficient amount of warm deionized water. The synthesized 1 mol. % W–Ni(OH)₂ precursors were dried overnight at 60 °C in a vacuum oven. Lithiation: The 1 mol. % W–Ni(OH)₂ precursor was mixed with LiOH using a Li:Ni:W molar ratio of 1:1. This was followed by calcination at 690 °C. Excess LiOH was used to compensate for lithium loss during high-temperature calcination and to ensure homogeneous lithiation. The mixture was calcined with a 2.5 °C min⁻¹ heating ramp and a 5 h holding time at 690 °C in an oxygen atmosphere and subsequently milled and sieved to less than 40 μm under dry room conditions; see Välikangas et al.³ for details. Inductively Coupled Plasma (ICP) spectroscopy measurements of the WLNO samples found an excess of lithium with a Li/Ni ratio of about 1.02, but we estimate that this ratio could be as large as 1.03 since excess of Li was used in our synthesis process.

The spherically averaged WLNO Compton profiles were measured at the high-energy inelastic x-ray scattering beamline BL08W at the Japanese synchrotron facility SPring-8^33,34 with the same experimental setup as in our previous LNO study³⁵ to maintain consistency in analyzing the results. Circularly polarized x rays with an energy of 182.6 keV, emitted from an elliptical multipole wiggler, were directed onto the sample. The incident x-ray beam at the sample position had a cross-sectional area of 1 mm². The Compton scattered x rays were detected using a pure Ge solid-state detector. The scattering angle was fixed at 178°, and the Compton profiles were measured at room temperature. The overall momentum resolution of the equipment was 0.5 a.u., sufficient to reveal the details of valence orbitals in cathode materials.³⁵ To suppress Compton scattering by air, the sample was arranged in a vacuum chamber and the measurements were performed under vacuum conditions. The sample structure was verified by XRD (Rigaku SmartLab) measurements using CuKa radiation. The diffraction patterns of the sample display a high integrated intensity ratio of I(003)/I(104), which confirms the layered character of the crystal structure.³ 

The experimental Compton profiles are shown in Fig. 2. The explanation of the normalization of these curves is as follows: Ideally, the pristine LNO profile is normalized to 47 electrons (core plus valence) across the entire momentum range. However, in our case, the momentum range is limited to a 10 a.u. cutoff. Therefore, the normalization should be smaller than 47. This number can be estimated using the Biggs’ atomic profiles since these profiles become exact at high momenta. For WLNO, the normalization on the complete momentum range yielded 47.55 electrons, corresponding to 1% W concentration (Z = 74) replacing Ni (Z = 28) and an excess of Li, resulting in a Li/Ni ratio of 1.03, optimizing the shape of the Compton profile tail. In addition, for pristine LNO, the total number of electrons is slightly lower than 47 due to a small deficit of Li atoms, detected via ICP (see the supplementary material of Ref. 35), now estimated at 2% through fitting with the Compton profile tails. The inset of Fig. 2 shows ΔJ(p) defined as the difference between the Compton profiles of the WLNO doped material and the Compton profile of the pristine LNO material. The profile difference ΔJ(p) facilitates the interpretation of electronic structure changes induced by W doping because many irrelevant contributions are thus subtracted out. Interestingly, the amplitude of ΔJ(p) can be explained with a simple model of the Compton profile based on Hartree–Fock atomic orbitals;²² see also Suzuki et al.³⁶ for a discussion of Compton profile changes due to lithiation in cathode materials.

Patel et al.³⁷ observed that the atomic model used to describe the Compton profile of LIBs can be refined by substituting the Li 2s valence electron with an electron in the O 2p orbital. This replacement reflects the transfer of one electron from the Li atom to a neighboring oxygen atom, resulting in Li atoms becoming a cation Li⁺. Here, we investigate whether a similar transfer occurs for the six valence electrons of W dopants in WLNO, leading to the formation of W⁶⁺ ion¹² with W valence electrons moving into the O 2p orbital. Recently, it has been shown⁴ that the ionic charges in LNO can be rationalized as Ni²⁺, O^1.5−, and Li⁺. If the W impurity goes into a Ni site, then a W⁶⁺ ion replaces a Ni²⁺ ion, and the four extra valence electrons occupy the O 2p states. In our VASP calculation, Ni has 10 valence electrons, but eight Ni 3d electrons are more localized and less active.⁴ Therefore, the contribution of these eight atomic-like 3d electrons of Ni can be approximately removed from the Compton valence profile of Ni by using their Compton profile based on atomic orbitals.²² 

To analyze the valence Compton profile differences ΔJ_v(p) between WLNO and LNO, we consider the case with a W concentration of 12.5% since a simple picture emerges. In the main panel of Fig. 3, we consider the profile in which the contributions of Ni 3d states are excluded by treating them as core states. In this case, one can easily justify that the four extra valence electrons are transferred to the O 2p states since the curve can be fitted by the profile of an O 2p Slater orbital used previously by Kothalawala et al.³⁵ In this connection, we note that the sample is fully metallic at 12.5% W doping and the LNO energy gap vanishes already with 3% W doping as shown Fig. 4. Our results are consistent with the finding of Gao et al.¹² that W doping improves the conductivity of LNO. Finally, the inset of Fig. 3 shows that the agreement between theory and experiment ΔJ(p) shows an overall improvement if the contribution of the Li excess and W is replaced by the Slater O 2p orbital. As expected, the resulting “modified atomic model” (green solid line in the inset) rationalizes the strong O 2p weight in the partial DOS (PDOS), Fig. 4(b) (red line); a significant Ni weight in the PDOS is also visible due to a strong nickel–oxygen hybridization.³⁵ If we incorporate the results of the DFT calculation at 1% W concentration for the W–Ni contribution instead of using the simple Slater O 2p orbital contribution used in the modified atomic model, we observe an impressive agreement with experiment also at low momenta.

Another DFT calculation³⁸ involving W doping, which considers charge density changes in LNO, shows that the introduction of W increases the charge density around O atoms more than around Ni atoms. Our simulations reflect this trend since our Bader analysis results show that the oxygen ions acquire on average 1.2 electrons around Ni and 1.7 electrons around W sites. We thus infer that the charge density around the O atoms plays an important role in enhancing electron conductivity in WLNO by generating an increase in the number of itinerant electrons.

Our calculations confirm the ferromagnetic phase of LNO [Fig. 4(a)], which was reported by Goodenough et al. as far back as 1958.³⁹ W doping increases the average magnetic moment per Ni atom. Interestingly, the magnetization density distribution in WLNO (Fig. 5) shows the appearance of significant magnetization on the oxygen ions neighboring W impurities. In our calculations, we determined that the total magnetic moment per Ni ion in LNO amounts to 1 μ_B.^29,35 Upon introducing a 1% concentration of W into the sample, this average magnetic moment increases to 1.02 μ_B. With a 3% concentration of W, the average magnetic moment further rises to 1.08 μ_B per Ni ion. Interestingly, at a concentration of 12.5%, the magnetic moment reaches 1.29 μ_B per Ni ion. Moreover, Fig. 5 shows that the magnetic moment mostly changes on the oxygen sites, while remains similar on the Ni ions.

Our first-principles calculations, coupled with Compton scattering experiments, show that the substitution of W by Ni in LNO enhances the electronic conductivity and the ferromagnetic phase. Hafiz et al.⁴⁰ have pointed to the role of itinerant ferromagnetic phases for obtaining an improved battery performance. Our analysis reveals that W valence electrons in WLNO undergo transfer to the O 2p itinerant states and gives an insight into the chemistry of W incorporation on LNO. Our study shows how, by combining x-ray Compton scattering measurements with parallel first-principles modeling, an atomic level understanding of doping effects can be obtained toward the rational design of battery materials.

J.N. was supported by the INERCOM LUT platform. K.S. was supported by JSPS KAKENHI Grant Nos. JP19K05519, 21KK0095, and 22H02103. The Compton scattering experiments were performed with the approval of JASRI (Proposal No. 2022A1454 and 2023A1475). The magnetization measurements were performed with the approval of the National Institute of Material Science (NIMS) Open Facility (Proposal Nos. 22NM8130 and 22NM8141). K.S. and Y.A. thank Shigeki Nimori for his support of the magnetization measurements. This work benefited from Northeastern University’s Advanced Scientific Computation Center and the Discovery Cluster, and part of the work at Northeastern University was supported by the Office of Naval Research Grant No. N00014-23-1-2330. The authors (P.T., P.L., J.V., T.H., and U.L.) acknowledge Business Finland for research funding 2021–2024 (University of Oulu, BATCircle2.0, Grant No. 44612/31/2020). The authors wish to acknowledge CSC—IT Centre for Science, Finland, for computational resources.

The authors have no conflicts to disclose.

Veenavee Nipunika Kothalawala: Conceptualization (equal); Formal analysis (lead); Investigation (equal); Methodology (equal); Visualization (lead); Writing – original draft (equal). Kosuke Suzuki: Data curation (equal); Formal analysis (equal); Methodology (equal); Validation (equal); Writing – review & editing (equal). Xin Li: Methodology (supporting); Writing – review & editing (supporting). Bernardo Barbiellini: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (lead); Writing – original draft (equal); Writing – review & editing (lead). Johannes Nokelainen: Data curation (equal); Investigation (equal); Methodology (equal); Supervision (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (supporting). Ilja Makkonen: Conceptualization (equal); Formal analysis (equal); Methodology (lead); Supervision (equal); Validation (equal); Writing – review & editing (lead). Rafael Ferragut: Writing – review & editing (equal). Pekka Tynjälä: Conceptualization (equal); Investigation (equal); Methodology (equal); Validation (equal). Petteri Laine: Conceptualization (equal); Investigation (equal); Methodology (equal); Validation (equal). Juho Välikangas: Conceptualization (equal); Investigation (equal); Methodology (equal); Validation (equal). Tao Hu: Conceptualization (equal); Investigation (equal); Methodology (equal); Validation (equal). Ulla Lassi: Conceptualization (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal) Kodai Takano: Methodology (supporting); Resources (equal); Validation (equal). Naruki Tsuji: Methodology (supporting); Resources (equal); Validation (equal). Yosuke Amada: Data curation (equal); Methodology (equal); Resources (equal). Assa Aravindh Sasikala Devi: Investigation (equal); Writing – review & editing (equal). Matti Alatalo: Investigation (equal); Writing – review & editing (equal). Yoshiharu Sakurai: Methodology (equal); Resources (equal); Writing – review & editing (equal). Hiroshi Sakurai: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Mohammad Babar: Writing – review & editing (supporting). Venkat Vishwanathan: Writing – review & editing (supporting). Hasnain Hafiz: Writing – review & editing (supporting). Arun Bansil: Investigation (equal); Methodology (equal); Writing – review & editing (lead).

The data that support the findings of this study are openly available in the GitHub repository at https://github.com/Veenavi92/W-doped-LiNiO2.