Spectral evidence for NiPS₃ as a Mott-Hubbard insulator

Abstract

The layered van der Waals trichalcogenide NiPS₃ has attracted widespread attention due to its unique optical, magnetic, and electronic properties. The complexity of NiPS₃ itself, however, has also led to ongoing debates regarding its characteristics such as the existence of self-doped ligand holes. In this study, X-ray absorption spectroscopy and resonant inelastic X-ray scattering have been applied to investigate the electronic structure of NiPS₃. With the aid of theoretical calculations using the charge-transfer multiplet model, we provide experimental evidence for NiPS₃ being a Mott-Hubbard insulator rather than a charge-transfer insulator. Moreover, we explain why some previous XAS studies have concluded that NiPS₃ is a charge-transfer insulator by comparing surface and bulk sensitive spectra.

I Introduction

Van der Waals (vdW) materials, including grapheneNovoselov et al. (2012), transition metal dichalcogenidesNicolosi et al. (2013); Xu et al. (2015), and boron nitrideWatanabe et al. (2004) are a promiment class of two-dimensional (2D) crystals. The reduced dimensionality of these systems often leads to exceptional physical properties, resulting from enhanced quantum effects and increased correlations due to the reduction of available phase space and diminished screeningAjayan et al. (2016). In this study, we focus on NiPS₃, a type of 2D vdW metal phosphorus trichalcogenides (MPX₃). NiPS₃ is monoclinic, with space group $C2/m$; it can be visualized as layered NiS₂ crystals with P-P pairs substituting for one third of the Ni sitesWang et al. (2018); Klingen et al. (1973) as shown in FIG. 1. NiPS₃ exhibits intriguing properties in electronicsKim et al. (2018); Lane and Zhu (2020); Tan et al. (2022),spintronicsWang et al. (2021); Basnet et al. (2021); Scheie et al. (2023), magnetismWildes et al. (2015); Kim et al. (2019a); He et al. (2024), and opticsLiu et al. (2019); Wang et al. (2022). NiPS₃ stands out among the 2D crystals due to its ultrasharp photoluminescence (PL) peak at 1.47 eV, induced by a spin-orbit-entangled exciton stateKang et al. (2020); Wang et al. (2021); Hwangbo et al. (2021). Understanding the properties of the exciton state requires an investigation of both the occupied and unoccupied states of NiPS₃; although the exciton of NiPS₃ has been extensively studiedBelvin et al. (2021); Ho et al. (2021); Klaproth et al. (2023); Kim et al. (2023), many fundamental electronic properties of NiPS₃ have still remained controversial.

In X-ray absorption spectroscopy (XAS), a core electron is excited to an empty state, thereby probing the unoccupied states of the system, with information about the occupied states being inferred indirectlyFuggle et al. (1992). Meanwhile, in resonant inelastic X-ray scattering (RIXS), photons are scattered inelastically off matter, and we measure the excitations corresponding to the changes that occur during the scatteringAment et al. (2011). Both XASKim et al. (2018, 2019b); Kang et al. (2020); Yan et al. (2021) and RIXSKang et al. (2020); DiScala et al. (2023); He et al. (2024) have been applied to NiPS₃, yet they have yielded differing results. The major controversy is whether NiPS₃ is a charge transfer insulator or Mott-Hubbard insulator. A charge-transfer insulator is defined by a charge-transfer energy $\Delta$ (the energy cost of transferring an electron from the $p$ level of a ligand to the metal ion) that is smaller than the Coulomb repulsion energy between two $d$ electrons $U_{dd}$ (the energy cost of a $d$ electron hopping to another already occupied $d$ site); by contrast, when $\Delta$ is greater than $U_{dd}$, the standard Hubbard model will be applied, resulting in a Mott-Hubbard insulatorKhomskii (2014). The simplified energy levels for these two types of insulatorsSahiner et al. (1996) are shown in FIG. 2: although both have a dominant ground state of $d^{8}$, the Mott-Hubbard insulator’s lowest excited state is $d^{9}$, while the charge-transfer insulator’s lowest excited state is $d^{9}\underline{L}$, where $\underline{L}$ represents a ligand hole. Kim et al. first applied XAS to the Ni L-edge of NiPS₃ and compared with calculated spectra corresponding to different values of $\Delta$Kim et al. (2018). They found that NiPS₃ is a self-doped charge transfer insulator with a negative value of $\Delta$. With a negative $\Delta$, the charge transfer from the ligand will occur spontaneously, resulting in a dominant ground state $d^{9}\underline{L}$. This claim, however, was directly challenged by a subsequent calculation which showed that NiPS₃ would be a metal rather than an insulator with $d^{9}\underline{L}$ ground stateKim et al. (2019b). They confirmed the presence of the paramagnetic Mott phase and the dominant ground state $d^{8}$ for NiPS₃. Following XAS studiesKang et al. (2020); Yan et al. (2021), although challenging NiPS₃ as having negative $\Delta$ with self-doped behavior, they still concluded that NiPS₃ has a small but positive $\Delta$ and belongs to the category of charge-transfer insulator. The exciton of NiPS₃ probed by RIXS is also a subject of debate. Initially, the exciton has been described as a Zhang-Rice singletKang et al. (2020). However, this does not explain the extremely narrow PL peak of the excitonHe et al. (2024) and its mismatch in response to an applied magnetic fieldJana et al. (2023). Instead of the Zhang-Rice mode, a most recent RIXS studies have revealed the dominant Hund’s character in NiPS₃He et al. (2024).

In this study, we compared XAS measurements of an as-grown single crystal of NiPS₃ to an exfoliated sample of NiPS₃ in order to reconcile the controversy regarding the value of $\Delta$ for NiPS₃. XAS was applied using both total electron yield (TEY) mode and total fluourescence yield (TFY) mode for surface sensitive ($<10$ nm) and bulk sensitive ($>100$ nm) measurements, respectivelyYang et al. (2021). Therefore, the surface and bulk features of NiPS₃ were simultaneously collected, enabling us to capture the most subtle details of the electronic structure. By comparing the XAS results in TEY mode and TFY mode, we obtained different spectra from previous XAS studiesKim et al. (2018); Kang et al. (2020); Yan et al. (2021) on NiPS₃. Furthermore, by comparison with calculations using the charge-transfer multiplet model we are led to conclude that, contrary to the previously held belief that NiPS₃ has a very small value of $\Delta$, it actually has a relatively large value of $\Delta$. Further analysis was conducted to explain why previous XAS results concluded that NiPS₃ belongs to the charge-transfer insulator. Moreover, we also applied RIXS to NiPS₃ and found that $dd$ excitations are dominant, while charge-transfer excitations were too weak to be observed. Combined with the conclusion from XAS that NiPS₃ has a relatively large $\Delta$, we provide strong evidence that NiPS₃ is a Mott-Hubbard insulator.

II Methodology

NiPS₃ single crystals were synthesized by the chemical vapor transport method. Pure elements, in a stoichiometric ratio of Ni:P:S $=$ 1:1:3 (2 g in total), along with 40 mg iodine as transport agent, were enclosed in quartz ampules under vacuum of 1$\times$10^$-$4 Torr. Subsequently, the ampules underwent heating in a two-zone furnace with the temperature range of 650 $-$ 600 ^∘C for 1 week, followed by cooling to room temperature. Bulk crystals were harvested from the lower temperature zone of the ampules. The as-grown NiPS₃ single crystal was directly mounted on the sample holder with conductive copper tape. The exfoliated sample was prepared by removing the top layers using Scotch tape under an argon atmosphere in a glovebox, and was transferred to the analysis chamber using a vacuum suitcase to prevent oxidation.

The nickel L_3,2-edge XAS spectra were measured at Advanced Light Source (ALS) beamline 7.3.1 at Lawrence Berkeley National Laboratory (LBNL). Both TEY and TFY modes were collected simultaneously. The XAS photon energy was calibrated using a Li(Ni_1/3Mn_1/3Co_1/3)O₂ reference sample. O K-edge XAS spectra with both TEY and TFY modes were also collected to determine the extent of O contamination in the sample.

The theoretically calculated XAS spectra were obtained using the Crispy software for the Quanty libraryRetegan (2019). A charge-transfer multiplet model is applied to the system, described by a semi-empirical Hamiltonian that includes atomic and crystal-field interactionsde Groot et al. (2021). In this study, nickel L_3,2-edge XAS spectra were calculated by Crispy for various values of $\Delta$. For simplicity, the symmetry is set to $O_{h}$. The applied parameters are: $F_{k}$=0.8, $G_{k}$=0.8, $\zeta$=1.0, $U(3d,3d)$=4.0 eV, $U(2p,3d)$=6.0 eV, $10Dq(3d)$=1.0 eV, $Ve_{g}(3d,L1)$=1.4 eV, $Vt_{2g}(3d,L1)$=1.0 eV. The broadening full width at half maximum parameters (eV) are set to be 0.48, 0.52 for Lorentzian and 0.3 for Gaussian. The values of $\Delta$ range from -1 eV to 5 eV.

The RIXS spectra of nickel L₃-edge were obtained at beamline 8.0.1 at the ALS. The RIXS photon energy was calibrated by an undulator.

III Results and discussion

XAS was first applied to nickel L_3,2-edge of as-grown NiPS₃ single crystal as shown in FIG. 3(a). Surprisingly, the spectrum from TEY mode (blue solid line) and TFY mode (red solid line) exhibited different features. In TEY mode, both the L₃-edge and L₂-edge show a single peak feature with a satellite peak indicated by the blue arrow around 856.6 eV. This is precisely the feature corresponding to the small $\Delta$ observed in previous literature through experiments and calculations, indicating a charge-transfer insulator. However, as mentioned earlier, the TEY signal primarily reflects the surface electronic structure of a sample. In contrast, the TFY spectra reveal that the satellite splits into two peaks at 855.8 eV and 857.2 eV. Furthermore, distinguishable double peak features appear at both L₃-edge (854.4 eV) and L₂-edge (871.0 eV), as indicated by the red arrows.

Although many studies indicate that NiPS₃ is stable in airLu et al. (2020); Fang et al. (2021), exposure to air may still cause surface oxidation and affect its TEY signalFrati et al. (2020). This can be inferred from the different features exhibited by the TEY and TFY signals in FIG. 3(a). To chracterize the surface oxidation, we also applied XAS in the oxygen K-edge region for the same sample used in FIG. 3(a) as presented in FIG. 4. As we previously inferred, the surface-sensitive TEY signal in the oxygen range exhibited very distinct signals, with characteristic peaks appearing in both the X-Ray absorption near edge structure (XANES) region (532 eV and 537 eV) and extended X-ray absorption fine structure (EXAFS) region (558 eV). This indicates that the oxygen on the surface of NiPS₃ is not merely adsorbed free oxygen from the air, but has chemically bonded with the Ni cations. The spectrum in TFY mode, on the other hand, shows a very weak signal-to-noise ratio (SNR) in oxygen range compared to the spectrum in TEY mode. The appearance of an oxygen peak in the bulk-sensitive TFY mode does not mean that the entire NiPS₃ sample is oxidized. Instead, it is because the TFY signal detection range covers the entire region to a depth of about 100 nanometers, including the few nanometers of the oxidized surface. It is precisely because the SNR of the TFY signal is significantly lower than that of the TEY signal that it proves the oxidation is only present on the surface of the NiPS₃ sample (occupying only a small portion of the TFY detection depth). This is reasonable, as the bulk NiPS₃ should not retain oxygen and should be air-stable.

Further investigation into the structure of the O K edge XAS has found that it is likely that nanoscaled NiO appear on the surface of NiPS₃, evidenced by the similarity to the O K edge XAS from NiO nanowires as shown by Wu et al.Wu et al. (2005); these nanoscaled NiO exhibit very different features in the XAS O K-edge region compared to bulk NiO. By comparing with their XAS results, we find that the characteristic peaks in the TEY mode in FIG. 4 are perfectly consistent with those of NiO nanowires in both the XANES and EXAFS regions. The TEY probe depth shows that the nanoscaled NiO once again demonstrates that oxidation occurs only on the surface of the NiPS₃ sample.

To eliminate the impact of surface oxidation on the experimental results, the most straightforward method is to directly remove the outermost few layers. As a vdW material, NiPS₃ can be easily exfoliated using Scotch tape. The XAS spectrum was remeasured for the Ni L_3,2-edge of exfoliated NiPS₃ sample [FIG. 3(b)]. After exfoliation, we observed the following: (1) The TEY and TFY spectra exhibited excellent consistency, as indicated by the blue and red arrows, respectively; compared to the untreated sample, (2) the TEY signal showed significant changes, with the most notable being that both the Ni L₂ peak and L₃ peak split into a double peak structure, with more distinct features appearing in the satellite peak next to the L₃ peak; (3) the TFY signal did not show obvious changes. These observations indicate that: (1) Exfoliation successfully removed the surface oxidation layers, making the TEY signal after exfoliation more likely to reflect the true sample signal; (2) The presence of the surface oxidation layers have a significant impact on the TEY signal; (3) The oxidation layers have a minimal impact on the TFY signal, allowing the TFY signal to remain reliable throughout the entire study.

To determine the $\Delta$ of NiPS₃, we performed theoretical calculations using the charge-transfer multiplet model. The comparison between the experimental data for the exfoliated sample and the theoretical calculation is presented in FIG. 5. From the calculated results, it can be observed that with the increase in $\Delta$, the most noticeable change is that the Ni L₃ and L₂ absorption edges transition from a single peak structure to a double peak structure, accompanied by a shift in the peak positions and changes in the satellite peaks. This trend is consistent with previous calculationsKim et al. (2018); Kang et al. (2020); Yan et al. (2021). When determining the $\Delta$ of NiPS₃ by comparing the features with calculations, however, the main focus and point of contention has been around the L₃ edge. By contrast, the most noticeable feature here, which differs from previous studies, is that the L₂ peak also splits into a double peak structure. The best match to the experimental results is the calculated spectra with $\Delta$=3 eV, yet the spectra with $\Delta$=2.5 eV and $\Delta$=3.5 eV also match some features of the experimental results very well. Therefore, we have obtained a relatively large $\Delta$ for NiPS₃, suggesting it is highly likely that NiPS₃ falls within the regime of a Mott-Hubbard insulator.

Thus far, we have determined the $\Delta$ of NiPS₃ with XAS and theoretical calculation. To further confirm that such a $\Delta$ categorizes NiPS₃ as a Mott-Hubbard insulator, we employed mapping of RIXS as shown in FIG. 6. It can be observed that parallel to the elastic peak (yellow solid line), two resonant X-ray Raman features (red dashed line) were found with energy-loss of 1.2 eV and 1.8 eV. Based on the scale of energy-lossAment et al. (2011) and previous optical spectra resultsKim et al. (2018), both resonant X-ray Raman features were assigned to $dd$ excitations. Therefore, we found that the $dd$ excitations are dominant, with the charge-transfer features being too weak to be observed. This indicates that standard Hubbard model applies to NiPS₃, with transitions occurring only among $d$-electrons themselves, rather than the $p-d$ model as in the case of charge-transfer insulator. Moreover, these energy-loss features match the previous RIXS resultHe et al. (2024), except that we did not find the ultrasharp PL peak due to the limitation of resolving power. With the aid of calculations, they determined the expectation values describing the NiPS₃ wavefunction and revealed that NiPS₃ has a dominant Hund’s character with a ground state $d^{8}$, instead of $d^{9}\underline{L}$ as in the Zhang-Rice scenario. This further proves that NiPS₃ belongs to a Mott-Hubbard insulator rather than a charge-transfer insulator.

IV Conclusion

In this study, we used XAS and RIXS to investigate the electronic structure of NiPS₃. By comparing experimental data from as-grown and exfoliated samples with theoretical calculations, we found that NiPS₃ has a relatively large $\Delta$ (about 3 eV), indicating it is a Mott-Hubbard insulator. Additionally, we observed surface oxidation on as-grown samples, which may have contributed to the previous debates over the XAS results and was mitigated by exfoliation. RIXS data showed dominant $dd$ excitations with weak charge-transfer features, further supporting this classification. These findings resolve previous controversies, contribute to the broader knowledge of NiPS₃, and emphasize the importance of controlling surface adsorption in studies of the electronic structure of vdW materials.

V Acknowledgement

This research used resources of the Advanced Light Source, which is a Department of Energy (DOE) Office of Science User Facility under contract no. DE-AC02-05CH11231. Work done by Q.T. and X.L. was supported by National Science Foundation (NSF) under Grant No. 2216008 and No. 1945364, and the U.S. DOE, Office of Science, Basic Energy Science (BES) under Award Number DE-SC0021064. Q.T. also acknowledges the support of the Laursen Fellowship at Boston University.

VI Figures

Figure 1: The crystal structure of NiPS₃ seen along the $c$ axis (top) and $a$ axis (bottom). Generated by VESTAKlingen et al. (1973).

Figure 2: Simplified electronic energy levels for NiPS₃ assuming it is a charge-transfer insulator (left) and assuming it is a Mott-Hubbard insulator (right)Sahiner et al. (1996).

Figure 3: Nickel L_3,2-edge XAS spectra with TEY mode (blue) and TFY mode (red) for (a) an untreated NiPS₃ sample and (b) a NiPS₃ sample after exfoliation of the surface layers in an air-isolated environment.

Figure 4: Surface oxidation check: O K-edge XAS spectra with TEY mode (blue) and TFY mode (red) for the untreated NiPS₃ sample used in FIG. 3(a).

Figure 5: Experimental (Exp.) nickel L_3,2-edge XAS spectra and calculated spectra from the charge-transfer multiplet model by Crispy softwareRetegan (2019). The top of the figure shows the experimental XAS data for the exfoliated NiPS₃ sample with TEY signal (blue) and TFY signal (red). The subsequent spectra from top to bottom represent the calculated spectra corresponding to $\Delta$ decreasing from 5.0 eV to -1.0 eV.

Figure 6: Nickel L₃-edge mapping of RIXS of NiPS₃. Yellow solid line: elastic peak. Red dashed line: energy lost features.