Molecular intercalation in the van der Waals antiferromagnets FePS₃ and NiPS₃

Low-dimensional magnetic systems have played a pivotal role not only in advancing our understanding of the quantum properties of materials but also in the realization and exploration of new concepts in many-body physics. Still, with the exception of certain geometrically discretized or strong-frustration scenarios, which lead to quantum disordered magnetic states including quantum spin liquids, most three-dimensional (3D) magnetic systems exhibit long-ranged order. Conversely, as emphasized by Bethe in his seminal work, low dimensions amplify quantum fluctuations and destabilize the conventional order parameters Bethe (1931). In the case of layered compounds, increasing the interlayer spacing by the insertion of large molecules is a very literal means of modulating the dimensionality, and hence the properties, from potentially 3D towards the 2D limit Alias and Sukumaran (1992); Pattayil and Sukumaran (1993). The discovery of 2D magnetic materials suitable for this type of control would hold significant promise for applications in nanoelectronics and spintronics Soumyanarayanan et al. (2016).

For this reason, magnetic van der Waals materials have attracted extensive attention in recent years Burch et al. (2018), although efforts at dimensional manipulation have so far been limited largely to approaching the monolayer limit by exfoliation. In this context, the Ising-type van der Waals magnets Cr₂Ge₂Te₆ and CrI₃ exhibit the emergence of intrinsic ferromagnetism with a high transition temperature in few- or monolayer films Gong et al. (2017); Huang et al. (2017); Tian et al. (2016); in particular, CrI₃ shows a systematic evolution of the ordering temperature and even the type of magnetic order with the number of layers Huang et al. (2017). MnSe₂ remains ferromagnetic (FM) at room temperature in its monolayer form O’Hara et al. (2018), and most surprisingly room-temperature FM order emerges in monolayer VSe₂ despite the bulk material being paramagnetic Bonilla et al. (2018).

The transition-metal trisulfide $M$PS₃ ($M$ = Mn, Fe, Co, Ni) is a class of antiferromagnetic (AFM) van der Waals materials that has also been studied extensively Le Flem et al. (1982); Ouvrard et al. (1985); Joy and Vasudevan (1992); Chittari et al. (2016); Du et al. (2016); Wang et al. (2018). As Fig. 1(a) shows, the $M$PS₃ compounds are isostructural, with the magnetic ions ($M^{2+}$) forming a honeycomb lattice. All exhibit the characteristics of a Mott insulator, displaying high resistivity at room temperature and a band gap well in excess of 1 eV (1.5 eV for FePS₃ and 1.6 eV for NiPS₃). This gap varies widely on exfoliation, offering an optoelectronic response over a broad frequency range for device applications Du et al. (2016). FePS₃ exhibits a phase transition under pressure from insulating to metallic Haines et al. (2018) and NiPS₃ in its AFM phase exhibits coherent excitonic states Kang et al. (2020).

Differences in the trigonal distortion of the $M$S₆ octahedra and in spin-orbit coupling mean that the anisotropy of intralayer magnetic interactions differs in the $M$PS₃ materials. MnPS₃ appears to have Heisenberg spin interactions, NiPS₃ shows a weak and CoPS₃ a stronger easy-plane (XY) anisotropy, and FePS₃ has a strong Ising anisotropy Kurosawa et al. (1983); Joy and Vasudevan (1992); Chandrasekharan and Vasudevan (1994); Chatterjee (1995); Rule et al. (2007); Lee et al. (2016); Wildes et al. (2017). Upon cooling, they all order in an AFM pattern, with respective transition temperatures, $T_{N}$, of 78 K, 118 K, 122 K, and 155 K for MnPS₃, FePS₃, CoPS₃ and NiPS₃ Le Flem et al. (1982); Brec (1986); Wang et al. (2018).

The weak van der Waals interlayer coupling leads to a cleavage energy close to that of graphite, and thus the $M$PS₃ materials are easy to exfoliate. Multiple efforts to thin $M$PS₃ samples have shown them to remain structurally stable down to a single layer Du et al. (2016); Kuo et al. (2016). Recent Raman scattering studies of monolayer FePS₃ have reported that $T_{N}$ either drops from 117 K to 104 K Wang et al. (2016) or remains unchanged Lee et al. (2016), which raises the possibility of substrate effects on this 2D Ising magnet McCreary et al. (2020). Raman investigation of monolayer NiPS₃ indicates a suppression of long-range order relative to the bulk material, which was interpreted in the framework of the Berezinskii-Kosterlitz-Thouless (BKT) phase transition Hu et al. (2023).

An alternative approach to dimensionality reduction is the incorporation of organic cations as spacers in the bulk materials, such that the altered layer separation enables control over the interlayer interactions Zhang et al. (2020, 2022). This is complementary to electrochemical treatments with ionic liquids whose primary effect is to induce protonation, or other electron doping effects, when the sample is placed on the cathode side Cui et al. (2018, 2019). It was reported in the early literature that pyridine (C₅H₅N) can be intercalated into MnPS₃, leading to a transition from an AFM to a weakly FM ground state Alias and Sukumaran (1992), and that the intercalation of alkylamines (C_nH_2n+1NH₂) into FePS₃ causes a strong reduction of $T_{N}$ Pattayil and Sukumaran (1993). Much more recently, the intercalation of tetraheptylammonium-bromide (C₂₈H₆₀BrN) into NiPS₃ was found to cause a transition from AFM to ferrimagnetic (FIM) order, followed by another transition to AFM order, in effects ascribed to the electron doping of the layers Mi et al. (2022).

In this work, we report the successful interlayer intercalation of EMIM⁺ into FePS₃ and NiPS₃, using the ionic liquid C₆H₁₁N₂BF₄ (EMIM-BF₄). By measuring changes in the mass and interlayer spacing, we estimate the intercalation levels to be approximately 0.27 EMIM⁺/f.u. in FePS₃ and 0.37 EMIM⁺/f.u. in NiPS₃, and that both procedures dilate the $c$-axis lattice parameter by over 50%. We investigated the magnetic properties of EMIM_x-FePS₃ and EMIM_x-NiPS₃, and compared them with FePS₃ and NiPS₃, by magnetic susceptibility and ³¹P nuclear magnetic resonance (NMR) measurements.We found that intercalation in FePS₃ causes changes in the Curie-Weiss temperatures determined above $T_{N}$, a suppression of $T_{N}$ itself, and a reduced spin gap at the lowest temperatures. For intercalated NiPS₃, the magnetic transition temperature barely changes and the spin fluctuations become very isotropic, suggesting an evolution from weakly XY toward Heisenberg-type magnetism. Taken together with a dramatic suppression of any interlayer magnetic interactions, these data provide strong evidence for the systematic modulation of intralayer magnetic interactions by intercalation, for which we deduce that charge doping on the P site should be taken into account.

The structure of this article is as follows. In Sec. II we describe our $M$PS₃ samples and intercalation procedure, and in Sec. III report our structural characterization. Section IV reports our measurements of the magnetic properties of pure and intercalated FePS₃ and Sec. V our data for pure and intercalated NiPS₃. In Sec. VI we discuss the consequences of our results for the understanding of magnetism in $M$PS₃ materials and provide a brief conclusion.

Single crystals of FePS₃ and NiPS₃ were grown by the method of chemical vapor transport (CVT) Brec (1986). As Fig. 1(a) shows, the $M$PS₃ layer is composed of covalently bonded $M$S₆ octahedra and double-cone [P₂S₆]^4- units in a 2:1 ratio Brec et al. (1985). The $M$ ions are thought to be in a robustly divalent $M^{2+}$ state in all compounds, as shown for FePS₃ by X-ray photoelectron spectroscopy (XPS) Yu et al. (2019). These ions form the honeycomb lattice, with the P-P pairs located at the centers of the $M$ hexagons Brec (1986); Joy and Vasudevan (1992); Le Flem et al. (1982). The S^2- layers stack in an ABC configuration along the $c$ direction to form a monoclinic structure with space group C2/m Ouvrard et al. (1985).

Both FePS₃ and NiPS₃ order magnetically with a “zig-zag” pattern, as shown in Fig. 1(b). The Fe²⁺ ions in FePS₃ take their high-spin, $S=2$ state and order with the moments normal to the $ab$ plane Joy and Vasudevan (1992). These moments have FM alignment parallel to the $a$ axis and AFM alignment along the $b$ and $c$ axes, with propagation wave vector $\bm{k}=[0~{}1~{}\frac{1}{2}]$ Kurosawa et al. (1983); Rule et al. (2007); Wildes et al. (2012); Lançon et al. (2016); Coak et al. (2021); Paul et al. (2023). For NiPS₃, the Ni²⁺ ($S=1$) moments are oriented primarily along the $a$ axis with a small component perpendicular to the $ab$ plane; the zig-zag chains also lie along the $a$ axis, with AFM alignment along $b$ but an FM repeat along the $c$ axis, resulting in $\bm{k}=[0~{}1~{}0]$ Kurosawa et al. (1983); Wildes et al. (2015); Kim and Park (2021); Wildes et al. (2022).

Figure 2(a) shows the configuration that we adopt for electrochemical treatment in order to achieve interlayer intercalation. The ionic liquid EMIM-BF₄ was packed in a container with two platinum electrodes placed approximately 20 mm apart and subjected to a 3 V potential difference Cui et al. (2019, 2018). A single crystal of FePS₃ or NiPS₃ was attached to the cathode and covered with silver paint. The ionic liquid was heated to a temperature around 60^∘C. After 24 hours the mass of the samples had changed significantly, but longer treatment times led to no further change, and so one could declare the intercalation process to be complete.

X-ray diffraction (XRD) measurements were performed using Cu_α and Cu_β radiation to determine the $c$-axis lattice parameters. The d.c. magnetic susceptibility was measured in a Magnetic Property Measurement System (MPMS) with a field of 100 Oe and at temperatures down to 1.8 K in field-cooled (FC) and zero-field-cooled (ZFC) conditions. ³¹P has nuclear spin $I=1/2$ and a Zeeman factor of ${}^{31}\gamma=17.235$ MHz/T. The ³¹P NMR measurements were performed on single crystals using a top tuning circuit and the spectra collected by spin-echo pulse sequences. For broad spectra, frequency sweeps were used to acquire the full spectrum. NMR Knight shifts were calculated from $K_{n}=(f/^{31}\gamma H-1){\times}100\%$, where $f$ is the resonance frequency of the spectral peaks in the paramagnetic phase and the average frequency in the ordered phase. Spin-lattice relaxation rates, $1/^{31}T_{1}$, were measured by the standard magnetization inversion-recovery method. These constitute a sensitive probe of low-energy spin fluctuations, because $1/T_{1}T={\rm lim}_{\omega\to 0}\Sigma_{q}A^{2}_{\rm hf}(q)\,{\rm Im}\,\chi(q,\omega)/\omega$, where $\chi(q,\omega)$ is the dynamical susceptibility, $A_{\rm hf}(q)$ is the hyperfine coupling, and $\omega$ is the NMR measurement frequency, which for electronic spins lies in the zero-energy limit.

Figure 2(b) shows single-crystal XRD data for our FePS₃ samples before and after 24-hour electrochemical treatment with the ionic liquid EMIM-BF₄, and Fig. 2(c) the analogous data for NiPS₃. We focus on the (0 0 $L$) Bragg peaks in order to extract the interlayer ($c$-axis) dimension of the unit cell. Studying the intralayer structure, meaning the $a$ and $b$ parameters, requires powder XRD measurements, but it has been found that the grinding process introduces disorder within the layers that precludes a meaningful analysis. The positions of the (0 0 $L$) XRD peaks change appreciably after the intercalation treatment, and the $c$-axis lattice parameters are shown and compared in Table I. The $c$-axis dimension increases from 6.72 Å to 10.40 Å for FePS₃ and from 6.64 Å to 10.34 Å for NiPS₃. Such an extremly large (over 50%) increase in interlayer spacing can be expected to have a very strong effect on reducing the dimensionality to the 2D limit.

Because the sample is attached to the cathode, and because of the large change in $c$-axis parameter, we deduce that EMIM⁺, rather than H⁺, is intercalated into the materials. A similar observation of intercalation by a large organic molecule has also been reported in NiPS₃ using a different type of ionic liquid Mi et al. (2022). We note that our intercalation in EMIM_x-NiPS₃ was not complete by volume, in that a small portion of the XRD pattern of EMIM_x-NiPS₃ coincided with that of pristine NiPS₃ [visible in Fig. 2(c)]. However, the volume ratio of residual NiPS₃ is rather small, as we establish later from the absence of detectable signals in the magnetic susceptibility and NMR spectra.

The mass of the FePS₃ crystal increased from $0.18\pm 0.01$ mg before to $0.21\pm 0.01$ mg after intercalation. From this we estimate the intercalant concentration and the charge doping level to be $x=27\pm 3\%$, given the molecular mass of 183 g/mol for FePS₃ and 111 g/mol for EMIM⁺. For the NiPS₃ crystal, the mass changed from $3.86\pm 0.01$ mg to $4.71\pm 0.01$ mg, and thus we extracted the doping level as $x=37\pm 0.4\%$, using the molecular mass of 186 g/mol for NiPS₃. Our NMR studies revealed in addition the presence of F^- ions in the intercalated samples, but at an unknown small concentration (data not shown), which suggests that the actual doping $x$ may vary with position across the samples.

Our susceptibility data show that the intercalation of FePS₃ and NiPS₃ does not cause any depletion of the available magnetic moments. Thus the changes we observe, specifically the reduction of $T_{N}$ in EMIM_x-FePS₃ and the isotropic spin response of EMIM_x-NiPS₃, should be attributed to changes of the magnetic interactions. To discuss these, in Table II we list the values of the magnetic interactions deduced by INS experiments for FePS₃ Lançon et al. (2016) and NiPS₃ Wildes et al. (2022). The authors of both studies adopted a minimal model consisting of the three intralayer Heisenberg interactions illustrated in Fig. 1(b), an interlayer Heisenberg interaction ($J^{\prime}$), and single-ion anisotropy terms, of which only one was found to be significant. The $J^{\prime}$ term was found to be very small in FePS₃ but potentially significant in NiPS₃. The $J_{3}$ term, for third-neighbor $M^{2+}$ ions, was found to be completely dominant in NiPS₃ and to play an essential role in establishing the zig-zag order of FePS₃.

Considering the increase of the $c$-axis lattice parameter that we achieve, in excess of 50%, it is natural to expect that the interlayer couplings should be reduced dramatically. Thus our primary conclusion is that the $J^{\prime}$ terms have no role in the magnetism of either material, where 2D ordering remains well established in both in EMIM_x-FePS₃ and in EMIM_x-NiPS₃. In the light of theoretical studies demonstrating that quasi-2D magnets with very small $J^{\prime}$ may nevertheless have significant $T_{N}$ values Yasuda et al. (2005), this result was not fully clear for FePS₃ (Table II). Our study demonstrates that it is also true for NiPS₃, despite the 3 K interlayer coupling scale extracted in Ref. Wildes et al. (2022).

It is then important to understand how the intralayer magnetic interactions are modified by intercalation. In principle, the addition of EMIM⁺ should lead to electron doping in both materials, with the electrons presumably located on the Fe or P ions, and reaching a doping around 1/3 e^-/f.u. (0.27 in EMIM_x-FePS₃, 0.37 in EMIM_x-NiPS₃). Considering the unchanged paramagnetic moments we extracted for EMIM⁺-FePS₃ and the unchanged magnetic properties of EMIM⁺-NiPS₃, a change of the $M^{2+}$ valence seems to be excluded. Thus we propose that the electronic doping affects the P sites, by transforming the less favorable P⁴⁺ ions Yu et al. (2019) to P³⁺. With the P ions located at the centers of the $M$ hexagons, we expect that this doping should be very effective in modifying the intralayer superexchange interactions, particularly the strong $J_{3}$ terms. We remark in this context that we found the intercalated samples to remain highly insulating, which is a prerequisite to regard the doped charges as localized, such that they can affect the magnetic interactions rather than drive a metallic transition.

Considering FePS₃ in more detail, the $T_{N}$ we measure for bulk EMIM_x-FePS₃ is lower than the value $T_{N}=104$ K obtained for monolayer FePS₃ Wang et al. (2016). Thus our intercalation is more than a simple isolation of the 2D units, and its effects on the competing intralayer interactions require more detailed microscopic modelling. In fitting their INS results, the authors of Ref. Lançon et al. (2016) assumed an extremely minimal model where all of the anisotropy was ascribed to one single-ion term, and further anisotropic interactions such as Dzyaloshinskii-Moriya or Ising-type spin-spin interactions were neglected. The modelling of these terms is a complex task that was approached only recently by correlated density functional theory (DFT) methods Amirabbasi and Kratzer (2023), while further experimental information on the magnetic anisotropy has been obtained by X-ray photoelectron microscopy Lee et al. (2023). With the ongoing interest in FePS₃ as a candidate spintronic material Soumyanarayanan et al. (2016), and for band engineering by ultrafast coherent light Mertens et al. (2023), a deeper understanding of the key interaction terms remains a priority.

In EMIM_x-NiPS₃, the fact that $T^{*}$ is only 3 K lower than $T_{N}$ in NiPS₃ demonstrates immediately that $J^{\prime}$ terms are irrelevant. Nevertheless, the absence of line splitting at low temperatures in the ³¹P spectra of EMIM_x-NiPS₃ [Fig. 8(d)], in contrast to the behavior in NiPS₃ [Fig. 8(b)], and the enhancement of $1/^{31}T_{1}$ (Fig. 10), lead us to suggest that the ordering at $T^{*}$ may be of a different type, such as incomplete (short-ranged) or incommensurate. Recent theoretical work has indeed proposed that NiPS₃ is close to incommensurate magnetic order Mellado (2023), and the large overlap of the two NMR peaks we observe in Fig. 8(b) may support this scenario. Another possibility could be weak anisotropy caused by intralayer strain arising from a lattice mismatch with the size of the intercalated molecules. Finally, a BKT transition is also possible if the system retains an easy-plane anisotropy, however weak, and in the intercalated system it is likely that quenched disorder would help to pin true long-range order from the quasi-long-ranged BKT order. We note here that a BKT phase has already been reported in monolayer NiPS₃ Hu et al. (2023).

In summary, we performed electrochemical treatment of FePS₃ and NiPS₃ with the ionic liquid EMIM-BF₄, obtaining a significant intercalation of the large EMIM⁺ ions into both compounds. The most important consequence of intercalation is a large and uniform increase of the $c$-axis lattice parameters, which unavoidably causes a strong reduction of interlayer couplings. The fact that we do not observe order-of-magnitude changes in the magnetic properties of either system demonstrates conclusively that the magnetic properties of the pristine materials are intrinsically 2D, being determined almost exclusively by intralayer physics. The changes we do observe are a consequence of modulated intralayer magnetic interactions. In FePS₃, the drop in magnetic ordering temperature suggests a change in the competing FM and AFM interactions while the system remains in the Ising limit; the effects in NiPS₃ are far more subtle, with the interactions changing from weakly XY-type to almost isotropic (Heisenberg-type). We propose the charge doping that accompanies intercalation, which appears to be localized around the P ions and reaches a level of 0.3-0.4 e^-/f.u., as the mechanism most likely to drive these changes. Our study of chemical intercalation with ionic liquids paves the way not only for approaching the 2D limit in van der Waals magnetic materials but also for tuning their intralayer magnetic interactions.