betaFeedback

WorksPhysics and Astronomy

Magnetic properties of bilayer VI3: Role of trigonal crystal field and electric-field tuning

Abstract

The magnetic properties of two-dimensional VI3 bilayer are the focus of our first-principles analysis, highlighting the role of trigonal crystal-field effects and carried out in comparison with the CrI3 prototypical case, where the effects are absent. In VI3 bilayers, the empty a1g state - consistent with the observed trigonal distortion - is found to play a crucial role in both stabilizing the insulating state and in determining the inter-layer magnetic interaction. Indeed, an analysis based on maximally localized Wannier functions allows to evaluate the interlayer exchange interactions in two different VI3 stackings (labelled AB and AB'), to interpret the results in terms of virtual-hopping mechanism, and to highlight the strongest hopping channels underlying the magnetic interlayer coupling. Upon application of electric fields perpendicular to the slab, we find that the magnetic ground-state in the AB' stacking can be switched from antiferromagnetic to ferromagnetic, suggesting VI3 bilayer as an appealing candidate for electric-field-driven miniaturized spintronic devices.

Introduction

Boosted by the experimental discovery of intrinsic magnetism in atomically thin layers of CrI3_3, Cr2_2Ge2_2Te6_6 and Fe3_3GeTe2_2, [1, 2, 3] two-dimensional (2D) van der Waals (vdW) magnets have recently received an increasing attention. Interestingly, the control of 2D magnetism in few atomic layers is enabled by external electric field[4, 5, 6] or by electron-hole doping,[3, 7] making them particularly appealing for potential spintronic applications.

Among those materials, VI3_3, belonging to the family of transition-metal (M) trihalides MX3_3 (X = Cl, B, and I) with honeycomb arrangement of the metal cations (similar to the most studied CrI3_3[1, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]), has recently emerged as a potential 2D ferromagnet[14, 15, 16, 17, 18, 19, 20]. It is known since more than 30 years ago that, in the bulk form, VI3_3 becomes ferromagnetic below a Curie temperature of Tc≃55T_c \simeq55 K, similar to CrI3_3 (with Tc≃68T_c \simeq68 K)[21]. Conversely, the structural properties are still under debate. Experimental characterizations of the crystal structure have in fact reported that VI3_3 undergoes a structural phase transition around 79 K, changing its symmetry across still unclear phases: the high-temperature (HT) crystal structure was proposed to be either trigonal P3112P3_112[14] or rhombohedral R3ˉR\bar{3}[20], or monoclinic C2/mC2/m structure[15]; the low-temperature (LT) crystal structure was proposed to be either C2/cC2/c[14] or R3ˉR\bar{3} structure[15]. Optical and electrical transport measurements have clearly showed bulk VI3_3 to be a semiconductor with an optical band gap of ∼0.67\sim0.67 eV[14]. However, from the theoretical point of view, the understanding and modeling of the electronic properties are the focus of present debate. Some studies have in fact reported that bulk VI3_3 is a Mott-insulator with a bandgap of about 1 eV[14, 15], whereas others have claimed a half-metallic character[19, 20]. In a thin-film limit, to the best of our knowledge, no experimental studies have been reported for atomic layers of VI3_3. On the theoretical side, current studies are controversial with respect to the electronic properties, in analogy to the situation for the corresponding bulk phase. In particular, by analysing electronic properties in the VI3_3 monolayer, in Ref.[16] a Mott-insulator ground state is proposed, reported to be lower in energy than the half-metallic state (by ∼0.3\sim0.3 eV/f.u.). Other attempts to explain the Mott-insulator ground state have also been reported: the authors in Ref.[17] have proposed orbital-ordered phases accompanying the lattice distortion, while the authors of Ref.[18] ascribed the gap opening to combined effects of spin-orbit coupling (SOC) and Hubbard UU correlations. On the other hand, a consensus is reached with respect to the magnetic properties of VI3_3 single layer: current theoretical characterizations report a ferromagnetic (FM) exchange coupling between first V-site neighbors, i.e. FM intra-layer coupling. The inter-layer magnetic stability has also been investigated for bilayer VI3_3. In particular in Ref.[16] it is reported that the inter-layer magnetic stability is sensitive to the layer stacking, in line with previous works on bilayer CrI3_3[9, 10, 11]. For the sake of completeness, we mention that in Ref.[22] it is claimed bilayer VI3_3 to show a stacking-independent ferromagnetic ground state, but considering the half-metallic state rather than the Mott-insulating one.

A deep understanding of the VI3_3 electronic structure is needed to interpret the related magnetic properties, both in the monolayer and in the bilayer case. In this study, we therefore focus on the crucial role of trigonal crystal field effects in determining the VI3_3 insulating behaviour and the related magnetic properties. As a counterexample, we consider the prototypical 2D magnet, i.e. CrI3_3 (where trigonal crystal field effects are absent), and we carry out a one-to-one comparison between VI3_3 and CrI3_3. In particular, we first focus on the monolayer and discuss crystal field effects (§ 1.2 A) and density of states (§ 1.2 B). Then we move to the magnetic properties of bilayer halides, by considering two different stacking arrangements. We concentrate on the interlayer exchange coupling, interpreting it in terms of virtual hopping mechanisms, highlighting the most efficient hopping paths (§ 1.3 A) and addressing the effects of an external electric field in tuning the magnetic stability (§ 1.3 B). Finally, we draw our conclusions in Section V.

Computational method

Density-functional theory (DFT) calculations were performed using the VASP code[23] within the generalized gradient approximation (GGA)[24]. The van der Waals (vdW) interactions were included for bilayer structure calculations. The rotationally invariant GGA + UU method was employed to account for correlation effects[25]. On-site Coulomb interaction for transition-metal 3d3d orbitals was considered with an effective UU of 2.0 eV[5, 26]. Semiconducting and metallic states are initialized via the density matrix within the GGA+UU scheme[26]. Brillouin zone integrations were performed using a kk-point grid of 6×6×16 \times6 \times1 for the structure optimization. Band structures and density of states (also including electric fields) were calculated by using 12×12×112 \times12 \times1 kk-points mesh. Electric fields are applied perpendicularly to the surface by saw-tooth-like potential with dipole correction[27]. The maximally-localized Wannier functions (MLWFs) were calculated by using WANNIER90 tool[28] interfaced with the VASP code.

Results for Monolayer VI3_3

Crystal-field effects

First we focus on the difference between VI3_3 and CrI3_3 monolayers, as far as crystal field effects are concerned. In particular, both VI3_3 and CrI3_3 show the magnetic atom coordinated to six I atoms, forming edge-sharing octahedra and resulting in octahedral crystalline electric field (CEF) splitting of the dd orbitals into the two-fold ege_g and three-fold t2gt_{2g} states. The t2gt_{2g} state can be further split into a doublet eg′e'_g and a singlet a1ga_{1g} state to reduce the band energy, i.e. a Jahn-Teller (JT) effect in trigonal symmetry[18, 29]. This effect leads to the trigonal lattice distortion, but such JT-distortion is not remarkable in this system. The I-Cr-I bond angle in CrI3_3 is almost 90° (i.e. a cubic octahedron), while the I-V-I angle in VI3_3 approaches 89°. As such, it exhibits a trigonal distortion – elongation along the zz axis – still preserving spatial inversion symmetry [see Figure 1(a) and (b)]. Figure 1(c) and (d) show the difference in CEF splitting for VI3_3 and CrI3_3 respectively: as schematically represented, the JT-induced splitting allows for the band gap opening in VI3_3 by half-filling the majority eg′e'_g channel and leaving the a1ga_{1g} empty in the case of V d2d^2. On the other hand, CrI3_3 is unaffected both because of the Cr d3d^3 valence and the almost cubic CEF.

Figure 1. Distortion of the crystal structure in (a) VI3_3 and (b) CrI3_3. The distortion does not change the bond length but alters the bond angle, leading to a trigonal elongation along the zz-direction. (c) and (d) show the crystal field splitting of dd levels in V 3d23d^2 and Cr 3d33d^3, respectively. Five 3d3d Wannier functions reflect the trigonal CEF states in monolayer VI3_3 (e) and cubic CEF states in monolayer CrI3_3 (f). The isosurface levels of the Wannier functions were set at 1.5 a0−3/21.5\,a_0^{-3/2} (yellow) and −1.5 a0−3/2-1.5\,a_0^{-3/2} (blue), where a0a_0 is the Bohr radius.

Within the global Cartesian {xyz}\{xyz\} coordinate system, Figure 1(e), the a1ga_{1g} and eg′e'_g states are written in form[30]

∣a1g⟩=3z2−r2∣eg1′⟩=13(2(x2−y2)−zx)∣eg2′⟩=13(2xy+yz)(1)\begin{aligned} \lvert a_{1g}\rangle&= 3z^2-r^2 \\ \lvert e'_{g1}\rangle&= \frac{1}{\sqrt{3}}\left(\sqrt{2}(x^2-y^2)-zx\right) \\ \lvert e'_{g2}\rangle&= \frac{1}{\sqrt{3}}\left(\sqrt{2}xy+yz\right) \tag*{(1)} \end{aligned}

where the zz axis is parallel to the out-of-layer direction (i.e. perpendicular to the slab). According to the different local crystal field effects, we projected the Bloch functions onto the local M6M_6 octahedral coordinate system {x′y′z′}\{x'y'z'\} for CrI3_3 (with basis axes directed along the Cr-I bonds), and onto the Cartesian {xyz}\{xyz\} system for VI3_3 [see Figure 1(e) and (f)]. After the projection and maximally localization process, the Wannier functions converged into localized orbitals, as shown in Figure 1 (e) and (f): orbitals shapes are in agreement with the ege_g-(a1g,eg′)(a_{1g},e^\prime_g) states splitting induced by the trigonal CEF in VI3_3, and the ege_g-t2gt_{2g} states induced by the cubic CEF in CrI3_3. In particular, for the latter, it is possible to recognize the ∣3z2−r2⟩\lvert3z^2-r^2\rangle orbital shape for the a1ga_{1g} state (occupying the empty space at the center of the I3_3 triangle and pointing along the zz-direction) and mixed shapes from the (∣x2−y2⟩,∣zx⟩)(\lvert x^2-y^2\rangle,\lvert zx\rangle) and (∣xy⟩,∣yz⟩)(\lvert xy\rangle,\lvert yz\rangle) orbitals for the eg1′e^\prime_{g1} and eg2′e^\prime_{g2} states respectively, according to Eq. (1).

Density of states

In line with previous works, our DFT calculations on monolayer VI3_3 converged to two different electronic states, corresponding to half-metallic [17] and Mott-insulator states [16, 18]: the two V d2d^2 electrons occupy t2gt_{2g} states as a1g1eg1a^1_{1g}e^1_g and eg′2a1g0e^{\prime2}_g a^0_{1g}, respectively. We found that the insulating state (with a direct band gap of ∼0.39\sim0.39 eV at the Γ\Gamma point for U=2.0U=2.0 eV) is lower in energy than the half-metallic state by 1.8 eV/f.u. Therefore, we will focus hereafter on the Mott insulator state as the ground state. Although Yang et al. proposed that the insulating ground state of monolayer VI3_3 is stabilized by spin-orbit coupling (SOC) splitting rather than by CEF splitting, here we remark that the crystal field splitting is energetically dominant over the SOC effect, as usually seen in 3d3d transition-metal compounds [31, 32]. In fact, when including SOC in our band structure calculations, we observed that it affects the width and energy bands related to the I-pp states, but it does not significantly change the V 3d3d band structure. Moreover, as we focus here on the eg′a1g0e^\prime_g a^0_{1g} state, eventual SOC induced splitting would not affect the mechanisms and conclusions presented in this study.

The partial density of states (DOS) for monolayer CrI3_3 and VI3_3 are shown in Figure 2; DOS are resolved for each MLWF state, clearly showing the D3dD_{3d} and OhO_h CEF splitting for VI3_3 and CrI3_3 respectively, therefore validating our basis functions choice for the Wannier projection as explained in the following.

Figure 2. The partial DOS (within U=2.0U=2.0 eV) projected onto (a) V-3d3d-D3dD_{3d} and (b) Cr-3d3d-OhO_h CEF states, respectively, via Wannier functions. Blue (red) color represents for majority spin (minority spin). Fermi level is set as energy origin.

We note that the Cr d3d^3 partial DOS, projected on the trigonal D3dD_{3d} basic set, so as to allow a direct comparison with the V d2d^2 case, shows an overlap in the energy range of the a1ga_{1g} and eg′e^\prime_g; this reflects the absence of splitting induced by the trigonal CEF, in line with the OhO_h local octahedral symmetry of CrI3_3. In particular, Cr-dd orbital states are split into empty ege_g and occupied t2gt_{2g} states with a gap of about 0.9 eV in the majority spin channel (up-spin states); the minority spin channel (down-spin states), unoccupied for both orbital types, does not display any relevant splitting. Such a different behavior between the majority and minority spin channels can be ascribed to the pdpd hybridization: the up-spin dd-states strongly hybridize with I-pp states located at the top of the valence band, causing the large CEF splitting supported

by the ege_g-pp bonding-antibonding splitting; the down-spin dd-states are higher in energy, i.e. away from I-pp levels, thus not showing any significant CEF splitting.

In VI3_3, the spin-up channel of the V dd states are clearly split into ege_g, a1ga_{1g}, eg′e^\prime_g trigonal CEF states. In particular, the eg′e^\prime_g is the lowest energy state with a broad distribution due to the pdpd hybridization below the Fermi level, similar to CrI3_3. On the other hand, the CEF splitting shows a different behaviour of the empty dd states: the a1ga_{1g} state becomes the lowest energy state, while the eg′e^\prime_g state still lies in the same energy region as the minority spin Cr-t2gt_{2g} state. This is related to the fact that the eg′e^\prime_g state has more bonding character with surrounding I pp state than the a1ga_{1g} state (compare the orbital shapes in Figure 1(e)); therefore, the pdpd hybridization shifts unoccupied eg′e^\prime_g level up and occupied I-pp level down.

Results for bilayer VI3

Electronic properties and inter-layer magnetic stability

Let us now consider the case of bilayer VI3_3, showing a similar atomic structure to its bulk counterpart. In particular, we studied two structures corresponding to R3ˉR\bar{3} and C2/cC2/c phases in bulk VI3_3 [15]. The difference between these two structures is related to the different stacking of the two VI3 single layers. As shown in Figure 3 (a) and (c), the AB-stacking is characterized by the top layer showing one V atom sitting above the hexagon center of the bottom layer, similar to bilayer graphene. Indeed, when comparing equivalent atoms in two layers, the top layer is horizontally shifted from the bottom layers by (2a+b)/3(2\mathbf{a}+\mathbf{b})/3 (shown by a black arrow), where a\mathbf{a} and b\mathbf{b} denote the lattice vectors. This structure has Rˉ3\bar{R}3 symmetry. On the other hand, the AB′ stacking, with C2/mC2/m space-group symmetry, is characterized by the shift of the top layer by −(a+b)/2-(\mathbf{a}+\mathbf{b})/2, as shown in Figure 3 (b) and (d). We employed the experimental values for the in-plane lattice constants of both materials: in bulk VI3, we used a=b=6.84a=b=6.84 Å[15] and for CrI3, we used a=b=6.87a=b=6.87 Å[8]. A 20 Å-thick vacuum is contained in the supercell for 2D slab simulation.

Figure 3. (a, b) Top views and (c, d) side views of atomic structure in bilayer VI3 in AB and AB′ stacking. The red and orange hexagons represents honeycomb structure of V atoms in top and bottom layers, and gray balls represent I atoms. The black arrow indicates the vector which connects equivalent atoms located in two layers and shows how the top layer is sliding with respect to the bottom layer. Inter-layer exchange coupling JijJ_{ij} in bilayer VI3 for (e) AB and (f) AB′ stacking.

ABAB′
VI3FM02.44
AFM0.692.40
CrI3FM034.8
AFM14.736.6

Table 1. Relative total energy (meV/f.u) for inter-layer FM and AFM spin configurations in bilayer VI3 and CrI3 in AB and AB′ stacking. The lowest energy state is highlighted.

First, we remark that the robustness of the intra-layer FM spin ordering is demonstrated by calculating the energy difference between FM and Néel type antiferromagnetic AFM spin configurations in VI3 monolayer, i.e. ΔE=EAFM−EFM=12.8\Delta E=E_{\mathrm{AFM}}-E_{\mathrm{FM}}=12.8 meV/f.u. The V magnetic moment was calculated as 2.16 μB2.16\,\mu_{\mathrm{B}}. We then address the magnetic properties of bilayers, by calculating the total energy between inter-layer FM and AFM orders for the two stackings in bilayer VI3 and compared the results with those obtained in CrI3. In bilayer VI3, the inter-layer FM and AFM spin configurations are very close in energy; nevertheless, the FM order is favored in the AB-stacking, while the AFM order is favored in AB′ stacking, as reported in Table I. Differently, in bilayer CrI3, the FM order is favored in both AB and AB′ stacking pattern. Noteworthy, the inter-layer FM order in AB′ stacking is only slightly more stable than the AFM order, the energy differences being rather sensitive to used on-site Coulomb UU values, therefore not allowing a direct comparison with previous works on bilayer CrI3. In any case, according to the energy differences reported in Table I, the magnetic stability in bilayers VI3 and in CrI3 AB′-stacking results to be weak. As such, it may lead to an easy control of the magnetism by either external electric fields or electrostatic doping.

To understand the magnetic stability, we evaluated the magnetic exchange interactions between V atoms by fitting total energies calculated in AB and AB′ stacking to the Heisenberg Hamiltonian. Here we assume the Heisenberg Hamiltonian,

H=∑⟨i,j⟩Jijsi⋅sj,(2)H=\sum_{\langle i,j\rangle}J_{ij}\mathbf{s}_i\cdot\mathbf{s}_j, \tag*{(2)}

where JijJ_{ij} are the isotropic Heisenberg coupling constant between spin sites ii and jj and si\mathbf{s}_i is the unit vector pointing to the direction of the spin at site ii. A parallel spin (FM) configuration is favored when J<0J<0 and an anti-parallel (AFM) spin configuration is favored when J>0J>0.

In addition to the intra-layer (in-plane) first nearest-neighbor coupling (J1J_{1}), we thus considered inter-layer couplings (J1J_{1} and J2J_{2} in AB stacking; J1′J_{1}^{\prime}, J2′J_{2}^{\prime}, J3′J_{3}^{\prime} and J4′J_{4}^{\prime} in AB′ stacking) as schematically illustrated in Figure 3 (e) and (f); associated atomic pairs distances are reported in Table II. In particular, we performed calculations to estimate J2′J_{2}^{\prime}, J3′J_{3}^{\prime} and J4′J_{4}^{\prime} in a 2×1×12\times1\times1 supercell via the four-state energy mapping method[33, 34, 35, 36, 37]. This method allows to consider one specific pair of spins and remove the background interactions, therefore allowing the calculation of the inter-layer magnetic exchange coupling constants of interest.

In Table II we report the estimated exchange coupling constants for bilayer VI3 and CrI3.

AB stackingAB′ stacking
VI3_3NN61962222
dd (Å)3.956.667.743.957.047.058.078.95
JijJ_{ij} (meV)-3.200.81-0.24-4.460.100.210.04-0.04
CrI3_3NN61962222
dd (Å)3.956.577.683.957.007.028.038.92
JijJ_{ij} (meV)-7.03-0.82-0.69-8.11-0.18-0.23-0.290.25

Table II. Number of equivalent bonds per unit cell NN, bond distance between transition-metal sites dd, and calculated exchange coupling constants JijJ_{ij} in AB and AB′ stacking for bilayer VI3_3 and CrI3_3.

For VI3, the intra-layer exchange coupling favors parallel spin state, while the inter-layer coupling eventually favors parallel spin state in AB-stacking and anti-parallel spin states in AB′-stacking, (cf Table I). In closer detail, in AB stacking, J1J_{1} favours anti-parallel coupling (0.81 meV), while J2J_{2} favours parallel coupling (−0.24 meV).

Since there is one J1J_1 bond and nine J2J_2 bonds per unit cell, overall the ferromagnetic configuration is more stable. In AB′ stacking, both J1′J'_1 and J2′J'_2 favour anti-parallel coupling (0.1 meV and 0.21 meV), thus dominantly contributing to the inter-layer AFM coupling stability. For CrI3_3, the intra-layer and inter-layer exchange coupling basically favor parallel spin states in both AB stacking and AB′ stacking.

In order to shed light on the microscopic mechanism behind the stacking-dependent magnetic couplings, we recall the “virtual hopping” idea based on the Hubbard model[30]. In particular, we discuss the results in terms of the virtual inter-layer hopping of ege_g-t2gt_{2g} states supported by the inter-layer M-I-I-M super-exchange effect. In the weak hopping limit, the inter-site hopping can be treated as a perturbation to the ground state in which magnetic ordering does not affect the energy. When the hopping process is allowed between occupied and unoccupied states, it in turn contributes to the ground state energy through the second-order contribution as the effective exchange energy Jeff=2t2/UJ_{\mathrm{eff}} = 2t^2/U with hopping integral tt and Coulomb repulsion UU, the process being called “virtual hopping”. If we consider the direct hopping between occupied and unoccupied 3d3d states at the transition metal sites, the parallel-spin configuration is favored if the hopping is strong between majority- and majority-spin states; on the other hand, the anti-parallel spin configuration is favored if the hopping between majority- and minority-spin states is strong. In order to discuss the virtual hopping process, we extracted the hopping parameters by using a MLWF basis set, as illustrated in Figure 1 (e) and (f). Note that the Wannier functions are centered at V and Cr sites and spreading the tail to I sites, so that our virtual hopping process implicitly includes the pdpd hybridization process. The same concept can be found in Anderson’s original work on super-exchange interaction[38].

Figure 4 shows the inter-layer hopping paths with the corresponding MLWFs which are responsible for the exchange energy in bilayer VI3_3 and CrI3_3. Here we select three types of inter-layer exchange couplings: first neighbor and second neighbor (J1J_1 and J2J_2) interactions in AB stacking; first neighbor (J1′J'_1) interaction in AB′ stacking. The calculated hopping integrals corresponding to these exchange couplings are shown in Table III.

Hopping t↑↑t_{\uparrow\uparrow}Hopping t↑↓t_{\uparrow\downarrow}
VI3_3a1g0a_{1g}^{0}-a1g0a_{1g}^{0}eg0e_g^{0}-eg0e_g^{0}eg′2e_g^{\prime2}-a1g0a_{1g}^{0}eg′2e_g^{\prime2}-eg0e_g^{0}a1g0a_{1g}^{0}-a1g0a_{1g}^{0}eg0e_g^{0}-eg0e_g^{0}eg′2e_g^{\prime2}-a1g0a_{1g}^{0}eg′2e_g^{\prime2}-eg0e_g^{0}eg′2e_g^{\prime2}-eg′0e_g^{\prime0}
Δε\Delta\varepsilon (eV)001.41.51.42.12.93.43.5
t1t_1 (meV)-3.62.10.0-0.7-4.62.70.01.00.6
t2t_2 (meV)-0.71.10.3-1.6-0.5-0.9-0.62.6-1.0
t1′t'_1 (meV)3.61.0-1.11.64.91.1-3.13.1-3.3

Table III. Hopping integrals calculated by MLWF basis set between occupied and unoccupied dd orbital states in parallel-(t↑↑t_{\uparrow\uparrow}) or anti-parallel (t↑↓t_{\uparrow\downarrow}) spin configurations. Three types of hopping integrals, t1t_1, t2t_2, and t1′t'_1, corresponding with inter-layer exchange couplings J1J_1, J2J_2, and J1′J'_1 are listed. Δε\Delta\varepsilon labels the difference between two eigenenergy for the MLWFs relevant to the hopping process, corresponding to Δε↑↑\Delta\varepsilon_{\uparrow\uparrow} and Δε↑↓\Delta\varepsilon_{\uparrow\downarrow} in the main text. The dominant hopping values relevant to the exchange couplings and those illustrated in Figure 4 are highlighted in bold.

In AB-stacking bilayer VI3_3, the trigonal CEF levels

and the two-electron occupation make J1J_1 positive (anti-parallel-spin-favored). The hopping between eg2′e'_{g2} and a1g′a'_{1g} states are calculated to be negligible (t∼0.0t \sim0.0 meV), thus not contributing to the magnetic interaction. On the other hand, the hopping between eg2′e'_{g2} and eg′e'_g states is sizable (t=1.0t=1.0 meV), which may be responsible for the anti-parallel-spin-favored exchange interaction. In AB-stacking bilayer CrI3_3, in contrast, the negative (parallel-spin-favored) J1J_1 can be explained by a sizeable hopping between occupied dx2d_{x^2} state and unoccupied dx2−y2†d^{{\dagger}}_{x^2-y^2} state (t=0.6t=0.6 meV). This is consistent with previous works, claiming that the ege_g-t2gt_{2g} hopping leads to the FM coupling[9, 39]. As shown in Figure 4(d), the diagonally elongated lobes of dxzd_{xz} and dx2−y2d_{x^2-y^2} orbitals show a path through Cr-I-I-Cr sites with dzx†d^\dagger_{zx}-pp-dx2−y2†d^\dagger_{x^2-y^2} hybridization, where the first pdpd hybridization shows π\pi-like and the second shows σ\sigma-like bonding. The second neighbor interaction J2J_2 is negative both for VI3_3 and CrI3_3. This can be explained by a large hopping integral between eg2′↑e_{g2}^{\prime\uparrow} and eg1′↑e_{g1}^{\prime\uparrow} and that between dxy↑d_{xy}^{\uparrow} and dz2↑d_{z^2}^{\uparrow} states. In Figure 4(b), we can recognize a eg2′↑e_{g2}^{\prime\uparrow}-pp-eg1′↑e_{g1}^{\prime\uparrow} hybridization, where the pdpd hybridization shows σ\sigma bonding. For the CrI3_3 case, a similar picture holds (cfr Figure 4(e)), where the eg2′e_{g2}^{\prime} orbital is replaced by dxyd_{xy} orbital.

Figure 4. MLWFs relevant to inter-layer exchange coupling in bilayer VI3_3 (a–c) and (d–f) CrI3_3. ↑ and ↓ denote the majority and minority spin state, respectively. The arrows show the electron hopping from an occupied orbital state to an unoccupied orbital state; the dashed and solid lines denote the parallel-spin and anti-parallel-spin configurations, respectively. Values of the hopping integrals (meV) are also shown nearby the arrows. Isosurface levels were set at ±0.45 a0−3/2a_0^{-3/2} for (a-c) and ±0.3 a0−3/2a_0^{-3/2} for (d-f).

The AB’-stacked bilayer VI3_3 shows an interesting interplay between the atomic arrangement and the specific character of the a1ga_{1g} orbital of V, related to the VI3_3 trigonal CEF: since a vanadium atom in the bottom layer is located right under an iodine atom in the upper layer, the V-a1ga_{1g} orbital strongly overlaps with the I-pzp_z orbital and form the σ\sigma bonding (Figure 4(c)). This makes the eg2′e_{g2}^{\prime}-a1ga_{1g} hopping relevant; in particular, the eg2′↑e_{g2}^{\prime\uparrow}-pp-a1g↑a_{1g}^{\uparrow} hopping is very strong (t=−3.1t=-3.1 meV), making J1′J_1^{\prime} positive. In AB’ stacked bilayers, the inter-layer exchange interactions are weaker than those in AB stacked bilayers, since several possible hoppings between multiple orbital states tend to cancel each other due to the atomic arrangement. In evident contrast, AB’-stacked bilayer CrI3_3 shows results similar to the AB-stacked case: the dyz↑d_{yz}^{\uparrow}-pp-pp-dx2−y2↓d_{x^2-y^2}^{\downarrow} hybridization path, where the former (latter) pdpd hybridization shows π\pi (σ\sigma) bonding, makes J1′J_1^{\prime} negative (Figure 4(f)).

Electric field control of magnetic stability

Finally, we discuss the effect of an applied electric field on the magnetic stability in bilayer VI3_3. This is indeed relevant, since a magnetic phase transition upon electric-field application has been reported in bilayer CrI3_3[4, 5, 6]. The energy difference between inter-layer AFM and FM states in AB and AB’ stacked with applied electric fields is shown in Figure 5. In both stacking cases, an applied electric field promotes the FM ordering. Remarkably, in the AB’ stacking, the ground state switches from AFM to FM ordering when the electric fields exceed a threshold value of ∼0.1\sim0.1 V/Å.

Figure 5. The energy difference between the FM and AFM ordering for (a) AB and (b) AB’ stacking as a function of an external electric field. The positive value of ΔE\Delta E means FM is favored and negative means AFM is favored. dd-orbital projected density of state for top (solid filled) and bottom (solid line) layer of bilayer VI3_3 with AFM AB’ stacking in AFM ordering (c) without and (d) with external electric field. Black arrow presents for the energy shifted by applied electric field. Vertical dash line denote the Fermi energy.

The microscopic mechanism of the tunable magnetic stability can be explained by invoking again the virtual hopping idea. The DOS projected onto V-dd orbital state of top and bottom layers in AB’ stacked bilayer VI3_3 is shown in Figure 5(c) and (d). Without electric field, the DOS relative to the top and bottom layer lie in the same energy range. As discussed above, there is a competition between parallel-spin hopping and anti-parallel-spin hopping in determining the first-neighbor exchange coupling J1′J_1^{\prime}. Since the energy difference between eg′↑e_g^{\prime\uparrow} and a1g↓a_{1g}^{\downarrow} state (Δϵ↑↓=2.9\Delta\epsilon_{\uparrow\downarrow}=2.9 eV) is much larger than the energy difference between eg′↑e_g^{\prime\uparrow} and a1g↑a_{1g}^{\uparrow} state (Δϵ↑↑=1.4\Delta\epsilon_{\uparrow\uparrow}=1.4 eV), one may think that a parallel-spin configuration is favored. However, J1′J_1^{\prime} is found to be slightly AFM-favored. This is because the anti-parallel spin hopping (t↑↓=−3.1t_{\uparrow\downarrow}=-3.1 meV) is stronger than the parallel-spin hopping (t↑↑=−1.1t_{\uparrow\uparrow}=-1.1 meV), resulting in a stronger AFM effective exchange coupling Jeff∝t2/ΔϵJ_{\mathrm{eff}}\propto t^2/\Delta\epsilon. Upon electric fields, the a1ga_{1g} orbital state of the top layer is shifted up, while it is shifted down in the bottom layer (cfr Figure 5). The band gap becomes narrower due to the shift of DOS and in turn decreases the difference of orbital energy levels, while the hopping integral is not significantly affected. Overall, this increases the tendency toward FM stability, eventually switching the favored magnetic configuration from AFM to FM.

Conclusions

By means of first-principles calculations, we investigated the magnetic stability in bilayer VI3_3 and compared our results with the corresponding well-studied case of CrI3_3. In particular, the magnetic exchange interactions have been analyzed by evaluating the hopping integrals between MLWFs projected onto 3d3d orbital states at V and Cr sites. We found out that the trigonal crystal field associated to the relevant JT distortion within a single-layer of VI3_3 (and absent in CrI3_3), plays an important role for the inter-layer magnetic exchange interaction. The t2gt_{2g} orbital states are in fact split into eg′e'_g and a1ga_{1g} states; the latter shows the typical lobe shape pointing along the out-of-plane direction and the strong hopping between bottom-layer a1ga_{1g} and top-layer eg′e'_g states determine the antiferromagnetic inter-layer coupling in AB′ stacking bilayer VI3_3. Nevertheless, since the hoppings favoring parallel-spin and anti-parallel-spin configuration are highly competing, the application of electric fields allows the switching of the inter-layer magnetic ordering from AFM to FM, paving the way to spintronic applications of VI3_3-based 2D magnets.

Hopping t↑↑t_{\uparrow\uparrow}Hopping t↑↓t_{\uparrow\downarrow}
CrI3_3eg0e_g^{0}-eg0e_g^{0}t2g3t_{2g}^{3}-eg0e_g^{0}t2g3t_{2g}^{3}-t2g3t_{2g}^{3}eg0e_g^{0}-eg0e_g^{0}t2g3t_{2g}^{3}-eg0e_g^{0}t2g3t_{2g}^{3}-t2g3t_{2g}^{3}
Δε\Delta\varepsilon (eV)01.502.74.34.6
t1t_1 (meV)0.90.60.71.3-1.11.7
t2t_2 (meV)0.61.50.80.62.7-1.3
t1′t'_1 (meV)2.11.31.43.0-3.14.6

Table 2.

Acknowledgments

This work was supported by “Center for Spintronics Research Network”, Osaka University. The numerical computation was performed on the Supercomputing Facilities at the Institute for Solid State Physics, University of Tokyo. KY also acknowledges Center for Computational Materials Science, Institute for Materials Research, Tohoku University for the use of MASAMUNE-IMR (Project No.20K0045). SP and DA acknowledge financial support from the Italian Ministry for Research and Education through the Progetto Internazionale Nanoscience Foundry and Fine Analysis (NFFA-MIUR) facility and through the PRIN- 2017 project “TWEET: Towards Ferroelectricity in two dimensions” (IT-MIUR Grant No. 2017YCTB59).

References

  1. [1]B. Huang, G. Clark, E. Navarro-Moratalla, D. R. Klein, R. Cheng, K. L. Seyler, D. Zhong, E. Schmigall, M. A. McGuire, D. H. Cobden, W. Yao, D. Xiao, P. Jarillo-Herrero, and X. Xu, Nature 546, 270 (2017).
  2. [2]C. Gong, L. Li, Z. Li, H. Ji, A. Stern, Y. Xia, T. Cao, W. Bao, C. Wang, Y. Wang, Z. Q. Qiu, R. J. Cava, S. G. Louie, J. Xia, and X. Zhang, Nature 546, 265 (2017).
  3. [3]Y. Deng, Y. Yu, Y. Song, J. Zhang, N. Z. Wang, Z. Sun, Y. Yi, Y. Z. Wu, S. Wu, J. Zhu, J. Wang, X. H. Chen, and Y. Zhang, Nature 563, 94 (2018).
  4. [4]B. Huang, G. Clark, D. R. Klein, D. MacNeill, E. Navarro-Moratalla, K. L. Seyler, N. Wilson, M. A. McGuire, D. H. Cobden, D. Xiao, W. Yao, P. Jarillo-Herrero, and X. Xu, Nature Nanotechnology 13, 544 (2018).
  5. [5]S. Jiang, J. Shan, and K. F. Mak, Nature Materials 17, 406 (2018).
  6. [6]E. S. Morell, A. León, R. H. Miwa, and P. Vargas, 2D Materials 6, 025020 (2019).
  7. [7]S. Jiang, L. Li, Z. Wang, K. F. Mak, and J. Shan, Nature Nanotechnology 13, 549 (2018).
  8. [8]M. A. McGuire, H. Dixit, V. R. Cooper, and B. C. Sales, Chemistry of Materials, Chemistry of Materials 27, 612 (2015).
  9. [9]N. Sivadas, S. Okamoto, X. Xu, C. J. Fennie, and D. Xiao, Nano Letters, Nano Letters 18, 7658 (2018).
  10. [10]D. Soriano, C. Cardoso, and J. Fernández-Rossier, Solid State Communications 299, 113662 (2019).
  11. [11]P. Jiang, C. Wang, D. Chen, Z. Zhong, Z. Yuan, Z.-Y. Lu, and W. Ji, Phys. Rev. B 99, 144401 (2019).
  12. [12]J. Kim, K.-W. Kim, B. Kim, C.-J. Kang, D. Shin, S.-H. Lee, B.-C. Min, and N. Park, Nano Letters, Nano Letters 929 (2020).
  13. [13]D.-H. Kim, K. Kim, K.-T. Ko, J. H. Seo, J. S. Kim, T.-H. Jang, Y. Kim, J.-Y. Kim, S.-W. Cheong, and J.-H. Park, Phys. Rev. Lett. 122, 207201 (2019).
  14. [14]S. Son, M. J. C. Koak, N. Lee, J. Kim, T. Y. Kim, H. Hamidov, H. Cho, C. Liu, D. M. Jarvis, P. A. C. Brown, J. H. Kim, C.-H. Park, D. I. Khomskii, S. S. Saxena, and J.-G. Park, Phys. Rev. B 99, 041402(R) (2019).
  15. [15]S. Tian, J.-F. Zhang, C. Li, T. Ying, S. Li, X. Zhang, K. Liu, and H. Lei, Journal of the American Chemical Society, Journal of the American Chemical Society 141, 5326 (2019).
  16. [16]Y.-P. Wang and M.-Q. Long, Phys. Rev. B 101, 024411 (2020).
  17. [17]C. Huang, F. Wu, S. Yu, P. Jena, and E. Kan, Phys. Chem. Chem. Phys. 22, 512 (2020).
  18. [18]K. Yang, F. Fan, H. Wang, D. I. Khomskii, and H. Wu, Phys. Rev. B 101, 100402(R) (2020).
  19. [19]J. He, S. Ma, P. Lyu, and P. Nachtigall, J. Mater. Chem. C 4, 2518 (2016).
  20. [20]T. Kong, K. Stolze, E. I. Timmons, J. Tao, D. Ni, S. Guo, Z. Yang, R. Prozorov, and R. J. Cava, Advanced Materials 31, 1808074 (2019).
  21. [21]J. A. Wilson, C. Maule, P. Strange, and J. N. Tothill, Journal of Physics C: Solid State Physics 20, 4159 (1987).
  22. [22]C. Long, T. Wang, H. Jin, H. Wang, and Y. Dai, The Journal of Physical Chemistry Letters, The Journal of Physical Chemistry Letters 11, 2158 (2020).
  23. [23]G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169 (1996).
  24. [24]J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996).
  25. [25]S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, and A. P. Sutton, Phys. Rev. B 57, 1505 (1998).
  26. [26]A. I. Liechtenstein, V. I. Anisimov, and J. Zaanen, Phys. Rev. B 52, R5467 (1995).
  27. [27]J. Neugebauer and M. Scheffler, Phys. Rev. B 46, 16067 (1992).
  28. [28]A. A. Mostofi, J. R. Yates, Y.-S. Lee, I. Souza, D. Vanderbilt, and N. Marzari, Computer Physics Communications 178, 685 (2008).
  29. [29]W. B. Wu, D. J. Huang, J. Okamoto, A. Tanaka, H.-J. Lin, F. C. Chou, A. Fujimori, and C. T. Chen, Phys. Rev. Lett. 94, 146402 (2005).
  30. [30]D. I. Khomskii, Transition Metal Compounds (Cambridge University Press, 2014).DOI
  31. [31]J. B. Goodenough, Phys. Rev. 171, 466 (1968).
  32. [32]P. Bruno, Phys. Rev. B 39, 865 (1989).
  33. [33]H. J. Xiang, E. J. Kan, S.-H. Wei, M.-H. Whangbo, and X. G. Gong, Phys. Rev. B 84, 224429 (2011).
  34. [34]H. Xiang, C. Lee, H.-J. Koo, X. Gong, and M.-H. Whangbo, Dalton Trans. 42, 823 (2013).
  35. [35]D. Šabani, C. Bacaksi̇z, and M. V. Milošević, Phys. Rev. B 102, 014457 (2020).
  36. [36]C. Xu, J. Feng, S. Prokhorenko, Y. Nahas, H. Xiang, and L. Bellaiche, Phys. Rev. B 101, 060404(R) (2020).
  37. [37]C. Xu, J. Feng, H. Xiang, and L. Bellaiche, npj Computational Materials 4, 57 (2018).
  38. [38]P. W. Anderson, Phys. Rev. 115, 2 (1959).
  39. [39]S. W. Jang, M. Y. Jeong, H. Yoon, S. Ryee, and M. J. Han, Phys. Rev. Materials 3, 031001(R) (2019).

Paper details

vlnc:W_b994a85c
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents