This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Vibronic effect on resonant inelastic x-ray scattering in cubic iridium hexahalides

Naoya Iwahara [email protected] Graduate School of Engineering, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba-shi, Chiba 263-8522, Japan    Wataru Furukawa Department of Materials Science, Faculty of Engineering, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba-shi, Chiba 263-8522, Japan
Abstract

In resonant inelastic x-ray scattering (RIXS) spectra of K2IrCl6, the peak for the j=3/2j=3/2 multiplet states shows a splitting that resembles non-cubic crystal-field effect although the compound is cubic down to 0.3 K. Here we theoretically describe the RIXS spectra concomitantly treating the spin-orbit and vibronic interactions. We found that the dynamic Jahn-Teller effect in the j=3/2j=3/2 multiplet states gives rise to the splitting of the RIXS spectra and the broadening of the spectra in raising the temperature. The validity of the interaction parameters for the simulations is supported by our ab initio calculations. Our results suggest that, in cubic iridium compounds, the dynamic Jahn-Teller effect induces the splitting of RIXS spectra without lowering the symmetry.

I Introduction

Strong spin-orbit coupling gives rise to rich quantum phenomena in 4d/5d4d/5d transition metal compounds [1, 2, 3]. One of the most investigated phenomena is the Kitaev spin liquid phase in Mott-insulating honeycomb compounds [4]. The candidate honeycomb materials consist of edge-sharing octahedra containing heavy d5d^{5} ions. The spin-orbit coupling of the d5d^{5} ion in the octahedral environment makes the local quantum states spin-orbit entangled j=1/2j=1/2 type and the exchange interaction between the d5d^{5} sites bond-dependent (Kitaev-type anisotropy). The Kitaev exchange interaction appears in other lattices such as triangular and face centered cubic (fcc) lattices [4], which could induce various magnetic phases [5, 6].

Motivated by the theoretical predictions, much experimental effort has been made to search Kitaev and related systems, while realizing pure j=1/2j=1/2 states in materials is far from trivial. Contrary to the assumption of a perfect octahedral environment in theories, candidate materials exhibit symmetry lowering, resulting in the hybridization of the j=1/2j=1/2 and the excited j=3/2j=3/2 states.

Recently, K2IrCl6 was reported as the first iridium compound with pure j=1/2j=1/2 states on metal sites [7]. The neutron powder diffraction data show that the compound is cubic down to 0.3 K [8]. The gg-tensor measured by electron paramagnetic resonance is consistent with the gg for j=1/2j=1/2 in an octahedral environment, suggesting no symmetry lowering [9].

Nevertheless, spectroscopic data look controversial to the cubic structure of K2IrCl6. Within the octahedral environment, the excited states of a single Ir site are four-fold degenerate j=3/2j=3/2 multiplet. In contrast, the resonant inelastic x-ray scattering (RIXS) measurement of the j=3/2j=3/2 states shows a splitting of 50\approx 50 meV down to low temperature (10 K) [8]. The RIXS spectra of cubic K2IrBr6 (in high-temperature phase) and Cs2IrCl6 resemble those of K2IrCl6 too [8, 10, 11]. The Raman scattering spectra of K2IrCl6 also show a similar splitting in the same range of energy [12]. As the origin of the splitting, non-cubic deformation of each octahedron and crystal-field splitting by rotational modes of the octahedra were discussed, whereas both interpretations require symmetry lowering of the system. Moreover, the latter leads to the splitting of only a few meV according to the previous experiments and calculations [10].

We recently proved ab initio that the vibronic excitations induce the splittings of peaks in the RIXS spectra of a similar compound, K2RuCl6 [13, 14]. In this compound, the dynamic Jahn-Teller (JT) effect develops in the excited spin-orbit multiplet states and determines the shape and temperature evolution of the RIXS spectra. The dynamic JT effect is characterized by the superpositions of many electron-lattice configurations with JT-deformed structures, and hence, structural anisotropy does not appear in K2RuCl6. Since the vibronic coupling tends to be stronger in 5d5d systems than in 4d4d systems [15], the dynamic JT effect would give nonnegligible changes in the RIXS spectra of the Ir compounds.

In this work, we examine the presence of the dynamic JT effect and reveal its impact on the RIXS spectra of the Ir4+ compounds. We calculate the vibronic states of the dynamic JT model and, based on them, the RIXS spectra. We confirm the validity of the derived interaction parameters by a series of post Hartree-Fock calculations.

II Vibronic model for 5d55d^{5} ion

Here we explain the theoretical framework for the quantum states and the RIXS spectra of a single Ir site in the cubic K2IrX6X_{6} (X=X= Cl, Br). We discuss only the phenomena in a single Ir octahedron because the energy scale of the intersite exchange interaction (about 1 meV) [7, 9, 10] is several tens times smaller than the splitting (about 35-60 meV) in the RIXS spectra [8, 10].

The quantum states of each 5d55d^{5} site are determined by the interplay of the ligand field, spin-orbit, and vibronic interactions. We consider the interactions in the order because their energy scales decrease from several eV by about one order of magnitude. The octahedral ligand field splits the 5d5d orbitals into ege_{g} doublet and t2gt_{2g} triplet. The t2gt_{2g} orbitals are lower in energy and have five electrons (one hole). Hereafter, we use the hole picture. Within the t2gt_{2g} orbitals, the spin-orbit coupling is unquenched [16]:

H^SO\displaystyle\hat{H}_{\text{SO}} =λ𝒍~𝒔^,\displaystyle=-\lambda\tilde{\bm{l}}\cdot\hat{\bm{s}}, (1)

where 𝒍~\tilde{\bm{l}} is the effective l~=1\tilde{l}=1 orbital operator, 𝒔^\hat{\bm{s}} is the spin operator, and λ\lambda (>0)(>0) is the spin-orbit coupling parameter. The t2gt_{2g} hole couples to the JT active EgE_{g} vibrations too: the JT model including the harmonic oscillator Hamiltonian for the EgE_{g} modes is [17]

H^JT\displaystyle\hat{H}_{\text{JT}} =γ=u,vω2(p^γ2+q^γ2)ωg[(12q^u+32q^v)P^yz\displaystyle=\sum_{\gamma=u,v}\frac{\hslash\omega}{2}\left(\hat{p}_{\gamma}^{2}+\hat{q}_{\gamma}^{2}\right)-\hslash\omega g\left[\left(-\frac{1}{2}\hat{q}_{u}+\frac{\sqrt{3}}{2}\hat{q}_{v}\right)\hat{P}_{yz}\right.
+(12q^u32q^v)P^zx+q^uP^xy].\displaystyle+\left.\left(-\frac{1}{2}\hat{q}_{u}-\frac{\sqrt{3}}{2}\hat{q}_{v}\right)\hat{P}_{zx}+\hat{q}_{u}\hat{P}_{xy}\right]. (2)

Here q^γ\hat{q}_{\gamma} are the dimensionless normal coordinates for the EgE_{g} modes (γ=u,v\gamma=u,v, where uu and vv transform as z2z^{2} and x2y2x^{2}-y^{2}, respectively), p^γ\hat{p}_{\gamma} the conjugate momenta of q^γ\hat{q}_{\gamma}, ω\omega the frequency of the EgE_{g} modes, gg the dimensionless vibronic coupling parameter, and P^γ\hat{P}_{\gamma} (γ=yz,zx,xy\gamma=yz,zx,xy) is the projection operator into the t2gγt_{2g}\gamma orbital. We ignore the vibronic coupling to the T2gT_{2g} modes because it is much weaker than that to the EgE_{g} modes.

The spin-orbit interaction splits the t2g1t_{2g}^{1} hole configurations into effective j=1/2j=1/2 and j=3/2j=3/2 multiplet states. These states belong to the eigen energies of H^SO\hat{H}_{\text{SO}}, λ2-\frac{\lambda}{2} and λ\lambda, respectively. Using the spin-orbit multiplet states, the vibronic coupling part of H^JT\hat{H}_{\text{JT}} becomes

ωg2(0002q^v02q^u002q^u02q^v002q^uq^u0q^v02q^v00q^u0q^v02q^vq^v0q^u02q^u00q^v0q^u).\displaystyle\frac{\hslash\omega g}{2}\begin{pmatrix}0&0&0&\sqrt{2}\hat{q}_{v}&0&-\sqrt{2}\hat{q}_{u}\\ 0&0&\sqrt{2}\hat{q}_{u}&0&-\sqrt{2}\hat{q}_{v}&0\\ 0&\sqrt{2}\hat{q}_{u}&-\hat{q}_{u}&0&-\hat{q}_{v}&0\\ \sqrt{2}\hat{q}_{v}&0&0&\hat{q}_{u}&0&-\hat{q}_{v}\\ 0&-\sqrt{2}\hat{q}_{v}&-\hat{q}_{v}&0&\hat{q}_{u}&0\\ -\sqrt{2}\hat{q}_{u}&0&0&-\hat{q}_{v}&0&-\hat{q}_{u}\\ \end{pmatrix}. (3)

The basis of the matrix is in the order of j=12j=\frac{1}{2} (mj=12,12m_{j}=-\frac{1}{2},\frac{1}{2}) and j=32j=\frac{3}{2} (mj=32,12,12,32m_{j}=-\frac{3}{2},-\frac{1}{2},\frac{1}{2},\frac{3}{2}). Eq. (3) indicates the presence of the JT effect in the j=3/2j=3/2 multiplet states and the pseudo JT effect between the j=1/2j=1/2 and j=3/2j=3/2 spin-orbit multiplets.

The energy eigenstates of 5d55d^{5} sites take the form of spin-orbit and lattice entanglement. Since the vibronic interaction, Eq. (3), hybridizes the spin-orbit multiplet states via the quantized normal coordinates, the energy eigenstates (vibronic states) of H^SO+H^JT\hat{H}_{\text{SO}}+\hat{H}_{\text{JT}} are

|ψν=jmj|jmj|χjmj;ν,\displaystyle|\psi_{\nu}\rangle=\sum_{jm_{j}}|jm_{j}\rangle\otimes|\chi_{jm_{j};\nu}\rangle, (4)

where |χjmj;ν|\chi_{jm_{j};\nu}\rangle include the information of the lattice degrees of freedom. The vibronic states reduce to the direct products of the spin-orbit multiplet and vibrational states in the limit of g0g\rightarrow 0.

Now, we turn to the Ir L3L_{3} RIXS spectra of the vibronic system. The L3L_{3} RIXS is a photon-in photon-out process involving electron transitions from the 2p3/22p_{3/2} states to the t2gt_{2g} (5d5d) orbitals, and then, from the t2gt_{2g} orbitals to the 2p3/22p_{3/2} within an Ir ion. Assuming that the initial and final states of the system are of vibronic type, Eq. (4), the cross-section for the L3L_{3} RIXS process in a single Ir center is [14]

d2σdΩdk\displaystyle\frac{d^{2}\sigma}{d\Omega dk^{\prime}} νρνkk|ααϵ𝒌λ,αϵ𝒌λ,αν|d^αP^ehd^α|ν|2\displaystyle\propto\sum_{\nu}\rho_{\nu}\frac{k^{\prime}}{k}\left|\sum_{\alpha^{\prime}\alpha}\epsilon_{\bm{k}^{\prime}\lambda^{\prime},\alpha^{\prime}}\epsilon_{\bm{k}\lambda,\alpha}\langle\nu^{\prime}|\hat{d}_{\alpha^{\prime}}\hat{P}_{\text{eh}}\hat{d}_{\alpha}|\nu\rangle\right|^{2}
×δ(Eν+ωEνω),\displaystyle\times\delta\left(E_{\nu}+\hslash\omega-E_{\nu^{\prime}}-\hslash\omega^{\prime}\right), (5)

by using the Kramers-Heisenberg formula [18, 19] and the dipole and fast collision approximations [20, 21]. In Eq. (5), ν\nu (ν\nu^{\prime}) indicate the initial (final) vibronic states, ρν\rho_{\nu} the canonical distribution for the Ir center, d^α\hat{d}_{\alpha} the α\alpha (=x,y,z=x,y,z) component of the electric dipole operator, P^eh\hat{P}_{\text{eh}} the projection operator into the electron-hole states with (2p3/2)3(t2g)6(2p_{3/2})^{3}(t_{2g})^{6} electron configurations, and ωi\hslash\omega_{i} (ωf\hslash\omega_{f}) and ϵ𝒌λ\bm{\epsilon}_{\bm{k}\lambda} (ϵ𝒌λ)\bm{\epsilon}_{\bm{k}^{\prime}\lambda^{\prime}}) are, respectively, the energy and the polarization of the incident (scattered) photon. We used the fast collision approximation for the Ir compounds because the energy scale of the lifetime broadening of Ir ion (\approx 4-5 eV) [22] is one order of magnitude larger than that of the electron-hole dynamics measured by the j=3/2j=3/2 band width (0.7\approx 0.7 eV) [7].

III Computational methods

For the derivation of the vibronic states, Eq. (4), we performed a numerical diagonalization of the model Hamiltonian [see e.g. Refs. [14, 15, 23]]. We expanded the lattice parts of the vibronic states, Eq. (4), as

|χjmj;ν\displaystyle|\chi_{jm_{j};\nu}\rangle =nu,nv|nu,nvCjmj,nu,nv;ν.\displaystyle=\sum_{n_{u},n_{v}}|n_{u},n_{v}\rangle C_{jm_{j},n_{u},n_{v};\nu}. (6)

To determine the coefficients Cjmj,nu,nv;νC_{jm_{j},n_{u},n_{v};\nu}, we calculated the Hamiltonian matrix of the JT model (H^SO+H^JT\hat{H}_{\text{SO}}+\hat{H}_{\text{JT}}) and diagonalized it. The basis of the matrix was a set of the direct products of the |j,mj|j,m_{j}\rangle (j=12,32j=\frac{1}{2},\frac{3}{2}) and the eigenstates of the harmonic oscillator part of H^JT\hat{H}_{\text{JT}}, |nu,nv|n_{u},n_{v}\rangle (nu,nv=0,1,2,)(n_{u},n_{v}=0,1,2,\cdots). We truncated our basis so that 0nu+nv150\leq n_{u}+n_{v}\leq 15.

The dynamic JT Hamiltonian contains interaction parameters, λ\lambda, ω\omega, and gg. We took λ\lambda from the RIXS data and ω\omega from the Raman scattering data and treated gg as a variable. We determined gg so that we can reproduce the experimental RIXS spectra.

To check the validity of the chosen parameters, we also derived all the interaction parameters by a series of post Hartree-Fock calculations as in Refs. [23, 15, 14]. We made clusters of K2IrCl6 and K2IrBr6 using the neutron diffraction data at 0.3 K for the former and 200 K for the latter, respectively [8]. Each cluster consists of three groups of atoms. The first group corresponds to K8IrX6X_{6} (X=X= Cl, Br) cluster, which we treated ab initio with atomic-natural-orbital relativistic-correlation consistent-valence quadruple-ζ\zeta polarization basis set. The second group contains the nearest 12 octahedra and the neighboring 48 K atoms, which are replaced by ab initio embedded potential [24]. The third part consists of point charges. The total charge of the entire cluster is neutral. We calculated T2g2{}^{2}T_{2g} states of the clusters with the complete active space self-consistent field (CASSCF) method [25] followed by a correction within the extended multistate complete active space second-order perturbation theory (XMS-CASPT2) [26, 27], and then, the spin-orbit jj multiplet states with spin-orbit restricted active space state interaction (SO-RASSI) method [25] implemented in OpenMolcas [28, 29]. In these calculations, we set three 5d5d t2gt_{2g} orbitals with one hole as the active space. We determined λ\lambda by using the SO-RASSI energy gap.

For the calculations of the vibronic coupling parameters and the frequencies, we fit the T2g2{}^{2}T_{2g} adiabatic potential energy surfaces with respect to the JT deformation to the JT model, Eq. (2) [See Refs. [23, 15, 14]]. We generated several JT-deformed Cartesian coordinates of an octahedron by using 𝑹A=𝑹A(0)+Qu(𝒆Eu)A/M\bm{R}_{A}=\bm{R}_{A}^{(0)}+Q_{u}\left(\bm{e}_{Eu}\right)_{A}/\sqrt{M} and calculated the term energies UγU_{\gamma} (γ=yz,zx,xy\gamma=yz,zx,xy) at each structure. Here 𝑹A\bm{R}_{A} are the Cartesian coordinates of ligand atom AA, 𝑹(0)\bm{R}^{(0)} indicates the experimental octahedral configurations, 𝒆Eu\bm{e}_{Eu} is the JT-active EuEu polarization vectors of the neighboring six Cl’s or Br’s, QuQ_{u} the mass-weighted normal coordinates (Qu=/ωquQ_{u}=\sqrt{\hslash/\omega}q_{u}), and MM the mass of ligand atoms. The gradient of UxyU_{xy} with respect to QuQ_{u} is the vibronic coupling parameter, V=Uxy/QuV=-\partial U_{xy}/\partial Q_{u}. The curvature of UU is mainly ω2\omega^{2} with a small correction from the quadratic vibronic coupling. The quadratic vibronic coupling has a similar form as the linear one in Eq. (2): the former is obtained by replacing q^γ\hat{q}_{\gamma} and gg in the latter with the symmetrized quadratic form of q^\hat{q} [q^u12(q^u2q^v2)\hat{q}_{u}\rightarrow-\frac{1}{\sqrt{2}}(\hat{q}_{u}^{2}-\hat{q}_{v}^{2}) and q^v2q^uq^v\hat{q}_{v}\rightarrow\sqrt{2}\hat{q}_{u}\hat{q}_{v}] and ww, respectively (see e.g., Ref. [15]). With VV and ω\omega, gg is defined by g=V/ω3g=V/\sqrt{\hslash\omega^{3}}.

(a) (b)
Refer to caption Refer to caption
(c) (d)
Refer to caption Refer to caption
Figure 1: Vibronic levels and RIXS spectra for (a)-(c) K2IrCl6 and (d) K2IrBr6. (a) The vibronic energy spectra with respect to gg. EJT=ω(g/2)2/2E_{\text{JT}}=\hslash\omega(g/2)^{2}/2. (b) The RIXS spectra with respect to gg at T=10T=10 K. (c) Theoretical (g=1.2g=1.2) and the experimental [8] RIXS spectra at T=10T=10 K and T=300T=300 K. The vertical lines indicate the cross-sections (×10\times 10) at 10 K (blue) and at 300 K (red). (d) Theoretical (g=1.5g=1.5) and the experimental [10] RIXS spectra at T=270T=270 K.

IV Results and discussions

IV.1 Vibronic energy spectra

We calculated the vibronic levels of the dynamic JT model for a single Ir4+ site. For the calculations, we used the interaction parameters from the experiments of K2IrCl6: λ=0.436\lambda=0.436 eV [8] and ω=37.1\omega=37.1 meV [12]. We reduced the reported λ\lambda by 0.04 eV for a better agreement between the theoretical and experimental RIXS spectra (Sec. IV.2).

The vibronic energy spectra around the excited spin-orbit j=3/2j=3/2 multiplet level show nontrivial distribution. Figure 1(a) indicates the energy spectra with respect to gg. Each level is either two-fold (the points) or four-fold (the open circles) degenerate. At g=0g=0 (no JT effect), the energy eigenstates in the figure correspond to the product states of |j=3/2,mj|nu,nv|j=3/2,m_{j}\rangle\otimes|n_{u},n_{v}\rangle type (nu+nv=0,1,2,n_{u}+n_{v}=0,1,2,...). Turning on gg, the vibronic coupling hybridizes these product states as Eq. (4). Consequently, the distribution of the vibronic levels for |g|>0|g|>0 differs from that without the JT effect (g=0g=0). Although the vibronic coupling hybridizes the ground j=1/2j=1/2 states with the excited j=3/2j=3/2 states, the ground states remain close to the j=1/2j=1/2 multiplet states, |j=1/2,mj|0,0|j=1/2,m_{j}\rangle\otimes|0,0\rangle, in the present range of gg: the weight (probability) of the pure j=1/2j=1/2 components in the ground state is 0.998 at g=1.0g=1.0 and 0.990 at g=2g=2.

We also calculated the vibronic states of K2IrBr6. We used λ=0.395\lambda=0.395 eV [10] and ω=24.7\omega=24.7 meV. We reduced λ\lambda from the experimental value by 0.05 eV. We estimated ω\omega by assuming that the elastic couplings KK for the EgE_{g} modes are common for K2IrCl6 and K2IrBr6 and using ωK/M\omega\propto\sqrt{K/M}. The pattern of the calculated vibronic states of K2IrBr6 resembles that of K2IrCl6. The ground state is well described with |j=1/2,mj|0,0|j=1/2,m_{j}\rangle\otimes|0,0\rangle: the weight 0.999 at g=1.0g=1.0 and 0.995 at g=2.0g=2.0.

IV.2 Vibronic peaks in RIXS spectra

With the derived vibronic states, we calculated the RIXS spectra around the j=3/2j=3/2 multiplet level at 10 K using Eq. (5). We set the incident and scattered photons following the experimental condition [8]. We convoluted the cross-section, Eq. (5), with the Lorentzian function with width Γ=50\Gamma=50 meV.

Varying gg, the shape of the RIXS spectrum changes [Fig. 1(b)]. When g=0g=0, the RIXS spectrum has only one peak corresponding to the gap between the ground j=1/2j=1/2 level and the j=3/2j=3/2 level with no vibrational excitations (nu=nv=0n_{u}=n_{v}=0). The transitions from the ground j=1/2j=1/2 states [|j=1/2,mj|0,0|j=1/2,m_{j}\rangle\otimes|0,0\rangle] to the vibrationally excited j=3/2j=3/2 multiplets [|j=3/2,mj|nu,nv|j=3/2,m_{j}\rangle\otimes|n_{u},n_{v}\rangle, nu+nv>0n_{u}+n_{v}>0] are zero due to the orthogonality of the vibrational states. As gg increases, the peak splits into a few originating from the transitions between the ground j=1/2j=1/2 states and several excited vibronic states (see the blue vertical lines): the latter states contain non-negligible contributions from |j=3/2,mj|0,0|j=3/2,m_{j}\rangle\otimes|0,0\rangle states.

The calculated RIXS spectrum with g=1.2g=1.2 agrees well with the experimental one from Ref. [8] [Fig. 1(c)]. The calculated spectrum has two strong peaks at about 0.645 eV and 0.690 eV corresponding to the transitions between the ground j=1/2j=1/2 multiplet and the four-fold degenerate vibronic states [The lowest two four-fold degenerate levels in Fig. 1(a)]. The gap between the peaks is about 45 meV, which is close to the experimental estimation of 48 meV.

Raising the temperature from 10 K to 300 K, the shape of the RIXS spectrum changes. To calculate the RIXS spectrum at 300 K, we only vary the temperature in the convoluted cross-section: the other parameters remain the same. The height of the main peak at 0.645 eV decreases, and the width increases as the temperature rises. These temperature dependencies agree with those of the experimental data. The temperature evolution of the spectra occurs due to the transitions from the thermally populated states with vibrational excitations to the vibronic states [the red lines in Fig. 1(c)]. Increasing the temperature from 10 K to 300 K, the statistical weight of the ground j=1/2j=1/2 states reduces from 1.000 to 0.558, which lowers the height of the main peak. The slight enhancement of the second peak originates from the transitions between the thermally populated first excited states to the excited vibronic states.

We also calculated the RIXS spectrum of cubic K2IrBr6 at 270 K. For the calculation, we used the vibronic states with g=1.5g=1.5 and the Lorentzian function with Γ=\Gamma= 40 meV. We estimated gg for K2IrBr6 from gg for K2IrCl6 by using gM1/4g\propto M^{1/4}. The relation is obtained by using the definition of g=V/ω3g=V/\sqrt{\hslash\omega^{3}} and V,ωM1/2V,\omega\propto M^{-1/2}. Figure 1(d) shows a good agreement between the calculated and experimental RIXS spectra [10]. The calculated spectrum has two peaks at 0.587 eV and 0.617 eV, and the gap between them (30 meV) is smaller than that of K2IrCl6, which is in line with the experimental observation (the gap is 33 meV) [10]. The theoretical spectrum underestimates the height of the peak at 0.617 eV. The heights of the peaks could vary by a hybridization of the corresponding j=3/2j=3/2 vibronic states due to the intersite interactions in the intermediate states. The gap between the peaks and the intersite interactions are smaller [8, 10] and larger [7] in K2IrBr6 than in K2IrCl6, respectively, and thus, the hybridization effect should be enhanced in K2IrBr6.

Our calculations suggest that the ground states are close to the pure spin-orbit j=1/2j=1/2 multiplet states in K2IrCl6 down to low-temperature and K2IrBr6 at high-temperature. The weight of the j=1/2j=1/2 multiplets in each of the ground vibronic doublets is 0.997 in K2IrCl6 and 0.998 in K2IrBr6.

Refer to caption
Figure 2: The T2g2{}^{2}T_{2g} term energies of K2IrCl6 with respect to the EguE_{g}u deformations. The red points are the ab initio data, and the solid (blue) and dashed (black) lines are the model adiabatic potential energy surfaces with and without quadratic vibronic coupling, respectively.

To confirm the validity of our interaction parameters, we performed ab initio calculations [Fig. 2]. The calculated interaction parameters are λ=0.469\lambda=0.469 eV, ω=\omega= 40.7 meV, g=0.94g=0.94, and w=0.11w=0.11 for K2IrCl6 and λ=0.472\lambda=0.472 eV, ω=24.5\omega=24.5 meV, g=1.03g=1.03, and w=0.13w=0.13 for K2IrBr6. The calculated λ\lambda and ω\omega are close to the experimental values: the calculated data are about 10 % larger than the experimental data [8, 10, 12]. The gg for K2IrBr6 is larger than that of K2IrCl6, which is in line with our estimation above. The overestimation of λ\lambda and ω\omega is due to insufficient delocalization of the t2gt_{2g} orbitals over the ligands within the post Hartree-Fock method. Due to the overestimation of ω\omega’s, the gg’s are underestimated. The calculated ww is about 10 times weaker than gg, which allows neglecting the quadratic vibronic coupling in the JT model, Eq. (2). Overall, we conclude that the ab initio calculations support the validity of the parameters extracted from the experimental data and the mechanism of the splitting.

Based on our interaction parameters, we rule out the non-cubic (or static JT effect) scenario. The JT deformations that reproduce the splitting of the j=3/2j=3/2 peak correspond to the shifts of the ligand atoms of about 0.007 Å for K2IrCl6 and 0.004 Å for K2IrBr6. Such large shifts were not observed in the x-ray and neutron diffraction data of K2IrCl6 down to low temperature.

We propose that the rotational modes of the octahedra enlarge the line width of the RIXS spectra. The rotational modes quadratically couple to the t2gt_{2g} orbitals [10] and, hence, they can contribute to the spin-orbit and lattice entanglement as the JT active modes. The rotational modes will give rise to new transitions between the levels around j=1/2j=1/2 and j=3/2j=3/2 with about one order of magnitude smaller gaps than the splitting of the j=3/2j=3/2 peak according to their frequencies: within the present resolution of energy, these transitions will enlarge the line width of the spectra. Due to the enlargement of the line width by raising the temperature, the peak positions of the spectra will shift, and hence, the energy gap will slightly increase.

Finally, we comment on the nature of the JT effect in the 4d4d and 5d5d spin-orbit Mott insulators. Our analysis shows that the vibronic coupling in K2IrCl6 (5d55d^{5}) is only about 10 % larger than that in K2RuCl6 (4d44d^{4}) [14]: the change is much smaller than that of spin-orbit coupling. Under this situation, the 4d4d and 5d5d compounds should exhibit different regimes of the interplay of the JT effect and spin-orbit coupling. Emergent phenomena induced by the interplay of the vibronic and spin-orbit couplings deserve future studies.

V Conclusion

In this work, we theoretically demonstrated that the dynamic JT effect determines the shape and the temperature dependence of the RIXS spectra of K2IrCl6. In particular, the dynamic JT effect in the j=3/2j=3/2 multiplet states develops and splits the peak for the j=3/2j=3/2 multiplet in the RIXS spectra. The interaction parameters used for the simulations are consistent with our ab initio interaction parameters. Thus, our results indicate that the system is cubic. The present results suggest that the RIXS spectra of spin-orbit Mott insulators with (quasi) degenerate multiple states on sites have to be interpreted fully considering the dynamic (pseudo) JT effect.

Acknowledgements.
This work was partly supported by the Iketani Science and Technology Foundation and Grant-in-Aid for Scientific Research (Grant No. 22K03507) from the Japan Society for the Promotion of Science.

References

  • Takagi et al. [2019] H. Takagi, T. Takayama, G. Jackeli, G. Khaliullin, and S. E. Nagler, Concept and realization of Kitaev quantum spin liquids, Nat. Rev. Phys. 1, 264 (2019).
  • Takayama et al. [2021] T. Takayama, J. Chaloupka, A. Smerald, G. Khaliullin, and H. Takagi, Spin-Orbit-Entangled Electronic Phases in 4d4d and 5d5d Transition-Metal Compounds, J. Phys. Soc. Jpn. 90, 062001 (2021).
  • Trebst and Hickey [2022] S. Trebst and C. Hickey, Kitaev materials, Phys. Rep. 950, 1 (2022).
  • Jackeli and Khaliullin [2009] G. Jackeli and G. Khaliullin, Mott Insulators in the Strong Spin-Orbit Coupling Limit: From Heisenberg to a Quantum Compass and Kitaev Models, Phys. Rev. Lett. 102, 017205 (2009).
  • Kimchi et al. [2014] I. Kimchi, J. G. Analytis, and A. Vishwanath, Three-dimensional quantum spin liquids in models of harmonic-honeycomb iridates and phase diagram in an infinite-DD approximation, Phys. Rev. B 90, 205126 (2014).
  • Cook et al. [2015] A. M. Cook, S. Matern, C. Hickey, A. A. Aczel, and A. Paramekanti, Spin-orbit coupled jeff=1/2{j}_{\mathrm{eff}}=1/2 iridium moments on the geometrically frustrated fcc lattice, Phys. Rev. B 92, 020417 (2015).
  • Khan et al. [2019] N. Khan, D. Prishchenko, Y. Skourski, V. G. Mazurenko, and A. A. Tsirlin, Cubic symmetry and magnetic frustration on the fcc spin lattice in K2IrCl6{\mathrm{K}}_{2}{\mathrm{IrCl}}_{6}Phys. Rev. B 99, 144425 (2019).
  • Reig-i Plessis et al. [2020] D. Reig-i Plessis, T. A. Johnson, K. Lu, Q. Chen, J. P. C. Ruff, M. H. Upton, T. J. Williams, S. Calder, H. D. Zhou, J. P. Clancy, A. A. Aczel, and G. J. MacDougall, Structural, electronic, and magnetic properties of nearly ideal Jeff=12{J}_{eff}=\frac{1}{2} iridium halides, Phys. Rev. Mater. 4, 124407 (2020).
  • Bhaskaran et al. [2021] L. Bhaskaran, A. N. Ponomaryov, J. Wosnitza, N. Khan, A. A. Tsirlin, M. E. Zhitomirsky, and S. A. Zvyagin, Antiferromagnetic resonance in the cubic iridium hexahalides (NH4)2IrCl6({\mathrm{NH}}_{4}{)}_{2}{\mathrm{IrCl}}_{6} and K2IrCl6{\mathrm{K}}_{2}{\mathrm{IrCl}}_{6}Phys. Rev. B 104, 184404 (2021).
  • Khan et al. [2021] N. Khan, D. Prishchenko, M. H. Upton, V. G. Mazurenko, and A. A. Tsirlin, Towards cubic symmetry for Ir4+{\mathrm{Ir}}^{4+}: Structure and magnetism of the antifluorite K2IrBr6{\mathrm{K}}_{2}{\mathrm{IrBr}}_{6}Phys. Rev. B 103, 125158 (2021).
  • [11] W. Tian, Q. Chen, A. A. Aczel, C. R. Wiebe, H. Zhou, and J. Cheng, Magnetic properties of Cs2IrCl6 single crystals, APS March Meeting 2023, G39.3.
  • Lee et al. [2022] S. Lee, B. H. Kim, M.-J. Seong, and K.-Y. Choi, Noncubic local distortions and spin-orbit excitons in K2IrCl6{\mathrm{K}}_{2}{\mathrm{IrCl}}_{6}Phys. Rev. B 105, 184433 (2022).
  • Takahashi et al. [2021] H. Takahashi, H. Suzuki, J. Bertinshaw, S. Bette, C. Mühle, J. Nuss, R. Dinnebier, A. Yaresko, G. Khaliullin, H. Gretarsson, T. Takayama, H. Takagi, and B. Keimer, Nonmagnetic J=0J=0 State and Spin-Orbit Excitations in K2RuCl6{\mathrm{K}}_{2}{\mathrm{RuCl}}_{6}Phys. Rev. Lett. 127, 227201 (2021).
  • Iwahara and Shikano [2023] N. Iwahara and S. Shikano, Vibronic excitations in resonant inelastic x-ray scattering spectra of K2RuCl6{\mathrm{K}}_{2}{\mathrm{RuCl}}_{6}Phys. Rev. Res. 5, 023051 (2023).
  • Iwahara et al. [2018] N. Iwahara, V. Vieru, and L. F. Chibotaru, Spin-orbital-lattice entangled states in cubic d1{d}^{1} double perovskites, Phys. Rev. B 98, 075138 (2018).
  • Sugano et al. [1970] S. Sugano, Y. Tanabe, and H. Kamimura, Multiplets of Transition-Metal Ions in Crystals (Academic Press, New York, 1970).
  • Bersuker and Polinger [1989] I. B. Bersuker and V. Z. Polinger, Vibronic Interactions in Molecules and Crystals (Springer-Verlag, Berlin and Heidelberg, 1989).
  • Sakurai [1967] J. J. Sakurai, Advanced Quantum Mechanics (Addison-Wesley, Massachusetts, 1967).
  • Ament et al. [2011] L. J. P. Ament, M. van Veenendaal, T. P. Devereaux, J. P. Hill, and J. van den Brink, Resonant inelastic x-ray scattering studies of elementary excitations, Rev. Mod. Phys. 83, 705 (2011).
  • Luo et al. [1993] J. Luo, G. T. Trammell, and J. P. Hannon, Scattering operator for elastic and inelastic resonant x-ray scattering, Phys. Rev. Lett. 71, 287 (1993).
  • van Veenendaal [2006] M. van Veenendaal, Polarization Dependence of LL- and MM-Edge Resonant Inelastic X-Ray Scattering in Transition-Metal Compounds, Phys. Rev. Lett. 96, 117404 (2006).
  • Clancy et al. [2012] J. P. Clancy, N. Chen, C. Y. Kim, W. F. Chen, K. W. Plumb, B. C. Jeon, T. W. Noh, and Y.-J. Kim, Spin-orbit coupling in iridium-based 5dd compounds probed by x-ray absorption spectroscopy, Phys. Rev. B 86, 195131 (2012).
  • Iwahara et al. [2017] N. Iwahara, V. Vieru, L. Ungur, and L. F. Chibotaru, Zeeman interaction and Jahn-Teller effect in the Γ8{\mathrm{\Gamma}}_{8} multiplet, Phys. Rev. B 96, 064416 (2017).
  • Seijo and Barandiarán [1999] L. Seijo and Z. Barandiarán, Computational Modelling of the Magnetic Properties of Lanthanide Compounds, in Computational Chemistry: Reviews of Current Trends, Vol. 4, edited by J. Leszczynski (World Scientific, Singapore, 1999) pp. 55–152.
  • Roos et al. [2016] B. O. Roos, R. Lindh, P.-Å. Malmqvist, V. Veryazov, and P.-O. Widmark, Multiconfigurational Quantum Chemistry (Wiley, New Jersey, 2016).
  • Granovsky [2011] A. A. Granovsky, Extended multi-configuration quasi-degenerate perturbation theory: The new approach to multi-state multi-reference perturbation theory, J. Chem. Phys. 134, 214113 (2011).
  • Shiozaki et al. [2011] T. Shiozaki, W. Győrffy, P. Celani, and H.-J. Werner, Extended multi-state complete active space second-order perturbation theory: Energy and nuclear gradients, J. Chem. Phys. 135, 081106 (2011).
  • Fdez. Galván et al. [2019] I. Fdez. Galván et al., OpenMolcas: From Source Code to Insight, J. Chem. Theor. Comput. 15, 5925 (2019).
  • Aquilante et al. [2020] F. Aquilante et al., Modern quantum chemistry with [Open]Molcas, J. Chem. Phys. 152, 214117 (2020).