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

Compass Impurity Model of Tb Substitution in Sr2IrO4

Long Zhang International Center for Quantum Materials, School of Physics, Peking University, Beijing, 100871, China    Fa Wang International Center for Quantum Materials, School of Physics, Peking University, Beijing, 100871, China Collaborative Innovation Center of Quantum Matter, Beijing, 100871, China    Dung-Hai Lee Department of Physics, University of California, Berkeley, CA 94720, USA Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
(September 28, 2025)
Abstract

We show that upon Tb substitution the interaction between the magnetic moments on the impurity Tb4+ ion and its surrounding Ir4+ ions is described by a “compass” model, i.e., Ising-like interaction favoring the magnetic moments across each bond to align along the bond direction. Such interaction nucleates quenched magnetic vortices near the impurities and drives a reentrant transition out of the antiferromagnetic ordered phase at low temperatures hence quickly suppresses the Néel temperature consistent with the experiment [Phys. Rev. B 92, 214411 (2015)]. As a by-product, we propose that the compass model can be realized in ordered double perovskites composed of the spin-orbital-coupled d5d^{5} ions and the half-closed-shell f7f^{7} ions.

Introduction.—The layered iridate compound Sr2IrO4 has attracted much attention recently partly due to its close resemblance to the cuprate superconductors Kim et al. (2008); Jackeli and Khaliullin (2009); Watanabe et al. (2010); Wang and Senthil (2011). It is described by a one-band (pseudo)spin-1/21/2 Hubbard model like the isostructural cuprate parent compound La2CuO4 Wang and Senthil (2011). Therefore, many interesting phenomena common to the cuprates are also found in Sr2IrO4. For example antiferromagnetic (AF) order exists in the parent compound Kim et al. (2009); Dhital et al. (2013); Ye et al. (2013), and upon doping Fermi arcs are seen in angle-resolved photoemission spectroscopy (ARPES) Kim et al. (2014). Moreover both scanning tunneling spectroscopy Yan et al. (2015) and ARPES Kim et al. (2015) suggest a low temperature nodal gap. Whether the latter is due to superconductivity is currently actively investigated.

The microscopic origins of the effective one-band description in Sr2IrO4 and cuprates, however, are quite different. For example in the parent compound the half filled band in cuprates has mixed copper dx2y2d_{x^{2}-y^{2}} and oxygen px,yp_{x,y} characters Zhang and Rice (1988). For most purposes spin-orbit coupling (SOC) is negligible. In contrast, the relevant band of Sr2IrO4 derives from an effective total angular momentum (pseudospin) 1/21/2 spin orbit coupled crystal field orbital. Due to the narrow bandwidth even the relatively weak on-site Coulomb correlation can render it Mott insulating Kim et al. (2008). At low temperatures the magnetic moments associated with this band become AF long range ordered Kim et al. (2009); Dhital et al. (2013); Ye et al. (2013).

In a recent experiment Wang et al. (2015) Tb4+ impurities are substituted into Sr2IrO4 to replace the Ir4+ ions. The Néel temperature is fully suppressed by less than 3%3\% Tb substitution. At low temperatures the hysteretic magnetic susceptibility and the linear-TT specific heat behaviors suggest the formation of a spin glass state. These phenomena are reminiscent of the insulating lightly hole-doped cuprates, in which the Néel order is also replaced by a spin glass state at low temperatures Keimer et al. (1992); Chou et al. (1993); Niedermayer et al. (1998); Coneri et al. (2010). Theoretically it has been shown the doped holes create nonlocal dipolar distortions, i.e., magnetic vortex-antivortex pairs around the holes, which quickly destroy the AF order Aharony et al. (1988); Glazman and Ioselevich (1989, 1990); Cherepanov et al. (1999); Shraiman and Siggia (1988); Kou and Weng (2003a, b); Mei (2013). However, we expect the microscopic mechanism for the Tb substituted Sr2IrO4 to be different because the isovalent Tb4+ substitution does not introduce extra charge carriers in the IrO2 plane.

Refer to caption
Figure 1: Left: Configuration of the Tb4+ impurity surrounded by four Ir4+ ions. The lattice is slightly distorted, θ11\theta\simeq 11^{\circ}. The magnetic interaction is Ising-like on each bond. Right: The bond geometries of the symmetry-allowed hopping processes t1t_{1} (upper) and t2t_{2} (lower). The signs of the wavefunctions are indicated by the filled (positive) and empty (negative) lobes.

In this work, we first show that the magnetic interaction between the Tb4+ impurity and its surrounding Ir4+ ions is given to a good approximation by

Hci=αiNN(Tb)S~γiJi,γieff,H_{\mathrm{ci}}=-\alpha\sum_{i\in\mathrm{NN(Tb)}}\tilde{S}_{\gamma_{i}}J^{\mathrm{eff}}_{i,\gamma_{i}}, (1)

in which the summation runs over the nearest neighboring Ir4+ sites (i=±x^,±y^i=\pm\hat{x},\pm\hat{y}) of the Tb site (see Fig. 1, left panel). S~γi\tilde{S}_{\gamma_{i}} is the spin operator of the Tb4+ ion (slightly rotated due to the lattice distortion as will be discussed later) and Ji,γieffJ^{\mathrm{eff}}_{i,\gamma_{i}} is the pseudospin operator of the Ir4+ ion at site ii. γi=x\gamma_{i}=x (yy) for i=±x^i=\pm\hat{x} (±y^\pm\hat{y}). The Ising-like interaction on each bond favors the magnetic moments aligning along the bond direction like the compass model Nussinov et al. (2004); Nussinov and Fradkin (2005); Douçot et al. (2005), so we call it the compass impurity model. As we shall show such highly anisotropic magnetic interaction is rooted in the spin-orbital coupled nature of the Ir4+ pseudospin-1/21/2 atomic levels. In previous studies such Ising-like magnetic interaction can arise only from edge-sharing IrO6 octahedra Jackeli and Khaliullin (2009). Our result opens a new route to strong exchange anisotropy in iridates with the corner-sharing IrO6 octahedron structure.

This realization leads us to propose that the (uniform) compass model, which has topologically protected double degeneracy and may serve in quantum computation as qubits protected against decoherence Douçot et al. (2005), can be realized in ordered double perovskites composed of spin-orbital-coupled d5d^{5} ions (Ir4+, Rh4+, Ru3+, etc.) and half-closed-shell f7f^{7} ions (Tb4+, Gd3+, Eu2+, etc.).

In an antiferromagnet with easy-plane anisotropy, e.g., Sr2IrO4, the “compass impurity” induces a distortion of the AF order parameter which decays as r2r^{-2} away from the impurity. In the dual Coulomb gas picture of the XY model the impurities induce quenched vortex quadrupoles. In the following we shall show the thermal vortices triggered by the impurity quadruple potential causes a low temperature reentrant transition to a disordered phase for arbitrarily small impurity concentration. Moreover, the Néel order is fully suppressed by only a few percent substitutions consistent with the experiment. Further experimental predictions shall also be discussed.

Compass impurity model.—In the dilute impurity limit we first consider a single Tb4+ ion embedded in the IrO2 as shown in the left panel of Fig. 1. The Tb4+ ion has electronic configuration 4f74f^{7}. Because the 4f4f orbitals are very localized the crystal field effects is negligible and all ff orbitals are nearly degenerate. The Hund’s rule coupling leads to a large spin S=7/2S=7/2 and an orbital singlet state on each Tb4+ ion.

The Tb-O-Ir bonds are slightly distorted due to the rotation of the IrO6 and the TbO6 octahedra around the zz axis. The rotation angle θi=±θ\theta_{i}=\pm\theta (θ11\theta\simeq 11^{\circ}) for iAi\in A and BB sublattices respectively. Because the electric field perpendicular to the bond (which spoils the spin conservation in the hopping process) and the wavefunction overlap are only affected to order sinθ\sin\theta, as a good approximation we consider the bond to be straight and perform symmetry analysis in the usual (orbital, spin) basis.

In the undistorted case the Tb-O-Ir bond along xx direction has reflection symmetries along yy and zz axes, which allows the following nearest neighbor hopping parameters to be non-zero:

t1=fy(z2x2)|Ht|dxyx^o,t2=fy3|Ht|dxyx^o;t1=fz(x2y2)|Ht|dzxx^o,t2=fz3|Ht|dzxx^o.\begin{split}t_{1}&={}_{o}\langle f_{y(z^{2}-x^{2})}|H_{t}|d_{xy}\rangle_{\hat{x}},\quad t_{2}={}_{o}\langle f_{y^{3}}|H_{t}|d_{xy}\rangle_{\hat{x}};\\ t^{\prime}_{1}&={}_{o}\langle f_{z(x^{2}-y^{2})}|H_{t}|d_{zx}\rangle_{\hat{x}},\quad t^{\prime}_{2}={}_{o}\langle f_{z^{3}}|H_{t}|d_{zx}\rangle_{\hat{x}}.\end{split} (2)

Here |dax^|d_{a}\rangle_{\hat{x}} and |fbo|f_{b}\rangle_{o} denote the Ir4+ orbitals at x^\hat{x} and Tb4+ orbitals at the origin respectively. We note that t3=fxyz|Ht|dyzx^ot_{3}={}_{o}\langle f_{xyz}|H_{t}|d_{yz}\rangle_{\hat{x}} is also allowed by symmetry, but it is much smaller than those in Eq. (2). The reasons are two fold: (1) the direct wavefunction overlap of the dyzd_{yz} and the fxyzf_{xyz} orbitals is much smaller due to the dyzd_{yz} orbital orientation, (2) all possible oxygen pp orbitals mediated hopping processes are prohibited. Therefore we neglect the t3t_{3} term in the rest of this work.

The 9090^{\circ} rotation around the xx axis is also an approximate symmetry of the Tb-O-Ir bond if the electrons are well localized on the Ir and Tb ions. It relates the hopping parameters in Eq. (2) such that t1t1t^{\prime}_{1}\simeq t_{1} and t2t2t^{\prime}_{2}\simeq t_{2}. This symmetry is well respected in the Ir-O-Ir bond of Sr2IrO4: the nearest-neighbor hopping parameters of the dxyd_{xy} and the dzxd_{zx} bands, which are also related by this rotation, are nearly equal: txy=0.36eVt_{xy}=0.36~\mathrm{eV} and tzx=0.37eVt_{zx}=0.37~\mathrm{eV} Watanabe et al. (2010). The Tb4+ 4f4f orbitals are more localized, so this approximate symmetry should also be respected. Therefore, the nearest neighbor hopping between the Tb4+ and the Ir4+ ions along the x^\hat{x} bond is described by

Ht,x^=σ(t1fy(z2x2),σcx^,xy,σ+t2fy3,σcx^,xy,σ+t1fz(x2y2),σcx^,zx,σ+t2fz3,σcx^,zx,σ)+H.c.\begin{split}H_{t,\hat{x}}=&\sum_{\sigma}\Big{(}t_{1}f_{y(z^{2}-x^{2}),\sigma}^{{\dagger}}c_{\hat{x},xy,\sigma}+t_{2}f_{y^{3},\sigma}^{{\dagger}}c_{\hat{x},xy,\sigma}\\ &+t_{1}f_{z(x^{2}-y^{2}),\sigma}^{{\dagger}}c_{\hat{x},zx,\sigma}+t_{2}f_{z^{3},\sigma}^{{\dagger}}c_{\hat{x},zx,\sigma}\Big{)}+\mathrm{H.c.}\end{split} (3)

Here σ\sigma is the spin component along the zz direction.

We then project Eq. (3) onto the Ir4+ pseudospin-1/21/2 atomic levels with the following replacement Wang and Senthil (2011): cj,xy,σiσ1/3eiθjσ/2dj,σc_{j,xy,\sigma}^{{\dagger}}\rightarrow-i\sigma\sqrt{1/3}e^{i\theta_{j}\sigma/2}d_{j,\sigma}^{{\dagger}} and cj,zx,σσ1/3eiθjσ/2dj,σc_{j,zx,\sigma}^{{\dagger}}\rightarrow\sigma\sqrt{1/3}e^{i\theta_{j}\sigma/2}d_{j,-\sigma}^{{\dagger}} and find

Ht,x^=13σ[(t1fy(z2x2),σ+t2fy3,σ)iσeiσθ/2dx^,σ+(t1fz(x2y2),σ+t2fz3,σ)σeiσθ/2dx^,σ]+H.c.\begin{split}H_{t,\hat{x}}=&\frac{1}{\sqrt{3}}\sum_{\sigma}\Big{[}\big{(}t_{1}f_{y(z^{2}-x^{2}),\sigma}^{{\dagger}}+t_{2}f_{y^{3},\sigma}^{{\dagger}}\big{)}i\sigma e^{i\sigma\theta/2}d_{\hat{x},\sigma}\\ &+\big{(}t_{1}f_{z(x^{2}-y^{2}),\sigma}^{{\dagger}}+t_{2}f_{z^{3},\sigma}^{{\dagger}}\big{)}\sigma e^{i\sigma\theta/2}d_{\hat{x},-\sigma}\Big{]}+\mathrm{H.c.}\end{split} (4)

In the second term the pseudospin is not conserved in the hopping process. The reason is that while the spin is conserved [see Eq. (3)] the spin of the dzxd_{zx} component in the pseudospin-1/21/2 states is antiparallel to the pseudospin due to the SOC. This leads to the anisotropic magnetic interaction in Eq. (1) as we shall see below.

Taking into account the onsite Coulomb repulsion on the Ir4+ and the Tb4+ ions we derive the second order perturbation Hamiltonian and project it onto the S=7/2S=7/2 subspace of the Tb4+ ion. The effective magnetic interaction on the x^\hat{x} bond is found to be Hx^=αS~xJx,x^effH_{\hat{x}}=-\alpha\tilde{S}_{x}J^{\mathrm{eff}}_{x,\hat{x}}, in which S~x=eiθSzSxeiθSz\tilde{S}_{x}=e^{i\theta S_{z}}S_{x}e^{-i\theta S_{z}} is the slightly rotated spin operator of the Tb4+ ion. The interaction strength α=421(Ud1+Uf1)(t12+t22)\alpha=\frac{4}{21}(U_{d}^{-1}+U_{f}^{-1})(t_{1}^{2}+t_{2}^{2}), in which UdU_{d} [UfU_{f}] is the energy difference between the (d6,f6)(d^{6},f^{6}) [(d4,f8)(d^{4},f^{8})] and the (d5,f7)(d^{5},f^{7}) electron configurations due to the onsite Coulomb repulsion. The magnetic interaction on the Tb-O-Ir bond along the yy direction is derived in the same fashion, Hy^=αS~yJy,y^effH_{\hat{y}}=-\alpha\tilde{S}_{y}J^{\mathrm{eff}}_{y,\hat{y}} with S~y=eiθSzSyeiθSz\tilde{S}_{y}=e^{i\theta S_{z}}S_{y}e^{-i\theta S_{z}}. Combining Hx^H_{\hat{x}} and Hy^H_{\hat{y}} gives the compass impurity model, Eq. (1). The Hund’s rule coupling on the Ir4+ gives an AF Heisenberg-type correction, which is smaller than the compass-type interaction by one order of magnitude.

If Ir4+ (or other ions with d5d^{5} electronic configuration and strong SOC) and Tb4+ (or other f7f^{7} ions) form ordered double perovskites with the chemical formula A4BBO8 (layered quasi-two dimensional structure) or A2BBO6 (three dimensional structure), in which B=d5\mathrm{B}=d^{5} ions and B=f7\mathrm{B}^{\prime}=f^{7} ions occupy different sublattices with unequal (pseudo)spin sizes in the 2D square or 3D cubic lattice, the effective magnetic interaction is given by the compass model,

Hc=αijSγij,iJγij,jeff,H_{\mathrm{c}}=-\alpha\sum_{\langle ij\rangle}S_{\gamma_{ij},i}J^{\mathrm{eff}}_{\gamma_{ij},j}, (5)

in which γij=x,y\gamma_{ij}=x,y (and zz in the 3D case) for the bond ij\langle ij\rangle along the x,yx,y (and zz) directions respectively.

Impurity-induced quadrupolar distortion.—The magnetic interaction of the pure Sr2IrO4 compound has an easy-plane anisotropy induced by the Hund’s rule coupling on the Ir4+ ions Jackeli and Khaliullin (2009); Watanabe et al. (2010), so the magnetic moments form inplane AF order at low temperature Kim et al. (2009); Dhital et al. (2013); Ye et al. (2013). With Tb substitution such an anisotropy is strengthened by the compass-type interaction. For example, the magnetic moments of Heisenberg model with a single compass impurity align in the xyxy plane in its classical ground state, which is obtained numerically. In experiments Wang et al. (2015), the uniform susceptibility along cc axis is larger than the inplane susceptibility at x=0.03x=0.03, which also shows that the inplane AF correlation is stronger. Therefore, we shall study the impact of the compass impurities on the antiferromagnetic XY model.

The compass impurity induces local frustration to the AF order. We numerically calculate the classical ground state of a single compass impurity embedded in the XY antiferromagnet described by

Hci-XY=ij(Jx,ieffJx,jeff+Jy,ieffJy,jeff)+Hci,H_{\text{ci-XY}}=\sum_{\langle ij\rangle^{\prime}}\big{(}J^{\mathrm{eff}}_{x,i}J^{\mathrm{eff}}_{x,j}+J^{\mathrm{eff}}_{y,i}J^{\mathrm{eff}}_{y,j}\big{)}+H_{\mathrm{ci}}, (6)

in which ij\langle ij\rangle^{\prime} excludes the bonds connected to the impurity site. The result is shown in Fig. 2, left panel, where the arrows indicate the direction of the staggered magnetic moments.

Refer to caption
Figure 2: Left: Staggered AF moment orientations in the classical ground state of Eq. (6) obtained by numerical minimization on a 17×1717\times 17-site lattice (only the central part is shown) for Jeff=1/2J^{\mathrm{eff}}=1/2, S=7/2S=7/2, and α=1\alpha=1. Vortices (+1+1) and antivortices (1-1) appear in the plaquettes adjacent to the impurity site in the center. Upper right: Long range decay of AF moment distortion δϕ(𝐫)\delta\phi(\mathbf{r}) away from the impurity site. Dashed curves are fitting with Eq. (8). Lower right: Fitted quadrupole strength AA versus compass impurity strength α\alpha.

Due to the ferromagnetic Ising-like interaction around the impurity, the AF moment on the impurity site lies antiparallel to the total AF moment of the system. The nearby AF moments are also distorted to gain the anisotropic interaction energy. This creates vortices and antivortices in the plaquettes adjacent to the impurity, i.e., the AF moment orientation changes by ±2π\pm 2\pi as one encircles the plaquette as shown in Fig. 2, left panel 111The vorticity in a single plaquette is counted by adding up the phase differences along the bonds ϕij=ϕiϕj\phi_{ij}=\phi_{i}-\phi_{j}, in which ϕij\phi_{ij} are restricted to the interval (π,π](-\pi,\pi] Jelić and Cugliandolo (2011).. Therefore, the compass impurity induces a vortex quadrupolar distortion of the AF order. The long range behavior of the AF moment orientation ϕ(𝐫)\phi(\mathbf{r}) is given by the solution of the following equation,

ϵijijϕ(𝐫)=AQijijδ(𝐫),\epsilon_{ij}\partial_{i}\partial_{j}\phi(\mathbf{r})=AQ_{ij}\partial_{i}\partial_{j}\delta(\mathbf{r}), (7)

in which QijQ_{ij} is a normalized (detQ=1\det Q=-1) traceless symmetric tensor indicating the orientation of the quadrupole moment and AA is the quadrupole strength.

The solution of Eq. (7) is ϕ(𝐫)=ϕ¯+δϕ(𝐫)\phi(\mathbf{r})=\bar{\phi}+\delta\phi(\mathbf{r}) with the distortion δϕ(𝐫)\delta\phi(\mathbf{r}) given by

δϕ(𝐫)=2AriQijϵjkrkr4,\delta\phi(\mathbf{r})=2A\frac{r_{i}Q_{ij}\epsilon_{jk}r_{k}}{r^{4}}, (8)

which fits the numerical results perfectly as shown in Fig. 2, upper right panel. Therefore, the impurity-induced distortion to the AF order extends nonlocally and decays as r2r^{-2}.

The quadrupole strength AA is obtained by fitting Eq. (8) for different compass impurity strength α\alpha (Fig. 2, lower right panel). It is of O(1)O(1) order and increases monotonically with α\alpha.

Suppression of AF order.—The classical XY model without impurities has the famous Kosterlitz-Thouless (KT) transition at finite temperature, which is driven by the unbinding of thermally activated vortex-antivortex pairs Kosterlitz and Thouless (1973). In the presence of quenched vortex dipoles a reentrant transition to a disordered phase occurs at low temperatures Rubinstein et al. (1983).

The compass impurities with quenched vortex quadrupoles turn out to have a similar impact on the AF order. In the presence of many quadrupolar impurities the continuum Hamiltonian is given by

Hϕ=ρsd2𝐫(12(iϕ(𝐫))2iϕ(𝐫)fi(𝐫)),H_{\phi}=\rho_{s}\int d^{2}\mathbf{r}\Big{(}\frac{1}{2}(\partial_{i}\phi(\mathbf{r}))^{2}-\partial_{i}\phi(\mathbf{r})f_{i}(\mathbf{r})\Big{)}, (9)

in which ρs\rho_{s} is the spin stiffness. The second term describes the interaction of the AF moment field with the quadrupolar impurity potential,

fi(𝐫)=AlQik(l)ϵkjjδ(𝐫𝐫lq).f_{i}(\mathbf{r})=A\sum_{l}Q_{ik}(l)\epsilon_{kj}\partial_{j}\delta(\mathbf{r}-\mathbf{r}^{\mathrm{q}}_{l}). (10)

In the dilute impurity limit, the quadrupole-quadrupole interaction decays as r4r^{-4} and can be neglected. Therefore, both the position 𝐫lq\mathbf{r}^{\mathrm{q}}_{l} and the orientation Qij(l)Q_{ij}(l) of each quadrupole are treated as quenched random variables without spatial correlation. Upon disorder average (d.a.) we have

[f(𝐫)]d.a.=0,[fi(𝐫)fi(𝐫)]d.a.=12xA2δii2δ(𝐫𝐫),\begin{split}&[f(\mathbf{r})]_{\mathrm{d.a.}}=0,\\ &[f_{i}(\mathbf{r})f_{i^{\prime}}(\mathbf{r}^{\prime})]_{\mathrm{d.a.}}=-\frac{1}{2}xA^{2}\delta_{ii^{\prime}}\partial^{2}\delta(\mathbf{r}-\mathbf{r}^{\prime}),\end{split} (11)

in which xx is the impurity concentration.

In the dual Coulomb gas picture vortices and antivortices are mapped to electric charges and the impurity vortex quadruples are mapped to electric quadrupoles. The Hamiltonian is given by

Hv=πρsllmlmllog|𝐫lv𝐫lva|+Eclml2+ρslmld2𝐫fi(𝐫)(𝐫lv𝐫)i(𝐫lv𝐫)2,\begin{split}H_{\mathrm{v}}=&-\pi\rho_{s}\sum_{l\neq l^{\prime}}m_{l}m_{l^{\prime}}\log\left|{\mathbf{r}^{\mathrm{v}}_{l}-\mathbf{r}^{\mathrm{v}}_{l^{\prime}}}\over a\right|+E_{c}\sum_{l}m_{l}^{2}\\ &+\rho_{s}\sum_{l}m_{l}\int d^{2}\mathbf{r}f_{i}(\mathbf{r})\frac{(\mathbf{r}^{\mathrm{v}}_{l}-\mathbf{r})_{i}}{(\mathbf{r}^{\mathrm{v}}_{l}-\mathbf{r})^{2}},\end{split} (12)

in which aa is the short-distance cutoff and mlm_{l} is the vorticity at 𝐫lv\mathbf{r}^{\mathrm{v}}_{l}. EcE_{c} is the vortex core energy. The last term is the Coulomb interaction between the vortices and the quenched quadrupolar impurities.

Following the standard KT renormalization group (RG) procedure José et al. (1977); Chaikin and Lubensky (1995) we define the reduced spin stiffness K=ρs/kBTK=\rho_{s}/k_{\mathrm{B}}T and find, from the dielectric function, the renormalized stiffness KRK_{\mathrm{R}} as

KR=K+π2K2ad2𝐫a2r2a2[m(𝐫)m(𝟎)T]d.a.,K_{\mathrm{R}}=K+\pi^{2}K^{2}\int_{a}^{\infty}\frac{d^{2}\mathbf{r}}{a^{2}}\frac{r^{2}}{a^{2}}[\langle m(\mathbf{r})m(\mathbf{0})\rangle_{T}]_{\mathrm{d.a.}}, (13)

in which m(𝐫)=lmlδ(𝐫𝐫lv)m(\mathbf{r})=\sum_{l}m_{l}\delta(\mathbf{r}-\mathbf{r}^{\mathrm{v}}_{l}) is the charge density of the Coulomb gas, and T\langle\cdot\rangle_{T} is the thermal average. Define the (thermal) vortex fugacity y=eEc/kBTy=e^{-E_{c}/k_{\mathrm{B}}T} we find, to O(y2)O(y^{2}),

m(𝐫)m(𝟎)T=2y2(r/a)2πKcoshI(𝐫),\langle m(\mathbf{r})m(\mathbf{0})\rangle_{T}=-2y^{2}(r/a)^{-2\pi K}\cosh I(\mathbf{r}), (14)

in which

I(𝐫)=2πKd2𝐫fi(𝐫)i(G(𝐫𝐫)G(𝐫)),I(\mathbf{r})=2\pi K\int d^{2}\mathbf{r}^{\prime}f_{i}(\mathbf{r}^{\prime})\partial^{\prime}_{i}(G(\mathbf{r}^{\prime}-\mathbf{r})-G(\mathbf{r}^{\prime})), (15)

where G(𝐫)=(1/2π)log(r/a)G(\mathbf{r})=(1/2\pi)\log(r/a).

The disorder average can be evaluated using the cumulant expansion Rubinstein et al. (1983),

[coshI(𝐫)]d.a.=e12[I(𝐫)2]d.a.=e2π2xA2K2δa,[\cosh I(\mathbf{r})]_{\mathrm{d.a.}}=e^{\frac{1}{2}[I(\mathbf{r})^{2}]_{\mathrm{d.a.}}}=e^{2\pi^{2}xA^{2}K^{2}\delta_{a}}, (16)

in which δa\delta_{a} is the short-range regularization constant for the δ\delta function, which comes from the core of the quenched quadruples.

From Eqs. (13)–(16) we find that the renormalized stiffness KRK_{\mathrm{R}} has exactly the same form as in the standard KT transition if yy is replaced by yxy_{x},

yx=yeπ2xA2K2δa.y_{x}=ye^{\pi^{2}xA^{2}K^{2}\delta_{a}}. (17)

The RG equations are given by José et al. (1977); Chaikin and Lubensky (1995)

ddlyx=(2πK)yx,ddlK1=4π3yx2.\begin{split}&{d\over dl}y_{x}=(2-\pi K)y_{x},\\ &{d\over dl}K^{-1}=4\pi^{3}y_{x}^{2}.\end{split} (18)

The RG flow in the yxy_{x}-K1K^{-1} parameter plane is shown in Fig. 3, left panel. The shaded region, in which the reduced stiffness KK flows to a non-zero value, is the KT phase with quasi-long range order (QLRO). The region outside is disordered because KK flows to zero. In the absence of impurities yxy_{x} reduces to the vortex fugacity yy. Its variation with the temperature is drawn as the dashed black curve – the pure system has QLRO at low temperature and becomes disordered at the KT transition, namely, the Néel temperature TN(0)T_{N}(0).

In the presence of quenched impurities the extra factor in yxy_{x} represents the nucleation of vortices near the quadrupoles. Approaching zero temperature it diverges faster than how yy vanishes so the system is disordered at zero temperature for any impurity concentration xx. The variation of yxy_{x} with temperature is illustrated as the dashed curves for different impurity concentrations. There is a critical concentration xcx_{c}. Below xcx_{c} the system shows QLRO at an intermediate temperature range and enters a reentrant disordered phase at low temperatures. Because this low-TT disordered regime is driven by the impurity potential we believe it can show the spin glass behavior seen in the experiment Wang et al. (2015); Fischer and Hertz (1991). Above xcx_{c} the intermediate ordered phase vanishes and the Néel temperature TNT_{N} abruptly drops to zero. This is schematically illustrated in the phase diagram (Fig. 3, right panel).

Refer to caption
Figure 3: Left: RG flow in the yxy_{x}-K1K^{-1} parameter plane. The dashed curves illustrate the variation of yxy_{x} with temperature for different impurity concentration xx. Right: Schematic phase diagram of the XY model with quenched quadrupolar impurities. The shaded regions in both panels indicate the quasi-long range order phase.

The critical concentration xcx_{c} is not universal. It depends on the vortex core energy EcE_{c}, the quadrupole strength AA and the inverse quadrupole core area δa\delta_{a}. If we take all these quantities to be of O(1)O(1) order, xcx_{c} is found to be only a few percent. For example, for Ec=2E_{c}=2 and A2δa=2A^{2}\delta_{a}=2 we find xc=0.027x_{c}=0.027, which is consistent with the quick suppression of the Néel temperature in the experiment Wang et al. (2015).

Summary.—To summarize, the magnetic interaction near Tb impurities in Sr2IrO4 is described by the planar compass impurity model. The strong in-plane anisotropy around the Tb site can be detected with nuclear magnetic resonance. The compass impurity induces a long range quadrupolar distortion to the antiferromagnetic order which drives a reentrant transition to a disordered phase at low temperature and quickly suppresses the Néel temperature. Motivated by this work we propose that the compass model can be realized in ordered double perovskites composed of spin-orbital-coupled d5d^{5} ions and half-closed-shell f7f^{7} ions.

Acknowledgements.
L.Z. is grateful to J. C. Wang for helpful discussions. This work was supported by the National Key Basic Research Program of China (Grant No. 2014CB920902) and the National Natural Science Foundation of China (Grant No. 11374018). DHL is supported by DOE Office of Basic Energy Sciences, Division of Materials Science, under Material Theory program, DE-AC02-05CH11231.

References