betaFeedback

WorksPhysics and Astronomy

Phase diagram of a distorted kagome antiferromagnet and application to Y-kapellasite

Abstract

We investigate the magnetism of a previously unexplored distorted spin-1/2 kagome model consisting of three symmetry-inequivalent nearest-neighbor antiferromagnetic Heisenberg couplings Jhexagon, J and J', and uncover a rich ground state phase diagram even at the classical level. Using analytical arguments and numerical techniques we identify a collinear Q = 0 magnetic phase, two unusual non-collinear coplanar Q = (1/3,1/3) phases and a classical spin liquid phase with a degenerate manifold of non-coplanar ground states, resembling the jammed spin liquid phase found in the context of a bond-disordered kagome antiferromagnet. We further show with density functional theory calculations that the recently synthesized Y-kapellasite Y3Cu9(OH)19Cl8 is a realization of this model and predict its ground state to lie in the region of Q = (1/3,1/3) order, which remains stable even after inclusion of quantum fluctuation effects within variational Monte Carlo and pseudofermion functional renormalization group. The presented model opens a new direction in the study of kagome antiferromagnets.

Introduction

The kagome lattice is arguably one of the most important two-dimensional (2D) lattices for the study of magnetic frustration. It is characterized by a complex phase diagram including magnetically ordered regimes and proposed quantum spin liquid phases [1], has rich magnetization dynamics [2], and supports some of the best studied quantum spin liquid candidates like herbertsmithite ZnCu3_3(OH)6_6Cl2_2 [3, 4, 5, 6]. In a more technical context, the study of the antiferromagnetic Heisenberg model on the kagome lattice has been a fertile ground for the development and benchmarking of theoretical methods. Notably, the competition between density matrix renormalization group (DMRG) [7, 8, 9], variational Monte Carlo (VMC) [10, 11], and tensor networks (TNs) [12] type methods, with the aim of resolving the nature of the spin liquids supported by the kagome lattice, has been a fervent area of research for many years. All these intense research activities have mainly focused on the ideal kagome structure. In contrast, distortions of this lattice have been studied much less, even though they are realized in some magnetic compounds, and their physical phenomenology may be even richer than for the standard kagome lattice. In some cases, like volborthite Cu3_3V2_2O7_7(OH)2⋅_2\cdot2H2_2O [13, 14], the distortion leads to a new 2D lattice which is still highly frustrated and possibly has a spin liquid ground state [15]. In Rb2_2Cu3_3SnF12_{12}, the deformed kagome lattice leads to a pinwheel valence bond solid [16]. Other kinds of distortions lower the rotational symmetry of the lattice and lead to kagome strips [17, 18]. Even the low temperature structure of herbertsmithite bears some signatures of distortion [19, 20, 21].

The focus of the present work lies on an unusual and previously unexplored distortion of the kagome lattice which is realized in the recently synthesized variant of herbertsmithite, namely Y3_3Cu9_9(OH)19_{19}Cl8_8. The distorted lattice structure consists of three symmetry-inequivalent nearest-neighbor kagome bonds forming a nine-site unit cell. Analyzing the corresponding Heisenberg model as a function of its two coupling ratios, using analytical arguments and numerical techniques, we find a surprisingly rich ground state phase diagram, even at the classical level. A first notable observation is that large parts of the phase diagram represent an unusual coplanar spin state with a commensurate magnetic wave vector Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3). This type of ordered state requires a 27 atom magnetic unit cell. Furthermore, in an extended regime around the standard undistorted kagome lattice, an even more complex classical spin liquid phase is identified, which cannot be characterized by any specific wave vector. It bears similarities with the well-known classical spin liquid on the undistorted kagome lattice in the sense that its low energy states follow from a set of spin constraints for each triangle [22]. In contrast, however, the ground state in the distorted case is found to be generally non-coplanar and, hence, resembles the jammed spin liquid investigated in Ref. [23]. After discussing in detail the general magnetic phenomena of this distorted kagome lattice, the second focus of this paper is on the specific case of Y3_3Cu9_9(OH)19_{19}Cl8_8 where this model is likely realized. This material was discovered in an attempt of electron doping the Cu 3d3d states in herbertsmithite with the aim of placing the Fermi level at the symmetry protected Dirac crossing of the Cu dd-bands in the kagome lattice [24, 25, 26]. In herbertsmithite-type copper hydroxy halides, however, the larger charge provided by Y3+Y^{3+} compared to Zn2+Zn^{2+} is always compensated by the incorporation of additional hydroxy or halide ions, preserving the antiferromagnetic insulator nature of the Cu2+Cu^{2+} layers. Even if charge doping remains elusive in these systems, the newly discovered by-product in form of Y3Cu9(OH)19Cl8Y_{3}Cu_{9}(OH)_{19}Cl_{8} appears to be of great interest in and of itself.

In Y3Cu9(OH)19Cl8Y_{3}Cu_{9}(OH)_{19}Cl_{8}, the Y3+Y^{3+} ions are placed in the center of the hexagon of the kagome lattice, making it a material which is structurally similar to kapellasite [27, 28, 29], haydeeite [30] or centennialite [31, 32]. We therefore name the system Y-kapellasite. Note that there is a closely related compound YCu3(OH)6Cl3YCu_{3}(OH)_{6}Cl_{3} with ideal kagome lattice but disordered Y positions [33]. The latter orders at TN=15T_N = 15 K in a Q⃗=0\vec{Q} = 0 structure with negative spin chirality [34, 35] which has been attributed to a strong Dzyaloshinskii-Moriya (DM) interaction [36]. In contrast, Y-kapellasite remains dynamical down to much lower temperatures than YCu3(OH)6Cl3YCu_{3}(OH)_{6}Cl_{3} [37]; it has a broad feature at T=2T = 2 K in the specific heat but muon spin resonance (μ\muSR) on powder samples seems to indicate the absence of any static magnetic order, although disorder effects may play a role. Recently, coexistence of magnetic order and persistent spin dynamics has been suggested for different samples of Y-kapellasite [38].

By extracting the Heisenberg Hamiltonian of Y-kapellasite using total energy mapping from density functional theory (DFT) calculations, we find that the three couplings on the symmetry-inequivalent nearest-neighbor kagome bonds dominate, with negligible longer range interactions. We may, hence, place Y-kapellasite in the region of Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) order in the classical ground state phase diagram obtained here. Investigating the corresponding spin-1/2 model within variational Monte Carlo (VMC) and pseudofermion functional renormalization group (PFFRG) we argue that quantum fluctuations are not sufficiently strong to suppress the long-range magnetic order. Accordingly, our semiclassical spin-wave analysis provides a realistic approximation of the system’s excitation spectrum which will be useful for comparison with future experimental data.

Results

Spin Hamiltonian. The model investigated in this work is a variant of the standard nearest-neighbor kagome Heisenberg model, but with three distinct nearest-neighbor couplings, which we call JJ, J⋈J_{\bowtie}, and J′J' [see Fig. 1a]. We will later argue that this model approximates well the microscopic interactions in Y-kapellasite. The Heisenberg Hamiltonian can be written as

H=∑⟨i,j⟩Jij S⃗i⋅S⃗j,(1)\mathcal{H} = \sum_{\langle i,j\rangle} J_{ij}\,\vec{S}_{i}\cdot\vec{S}_{j}, \tag*{(1)}

where S⃗i\vec{S}_{i} are the spin degrees of freedom (which, below, are either chosen as spin-1/2 operators or as classical normalized vectors) and JijJ_{ij} is given by JJ, J⋈J_{\bowtie} or J′J', depending on the bond. All these couplings are assumed to be positive (antiferromagnetic). It is clear that J=J⋈=J′J = J_{\bowtie} = J' leads back to the standard undistorted nearest-neighbor kagome model. As a consequence of the broken translational symmetry of the kagome lattice the system’s periodic structure is described by a decorated triangular lattice with a unit cell of nine sites [see Fig. 1a]. We can distinguish two inequivalent sets of sites inside the unit cell, which are not connected by point group symmetries and form two distinct sublattices: sublattice AA is made of the six sites connected by J⋈J_{\bowtie} [the vertices of the red hexagons of Figure 1a]; sublattice BB is made of the remaining three sites. Also note that the model is invariant under exchanging JJ and J′J' followed by a reflection with respect to the a⃗1\vec{a}_{1} axis.

Figure 1. Distorted kagome lattice and classical Q~=(1/3,1/3)\widetilde{\mathbf{Q}}=(1/3,1/3) magnetic order. a Schematic illustration of the three exchange couplings characterizing the effective Heisenberg Hamiltonian for Y-kapellasite (J◊J_{\lozenge}, JJ and J′J') shown in red, blue and green, respectively. The presence of three different couplings breaks the translational symmetry of the kagome lattice and leads to a decorated triangular lattice with an enlarged unit cell of 9 sites, here represented by the black hexagon (Wigner-Seitz cell). The Hamiltonian of the system is periodic under translations along the Bravais vectors a⃗1\vec{a}_1 and a⃗2\vec{a}_2 and the sites within the unit cell are divided into sublattices AA (hollow symbols) and BB (solid symbols). Due to the different values of the three exchange terms, the D6D_6 point group symmetry of the kagome lattice is broken down to C6C_6. b Pictorial view of the reciprocal space. The blue arrows represent the unit vectors of the reciprocal space (b⃗1\vec{b}_1 and b⃗2\vec{b}_2). The gray hexagons tiling the reciprocal space depict the first Brillouin zone of the lattice, while the black dashed hexagon delimits the so-called extended Brillouin zone. Some of the high symmetry points are marked with black dots. Finally, red lines represent the path along which the magnon dispersion is plotted in Figure 3. c Classical Q~=(1/3,1/3)\widetilde{\mathbf{Q}}=(1/3,1/3) magnetic order for J>J′J>J' (red region of Figure 2). The orientations of the spins are fully specified by the angle ϕ\phi between neighboring spins in the J◊J_{\lozenge}-hexagons, as outlined in the main text (here we take the value of ϕ\phi for the case J′=0J'=0 and J◊=JJ_{\lozenge}=J). In this figure, the spins are arranged in the xyxy-plane and their orientation is represented by the angle with respect to the SxS_x axis. The red, blue and green colors of the spins of sublattice AA help visualizing the Q~=(1/3,1/3)\widetilde{\mathbf{Q}}=(1/3,1/3) pattern. d The classical Q~=(1/3,1/3)\widetilde{\mathbf{Q}}=(1/3,1/3) magnetic order of Figure 1 c in the J≫J◊J\gg J_{\lozenge} limit (J′=0J'=0). The spins form antiferromagnetic trimers along the JJ-bonds (depicted in black). The trimers are arranged in an effective kagome lattice structure and their orientations, highlighted by the three different colors, follow the 3×3\sqrt{3}\times\sqrt{3} pattern [69].

Classical phase diagram. In Figure 2 we summarize the classical ground state phase diagram of the Heisenberg model on the distorted kagome lattice [Eq. (1)] as a function of the ratios J/J⋈J/J_{\bowtie} and J′/J⋈J'/J_{\bowtie}, which has been obtained by combining analytical arguments, iterative minimization and classical Monte Carlo calculations (see Methods). At the classical level, we observe (i) a collinear Q⃗=0\vec{Q} = 0 magnetic phase, (ii) two non-collinear coplanar magnetic phases, both labelled as Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) order, separated by (iii) a classical spin-liquid phase with a degenerate manifold of non-coplanar ground states, which in the context of a bond-disordered kagome antiferromagnet was dubbed a ”jammed spin liquid” phase [23].

Figure 2. Classical phase diagram of the distorted kagome model. We note that the phase diagram is symmetric under the exchange of the axes (i.e., J↔J′J \leftrightarrow J'), as a consequence of the symmetry of the Hamiltonian. The Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetic order of the red region is depicted in Figure 1c, and related to the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order of the blue region by a mirror reflection with respect to a⃗1\vec{a}_1. Inside the gray area, the system features a classical spin-liquid phase with degenerate non-coplanar ground states, as discussed in the main text. The Q⃗=0\vec{Q}=0 magnetic order can be viewed as parallel spins on the same sublattice, while on different sublattices the spins are anti-aligned. We note that the axes change between the four quadrants of this plot. The empty and filled squares indicate two possible sets of couplings for Y-kapellasite.

Even without any prior knowledge of the precise nature of the different phases, there exists a simple argument that determines the location of the phase boundaries. To this end, we employ the analytical procedure of Ref. [23] where a bond disordered Heisenberg model on the kagome lattice was studied. In the first step, we rewrite the Hamiltonian of Eq. (1) as

H=12∑△(L⃗△)2+const.(2)\mathcal{H} = \frac{1}{2}\sum_{\triangle}\left(\vec{L}_{\triangle}\right)^{2}+\mathrm{const.} \tag*{(2)}

where the sum runs over all triangles formed by nearest-neighbor bonds of the kagome lattice (both up and down triangles are considered). We define

L⃗△=JijJikJjk S⃗i+JijJjkJik S⃗j+JikJjkJij S⃗k,(3)\vec{L}_{\triangle} = \sqrt{\frac{J_{ij}J_{ik}}{J_{jk}}}\,\vec{S}_{i} +\sqrt{\frac{J_{ij}J_{jk}}{J_{ik}}}\,\vec{S}_{j} +\sqrt{\frac{J_{ik}J_{jk}}{J_{ij}}}\,\vec{S}_{k}, \tag*{(3)}

where ii, jj, k∈△k \in\triangle are the three sites forming a triangle. In our distorted model, all triangles of the lattice are formed by one J⋈J_{\bowtie}, one JJ, and one J′J' coupling [see Fig. 1a]. Thus, by an appropriate choice of the i,j,ki,j,k labels of Eq. (3), we can write

L⃗△=JJ′J⋈ S⃗i+JJ⋈J′ S⃗j+J′J⋈J S⃗k(4)\vec{L}_{\triangle} = \sqrt{\frac{JJ'}{J_{\bowtie}}}\,\vec{S}_{i} +\sqrt{\frac{JJ_{\bowtie}}{J'}}\,\vec{S}_{j} +\sqrt{\frac{J'J_{\bowtie}}{J}}\,\vec{S}_{k} \tag*{(4)}

for all triangles. From Eq. (2) it immediately follows that any spin configuration that fulfills the condition L⃗△=0\vec{L}_{\triangle} = 0 ∀△\forall\triangle is a ground state of the system. However, depending on the values of the couplings JJ, J⋈J_{\bowtie}, and J′J', it may occur that L⃗△=0\vec{L}_{\triangle} = 0 is impossible for any triangle when one term on the right hand side of Eq. (4) dominates so strongly that it cannot be compensated by the other two terms.

Restricting to an isolated triangle, it is easy to show that L⃗△=0\vec{L}_{\triangle}=0 can only be fulfilled if

JJ◊≤J′J◊−J′,J′≤min⁡(J,J◊),(5)\frac{J}{J_{\lozenge}}\leq\frac{J'}{J_{\lozenge}-J'},\qquad J'\leq\min(J,J_{\lozenge}), \tag*{(5)}
JJ◊≥J′J◊+J′,J≤min⁡(J′,J◊),(6)\frac{J}{J_{\lozenge}}\geq\frac{J'}{J_{\lozenge}+J'},\qquad J\leq\min(J',J_{\lozenge}), \tag*{(6)}
JJ◊≥J′J′−J◊,J◊≤min⁡(J,J′).(7)\frac{J}{J_{\lozenge}}\geq\frac{J'}{J'-J_{\lozenge}},\qquad J_{\lozenge}\leq\min(J,J'). \tag*{(7)}

These conditions define the phase boundaries in Figure 2. In the regions where an isolated triangle cannot satisfy L⃗△=0\vec{L}_{\triangle}=0, the system realizes one of the aforementioned coplanar phases [Q~=(1/3,1/3)\widetilde{\mathbf{Q}}=(1/3,1/3) order and Q~=0\widetilde{\mathbf{Q}}=0 order]. On the other hand, in the region where an isolated triangle can fulfill Eq. (16) we observe a classical spin-liquid phase. We note that analogous phase boundaries characterize the classical phase diagram of the square-kagome antiferromagnet [39].

Coplanar orders

We start our discussion of the classical ground states with the coplanar phases where L⃗△=0\vec{L}_{\triangle}=0 is necessarily violated. The rewritten Hamiltonian in Eq. (2) still implies that these phases form in a way that minimizes (L⃗△)2(\vec{L}_{\triangle})^2. To simplify the investigation, we first restrict ourselves to the case J′=0J'=0 where the Q~=(1/3,1/3)\widetilde{\mathbf{Q}}=(1/3,1/3) phase is realized. In the phase diagram of Figure 2, this corresponds to the leftmost vertical axis and it will turn out to provide a good approximation of the exchange couplings of Y-kapellasite determined by the abab initio DFT calculations (marked with squares in the figure).

In the limit J′=0J'=0, the model consists of a lattice of hexagons, made of sublattice AA sites, which are connected to each other through the JJ-trimers involving sublattice BB sites [Figure 1a]. Note that the middle spin of each trimer is fixed in the direction opposite to the sum of the edge spins. The magnetic order realized along the J′=0J'=0 line is depicted schematically in Figure 1c. The spins are coplanar and form a periodic configuration with momentum Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) (in units of the reciprocal lattice vectors b⃗1\vec{b}_1 and b⃗2\vec{b}_2). This momentum corresponds to the KK point of the Brillouin zone of the lattice (and the symmetry related points), see Figure 1b. Within a given unit cell, the spins of sublattice AA form an alternating pattern around the J□J_{\Box}-hexagons: the spins on even and odd sites are ferromagnetically aligned along two different directions, which are rotated with respect to each other by an angle ϕ\phi. The orientations of the spins on the remaining sites (i.e., sublattice BB), which are only two-coordinated, are uniquely determined by the value of the angle ϕ\phi. Thus, in the limit J′=0J'=0, we can express the classical energy per site of the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order as a simple function of ϕ\phi:

E/N=23[J□cos⁡(ϕ)+Jcos⁡(ϕ2+π3)](8)E/N=\frac{2}{3}\left[J_{\Box}\cos(\phi)+J\cos\left(\frac{\phi}{2}+\frac{\pi}{3}\right)\right] \tag*{(8)}

Minimization yields optimal angles ϕ\phi that go from ϕ=π\phi=\pi in the strong hexagon limit J≪J□J\ll J_{\Box} to ϕ=4π3\phi=\frac{4\pi}{3} in the trimer limit J≫J□J\gg J_{\Box} (see Figure 1d and Supplementary Note 2). Going away from the J′=0J'=0 limit, the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order extends for a finite region along the J′/J□J'/J_{\Box} axis (red area in Figure 2), which is bounded by the onset of the classical spin-liquid phase. Within this region, the numerical minimization of the classical energy shows that the spin pattern is unchanged with respect to the one shown in Figure 1c and the orientation of the spins is still determined only by the angle ϕ\phi between the spins of sublattice AA. As already mentioned, the phase diagram is invariant under the exchange of JJ and J′J', and a corresponding Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order is observed also in the proximity of the J=0J=0 limit (blue area in Figure 2). The two Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) ordered phases, one for J>J′J>J' and the other for J<J′J<J', are transformed into each other by a mirror reflection with respect to a⃗1\vec{a}_1. In the numerical calculations, the two phases can be distinguished by their spin susceptibility in momentum space, defined as

χk⃗=1N∑i,jeik⃗⋅(r⃗i−r⃗j)⟨S⃗i⋅S⃗j⟩(9)\chi_{\vec{k}}=\frac{1}{N}\sum_{i,j}e^{i\vec{k}\cdot(\vec{r}_i-\vec{r}_j)}\left\langle\vec{S}_i\cdot\vec{S}_j\right\rangle \tag*{(9)}

where r⃗i\vec{r}_i is the position of site ii in the kagome lattice, NN is the total number of sites and the brackets ⟨…⟩\langle\ldots\rangle denote an appropriate average. In the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) phase at J>J′J>J', χk⃗\chi_{\vec{k}} displays high intensity peaks at the K2′K'_2 points, while in the ordered phase at J<J′J<J', its maxima are located at the K3′K'_3 points [see Figure 1b]. In the limit where J□≪J,J′J_{\Box}\ll J,J' the system is no longer frustrated since each coupled neighbor of an AA site is a BB site and vice versa. The ground state order is, hence, given by a simple collinear Q⃗=0\vec{Q}=0 state where the two sublattices have opposite spin orientations. This regime is marked green in Figure 2.

As it represents a previously unexplored magnetic state, it is interesting to study the classical spin wave dispersion of the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order. In Figure 3, we show the spin wave spectra for the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetic order in three paradigmatic regimes (with J′=0J'=0 in each case): J≪J□J\ll J_{\Box} (strong hexagon limit) [Figure 3a], J=J□J=J_{\Box} [Figure 3b], and J≫J□J\gg J_{\Box} (strong trimer limit) [Figure 3c]. In all three cases the spin wave spectrum has gapless modes at Γ\Gamma and KK points. For J≪J□J\ll J_{\Box}, J=J□J=J_{\Box}, we observe a sizable gap between the low-lying magnon band (which gives rise to three branches when folded) and the higher bands. In the strong hexagon limit, where the system is made of weakly coupled hexagons forming a triangular pattern, the excitation spectrum at low energies resembles that of the triangular lattice antiferromagnet (see Supplementary Note 2 and Supplementary Fig. 3). The gap between the low-lying branches and the higher ones closes when the ratio J/J□J/J_{\Box} is sufficiently large, as shown by the J≫J□J\gg J_{\Box} case [Figure 3c]. In this limit (strong trimer limit), the system is described by trimers of spins forming an effective kagome lattice, and the spin wave spectrum resembles that of the kagome 3×3\sqrt{3}\times\sqrt{3} magnetic order (see Supplementary Note 2 and Supplementary Fig. 3).

Spin-wave dispersion and intensity results

Figure 3. Spin wave theory results. Calculated spin-wave dispersion within linear spin wave theory along the (ξ,ξ)(\xi,\xi)-direction [from Γ\Gamma to Γ′′\Gamma'', in Figure 1b] for three different cases: a J/J□=1/5J/J_{\square}=1/5, b J/J□=1J/J_{\square}=1, c J/J□=25J/J_{\square}=25, where we set J′=0J'=0 in each case. For all the choices of couplings considered here, the system is in the Q~=(1/3,1/3)\tilde{\mathbf{Q}}=(1/3,1/3) ordered phase. The energy scale is set by J□J_{\square} in panel (a), and by JJ in panel (b) and (c). We note that in the J≪J□J\ll J_{\square} (a) and J=J□J=J_{\square} (b) cases the lowest band (three-times folded) is separated from the higher bands by an energy gap. This gap closes upon increasing the ratio J/J□J/J_{\square}. d Calculated spin-wave dispersion and intensity within linear spin wave theory along the path Γ\Gamma-KK-K1′K'_1-Γ′′\Gamma''-K2′K'_2-K3′K'_3-Γ′′\Gamma'' [see Figure 1b] for the ab initio estimated Heisenberg couplings J=154.4J=154.4 K, J□=134.2J_{\square}=134.2 K and J′=8.7J'=8.7 K (SXRD structure).The spectral intensity is given by the perpendicular component of the spin dynamical structure factor, which is related to the cross section of unpolarized neutron scattering experiments [70] (the overall color scale is in arbitrary units). The low-energy spectral intensity at the K1′K'_1 (K3′K'_3) point is approximately 40% (20%) of the maximal intensity, located at the K2′K'_2 point. We apply a Gaussian broadening with a standard deviation of 0.2 meV.

Figure 4. Local ground state constraint within the classical spin liquid phase. a Value of ⟨∣L⃗Δ∣⟩\langle|\vec{L}_{\Delta}|\rangle (averaged over triangles) in the optimal classical ground state found by iterative minimization. The results are obtained in the classical spin liquid phase on a N=27N=27 sites cluster (see Supplementary Figure 1). Most of the points fulfill the L⃗Δ=0\vec{L}_{\Delta}=0 constraint for the ground state. The small number of points with ⟨∣L⃗Δ∣⟩>0\langle|\vec{L}_{\Delta}|\rangle>0 are found to have a coplanar non-degenerate ground state which corresponds to the nearby Q~=(1/3,1/3)\tilde{\mathbf{Q}}=(1/3,1/3) order. b Phase diagram from classical Monte Carlo simulations. We depict the average of ∣L⃗Δ∣|\vec{L}_{\Delta}| for 10 simulated systems with N=675N=675 sites (type-II cluster, L=5L=5, cfr. Supplementary Note 1) at the temperature T=0.001 Jmax⁡T=0.001\,J_{\max}, where Jmax⁡=max⁡(J□,J,J′)J_{\max}=\max(J_{\square},J,J'). The Q~=(1/3,1/3)\tilde{\mathbf{Q}}=(1/3,1/3) (Q~=0\tilde{\mathbf{Q}}=0) ordered phase, with maximal spin susceptibility at the K2′K'_2 or K3′K'_3 (Γ\Gamma) points in the Brillouin zone, lies outside the region defined by Eqs. (16) where L⃗Δ\vec{L}_{\Delta} can never be zero. Inside the region where L⃗Δ\vec{L}_{\Delta} can potentially vanish, we still find finite but small values. The maximum value in the plot, ⟨∣L⃗Δ∣⟩max⁡≃2.83\langle|\vec{L}_{\Delta}|\rangle_{\max}\simeq2.83, is found at the points of maximal distortion, i.e. J′(J)=10 J□=100 J(J′)J'(J)=10\,J_{\square}=100\,J(J'). The logarithmic color function scales as ln⁡(100⟨∣L⃗Δ∣⟩+1)/ln⁡(100⟨∣L⃗Δ∣⟩max⁡+1)\ln(100\langle|\vec{L}_{\Delta}|\rangle+1)/\ln(100\langle|\vec{L}_{\Delta}|\rangle_{\max}+1). The empty and filled squares indicate two possible sets of couplings for Y-kapellasite.

Classical spin liquid phase. Finally, we consider the regime where the three couplings generally enable the fulfillment of L⃗Δ=0\vec{L}_{\Delta}=0 in Eq. (4) and we unveil the ground state nature of this intriguing phase. Even though satisfying L⃗Δ=0\vec{L}_{\Delta}=0 in an isolated triangle is possible, this does not immediately imply that achieving L⃗Δ=0\vec{L}_{\Delta}=0 in each individual triangle of the full system is a trivial task. In Ref. [23] a generic bond disordered kagome system was investigated for which it was shown that L⃗Δ=0\vec{L}_{\Delta}=0 can be satisfied in each triangle. Furthermore, the authors constructed global ground states where each triangle may realize up to two possible spin configurations that locally obey L⃗Δ=0\vec{L}_{\Delta}=0 resulting in an extensively, but discretely degenerate classical spin liquid forming a set of the ground states with the cardinality ℵ0\aleph_{0}, which was named ”jammed spin liquid”.

We performed iterative minimization of the classical energy to confirm the presence of a degenerate manifold of non-coplanar ground states with L⃗Δ=0\vec{L}_{\Delta}=0 within the gray region of Figure 2 (see Methods). The main results of the minimization are summarized in Fig. 4a, where we plot the value of ⟨∣L⃗Δ∣⟩\langle|\vec{L}_{\Delta}|\rangle (averaged over triangles) as a function of the exchange couplings, for a finite-size cluster with N=27N=27 sites (see Supplementary Figure 1, and Supplementary Figure 3 for analogous results on a N=36N=36 site cluster). In the optimal solutions, ∣L⃗Δ∣|\vec{L}_{\Delta}| is actually found to be identical for all triangles. Its square value yields the residual energy per triangle with respect to the ideal ground state with L⃗Δ=0\vec{L}_{\Delta}=0. We observe that in most of the region delimited by the boundaries of Eq. (16), we obtain a degenerate set of ground states with ∣L⃗Δ∣2=0|\vec{L}_{\Delta}|^2=0 for each triangle (within machine precision). Indeed, starting the iterative minimization with different initial points, we find several independent minima which cannot be connected to each other by lattice symmetries and global spin rotations, thus confirming the large degeneracy of the classical ground state, as predicted in Ref. [23]. We also note that the L⃗Δ=0\vec{L}_{\Delta}=0 solutions found by the minimization are, in general, non-coplanar, and can be exploited to construct L⃗Δ=0\vec{L}_{\Delta}=0 ground states for larger systems, by simply using the N=27N=27 sites cluster as effective unit cells for type-II clusters (see Supplementary Note 1).

However, close to the boundaries of the non-coplanar region, we find a number of points in the phase diagram where achieving L⃗Δ=0\vec{L}_{\Delta}=0 by numerical minimization was not possible. We are not able to provide a final statement whether the ⟨∣L⃗Δ∣⟩>0\langle|\vec{L}_{\Delta}|\rangle>0 points close to the boundaries are an artifact of the finite-size calculations (and boundary conditions [23]), or whether they belong to the neighboring Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) ordered region, which may extend slightly beyond the ideal boundaries of Eq. (16). Indeed, it is worth noting that the best Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) solution at the analytical phase boundary with the spin-liquid phase [i.e., when the equality of Eq. (16) or Eq. (16) holds] always has a finite residual energy, i.e. ⟨∣L⃗Δ∣⟩>0\langle|\vec{L}_{\Delta}|\rangle>0 (except for the extreme cases where one coupling is zero). Thus, a continuous deformation of the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order to the ∣L⃗Δ∣=0|\vec{L}_{\Delta}|=0 spin-liquid phase cannot take place at the precise location of the analytical boundaries. Therefore, either the transition is not continuous, or the position of the phase boundaries is slightly shifted with respect to the analytical conditions of Eqs. (16) and (6). Nevertheless, except for the precise location of the transition, our numerical minimization confirms the presence of an extended classical spin-liquid phase, characterized by degenerate non-coplanar ground states.

In addition to energy minimization, we performed classical Monte Carlo simulations in the low temperature limit (see Methods), computing the value of ⟨∣L⃗Δ∣⟩\langle|\vec{L}_{\Delta}|\rangle in the full phase diagram, as shown in Fig. 4b. The Monte Carlo results confirm the presence of a region of non-coplanar ground states within the boundaries of Eq. (16). In this region the value of ⟨∣L⃗Δ∣⟩\langle|\vec{L}_{\Delta}|\rangle is found to be clearly smaller than in the rest of the phase diagram, where the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) and Q⃗=0\vec{Q}=0 orders are observed. The finite value of ⟨∣L⃗Δ⟩\langle|\vec{L}_{\Delta}\rangle within the non-coplanar region can be ascribed to the effect of finite temperature. To further characterize the properties of this phase, we compute the spin susceptibility [Eq. (9)]. As shown in Figure 5, the spin-spin correlations in the non-coplanar phase cannot be described by any particular wave vector, but rather by a distribution of wave vectors.

Image not extracted

Figure 5. Classical Monte Carlo results. a to e: Spin susceptibility in momentum space at different temperatures from classical Monte Carlo simulations for the SXRD_{\mathrm{XRD}} structure on a cluster with N=7803N = 7803 spins (type-II cluster with L=17L = 17, see Supplementary Note 1), with γ=0.001\gamma= 0.001 (see Methods for details). We only consider finite real-space correlations within a circle with a radius of 50 nearest-neighbor distances around each spin. The extended Brillouin zone is depicted as a black hexagon, cfr. Figure 1 b. f Spin susceptibility for J/J◊=0.5J/J_{\lozenge} = 0.5 and J′/J◊=0.45J'/J_{\lozenge} = 0.45, within the non-coplanar phase. The susceptibility has been computed by classical Monte Carlo calculations for ten N=7803N = 7803 sites clusters (type-II, L=17L = 17, see Supplementary Note 1) at T=0.001J◊T = 0.001 J_{\lozenge}.

Magnetic Hamiltonian of Y-kapellasite. We now concentrate on the specific case of Y-kapellasite and investigate the magnetic properties of its spin Hamiltonian in more detail, also including the effects of quantum fluctuations. We start by performing ab initio density functional theory calculations to confirm that Y-kapellasite supports the spin model of Figure 1a and determine the precise values of the coupling constants. We used both published crystal structures, the one determined by X-ray diffraction of single crystals [26], and the structure of Y3Cu9(OD)19Cl8\mathrm{Y}_{3}\mathrm{Cu}_{9}(\mathrm{OD})_{19}\mathrm{Cl}_{8} determined by neutron diffraction on powder samples [37]. We consider the former structure more reliable than the latter (we follow the privately communicated assessment of P. Puphahl that the single crystal samples of Ref. [26] are more pure and less strained than the deuterated powders of Ref. [37]), and we will refer to the single crystal structure [26] as SXRDS_{\mathrm{XRD}} structure and to the powder structure [37] as SNDS_{\mathrm{ND}} structure (see Methods for more details). Note that our analysis below is valid for both structures.

The three largest couplings JJ, J⋈J_{\bowtie} and J′J' (all antiferromagnetic) are shown in Figure 6b, as obtained by energy mapping for the SXRDS_{\mathrm{XRD}} structure of Y-kapellasite. The couplings are tabulated in Supplementary Table 1. These three couplings form the distorted kagome lattice illustrated in the inset of Figure 6b. Our reasoning that the relevant exchange couplings for Y-kapellasite are just the three nearest neighbours of the distorted kagome lattice is based on extensive energy mapping for seven and, for additional confidence, 24 neighbours up to Cr–Cr distances of 8.148.14 Å. The determination of 24 couplings for a larger supercell as listed in Supplementary Table 3 fully confirms the three largest couplings, shown as empty symbols in Figure 6b. It also shows that Y-kapellasite is a very two-dimensional material, and in the following we neglect all interlayer couplings. Among the three second and six third nearest neighbor couplings of the

Image not extracted

Figure 6. Structure and exchange couplings of Y3_3Cu9_9(OH)19_{19}Cl8_8. a Crystal structure of Y-kapellasite [26] (space group 148, R3ˉR\bar{3}) with DFT relaxed hydrogen positions. b Exchange couplings of Y3_3Cu9_9(OH)19_{19}Cl8_8 as function of interaction strength UU, determined by energy mapping using GGA+UU at JH=1J_H = 1 eV. Solid symbols: P1P1 cell, 7 couplings extracted. Empty symbols: 2×2×1\sqrt{2} \times\sqrt{2} \times1 supercell, 24 couplings extracted. The inset shows the nearest-neighbor exchange paths of the perfect kagome lattice which differentiate into J◊J_{\lozenge}, JJ and J′J' in Y-kapellasite.

distorted kagome lattice, the largest is J7J_{7} with a value 2% of JJ, which is rather small. This means that it is justified to focus the study of Y-kapellasite on the nearest-neighbor Hamiltonian. The couplings for the SNDS_{\mathrm{ND}} structure of Y-kapellasite are given in Supplementary Figure 8 and Supplementary Table 2, respectively. There is one clear difference between SXRDS_{\mathrm{XRD}} and SNDS_{\mathrm{ND}}: J◊J_{\lozenge} is 13% smaller than JJ for the SXRDS_{\mathrm{XRD}} structure but 3% larger for the SNDS_{\mathrm{ND}} structure. We will discuss the implications for the Hamiltonian in the next section. We emphasize that, according to the ab initio calculations above, the value of the exchange coupling J′J' is considerably smaller than those of J◊J_{\lozenge} and JJ, which are of comparable size. In conclusion, by calculating and inspecting a large number of exchange couplings, we verified that the JJ, J◊J_{\lozenge}, J′J' Hamiltonian is not a simplifying assumption but a defining feature of the material Y-kapellasite.

Exchange couplings of Y$_3$Cu$_9$(OH)$_{19}$Cl$_8$

Supplementary Figure 8. Exchange couplings of Y3_3Cu9_9(OH)19_{19}Cl8_8 (SNDS_{\mathrm{ND}} structure) extracted with GGA+U at JH=1J_H = 1 eV as function of interaction strength UU. Solid symbols: P1P1 cell, 7 couplings extracted. Empty symbols: 2×2×1\sqrt{2} \times\sqrt{2} \times1 supercell, 24 couplings extracted.

In what follows, we concentrate on the magnetic properties of Y-kapellasite as described by the structure SXRDS_{\mathrm{XRD}} and the coupling constants JJ, J◊J_{\lozenge}, J′J' of Supplementary Table 1 which place the material in the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) ordered regime of the classical phase diagram (see Figure 2). This placement puts our calculations in agreement with very recent experimental observation of (partial) magnetic order in Y-kapellasite [38].

Classical Monte Carlo simulations. We start our investigation of the Heisenberg Hamiltonian of Y-kapellasite (for the SXRDS_{\mathrm{XRD}} structure) using the classical Monte-Carlo technique. Despite neglecting quantum fluctuations, this analysis allows us to study how thermal fluctuations impact the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order (see Methods for technical details). In Figure 5 we present results on the spin susceptibility in momentum space [see Eq. (9)] for different temperatures. At high temperatures [Figure 5 e], the response is almost homogeneous along the edges of the extended Brillouin zone. This response resembles the one of the standard undistorted nearest-neighbor kagome model, indicating that at these high temperatures details of the precise detuning between JJ, J◊J_{\lozenge} and J′J' do not yet affect the susceptibility. When TT is lowered [going from panel e to panel a in Figure 5], additional features become discernible such as three maxima around each corner of the extended Brillouin zone (3×3\sqrt{3}\times\sqrt{3} positions). Each such triad forms an equilateral triangle and with decreasing temperature the peaks become sharper. Simultaneously, the triangles show a slight rotation around their center points (3×3\sqrt{3}\times\sqrt{3} positions) until in the low-temperature limit the peaks reach the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order positions [K2′K'_2 points in Figure 1 b]. This shift of peaks roughly occurs along a line connecting the K2′K'_2 and Γ′′\Gamma'' points. Please see Figure 8 c for a trace of the peak positions at different temperatures. Note that as a result of the Mermin-Wagner theorem, real long-range magnetic order is possible only at strictly T=0T=0. However, the fact that the short-range correlations in the intermediate temperature regime manifest themselves in susceptibility peaks at incommensurate wave vectors away from Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) indicates that thermal fluctuation act in a non-trivial and unexpected way. At T=0T=0, where the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) sets in, the maxima of the susceptibility are located at the K2′K'_2 points, with smaller peaks appearing at the K1′K'_1 and K3′K'_3 points. The relative heights of the latter peaks are χK1′/χK2′≈40%\chi_{K'_1}/\chi_{K'_2}\approx40\% and χK3′/χK2′≈20%\chi_{K'_3}/\chi_{K'_2}\approx20\%, respectively.

Figure 8. Pseudofermion functional renormalization group results. a One- and two-loop spin susceptibility flows from PFFRG for SXRDS_{\mathrm{XRD}} and SSNDS_{\mathrm{SND}} structures. The arrows mark a kink or peak during the flow indicating the onset of magnetic order. b Spin susceptibility in k⃗\vec{k} space [Eq. (7)] from two-loop S=1/2S=1/2 PFFRG for SXRDS_{\mathrm{XRD}} structure at the critical cutoff (marked by red arrow in a). The extended Brillouin zone is indicated by the black hexagon. Maxima of χk⃗Λcrit\chi_{\vec{k}}^{\Lambda_{\mathrm{crit}}} appear at incommensurate positions. c Momentum-space position of the maximal susceptibility at the breakdown of the PFFRG flow together with the corresponding values obtained from classical Monte Carlo at finite temperature TT. We show a section of the first Brillouin zone (gray lines) close to one edge of the extended Brillouin zone (black line), confer Figure 1b. The PFFRG results for the SXRDS_{\mathrm{XRD}} parameters are shown as triangles (the peak positions for the SSNDS_{\mathrm{SND}} structure are similar). Red and green triangles represent one-loop and two-loop results, respectively. Dots represent the susceptibility maxima from classical Monte Carlo for the system with SXRDS_{\mathrm{XRD}} couplings and the color represents temperature.

Variational Monte Carlo results. We now analyze the ground state properties of the spin model in the quantum regime with variational Monte Carlo (VMC) calculations. As in the previous section, we focus on the set of exchange couplings obtained for the SXRDS_{\mathrm{XRD}} structure of Y-kapellasite, which lies in the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) ordered region of the classical phase diagram (see Figure 2). Our variational method is based on Gutzwiller-projected fermionic states (see Methods and Supplementary Note 5 for details). Optimizing the variational state, we obtain a finite value of the magnetic field variational parameter, which indicates the resilience of the classical Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetic order against quantum fluctuations. To corroborate this finding, we compute the spin susceptibility, Eq. (9), with ⟨⋯ ⟩=⟨Ψ0∣⋯∣Ψ0⟩\langle\cdots\rangle=\langle\Psi_{0}|\cdots|\Psi_{0}\rangle representing the expectation value over the optimal variational wave function. The results for a finite cluster of N=972N=972 sites (type-II, L=6L=6, cfr. Supplementary Note 1) are shown in Figure 7: the susceptibility is clearly dominated by sharp Bragg peaks at the K2′K'_2 points of the extended Brillouin zone, thus confirming the presence of Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetic order. We note that χk⃗\chi_{\vec{k}} is not significantly different from the classical result at zero temperature (i.e., no specific features are detected except for the Bragg peaks), despite the important contribution of the fermionic hoppings and the Jastrow factor to the variational energy. An almost identical susceptibility is obtained when considering the exchange couplings of the SNDS_{\mathrm{ND}} structure. Thus, according to our VMC results, the minimal Heisenberg model for Y-kapellasite has a Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetically ordered ground state.

Spin susceptibility in momentum space for the optimal variational state

Figure 7. Variational Monte Carlo result. Spin susceptibility in momentum space [Eq. (9)] computed by VMC. The results refer to the optimal variational state for the Hamiltonian of Eq. (1) with the exchange couplings of the SXRDS_{\mathrm{XRD}} structure of Y-kapellasite (see Supplementary Table 1). The calculation is performed on a finite cluster of N=972N=972 sites (type-II, L=6L=6, cfr. Supplementary Note 1).

Pseudofermion functional renormalization group calculations. Next, we employ the pseudofermion functional renormalization group (PFFRG) approach [40, 41, 42, 43, 44] to investigate ground state quantum effects in our distorted kagome Heisenberg model from a complementary methodological perspective. Within PFFRG, we compute the static spin susceptibility in momentum space χk⃗Λ\chi_{\vec{k}}^{\Lambda} as a function of the RG parameter Λ\Lambda (which acts as a low-energy frequency cutoff). We employ two variants of this technique, the one-loop and two-loop schemes, where the latter can be considered more accurate (but computationally more demanding) as it includes additional diagrammatic contributions to better account for the system’s fluctuations beyond mean-field (see Methods for details). Most importantly, an onset of magnetic order can be observed as an instability during the RG flow of the maximal k⃗\vec{k}-space component of χk⃗Λ\chi_{\vec{k}}^{\Lambda}. Such an instability is indeed evident in the RG flows for both schemes (one-loop, two-loop) and for both structures of Y-kapellasite (see Figure 8 a) confirming the findings from VMC. However, the fact that these instability features are quite weak and only detectable as small kinks rather than sharp peaks indicates the significance of quantum fluctuations, possibly associated with a small ordered moment. Note that the instability is observed at much smaller Λ\Lambda in the two-loop scheme as compared to the one-loop approach, which is a known property resulting from the better fulfillment of the Mermin-Wagner theorem in the former method [41]. The momentum resolved susceptibility χk⃗Λcrit\chi_{\vec{k}}^{\Lambda_{\mathrm{crit}}} at the critical RG scale Λcrit\Lambda_{\mathrm{crit}} from two-loop PFFRG for the SXRDS_{\mathrm{XRD}} structure is shown in Figure 8 b. The maxima are rather broad, again indicating strong effects of quantum fluctuations. Furthermore, the peaks do not exactly coincide with the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) positions, as in VMC results, but show a small shift along the K2′−Γ′′K_2^{\prime}-\Gamma^{\prime\prime}-line, resembling our above findings from classical Monte Carlo. This indicates that quantum fluctuations may have similar effects as thermal fluctuations. We emphasize that this behavior is rather unusual, since, typically, quantum fluctuations tend to lock magnetic orders at commensurate positions. In Figure 8 c, we summarize the peak positions from one-loop and two-loop PFFRG as well as from classical Monte Carlo at intermediate temperatures. As can be seen, all results show a shift along similar momentum-space directions, however, the displacement away from the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) point becomes smaller as we advance the approach from one-loop to two-loop. It is hence conceivable that the shift would completely disappear upon further improving the method towards multi-loop schemes [45, 46]. We leave this as an open question for future investigations. We remark, however, that VMC and PFFRG both find magnetic long range order for Y-kapellasite.

Linear Spin Wave Theory. We conclude the analysis of the magnetic properties of Y-kapellasite by showing in Figure 3 d the spin wave spectrum and intensities for the Heisenberg Hamiltonian corresponding to the SXRDS_{\mathrm{XRD}} structure. We observe that the spectrum is very similar to the simpler case of J=J□J=J_{\square} and J′=0J^{\prime}=0 (Figure 3), which can then be regarded as a reliable minimal approximation for the full model. The intensity is largest at the K2′K_2^{\prime} point [see Figure 1 b] corresponding to the Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) order, as also observed in the spin susceptibility results above.

Discussion

Summarizing, by a combination of DFT, effective spin models, classical (iterative minimization, classical Monte Carlo) and quantum approaches (VMC, PFFRG) we investigated the magnetic properties of a distorted kagome lattice as realized in the recently synthesized Y-kapellasite. We found an unexpectedly rich phase diagram already at the classical level which includes a collinear Q⃗=0\vec{Q}=0 magnetic phase, two unusual non-collinear coplanar Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetic phases, and a classical spin liquid phase that resembles the jammed spin liquid phase found in the context of a bond-disordered kagome antiferromagnet. Our analysis of the spin model for Y-kapellasite places this system in the region of Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) magnetic order with an excitation spectrum that lies halfway between that of an underlying triangular lattice of hexagons and a kagome lattice of trimers. While it is not experimentally settled whether Y-kapellasite orders magnetically, our theoretical results provide strong evidence in favor of a magnetic Q⃗=(1/3,1/3)\vec{Q}=(1/3,1/3) ground state. The presence of an extended classical spin liquid phase in the vicinity of our DFT couplings sheds additional interesting light on this compound. Possibly, through external perturbations such as pressure or strain one might be able to shift the couplings towards the classical spin liquid phase, which due to the large extent of this regime, may not require any fine-tuning. This opens the question about the fate of the classical spin liquid upon adding quantum fluctuations, which we did not tackle in this work. Given the complexity of this phase already on the classical level one may expect even richer phenomena in the quantum case, including a quantum spin liquid. The numerical investigation of this regime in the quantum limit will certainly be a challenging future task but also gives hope for rewarding insights. In total, this work demonstrates that a relatively simple but realistic distortion of the kagome lattice gives rise to a multitude of interesting and unexpected magnetic phenomena whose full investigation goes far beyond the scope of the present work. In the future, our investigation may inspire and guide both a deeper experimentally motivated investigation of Y-kapellasite, as well as a closer numerical analysis of the underlying spin model.

Methods

Density functional theory based energy mapping. We calculate the electronic structure of Y-kapellasite with DFT using the full potential local orbital (FPLO) basis set [47] and the generalized gradient approximation (GGA) to the exchange correlation functional [48]. We apply the GGA+U approximation [49] to correct for strong electronic correlations of the Cu 3d3d electrons. We set the Hund’s rule coupling to a typical value value [50, 51] JH=1J_{\mathrm{H}}=1 eV for Cu2+^{2+} and vary only the onsite interaction UU. Even though the SSNDS_{\mathrm{SND}} structure [37] nominally has 8/9 filling, there is no evidence that Y-kapellasite is charge doped, and therefore we treat the O1 position as occupied with a hydroxy group (or a chloride ion which leads to the same results). In this position, the SXRDS_{\mathrm{XRD}} structure [26] has an orientationally disordered OH−^{-} ion between two Y3+^{3+} ions and therefore a 1/6 occupation of the six symmetry equivalent H positions is consistent with the Rˉ3\bar{R}3 space group. We model the orientationally disordered OH−^{-} ion using the virtual crystal approximation [52], setting the nuclear charge of H in this position to 1/6. The hydrogen positions H2 to H4 are relaxed within GGA in both structures. We shift the partially occupied H1 hydrogen position to the equilibrium O-H distance. The resulting structure is shown in Figure 6a.

We use total energy mapping [53, 54] to determine the Heisenberg Hamiltonian parameters of Y-kapellasite. For that we calculate with DFT(GGA+U) the total energy for 24 out of the 47 unique spin configurations which are possible with the 9 inequivalent Cu2+^{2+} ions in the P1P1 unit cell of Y3_3Cu9_9(OH)19_{19}Cl8_8. Considering that third-neighbor couplings are important for some kapellasite type compounds [50], we also perform calculations for a 2×2×1\sqrt{2} \times\sqrt{2} \times1 supercell with 18 independent Cu sites; we calculate 44 out of nearly 30000 spin configurations with distinct energies.

Iterative minimization. To numerically determine the classical ground state of the spin model of Eq. (1), we employ the iterative minimization method [55, 56]. We initialize our system in a random configuration and we iteratively perform local moves to update the spins. In each move, we pick up a random spin, S⃗i\vec{S}_i, and we orient it antiparallel to the local field created by the neighboring spins, i.e.

S⃗i↦−B⃗i∣B⃗i∣,with B⃗i=∑jJijS⃗j.(10)\vec{S}_i \mapsto-\frac{\vec{B}_i}{\lvert\vec{B}_i\rvert}, \qquad\text{with } \vec{B}_i=\sum_j J_{ij}\vec{S}_j. \tag*{(10)}

The procedure is repeated for a sufficient number of steps until the energy converges. In order to reduce the risk of ending up in local minima, we repeat the calculations several times starting from different spin configurations and we keep the solution with the best energy. The calculations are performed on the small finite-size clusters shown in Supplementary Fig. 1 with N=27N=27 and N=36N=36 sites, and periodic boundary conditions. It is important to emphasize that finding a classical ground state with L⃗Δ=0\vec{L}_{\Delta}=0 on one of these small clusters (with periodic boundary conditions) implies that one can immediately define a L⃗Δ=0\vec{L}_{\Delta}=0 ground state for any larger cluster of the same type (see Supplementary Note 1).

Classical Monte Carlo simulations. We perform a Monte Carlo analysis using the Metropolis algorithm with over-relaxation protocol for better thermal convergence [57, 58, 59, 60]. For the investigations of the SXRDS_{\mathrm{XRD}} structure, the system that we simulate is a cluster of N=7803N=7803 spins with periodic boundary conditions (type-II cluster with L=17L=17 (Supplementary Fig. 1). It is seeded at T0=2Jmax⁡T_0=2J_{\max} with random spins and cooled down via T=T0e−γnT=T_0e^{-\gamma n} where n=0,1,2,…n=0, 1, 2, \ldots is the number of Metropolis steps. During each step, every spin is updated once on average by a new random spin. The update takes place either with certainty if the acquired energy ΔE≤0\Delta E\leq0 or with a probability p=e−ΔE/Tp=e^{-\Delta E/T}. For the different coupling regimes, we set γ=0.001\gamma=0.001 and cooled 100 (10) random systems down to T=0.1Jmax⁡T=0.1J_{\max} (T=0.001Jmax⁡T=0.001J_{\max}), where Jmax⁡=max⁡(J□,J,J′)J_{\max}=\max(J_{\square},J,J'). For the investigation of the classical phase diagram we restrict ourselves to type-II clusters with L=5L = 5 (N=675N = 675 spins), for a numerical speedup.

Linear spin wave theory. We performed linear spin wave calculations with the SpinW package [61], computing the classical ground state by energy minimization.

Variational Monte Carlo. We employ Gutzwiller-projected fermionic states as variational ansätze [62]. This class of wave functions has been shown to provide an accurate description of the ground state of several spin models [63], including state-of-the-art results for kagome lattice antiferromagnets [10, 29, 64, 65, 66]. The optimal variational ansatz for the Hamiltonian of Y-kapellasite is a Gutzwiller-projected Jastrow-Slater wave function possessing Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) magnetic order (see Supplementary Note 5 for the definition). For the optimization of the variational parameters we use the stochastic reconfiguration method [67].

Pseudofermion Functional Renormalization Group. The PFFRG method is based on the one-loop plus Katanin truncation PFFRG scheme first introduced in Ref. [40] and extended to the two-loop plus Katanin variant in Ref. [41]. It utilizes the Abrikosov pseudofermion representation of spin operators. This spin representation enlarges the Hilbert space by adding two unphysical S=0S = 0 states per site (unoccupied, doubly occupied) which, however, leave the ground state properties largely unaffected [42]. Within PFFRG the bare propagator of the fermions is regularized by a sharp cutoff function:

G0(ω)=1iω⟶G0Λ(ω)=θ(∣ω∣−Λ)iω.(11)G_0(\omega) = \frac{1}{\mathrm{i}\omega} \quad\longrightarrow\quad G_0^\Lambda(\omega) = \frac{\theta(|\omega|-\Lambda)}{\mathrm{i}\omega}. \tag*{(11)}

Here, ω\omega is a continuous Matsubara frequency at T=0T = 0 and the cutoff Λ\Lambda prohibits fermionic propagation if ∣ω∣≤Λ|\omega| \leq\Lambda. This insertion causes a cutoff dependence of the generating functional for the fermionic one-particle-irreducible vertex functions. Flow equations which describe the Λ\Lambda derivatives of all nn-particle vertex functions can be derived. These equations couple the nn-particle vertex to all mm-particle vertices with m≤n+1m \leq n+1 leading to an infinite hierarchy of equations.

Linear spin wave theory

Variational Monte Carlo

Pseudofermion Functional Renormalization Group

In principle, physical results in the cutoff-free limit Λ→0\Lambda\to0 can be obtained by solving the integro-differential flow equations starting from the limit Λ→∞\Lambda\to\infty where the initial conditions are set by the bare interactions from our spin Hamiltonian. For numerical solvability this hierarchy of equations needs to be truncated. In the one-loop scheme, the truncation occurs on the level of the three-particle vertex which is replaced by contributions from the Katanin scheme [68], particularly, the single-scale propagator is upgraded to SΛ(ω)=−ddΛGΛ(ω)S^\Lambda(\omega) = -\frac{\mathrm{d}}{\mathrm{d}\Lambda}G^\Lambda(\omega) where the full Green’s function is GΛ(ω)=[(G0Λ(ω))−1−ΣΛ(ω)]−1G^\Lambda(\omega) = \left[\left(G_0^\Lambda(\omega)\right)^{-1} - \Sigma^\Lambda(\omega)\right]^{-1}. The one-loop flow equations for the self energy ΣΛ\Sigma^\Lambda and the two-particle vertex ΓΛ\Gamma^\Lambda are depicted diagrammatically in Supplementary Fig. 7. In the two-loop approach further contributions of the three-particle vertex are included, which have the form of nested one-loop diagrams [41]. We solve the flow equations numerically with an Euler scheme in real space taking into account finite spin correlations on hexagonal clusters with an edge length of N≥7N \geq7 nearest-neighbor distances around reference sites from each sublattice. The Matsubara frequencies are discretized using a linear plus logarithmic mesh with Mω≥60M_\omega\geq60 points. We carefully analyzed that the qualitative PFFRG results are converged with respect to the number of frequency points and the finite correlation length. From the resulting two-particle vertex, we are able to compute the Λ\Lambda dependent static spin susceptibility χkΛ\chi_{\mathbf{k}}^\Lambda in momentum space. For more details we refer the reader to Refs. [40, 41, 42].

Data Availability

The datasets generated and/or analysed during the current study are available from the corresponding authors upon reasonable request.

Code Availability

The calculation codes used in this paper are available from the corresponding authors upon reasonable request.

References

Acknowledgments

We thank Q. Barthélemy, F. Bert, T. Biesner, M. Dressel, E. Kermarrec, P. Mendels, P. Pupha, S. Roh for useful discussions. F.F. acknowledges support from the Alexander von Humboldt Foundation through a postdoctoral Humboldt fellowship. A.R., R.V. and J.R. acknowledge support by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) for funding through TRR 288 - 422213477 (projects A05, B05) (A.R. and R.V.) and CRC 183 (project A04) (J.R.). I.I.M. acknowledges support from the U.S. Department of Energy through Grant No. DE-SC0021089 and from the Wilhelm and Else Heraeus Foundation.

Author contributions

M.H. and F.F. contributed equally to this work. R.V. and H.O.J. conceived the project. I.I.M., R.V., H.O.J. and J.R. supervised the project. The analytical calculations were performed by F.F., A.R. and I.I.M. The iterative minimization was done by F.F., the classical Monte Carlo by M.H. The variational Monte Carlo was performed by F.F., the pseudofemion functional renormalization group calculations by M.H. Density functional theory calculations were performed by H.O.J. All authors contributed to the manuscript.

Competing interests

The authors declare no competing interests.

Additional information

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/xxx.

Phase diagram of a distorted kagome antiferromagnet and application to Y-kapellasite – Supplementary Information –

Max Hering; Francesco Ferrari; Aleksandar Razpopov; Igor I. Mazin; Roser Valentí; Harald O. Jeschke; Johannes Reuther

Finite-size clusters

Most of the numerical calculations of this work are performed on finite-size clusters with periodic boundary conditions. We consider two classes of fully symmetric finite-size clusters, type-I and type-II, which are defined by two different choices for the boundary conditions. Type-I clusters are constructed by imposing periodic boundary conditions along the vectors T⃗1=La⃗1\vec{T}_1 = L\vec{a}_1 and T⃗2=La⃗2\vec{T}_2 = L\vec{a}_2, for a total of N=9L2N = 9L^2 sites. Only the Type-I clusters in which LL is a multiple of 3 can accommodate the Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) magnetic order. For type-II clusters, instead, we take periodic boundary conditions along T⃗1=L(a⃗1−a⃗2)\vec{T}_1 = L(\vec{a}_1 - \vec{a}_2) and T⃗2=L(2a⃗1+a⃗2)\vec{T}_2 = L(2\vec{a}_1 + \vec{a}_2), which are parallel to the axes of the ideal (undistorted) kagome lattice. Type-II clusters contain N=27L2N = 27L^2 sites, and they can always accommodate the Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) order, regardless of the value of LL. In Supplementary Figure 1 we show the small finite-size clusters employed for the iterative minimization calculations, i.e. the L=2L = 2 type-I cluster with N=36N = 36 sites and the L=1L = 1 type-II cluster with N=27N = 27 sites.

Finite-size clusters

Supplementary Figure 1. Finite-size clusters used for iterative minimization, classical Monte Carlo, and variational Monte Carlo calculations.

More on the Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) ordered phase

In the limit J′=0J' = 0, the energy of the classical spin configuration is given by Eq. (9) of the main text. The optimal value of ϕ\phi is obtained by the minimization of this energy and its dependency on the ratio J/J□J/J_{\square} is shown in Supplementary Figure 2.

Optimal value of the angle $\phi$

Supplementary Figure 2. Optimal value of the angle ϕ\phi for the Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) magnetic order of Figure 1c of the main text as a function of the J/J□J/J_{\square} ratio (J′=0J' = 0).

Further insight is gained by looking at two extreme limits of the ratio J/J□J/J_{\square}. When J≪J□J \ll J_{\square} (strong hexagons limit), the angle ϕ→π\phi\to\pi, namely the spins of AA sites form an antiferromagnetic pattern around the J□J_{\square}-hexagons. On the other hand, in the opposite limit, J≫J□J \gg J_{\square} (trimer limit), the angle ϕ→4π/3\phi\to4\pi/3 and the spins of sublattice BB are antiferromagnetically aligned with respect to the two nearest-neighboring AA sites, thus forming antiferromagnetic trimers as shown in the main text, Figure 1d. Considering the trimers as single effective spins, we note that they form a kagome lattice and they interact through an effective antiferromagnetic coupling (each trimer is connected to four other trimers by a J◊J_{\lozenge} bond). In this limit, the Q⃗=(1/3,1/3)\vec{Q} = (1/3,1/3) magnetic order of our model reduces to the 3×3\sqrt{3} \times\sqrt{3} order [3] of the effective kagome lattice made by the antiferromagnetic trimers.

namedCu−Cud_{Cu-Cu} (Å)assignment
J1≡J′J_1 \equiv J'3.25310inplane 1nn
J2≡JJ_2 \equiv J3.36068inplane 1nn
J3≡J◊J_3 \equiv J_{\lozenge}3.37591inplane 1nn
J4J_45.68423interplane 1nn
J5J_55.68831inplane 2nn
J6J_65.76630inplane 2nn
J7J_75.77832inplane 2nn
UU (eV)J1J_1 (K)J2J_2 (K)
411.2(7)151.4(8)
4.4610.3(7)139.8(8)
4.510.3(6)138.8(7)
59.4(5)127.4(6)
5.58.5(4)117.1(5)
67.7(3)107.7(4)
6.56.8(3)99.1(3)
76.1(2)91.1(3)
7.55.3(2)83.8(2)
84.5(2)77.0(2)

Table 2. Supplementary Table 2. Exchange couplings of Y3_3Cu9_9(OH)19_{19}Cl8_8 (SND_{\mathrm{ND}} structure), calculated within GGA+U and 6×6×66 \times6 \times6 kk points. The experimental value TCW=−100T_{\mathrm{CW}} = -100 K of the Curie Weiss temperature is matched by the line in bold face. Statistical errors are indicated.

As discussed in the main text, the linear spin wave dispersion of the Q⃗=(1/3,1/3)\vec{Q} = (1/3,1/3) order in the strong hexagon (J≪J◊J \ll J_{\lozenge}, Figure 3a) and trimer (J≫J◊J \gg J_{\lozenge}, Figure 3c) limits share similarities with those of the triangular lattice 120∘120^\circ order [1] and of the kagome lattice 3×3\sqrt{3} \times\sqrt{3} order [3], respectively. In Supplementary Figure 3 we show the linear spin wave dispersions for the latter two cases, for comparison.

Linear spin wave dispersions for a 120° magnetic order on the triangular lattice and a √3 × √3 order on the kagome lattice

Supplementary Figure 3. Linear spin wave dispersions for a 120° magnetic order in the antiferromagnetic Heisenberg model on the triangular lattice [1] and a √3 × √3 order in the antiferromagnetic Heisenberg model on the kagome lattice [2]. We note that in the latter case the true ground state of the S=1/2S = 1/2 system is not magnetically ordered. A zero-energy dispersionless mode is present in the linear spin-wave spectrum of the √3 × √3 order on the kagome lattice [2]. In both panels the spin wave energies are plotted along the Γ−K−K1′−Γ\Gamma- K - K'_1 - \Gamma line (see Figure 1b of the main text).

Supplementary Note 3. More on the classical spin liquid phase

It is worth mentioning that the J=J′J = J' line represents a special case within this region, where coplanar ground states with L⃗△=0\vec{L}_{\triangle} = 0 can be analytically defined. Indeed, in this special limit, for each triangle we can write

L⃗△=JS⃗b+S⃗a+S⃗a′(12)\vec{L}_{\triangle} = J\vec{S}_{b} + \vec{S}_{a} + \vec{S}_{a'} \tag*{(12)}

where aa and a′a' are two neighboring sites of sublattice AA, bb is a site of sublattice BB and we have set J◊=1J_{\lozenge} = 1 for simplicity. Since this expression is valid for all triangles of the system, simple solutions with L⃗△=0\vec{L}_{\triangle} = 0, ∀△\forall\triangle can be found by focusing on a single unit cell. There, we can set, e.g., S⃗b=(1,0,0)\vec{S}_b = (1,0,0) for all sites of sublattice BB and we can choose an alternating pattern for the sites of sublattice AA forming the J∘J_{\circ}-hexagons, where we have S⃗a=(cos⁡(θ),sin⁡(θ),0)\vec{S}_a = (\cos(\theta),\sin(\theta),0) for odd sites and S⃗a′=(cos⁡(θ),−sin⁡(θ),0)\vec{S}_{a'} = (\cos(\theta),-\sin(\theta),0) for even sites, with θ=arccos⁡(−J/2)\theta= \arccos(-J/2). The same analysis can be repeated for each unit cell separately and coplanar ground states can be constructed, as discussed also in Ref. [4], where the J=J′J = J' model is investigated in connection to the physics of Volborthite. We note that this solution is connected to the Q⃗=0\vec{Q} = 0 coplanar order, since for J=J′=2J∘J = J' = 2J_{\circ} we get θ=π\theta= \pi, i.e. the spins inside the J∘J_{\circ}-hexagons are ferromagnetically arranged, and antiparallel to the spins of sublattice BB.

Value of the averaged scalar chirality in the optimal classical ground state

Figure 4 (PDF p. 16). Value of ⟨∣L⃗△∣⟩\langle|\vec{L}_{\triangle}|\rangle (averaged over triangles) in the optimal classical ground state found by iterative minimization. The results are obtained in the classical spin liquid phase, on the N=36N = 36 sites cluster sketched in Supplementary Figure 1 (type-I, L=2L = 2, see Supplementary Note 1 for details). Close to the inner boundaries of the spin-liquid phase, we find a number of points where the system is characterized by two distinct non-coplanar degenerate ground states with ⟨∣L⃗△∣⟩>0\langle|\vec{L}_{\triangle}|\rangle> 0. We note that the N=36N = 36 sites cluster cannot accommodate the Q⃗=(1/3,1/3)\vec{Q} = (1/3, 1/3) order.

Diagrammatic representation of the PFFRG flow equations

Supplementary Figure 7. Diagrammatic representation of the PFFRG flow equations for ΣΛ\Sigma^\Lambda (circle) and ΓΛ\Gamma^\Lambda (square). Slashed (bare) arrows depict the single-scale propagator SΛS^\Lambda (dressed propagator GΛG^\Lambda). Since the two-particle vertex couples to the three-particle vertex (hexagon), the flow equations are truncated at this level. External legs contain one full set of quantum numbers each and internal quantum numbers have to be summed up or integrated.

Supplementary Note 4. Analytical approach to explain classical Monte Carlo data

In this Supplementary Note we discuss the origin of the unusual peak shift in the classical Monte Carlo data for the SXRDS_{\mathrm{XRD}} structure as a function of temperature, as observed in Figure 5 of the main text. A possible explanation for this behavior comes from the system’s coupling matrix in momentum space

Jρσ(k⃗)=∑r⃗a−r⃗beik⃗(R⃗a−R⃗b)Jaρbσ,(13)J_{\rho\sigma}(\vec{k}) = \sum_{\vec{r}_a-\vec{r}_b} e^{i\vec{k}(\vec{R}_a-\vec{R}_b)} J_{a\rho b\sigma}, \tag*{(13)}

which forms the basis for various analytical techniques for studying classical spin systems. Here, the sites ii are characterized by the two indices aa and ρ\rho, specifying the unit cell and the sublattice of site ii, respectively. Similarly, j→b,σj \to b,\sigma and Jij→JaρbσJ_{ij} \to J_{a\rho b\sigma}, yielding the 9×99 \times9 matrix Jρσ(k⃗)J_{\rho\sigma}(\vec{k}). Furthermore, R⃗a\vec{R}_a, R⃗b\vec{R}_b specify the positions of the unit cells aa and bb. Diagonalizing this matrix for the couplings of the SXRDS_{\mathrm{XRD}} structure one finds that the third band (when counting from the lowest one) is in an energy range from E∼0.6JE \sim0.6J to E∼1.3JE \sim1.3J relative to the bottom of the lowest band. This intermediate energy range roughly corresponds to the temperature regime at which the peak shift in Monte Carlo is observed, i.e., the third band can be expected to be connected to these unusual thermal fluctuations. Indeed, as shown in Supplementary Figure 5 the third band exhibits a pronounced valley between the KK and Γ\Gamma points (dashed line in Supplementary Figure 5) such that fluctuations preferably occur along this direction. Note that the magnetic band structure is defined in the first Brillouin zone such that all points Γ′\Gamma', Γ′′\Gamma'' are folded back onto Γ\Gamma and, equivalently, K1′K'_1, K2′K'_2, K3′K'_3 all map onto KK.

Third band of the coupling matrix in the first Brillouin zone

Supplementary Figure 5. Third band of the coupling matrix [see Eq. (2)] depicted in the first Brillouin zone (gray hexagon). Energies EE are indicated relative to the bottom of the lowest band. The dashed line marks a valley in the band structure which may explain the system’s unusual thermal fluctuations.

Details on the variational Monte Carlo calculations

The variational Monte Carlo (VMC) calculations for the Hamiltonian of Y-kapellasite make use of Gutzwiller-projected fermionic wave functions as variational Ansätze. The definition of these states is based on the Abrikosov fermion representation of S=1/2S = 1/2 spins,

S⃗i=12∑α,βci,α†σ⃗α,βci,β.(14)\vec{S}_i = \frac{1}{2} \sum_{\alpha,\beta} c_{i,\alpha}^{\dagger} \vec{\sigma}_{\alpha,\beta} c_{i,\beta}. \tag*{(14)}

Here σ⃗=(σx,σy,σz)\vec{\sigma} = (\sigma_x,\sigma_y,\sigma_z) is a vector of Pauli matrices and ci,αc_{i,\alpha} (ci,α†c_{i,\alpha}^{\dagger}) are fermionic annihilation (creation) operators. Within this formalism, suitable variational states for spin systems can be constructed by applying the Gutzwiller projector, PG=∑i(ni,↑−ni,↓)2P_G = \sum_i (n_{i,\uparrow} - n_{i,\downarrow})^2, to a fermionic wave function, e.g. a Slater determinant [5]. The Gutzwiller operator enforces the single occupation of each lattice site and projects the fermionic wave function onto the spin Hilbert space. In this work, we consider Gutzwiller-projected Jastrow-Slater variational Ansätze of the form

∣Ψ0⟩=JPG∣Φ0⟩.(15)|\Psi_0\rangle= \mathcal{J} P_G |\Phi_0\rangle. \tag*{(15)}

Here, ∣Φ0⟩|\Phi_0\rangle is the ground state of an auxiliary quadratic Hamiltonian of Abrikosov fermions, named H0\mathcal{H}_0, which includes first-neighbor hopping terms (ti,jt_{i,j}) and a site-dependent magnetic field of strength hh:

H0=∑⟨i,j⟩∑σti,jci,σ†cj,σ+h.c.+h∑in^iQ⋅S⃗i.(16)\mathcal{H}_0 = \sum_{\langle i,j\rangle}\sum_{\sigma} t_{i,j} c_{i,\sigma}^{\dagger} c_{j,\sigma} + \mathrm{h.c.} + h \sum_i \hat{n}_i^{Q} \cdot\vec{S}_i. \tag*{(16)}

The local orientation of the magnetic field, given by the unit vector n^iQ\hat{n}_i^{Q}, is chosen such that it induces the optimal single-Q⃗\vec{Q} classical order in the (Sx,Sy)(S_x,S_y)-plane. Since we focus on the exchange couplings derived for the SXRDS_{\mathrm{XRD}} structure of Y-kapellasite, our magnetic field corresponds to the best classical order with Q⃗=(1/3,1/3)\vec{Q} = (1/3,1/3) periodicity. Non-trivial quantum fluctuations are introduced by the hopping terms ti,jt_{i,j} of the auxiliary Hamiltonian H0\mathcal{H}_0, and by the application of a long-range spin-spin Jastrow

J=exp⁡[∑i,jv(i,j)SizSjz].(17)\mathcal{J} = \exp\left[\sum_{i,j} v(i,j) S_i^z S_j^z\right]. \tag*{(17)}

on top of the Gutzwiller-projected fermionic state [see Eq. (4)]. The optimal parametrization of the hoppings, which yields the best variational energy for the exchange couplings of the SXRDS_{\mathrm{XRD}} structure, is illustrated in Supplementary Figure 6.

Optimal hopping pattern of the best variational Ansatz

Supplementary Figure 6. Optimal hopping pattern of the best variational Ansatz [Eq. (16)] for the effective spin model of Y-Kapellasite. The empty (full) dots represent the sites of sublattice AA (BB). The arrows encode the convention for complex hoppings: i→ji \to j represents the hopping terms ti,jci,σ†cj,σt_{i,j}c^\dagger_{i,\sigma}c_{j,\sigma} (σ=↑,↓\sigma= \uparrow,\downarrow). A total of five real parameters is used to parametrize the hopping part of the variational Ansatz (t∘t_{\circ}, tt, t′t', φ∘\varphi_{\circ}, φ\varphi).

Additional DFT results

In this Supplementary Note we present the full set of Heisenberg exchange coupling constants up to nthn^{\mathrm{th}} neighbors. These values were extracted from spin-polarized ab initio DFT calculations by making use of the total energy mapping method as explained in the main text. Supplementary Table 1 shows the results for the SXRDS_{\mathrm{XRD}} structure (see Figure 6 of the main text for the definition of J◊J_{\lozenge}, JJ and J′J') up to 8th8^{\mathrm{th}} neighbors, and Supplementary Table 2 shows the results for the SNDS_{\mathrm{ND}} structure up to 7th7^{\mathrm{th}} neighbors, where the definitions of J◊J_{\lozenge}, JJ and J′J' are indicated in Supplementary Figure 8. As explained in the method section of the main text, we perform the energy mapping method by calculating far more total energies than would be minimally required to extract the exchange interactions; for example, for extracting 7 exchange interactions, we calculate 24 energies instead of the minimal number 8. Fitting the Hamiltonian using singular value decomposition yields a statistical error for the exchange couplings which is indicated in Table 1 to Table 3. These errors are an indication of the uncertainty of the Hamiltonian parameters calculated for a particular crystal structure; however, the impact of the uncertainties in the crystal structure on the exchange couplings is not included.

namedCu−Cud_{Cu-Cu} (Å)assignment
J1≡J′J_1 \equiv J'3.24978inplane 1nn
J2≡J◊J_2 \equiv J_{\lozenge}3.36832inplane 1nn
J3≡JJ_3 \equiv J3.37619inplane 1nn
J4J_45.67876interlayer
J5J_55.70735inplane 2nn
J6J_65.76750inplane 2nn
J8J_85.86065interlayer
UU (eV)J1J_1 (K)J2J_2 (K)
411.3(9)161.1(6)
4.510.2(7)149.3(5)
59.1(6)138.7(4)
5.238.7(6)134.2(4)
5.58.2(5)129.1(3)
67.4(4)120.2(3)
6.56.6(3)112.0(2)
75.9(3)104.4(2)
7.55.2(2)97.2(2)
84.5(2)90.5(1)

Table 1. Supplementary Table 1. Exchange couplings of Y3_3Cu9_9(OH)19_{19}Cl8_8 (SXRD_{\mathrm{XRD}} structure), calculated within GGA+U and 6×6×66 \times6 \times6 kk points. The experimental value TCW=−100T_{\mathrm{CW}} = -100 K of the Curie Weiss temperature is matched by the line in bold face. Statistical errors are indicated.

namedCu−Cud_{\mathrm{Cu-Cu}} (Å)JiJ_i (K)assignment
J1J_13.249788.4(9)inplane 1nn Cu1-Cu2
J2J_23.36832133.1(6)inplane 1nn Cu1-Cu1
J3J_33.37619152.0(9)inplane 1nn Cu1-Cu2
J4J_45.678760.3(8)interlayer
J5J_55.707351.0(5)inplane 2nn Cu1-Cu1
J6J_65.767500.6(6)inplane 2nn Cu2-Cu2
J7J_75.830403.0(5)inplane 2nn Cu1-Cu1
J8J_85.860650.0(6)interlayer
J9J_96.49956-0.6(5)inplane 3nn (2J12J_1, Cu1-Cu1)
J10J_{10}6.51314-0.5(6)interlayer
J11J_{11}6.57249-0.5(5)interlayer
J12J_{12}6.61523-1.7(7)inplane 3nn (J1+J2J_1+J_2, Cu1-Cu2)
J13J_{13}6.62488-0.3(7)inplane 3nn (J1,J3J_1,J_3 hex., Cu1-Cu2)
J14J_{14}6.63443-1.6(9)interlayer
J15J_{15}6.664440.2(7)interlayer
J16J_{16}6.69872-0.2(5)interlayer
J17J_{17}6.726990.4(6)interlayer
J18J_{18}6.733431.2(5)inplane 3nn (J2J_2 hex., Cu1-Cu1)
J19J_{19}6.740650.2(8)inplane 3nn (J2+J3J_2+J_3, Cu1-Cu2)
J20J_{20}6.75238-0.1(6)inplane 3nn (2J32J_3, Cu1-Cu1)
J21J_{21}8.046030.7(8)interlayer
J22J_{22}8.090023.2(5)interlayer
J23J_{23}8.097910.4(7)interlayer
J24J_{24}8.141621.0(7)interlayer

Table 3. Exchange couplings of Y3_3Cu9_9(OH)19_{19}Cl8_8 (SXRDS_{\mathrm{XRD}} structure), calculated within GGA+U and 4×4×44 \times4 \times4 kk points. The couplings are reported at U=5.28U = 5.28 eV; this UU value is obtained by demanding that the model matches the experimental Curie-Weiss temperature. Statistical errors are indicated.

In order to check the presence of possible long-ranged coupling constants, we also performed calculations for larger supercells of the SXRDS_{\mathrm{XRD}} structure and extracted values of the exchange coupling constants up to 24th24^{\mathrm{th}} neighbors, as shown in Table 3. This calculation is computationally very demanding and cannot be repeated for all structures and values of UU. However, it is an important sanity check because it is the nature of the energy mapping method that a calculation with a small cell yielding few exchange interactions effectively provides information on each coupling with some average of unresolved longer range exchange couplings added to it. By resolving more couplings as shown in Supplementary Table 3, we have verified that there are no important long range exchange couplings that could have introduced errors into the couplings shown in Supplementary Tables 1 and 2 or that would significantly modify the magnetic properties. Comparing the coupling J5J_5 at U=5.23 eVU = 5.23\,\mathrm{eV} obtained in the smaller cell (J5=3.2(4) KJ_5 = 3.2(4)\,\mathrm{K}) to that obtained in the large supercell (J5=1.0(5) KJ_5 = 1.0(5)\,\mathrm{K}), we find that the larger value 3.2 K3.2\,\mathrm{K} is not substantiated, and it is reasonable to focus all further analysis on the first three exchange couplings J◊J_{\lozenge}, JJ and J′J'.

Comparison of VMC and PFFRG results

While VMC and PFFRG both find magnetic long range order for Y-kapellasite, the small peak shift and the rather broad susceptibility profile in PFFRG are not seen in VMC, which detects sharp Bragg peaks at the ordering vectors (compare Figure 7 and Figure 8 b of the main text). Given that the two techniques are conceptually very different and rely on different approximations, such discrepancies are not surprising. For example, detecting incommensurate magnetic order within VMC is a hard task, because the calculations are restricted to finite-size clusters and thus the vectors of magnetic order are discretized. It is interesting to note that taking a pure hopping Ansatz within VMC, i.e. discarding the magnetic field variational parameter (cfr. Supplementary Note 5), we obtain a broad susceptibility profile with shifted peaks, which is very similar to the one provided by PFFRG. However, the corresponding variational state has a considerably higher energy. Possibly, this energy could be reduced if one were able to implement a magnetic field variational parameter for an incommensurate structure, while at the same time maintaining broad susceptibility peaks and thereby reaching better agreement with PFFRG. On the other hand, it is possible that PFFRG and VMC results will resemble each other more closely when promoting the PFFRG towards the aforementioned multi-loop scheme.

References

  1. [1]Bieri, S., Lhuillier, C. & Messio, L. Projective symmetry group classification of chiral spin liquids. Phys. Rev. B 93, 094437 (2016).
  2. [2]Nishimoto, S., Shibata, N. & Hotta, C. Controlling frustrated liquids and solids with an applied field in a kagome Heisenberg antiferromagnet. Nat. Commun. 4, 2287 (2013).DOI
  3. [3]Mendels, P. & Bert, F. Quantum kagome antiferromagnet ZnCu₃(OH)₆Cl₂. J. Phys. Soc. Jpn. 79, 011001 (2010).
  4. [4]Han, T.-H. et al. Fractionalized excitations in the spin-liquid state of a kagome-lattice antiferromagnet. Nature 492, 406–410 (2012).DOI
  5. [5]Norman, M. R. Colloquium: Herbertsmithite and the search for the quantum spin liquid. Rev. Mod. Phys. 88, 041002 (2016).DOI
  6. [6]Khuntia, P. et al. Gapless ground state in the archetypal quantum kagome antiferromagnet ZnCu₃(OH)₆Cl₂. Nat. Phys. 16, 469–474 (2020).DOI
  7. [7]Depenbrock, S., McCulloch, I. P. & Schollwöck, U. Nature of the spin-liquid ground state of the S = 1/2 Heisenberg model on the kagome lattice. Phys. Rev. Lett. 109, 067201 (2012).
  8. [8]Jiang, H.-C., Wang, Z. & Balents, L. Identifying topological order by entanglement entropy. Nat. Phys. 8, 902–905 (2012).
  9. [9]He, Y.-C., Zaletel, M. P., Oshikawa, M. & Pollmann, F. Signatures of Dirac cones in a DMRG study of the kagome Heisenberg model. Phys. Rev. X 7, 031020 (2017).DOI
  10. [10]Iqbal, Y., Becca, F., Sorella, S. & Poilblanc, D. Gapless spin-liquid phase in the kagome spin-1/2 Heisenberg antiferromagnet. Phys. Rev. B 87, 060405 (2013).
  11. [11]Iqbal, Y., Poilblanc, D. & Becca, F. Spin-1/2 Heisenberg J₁–J₂ antiferromagnet on the kagome lattice. Phys. Rev. B 91, 020402 (2015).
  12. [12]Liao, H. J. et al. Gapless spin-liquid ground state in the S = 1/2 kagome antiferromagnet. Phys. Rev. Lett. 118, 137202 (2017).
  13. [13]Yoshida, H. et al. Orbital switching in a frustrated magnet. Nat. Commun. 3, 860 (2012).DOI
  14. [14]Janson, O. et al. Magnetic behavior of volborthite Cu3V2O7(OH)2·2H2O determined by coupled trimers rather than frustrated chains. Phys. Rev. Lett. 117, 037206 (2016).DOI
  15. [15]Watanabe, D. et al. Emergence of nontrivial magnetic excitations in a spin-liquid state of kagomé volborthite. Proc. Natl. Acad. Sci. 113, 8653–8657 (2016).DOI
  16. [16]Matan, K. et al. Pinwheel valence-bond solid and triplet excitations in the two-dimensional deformed kagome lattice. Nat. Phys. 6, 865–869 (2010).DOI
  17. [17]Goto, M. et al. Various disordered ground states and magnetization-plateau-like behavior in the S = 1/2 Ti3+ kagome lattice antiferromagnets Rb2NaTi3F12, Cs2Na3Ti3F12, and Cs2KTi3F12. Phys. Rev. B 94, 104432 (2016).
  18. [18]Jeschke, H. O., Nakano, H. & Sakai, T. From kagome strip to kagome lattice: Realizations of frustrated S = 1/2 antiferromagnets in Ti(III) fluorides. Phys. Rev. B 99, 140410 (2019).
  19. [19]Zorko, A. et al. Symmetry reduction in the quantum kagome antiferromagnet herbertsmithite. Phys. Rev. Lett. 118, 017202 (2017).DOI
  20. [20]Laurita, N. J. et al. Evidence for a parity broken monoclinic ground state in the S = 1/2 kagomé antiferromagnet herbertsmithite. Preprint at https://arxiv.org/abs/1910.13606 (2019).
  21. [21]Li, Y. et al. Lattice dynamics in the spin-1/2 frustrated kagome compound herbertsmithite. Phys. Rev. B 101, 161115 (2020).
  22. [22]Moessner, R. & Chalker, J. T. Properties of a classical spin liquid: The Heisenberg pyrochlore antiferromagnet. Phys. Rev. Lett. 80, 2929–2932 (1998).
  23. [23]Bilitewski, T., Zhitomirsky, M. E. & Moessner, R. Jammed spin liquid in the bond-disordered kagome antiferromagnet. Phys. Rev. Lett. 119, 247201 (2017).DOI
  24. [24]Mazin, I. I. et al. Theoretical prediction of a strongly correlated Dirac metal. Nat. Commun. 5, 4261 (2014).DOI
  25. [25]Güterding, D., Jeschke, H. O. & Valenti, R. Prospect of quantum anomalous Hall and quantum spin Hall effect in doped kagome lattice mott insulators. Sci. Rep. 6, 25988 (2016).DOI
  26. [26]Puphal, P. et al. Strong magnetic frustration in Y3Cu9(OH)19Cl8: a distorted kagome antiferromagnet. J. Mater. Chem. C 5, 2629–2635 (2017).
  27. [27]Fäk, B. et al. Kapellasite: A kagome quantum spin liquid with competing interactions. Phys. Rev. Lett. 109, 037208 (2012).
  28. [28]Bieri, S., Messio, L., Bernu, B. & Lhuillier, C. Gapless chiral spin liquid in a kagome Heisenberg model. Phys. Rev. B 92, 060407 (2015).
  29. [29]Iqbal, Y. et al. Paramagnetism in the kagome compounds (Zn, Mg, Cd)Cu3(OH)6Cl2. Phys. Rev. B 92, 220404 (2015).DOI
  30. [30]Boldrin, D. et al. Haydeeite: A spin-1/2 kagome ferromagnet. Phys. Rev. B 91, 220408 (2015).
  31. [31]Doki, H. et al. Spin thermal Hall conductivity of a kagome antiferromagnet. Phys. Rev. Lett. 121, 097203 (2018).DOI
  32. [32]Iida, K. et al. q = 0 long-range magnetic order in centennialite CaCu3(OD)6Cl2·0.6D2O: A spin-1/2 perfect kagome antiferromagnet with J1 = J2 = Jd. Phys. Rev. B 101, 220408 (2020).
  33. [33]Sun, W., Huang, Y.-X., Nokhrin, S., Pan, Y. & Mi, J.-X. Perfect kagomé lattices in YCu3(OH)6Cl3: a new candidate for the quantum spin liquid state. J. Mater. Chem. C 4, 8772–8777 (2016).DOI
  34. [34]Zorko, A. et al. Coexistence of magnetic order and persistent spin dynamics in a quantum kagome antiferromagnet with no intersite mixing. Phys. Rev. B 99, 214441 (2019).DOI
  35. [35]Zorko, A. et al. Negative-vector-chirality 120° spin structure in the defect- and distortion-free quantum kagome antiferromagnet YCu3(OH)6Cl3. Phys. Rev. B 100, 144420 (2019).
  36. [36]Arh, T. et al. Origin of magnetic ordering in a structurally perfect quantum kagome antiferromagnet. Phys. Rev. Lett. 125, 027203 (2020).DOI
  37. [37]Barthélemy, Q. et al. Local study of the insulating quantum kagome antiferromagnets YCu3(OH)6OxCl3−x (x = 0, 1/3). Phys. Rev. Mater. 3, 074401 (2019).DOI
  38. [38]Sun, W. et al. Magnetic ordering of the distorted kagome antiferromagnet Y3Cu9(OH)18[Cl8(oh)] prepared via optimal synthesis. Phys. Rev. Materials 5, 064401 (2021).DOI
  39. [39]Morita, K. & Tohyama, T. Magnetic phase diagrams and magnetization plateaus of the spin-1/2 antiferromagnetic Heisenberg model on a square-kagome lattice with three nonequivalent exchange interactions. J. Phys. Soc. Jpn. 87, 043704 (2018).DOI
  40. [40]Reuther, J. & Wölfle, P. J1-J2 frustrated two-dimensional Heisenberg model: Random phase approximation and functional renormalization group. Phys. Rev. B 81, 144410 (2010).DOI
  41. [41]Rück, M. & Reuther, J. Effects of two-loop contributions in the pseudorefermion functional renormalization group method for quantum spin systems. Phys. Rev. B 97, 144404 (2018).DOI
  42. [42]Baez, M. L. & Reuther, J. Numerical treatment of spin systems with unrestricted spin length S: A functional renormalization group study. Phys. Rev. B 96, 045141 (2017).DOI
  43. [43]Büessen, F. L., Roscher, D., Diehl, S. & Trebst, S. Functional renormalization group approach to SU(N) Heisenberg models: Real-space renormalization group at arbitrary N. Phys. Rev. B 97, 064415 (2018).
  44. [44]Büssen, F. L., Noculak, V., Trebst, S. & Reuther, J. Functional renormalization group for frustrated magnets with nondiagonal spin interactions. Phys. Rev. B 100, 125164 (2019).DOI
  45. [45]Thoenniss, J., Ritter, M. K., Kugler, F. B., von Delft, J. & Punk, M. Multiloop pseudofermion functional renormalization for quantum spin systems: Application to the spin-1/2 kagome Heisenberg model. Preprint at https://arxiv.org/abs/2011.01268 (2020).
  46. [46]Kiese, D., Müller, T., Iqbal, Y., Thomale, R. & Trebst, S. Multiloop functional renormalization group approach to quantum spin systems. Preprint at https://arxiv.org/abs/2011.01269 (2020).
  47. [47]Köpernik, K. & Eschrig, H. Full-potential nonorthogonal local-orbital minimum-basis band-structure scheme. Phys. Rev. B 59, 1743–1757 (1999).DOI
  48. [48]Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
  49. [49]Liechtenstein, A. I., Anisimov, V. I. & Zaanen, J. Density-functional theory and spin interactions: Orbital ordering in Mott-Hubbard insulators. Phys. Rev. B 52, R5467–R5470 (1995).
  50. [50]Jeschke, H. O., Salvat-Pujol, F. & Valentí, R. First-principles determination of Heisenberg Hamiltonian parameters for the spin-1/2 kagome antiferromagnet ZnCu₃(OH)₆Cl₂. Phys. Rev. B 88, 075106 (2013).
  51. [51]Jeschke, H. O. et al. Barlowite as a canted antiferromagnet: Theory and experiment. Phys. Rev. B 92, 094417 (2015).DOI
  52. [52]Nordheim, L. Zur Elektronentheorie der Metalle. I. Ann. Phys. 401, 607–640 (1931).DOI
  53. [53]Gudering, D., Valentí, R. & Jeschke, H. O. Reduction of magnetic interlayer coupling in barlowite through isoelectronic substitution. Phys. Rev. B 94, 125136 (2016).DOI
  54. [54]Iqbal, Y. et al. Signatures of a gearwheel quantum spin liquid in a spin-1/2 pyrochlore molybdate Heisenberg antiferromagnet. Phys. Rev. Mater. 1, 071201 (2017).DOI
  55. [55]Walker, L. R. & Walstedt, R. E. Computer model of metallic spin-glasses. Phys. Rev. B 22, 3816–3842 (1980).
  56. [56]Sklan, S. R. & Henley, C. L. Nonplanar ground states of frustrated antiferromagnets on an octahedral lattice. Phys. Rev. B 88, 024407 (2013).DOI
  57. [57]Creutz, M. Overrelaxation and Monte Carlo simulation. Phys. Rev. D 36, 515–519 (1987).
  58. [58]Kanki, K., Loison, D. & Schotte, K. D. Efficiency of the microcanonical over-relaxation algorithm for vector spins analyzing first and second order transitions. Eur. Phys. J. B 44, 309–315 (2005).
  59. [59]Zhitomirsky, M. E. Octupolar ordering of classical kagome antiferromagnets in two and three dimensions. Phys. Rev. B 78, 094423 (2008).DOI
  60. [60]Pixley, J. H. & Young, A. P. Large-scale Monte Carlo simulations of the three-dimensional XY spin glass. Phys. Rev. B 78, 014419 (2008).DOI
  61. [61]Toth, S. & Lake, B. Linear spin wave theory for single-Q incommensurate magnetic structures. J. Phys.: Condens. Matter 27, 166002 (2015).DOI
  62. [62]Becca, F. & Sorella, S. Quantum Monte Carlo Approaches for Correlated Systems (Cambridge University Press, 2017).
  63. [63]Becca, F., Capriotti, L., Parola, A. & Sorella, S. Variational Wave Functions for Frustrated Magnetic Models, Models, in Springer Series in Solid-State Sciences 164, edited by C. Lacroix, P. Mendels, and F. Mila, 379–406 (Springer Berlin Heidelberg, 2011).
  64. [64]Iqbal, Y., Poilblanc, D., Thomale, R. & Becca, F. Persistence of the gapless spin liquid in the breathing kagome Heisenberg antiferromagnet. Phys. Rev. B 97, 115127 (2018).DOI
  65. [65]Zhang, C. & Li, T. Variational study of the ground state and spin dynamics of the spin-1/2 kagome antiferromagnetic Heisenberg model and its implication for herbertsmithite ZnCu₃(OH)₆Cl₂. Phys. Rev. B 102, 195106 (2020).
  66. [66]Ferrari, F., Parola, A. & Becca, F. Gapless spin liquids in disguise. Phys. Rev. B 103, 195140 (2021).
  67. [67]Sorella, S. Green function Monte Carlo with stochastic reconfiguration. Phys. Rev. Lett. 80, 4558–4561 (1998).
  68. [68]Katanin, A. A. Fulfillment of Ward identities in the functional renormalization group approach. Phys. Rev. B 70, 115109 (2004).
  69. [69]Reimers, J. N. & Berlinsky, A. J. Order by disorder in the classical Heisenberg kagomé antiferromagnet. Phys. Rev. B 48, 9539–9554 (1993).
  70. [70]Bramwell, S. T. Neutron Scattering and Highly Frustrated Magnetism, 45–78 (Springer Berlin Heidelberg, Berlin, Heidelberg, 2011).

Paper details

vlnc:W_mgqyq2fc
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