2021 Volume 62 Issue 3 Pages 367-373
Recent discovery of various magnetism in Tsai-type quasicrystal approximants, in whose alloys rare-earth ions located on icosahedral vertices are coupled with each other via the Ruderman-Kittel-Kasuya-Yosida interaction, indicates an avenue to find novel magnetism originating from the icosahedral symmetry. Here we investigate classical and quantum magnetic states on an icosahedral cluster within the Heisenberg interactions of all bonds. Simulated annealing and numerical diagonalization are performed to obtain the classical and quantum ground states. We obtain qualitative correspondence of classical and quantum phase diagrams. Our study gives a good starting point to understand the various magnetism in not only quasicrystal approximants but also quasicrystals.
Long-range magnetic orders in quasi-periodic lattice have been a fascinating and challenging target since the discovery of quasicrystals.1) First investigation of magnetism in quasicrystals was performed in Al–Mn based alloy in 1986.2–4) Next, Bergman-type quasicrystals were examined thanks to the discovery of Zn–Mg–RE quasicrystals5,6) (RE = rare earth). However, no magnetic long-range ordering has been observed so far.7,8) On the other hand, another type of quasicrystals with containing rare-earth elements was discovered by A. P. Tsai in 2000.9–12) To attain the target, at present, Tsai-type quasicrystals are intensively investigated13) due to a key observation of antiferromagnetism in an approximant Cd6Tb,14–16) which have the same local structure as the quasicrystals but have a periodicity.
The Tsai-type quasicrystal approximants in common with the quasicrystals, consist of rhombic triacontahedral clusters. The cluster includes a concentric tetrahedron, a dodecahedron, an icosahedron, and an icosidodecahedron from center out, whose vertices constituent ions are located on.17,18) Among polyhedra, the rare-earth ions placed on the icosahedron only contributes to magnetism helped by so-called the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction,19–21) resulting in novel magnetic orders, e.g., multifarious magnetism discovered in Tsai-type 1/1 approximants.22–24) Interestingly, the magnetism is controlled by constitutional ratio of ions in ternary alloys of the approximants via electron density, because the RKKY interaction depends on the Fermi wavenumber which is a function of the electron density in Fermi gas approximation. Actually, the Curie-Weiss temperature observed in the 1/1 approximants shows an oscillation as a function of estimated electron density indicating the RKKY interaction.22,23)
Surprisingly, the latest experimental study on a Tsai-type 2/1 approximants has reported almost the same behaviors of magnetism as the 1/1 approximants despite difference of crystal structure between 1/1 and 2/1 approximants.25,26) This result suggests importance of the common local structure, i.e., the rhombic triacontahedral cluster including the magnetic icosahedron. The numerical calculation of the magnetic ground state and physical properties in a single icosahedral cluster both in the classical and quantum Heisenberg model is already performed with nearest-neighbor exchange interaction.27–30) However, these does not consider the interaction between the 2nd and 3rd neighbor interactions. Therefore, we investigate magnetism of an isolated icosahedron within the long-ranged exchange interactions. Especially, we focus our examination on the magnetic ground states to understand low-temperature physics appearing in the Tsai-type approximants. Since magnitude of magnetic moment depends on rare-earth ions, we consider both quantum and classical spins corresponding to small and large magnetic moments, respectively.
In this paper, we examine magnetic ground states obtained with the following model Hamiltonian,
\begin{equation} \mathcal{H} = J_{1}\sum_{\langle i,j\rangle_{1}}\boldsymbol{S}_{i} \cdot \boldsymbol{S}_{j} + J_{2}\sum_{\langle i,j\rangle_{2}} \boldsymbol{S}_{i} \cdot \boldsymbol{S}_{j} + J_{3}\sum_{\langle i,j\rangle_{3}}\boldsymbol{S}_{i} \cdot \boldsymbol{S}_{j}, \end{equation} | (1) |
(a) Schematic icosahedral spin cluster and its connectivity of interactions. Balls (circles) denote vertices assigning spins. The red, green, and blue bonds correspond to 1st, 2nd, and 3rd neighbor interactions. An ellipse represents a pair of spins located on opposite vertices of icosahedron. (b), (c) Connectivity of 1st or 2nd neighbor interaction. These graphs are equivalent.
To obtain the ground state, we numerically apply simulated annealing method to the classical model and exact diagonalization method to the quantum model. The simulated annealing is a Monte-Carlo method where vector spins are updated one by one with a certain probability based on the statistical mechanics at each temperature. The temperature is gradually lowered to zero like the annealing process in heat treatment of real materials. On the other hand, the exact diagonalization is genuinely a quantum method at zero temperature to take quantum fluctuations into account. In this method, a matrix form of the Hamiltonian represented in the basis of spin wavefunctions is exactly diagonalized to obtain the eigenstate of the minimal energy. We have confirmed accordance of ground-state energies obtained by the numerical methods and analytical energies at exactly solvable points in the parameter space, e.g., (θJ, ϕJ) = (0, 0), (π, 0), ($\tan ^{ - 1}\sqrt{2} $, π/4), and so on.
In this section, we first show numerical results in the classical model with |Si| = 1. In this case, the total energy can be described as
\begin{align} E_{\mathrm{nc}}& = J_{1}N_{1}\langle\cos\alpha_{ij}\rangle_{\langle i,j\rangle_{1}} + J_{2}N_{2}\langle\cos\alpha_{ij}\rangle_{\langle i,j\rangle_{2}} \\ &\quad + J_{3}N_{3}\langle\cos\alpha_{ij}\rangle_{\langle i,j\rangle_{3}}, \end{align} | (2) |
Figure 2(a)–(c) shows the $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{n}}$ in the parameter space (θJ, ϕJ) obtained by the simulated annealing. The mean values $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{1}}$ and $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{2}}$ take only four values −0.44, −0.20, 0.44 and 1.00 in Fig. 2(a)–(b). In consideration of these mean values, the spin configuration can be classified principally into four ground-state phases. Besides, $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{3}} = \pm 1$ in Fig. 2(c) indicates only two configurations of spins on the opposite sites of icosahedral vertices, corresponding to parallel and anti-parallel spins.
Numerical results of the classical model: (a) $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{1}}$, (b) $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{2}}$ and (c) $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{3}}$. In the parameter space, we take 41 × 201 sample points for calculation. See Fig. 3 for the acronyms, HA, DHA, PPA, and F.
Figure 3(a)–(e) displays the five spin configurations in an icosahedral cluster obtained by the numerical calculation. Characteristics of these configurations are listed below. Note that the energy is invariant with respect to any global O(3) rotation.
\begin{equation} \langle \cos\alpha_{ij}\rangle_{\langle i,j\rangle_{1}} = -\langle\cos\alpha_{ij}\rangle_{\langle i,j\rangle_{2}} = \frac{\tau}{\tau + 2}=\frac{1}{\sqrt{5}} = 0.447.., \end{equation} | (3) |
(a)–(d) Examples of ground-state spin configuration in hedgehog antiferromagnetic (HA), dual hedgehog antiferromagnetic (DHA), ferromagnetic (F), parallel pairs’ antiferromagnetic (PPA) phase. (e) Spin configuration of antiparallel pairs’ antiferromagnetic (APA) state appearing at the boundary of the HA and DHA phases.
In the Ferromagnetic (F) phase, $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{n}} = 1$ for all n implies all spins are parallel as shown in Fig. 3(c). This numerical calculation does not reproduce the spin configuration of ferromagnetism determined by neutron diffraction in Au–Si–Tb 1/1 approximants.32) This can be an evidence that the anisotropy caused by total angular momentum of Tb ion dominates the low-temperature magnetism,33) though we do not consider the anisotropy in this study.
In the parallel pairs’ antiferromagnetic (PPA) phase, the spin configuration changes one by one in many trials of calculation, while two spins on opposite vertices of icosahedron, which are connected with a 3rd neighbor bond, are always parallel [see Fig. 3(d)]. Therefore, we call this PPA phase. Undoubtedly, Fig. 2(a)–(b) displays that the same mean values $\langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{1}} = \langle \cos \alpha _{ij}\rangle _{\langle i,j\rangle _{2}} = - 0.20$, implying no distinct orders. In fact, if a certain spin’s angles to 1st and 2nd neighboring spins are completely random values and the mean values are the same, the mean value reads,
\begin{align} \langle\cos\alpha_{ij}\rangle_{\langle i,j\rangle_{1}} &= \langle\cos\alpha_{ij}\rangle_{\langle i,j\rangle_{2}} \\ &= 120^{-1}\left[\left(\sum_{i = 0}^{11}\boldsymbol{S}_{i}\right)^{2} {}- \sum_{i = 0}^{11} \boldsymbol{S}_{i}^{2} - 2\sum_{\langle i,j\rangle_{3}} \boldsymbol{S}_{i} \cdot \boldsymbol{S}_{j}\right] \\ &= -\frac{1}{5}. \end{align} | (4) |
We show total energy Enc as a function of (θJ, ϕJ) in Fig. 4(a), which is obtained numerically. To evaluate the Enc, this result is compared with the energy function Eest, which is estimated by the spin configurations in four phases: 30 (τ/(τ + 2))(J1 − J2) − 6J3 for the HA phase, 30 (τ/(τ + 2))(J2 − J1) − 6J3 for the DHA phase, 30 (J1 + J2) + 6J3 for the F phase, and −6(J1 + J2) + 6J3 for the PPA phase. Note that the APA state appears only at J1 = J2, so that this is merged into the HA and DHA phases. Energy difference Enc − Eest displayed in Fig. 4(b) gives a good coincidence between Enc and Eest except for phase boundaries, where numerical accuracy is not enough. Furthermore, the phase boundaries are apparently obtained by the derivative of total energy with respect to ϕJ in Fig. 4(c).
(a) Numerical result of the total energy Enc. (b) Energy difference between Enc and Eest which is analytically estimated by the ground-state spin configurations in Fig. 3. (c) The derivative of Enc with respect to ϕJ.
To explain the ground-state phase transitions in the classical model, we consider two specific conditions, (I) J1 = J2 and (II) J1 > 0 (J2 > 0) with J2 = J3 = 0 (J1 = J3 = 0). Condition (I) corresponds to ϕJ = π/4 and 5π/4 lines. Condition (II) is (θJ, ϕJ) = (π/2, 0) or (π/2, π/2).
Condition (I)— At the symmetric line of J1 = J2, the classical Hamiltonian corresponds to the following form,
\begin{equation*} \mathcal{H} = \frac{J_{1}}{2}\left[\left(\sum_{i = 0}^{11} \boldsymbol{S}_{i}\right)^{2}{} - 12\right] + (J_{3} - J_{1})\sum_{\langle i,j\rangle_{3}}\boldsymbol{S}_{i} \cdot \boldsymbol{S}_{j}. \end{equation*} |
\begin{equation*} E = \frac{J_{1}}{2}[(12\cos\alpha)^{2} - 12] + 6(J_{3} - J_{1})\cos\alpha. \end{equation*} |
Condition (II)— We first consider the ground state with only 1st neighbor interaction. In this condition, we assume following spin configuration with vector spins Si = (sin αi cos βi, sin αi sin βi, cos αi), α0 = 0, β1 = 0, and,
\begin{align*} &\alpha_{1} = \alpha_{2} = \alpha_{3} = \alpha_{4} = \alpha_{5} (= \alpha),\ \\ &\beta_{2} - \beta_{1} = \beta_{3} - \beta_{2} = \beta_{4} - \beta_{3} = \beta_{5} - \beta_{4} = \beta_{1} - \beta_{5} (= \Delta). \end{align*} |
\begin{equation*} E_{0,i} = \cos\alpha,\quad E_{i,i + 1} = E_{5,1} = \cos^{2}\alpha + \sin^{2}\alpha \cos \Delta. \end{equation*} |
\begin{equation*} E_{i,j} = \cos \alpha = \begin{cases} 1 & (\Delta = 0)\\ 1/\sqrt{5} & (\Delta = \pm 2\pi/5)\\ -1/\sqrt{5} & (\Delta = \pm 4\pi/5) \end{cases} . \end{equation*} |
In this section, we show numerical results in the quantum model with spin-1/2 operators located at twelve vertices of icosahedron. The ground state only with antiferromagnetic 1st neighbor interaction has already been presented by N. P. Konstantinidis.27) However, effects of 2nd and 3rd neighbor interactions on the ground state remain unclear so far. Thus, we investigate these effects and compare quantum phase diagram with the classical one.
We first calculate ground-state energies in the Hilbert subspace limited by $S_{\text{tot}}^{z} = \sum\nolimits_{\text{i}}S_{\text{i}}^{\text{z}} $. By checking degeneracy of ground states between different subspaces, we can determine total S, e.g., if the ground-state energies of only $S_{\text{tot}}^{z} = 0$ and 1 are the same, the ground state is triplet (total S = 1). Figure 5(a)–(b) shows the ground-state energy and its derivative with respect to ϕJ in the parameter space. In the shaded area of Fig. 5(a), ground states are 13-fold degeneracy, i.e., total S = 6 ground state corresponding to the ferromagnetic (F) state. Except for the shaded area, we have confirmed no degeneracy between different subspaces, so that singlet ground state appears. In the singlet area, we can see anomalous lines in Fig. 5(b), which imply phase boundaries. To confirm the phase boundaries, we also calculate an overlap of ground states with neighboring sample points in the parameter space, so-called fidelity, defined by
\begin{equation*} \text{Fd}(\theta_{J},\phi_{J};\delta) = |\langle\text{gs}{:}\ \theta_{J},\phi_{J}|\text{gs}{:}\ \theta_{J},\phi_{J} + \delta\rangle|. \end{equation*} |
(a) Ground state energy and (b) its derivative with respect to ϕJ obtained by exact diagonalization of the quantum model. In the parameter space, we take 100 × 100 sample points for calculation. The shaded area in (a) represents total S = 6, i.e., ferromagnetism. (c) Fidelity of the ground state. The dotted lines drawn by hand represents the phase boundaries among hedgehog singlet (HS), dual hedgehog singlet (DHS), bonding pairs’ singlet (BPS), antibonding pairs’ singlet (APS), and ferromagnetic (F) phases.
The four singlet phases are understood as follows. The upper region of singlet phase includes the north pole θJ = 0 and its ground state is intuitively described by that at the north pole. Since only the 3rd neighbor antiferromagnetic interaction is non-zero at the north pole, two spins on opposite vertices of icosahedron compose a singlet (antibonding) pair and the ground state is the direct product of six singlet pairs. Hence, we call the upper region antibonding pairs’ singlet (APS) phase. On the other hand, the south pole θJ = π requires close attention because the south pole is a singular point between the ferromagnetic phase and the lower region of singlet phase. In fact, at the south pole, where J1 = J2 = 0 and J3 > 0, six triplet (bonding) pairs consisting of spins on opposite vertices of icosahedron are completely decoupled. With slight positive J1 = J2 > 0, which is located in the lower region of singlet phase, six triplets antiferromagnetically interact with each other, resulting in a singlet ground state. Thus, we call the lower region bonding pairs’ singlet (BPS) phase. In middle region, there are two singlet phases more, which include parameter points with only the 1st and 2nd neighbor antiferromagnetic interactions, i.e., (θJ, ϕJ) = (π/2, 0) and (π/2, π/2), respectively. In the classical model, the ground state with only the 1st or 2nd neighbor interaction is discussed in Sec. 3.2. Especially, with only the 2nd neighbor interaction, the classical spin configuration is hedgehog-like. Therefore, we call these two phases hedgehog singlet (HS) and dual hedgehog singlet (DHS) phases, respectively.
To distinguish the quantum phases, we introduce projection operators of singlet and triplet pairs of ith and jth sites given by,
\begin{equation*} \mathcal{P}_{i,j}^{s} = \frac{1}{4} - \boldsymbol{S}_{i} \cdot \boldsymbol{S}_{j},\quad \mathcal{P}_{i,j}^{t} = \boldsymbol{S}_{\boldsymbol{i}} \cdot \boldsymbol{S}_{\boldsymbol{j}} + \frac{3}{4}. \end{equation*} |
\begin{equation*} O_{\textit{APS}} = \left\langle \prod_{i\ (<\bar{\imath})} \mathcal{P}_{i,\bar{i}}^{s} \right\rangle,\quad O_{\textit{BPS}} = \left\langle \prod_{i\ (<\bar{\imath})} \mathcal{P}_{i,\bar{i}}^{t} \right\rangle. \end{equation*} |
\begin{equation*} O_{\textit{HS}} = \left\langle \prod_{(i,j) = (0,5),(1,6),(2,3)}(\mathcal{P}_{i,\bar{i}}^{s}\mathcal{P}_{j,\bar{j}}^{t} - \mathcal{P}_{i,\bar{i}}^{t}\mathcal{P}_{j,\bar{j}}^{s}) \right\rangle. \end{equation*} |
Order parameters for (a) OAPS (b) OBPS, and (c) OHS. The color boundaries correspond to the phase boundaries in Fig. 5.
In this paper, we have investigated magnetic ground states in both classical and quantum Heisenberg spin models on an icosahedral cluster, where all bonds are considered as ferromagnetic or antiferromagnetic exchange interactions. The ground-state phase diagrams have been numerically determined by using simulated annealing and exact diagonalization methods. Moreover, we have shown analytical explanations of spin configurations at specific points in the parameter space. Based on the numerical and analytical examinations, we have characterized four ground-state phases, i.e., the HA, DHA, PPA, and F phases with the APA state in the classical model. On the other hand, we have also classified the ground-state phases in the quantum model with numerical results on the analogy of the classical phases. In fact, we have successfully demonstrated the qualitative coincidence between the classical and quantum phases. Furthermore, we have found a distinctive quantum phase, the APS phase, in addition to four quantum analogs of classical phases, namely the BPS, HS, DHS, and F phases together with those order parameters. The icosahedral spin clusters are in general found in the Tsai-type quasicrystals and approximants. In these alloys, spins are coupled with each other via so-called the RKKY interactions, and therefore, the icosahedral spin clusters are not isolated but interact with each other. However, magnetic properties can strongly reflect characteristics of an isolated icosahedral spin cluster if intra cluster interactions are relatively large enough as compared with inter cluster interactions. Thus, our study can give a good starting point to understand the magnetic properties experimentally observed in the Tsai-type quasicrystals and approximants.
We would like to thank T. J. Sato and T. Hiroto for fruitful discussions. This work was partly supported by Challenging Research (Exploratory) (Grant No. JP17K18764), Grant-in-Aid for Scientific Research on Innovative Areas (Grant No. JP19H05821).