Subscriber access provided by Massachusetts Institute of Technology
J. Phys. Soc. Jpn. 91, 104603 (2022) [7 Pages]
FULL PAPERS

Small Spin Clusters Mimicking a Temperature-Induced Phase Transition: Spins on Vertices of Regular Octahedron and Icositetrachoron

+ Affiliations
Department of Chemistry, Faculty of Pure and Applied Sciences, University of Tsukuba, Tsukuba, Ibaraki 305-8571, Japan

Monte Carlo simulations yielded examples of small spin clusters that exhibit a qualitative change in their character upon temperature variation. Headless spins on vertices of the regular octahedron (6 spins) and icositetrachoron (24 spins), a hyper polyhedron in dimension four, communicate with the neighbors through interaction that prefers mutual twists. The qualitative change arises from the twist interaction and the triangular “lattice” geometry, often tempting to bring frustration.

©2022 The Physical Society of Japan
1. Introduction

Phase transitions are the most fundamental phenomena for switching materials properties. This statement applies even to nanomaterials while reserving the definitions of phases and phase transitions. This paper concerns how small systems exhibit a property change resembling a phase transition.

In the ordinary context, the phase is a concept characterizing a state of bulk materials. Theoretical treatments often assume the recovery of bulk properties in the so-called thermodynamic limit to an infinitely large system. Indeed, mathematical singularities in thermodynamic quantities, which characterize many phase transitions, can happen in statistical mechanical computations only within this limit. There seems no room to expect a property change mimicking a phase transition for a system consisting of a small number of particles in this sense.

Thermodynamic singularity and symmetry breaking are widely accepted as intrinsic properties to phase transitions,13) though there seems to remain room for whether the latter unavoidably brings the former with detectable strength. Besides, there have been many examples that their identifications came after discoveries in practice. Even if we reserve the limited size of actual bulk materials, singularities are always blurred because of limited sample purity and incomplete experiments. Besides, the broken symmetry can be hidden initially and revealed only after theoretical elucidation. For a remarkable example, the proper identification of the broken symmetry, U(1), for the isotropic superconductivity4) came later, about 50 years after the discovery.5) Thus, the practical criterion to distinguish “states” of nanomaterials (in place of phases) would be their distinguishability without referencing the outside. The criterion is precise enough to exclude a quantitative difference in liquid–gas transitions. It is also noteworthy that the cooperative conformational change of proteins is widely regarded as a kind of phase transition blurred by small system size, and the ability to exhibit such a transition may be an essential ingredient of the definition of “proteins.”6,7)

The magnetism of small clusters has long attracted considerable attention. Plenty of literature is available about theoretical813) and experimental aspects.1417) In the early days, theoretical/computational studies aimed to elucidate the properties of bulk systems.8,9) Experimental studies have grown significantly1417) after some techniques became available to produce clusters. There is a claim of a phase transition in clusters of about a hundred atoms.15) In recent years, the magnetism of small systems has been one of the central issues in nanotechnology.14,18) Moreover, various spin models have played a crucial role in the research field of phase transitions as essential yet tractable models extracted from a broad range of phenomena in liquid–gas transition and protein folding.13,19)

Recently, we have encountered a symptom of a phase transition, which seemingly contradicts the Hohenberg–Mermin–Wagner theorem,20,21) in a large ensemble of continuous headless spins on the two-dimensional triangular lattice under the periodic boundary condition.22) Although the interaction preferring the parallel alignment does not produce even the Kosterlitz–Thouless transition,2325) the preference for mutual twist results in an emergence of the transition. The appearing order is neither spin orientation nor lattice periodicity without uniaxial anisotropy. However, the change in the periodicity appears (the formation of three sublattices) under an external field, which scarcely affects the thermodynamic properties of the transition.26) The spin arrangement of the ordered state under fields resembles the ground state of the antiferromagnetic 3-state Potts model.27) It is noteworthy that the system tends to attain the absolute minimum energy (without frustration) in the low-temperature limit, irrespective of uniaxial anisotropy. The distinguishability of two phases based solely on the local property without field prompted us to consider how a small system is sufficient to exhibit such a “transition”? Then, apart from the torus implicitly assumed in the periodic boundary condition in dimension two, “isotropic” lattices on a sphere seem intriguing. Those in dimension three correspond to regular polyhedra (Platonic solids), of which all vertices are geometrically congruent.28)

The target of the present study is two spin clusters, of which geometrical properties are compatible with the ground state configuration of the antiferromagnetic 3-state Potts model, consisting of three equivalent sublattices. The required conditions are a triangle as the smallest bonding circuit and a prohibited odd-membered path surrounding a central spin. Only the octahedron fulfills the conditions in dimension three, as seen in Fig. 1. If we search such clusters within the geometrically regular polytopes (polyhedra in arbitrary dimension), we find only the icositetrachoron (also known as “24-cell”) in dimension four (Fig. 5) and none in any higher dimensions.28) Computed properties indicate these clusters exhibit a qualitative property change, which we need no outside reference to judge. The results demonstrate the crucial importance of the combination of the interaction and the lattice geometry.


Figure 1. (Color online) Ground state spin configurations of the octahedral cluster consisting of 6 spins interacting by Eq. (1). Deep and light blue spins (B and B′) on the foreground vertex participate in the regular RGB configuration and one of the B-flipped configurations, respectively.

2. Model and Method

We examine the property of the clusters consisting of spins on vertices of regular octahedron and icositetrachoron. The spins are headless and classical (continuous) with the unit length, i.e., \(|\boldsymbol{s}_{i}|=1\). The interaction between two neighboring spins, \(\boldsymbol{s}_{i}\) and \(\boldsymbol{s}_{j}\), is assumed to have the following form with \(r=3/10\):22,29) \begin{equation} V_{ij} = - V_{0} \left[(1-r)P_{2}(\boldsymbol{s}_{i}\cdot\boldsymbol{s}_{j})-r\frac{7}{3}P_{4}(\boldsymbol{s}_{i}\cdot\boldsymbol{s}_{j})\right] \end{equation} (1) where \(P_{n}(\cdot)\) is the Legendre polynomial of the n-th order. This potential changes its shape with increasing r from 0. The state with the minimum energy is unique at the parallel alignment of a spin pair for \(r\leq r_{\text{c}} =9/79\) but not for \(r>r_{\text{c}}\). For \(r>r_{\text{c}}\), including the present study, three nearest neighbor spins (“trio”) interacting with each other (forming a triangle) can find a suitable configuration satisfying the best conditions for all pairs. The description of the geometry of clusters and possible spin arrangement on them comes later in the relevant sections.

Monte Carlo (MC) simulations were performed using the Metropolis algorithm.30) A trial orientation of a spin was uniformly generated from 1% of the hemisphere around the original orientation. At each temperature, we performed \(2\times 10^{6}\) steps per spin. The first \(6\times 10^{5}\) steps were treated as the equilibration steps based on the analysis of sequential energy data, and the last \(1.4\times 10^{6}\) steps were used for calculating physical quantities. To reduce the error arising from the small number of spins in a cluster, we averaged the results of independent MC simulations to yield the presented data. The numbers are nine for the octahedron and eight for the icositetrachoron. Since the MC steps were not large enough at low temperatures, as seen in Fig. 3, we set a low-temperature limit for discussion in each case.

Because of the headless nature of our spins, we employ the so-called nematic order parameter to sense the orientational order. It is defined by \begin{equation} \sigma = \langle P_{2}(\cos\theta)\rangle, \end{equation} (2) where θ is the polar angle of the orientation of a spin. Practically, it is impossible (but unnecessary in practice) to identify the polar axis because the orientation of the polar axis fluctuates appreciably due to the smallness of clusters. Alternately, the nematic order parameter is calculated by analyzing \(\langle\boldsymbol{s}\otimes\boldsymbol{s}\rangle\), an instantaneous system average of the self Kronecker product of a spin.

3. Results and Discussion
Octahedron
Geometry and ground state

The octahedron bears the Schläfli symbol28) of \(\{3,4\}\) and satisfies both conditions of the triangles and even-membered paths around each spin. The ground state configuration of the “highest order” realizes by arranging three orientations (say, R, G, and B), which satisfy the optimum condition, on three diagonal pairs. This configuration is equivalent to the ground state of the antiferromagnetic 3-state Potts model.27) However, each spin can flip without energy penalty to the image reflected by the mirror plane containing four bonding spins. This mechanism is the same as the origin of the macroscopic degeneracy of the triangular lattice case.22) The regular (highest-order) configuration and an example of flipped ones are illustrated in Fig. 1. While fixing R and G, the two remaining spins can independently take either B or B′. Thus, the ground states are \(2^{2}\)-times degenerate for a fixed pair of R and G. Note that the spin configuration can continuously change orientation while maintaining optimum conditions among all spins.

3-State Potts models

Before proceeding into the simulation results of the present model, we see what happens in ordinary cases by looking at the behaviors of the 3-state Potts models27) on the same cluster geometry. The interaction is written as \begin{equation} V_{ij} = - J\boldsymbol{s}_{i}\cdot\boldsymbol{s}_{j}, \end{equation} (3) where the three states are expressed by mutually orthogonal unit vectors in dimension three. Note that an antiferromagnetic (neighbor-non-favoring) model (\(J<0\)) does not suffer from any frustration for the octahedral cluster, similar to the present model. Under this setting, we can use the nematic order parameter as an indicator of an average order between two spins, i and j. Irrespective of the sign of J, the local order \(\langle P_{2}(\boldsymbol{s}_{i}\cdot\boldsymbol{s}_{j})\rangle\) is 1 for the diagonals in the ground states.

We computed the properties analytically while considering all (\(= 3^{6}\)) spin configurations. Figure 2 shows the temperature dependences of \(\langle P_{2}\rangle_{\text{e}}\) (\(|\boldsymbol{d}|=1\)) and \(\langle P_{2}\rangle_{\text{d}}\) (\(|\boldsymbol{d}|=\sqrt{2}\)) for ferromagnetic and antiferromagnetic 3-state Potts models. Since the octahedral cluster is compatible with the optimal spin configurations without frustration in both cases, the limiting magnitudes at \(T=0\) are either \(-1/2\) or 1. With increasing temperature, they gradually and monotonously reach 0 (vanishing the spin order). The heat capacity exhibits a broad maximum around the temperature of the maximum where \(|d\langle P_{2}\rangle/dT|\) exhibits a maximum. An essential feature is that they retain the order of averaged magnitudes independently of temperature. The ferromagnetic model always has a larger average for the neighbors \(\langle P_{2}\rangle_{\text{e}} >\langle P_{2}\rangle_{\text{d}}\), while the antiferromagnetic one keeps the counter relation. We can translate the latter fact that the sublattice order is always in the antiferromagnetic case and only weakens with increasing temperature. Thus, irrespective of temperature, the point group assignable to the cluster while considering the sublattice order is \(D_{2h}\) of cuboids, of which three pairs of faces are distinct. In contrast, that of the ferromagnetic case is \(O_{h}\) of cubes with six equivalent faces.


Figure 2. (Color online) Temperature dependence of two-point correlation functions, \(\langle P_{2}\rangle_{\text{e}}\) (\(|\boldsymbol{d}|=1\)) and \(\langle P_{2}\rangle_{\text{d}}\)) (\(|\boldsymbol{d}|=\sqrt{2}\)), for ferromagnetic (superscript F) and antiferromagnetic (AF) 3-state Potts models [defined by Eq. (3)] on an octahedral cluster consisting of 6 spins on vertices.

Simulation results

Figure 3 shows the heat capacity C and susceptibility χ of the octahedral cluster together with additional data. Both quantities directly reflect fluctuation in, respectively, energy and nematic order parameter. Consequently, they serve as indicators of phase transitions.13) They exhibit peaks at different temperatures in the present case, resulting in three temperature regions. The highest temperature region corresponds to the isotropically disordered state of spin alignment. The decreasing trend in heat capacity on heating and roughly constant magnitude of σ (Fig. 4) on the low-temperature side are consistent with this assignment. Although the crossover between the highly disordered and locally ordered states is widely seen in usual spin models, irrespective of ferromagnetic or antiferromagnetic interactions, only a broad peak appears in such cases. A distinction in the present case from such cases manifests in the peak at the lower temperature. Indeed, a high-temperature peak of the triangular lattice is due to the crossover, whereas a low-temperature one corresponds to the putative phase transition exhibiting exciting properties.22,26)


Figure 3. (Color online) Heat capacity C per spin (a) and susceptibility χ (b) of the clusters of spins on the vertices of the regular octahedron (cross, \(z=4\)) and icositetrachoron (circle, \(z=8\)) with error bars estimated from the statistics of separate simulations. The heat capacity of the triangular lattice22) (\(z=6\)) is shown for reference by a dotted curve in the upper panel.


Figure 4. (Color online) Nematic order parameter σ (circle) and two-point correlation functions [plus-sign for edge \(\langle P_{2}\rangle_{\text{e}}\) (\(|\boldsymbol{d}|=1\)), cross for diagonal \(\langle P_{2}\rangle_{\text{d}}\) (\(|\boldsymbol{d}|=\sqrt{2}\))] for the octahedral cluster consisting of 6 spins on vertices. The horizontal line indicates the magnitude calculated from the optimal condition for the neighbors. In both, error bars indicate the standard deviation in 9 simulations.

Because of the smallness of the cluster, we can define only two local quantities: \(\langle P_{2}(\boldsymbol{s}_{i}\cdot\boldsymbol{s}_{j})\rangle\) averaged over bonding edges \(\langle P_{2}\rangle_{\text{e}}\) (\(|\boldsymbol{d}|=1\) with a vector \(\boldsymbol{d}\) between spins) and that over non-bonding diagonals \(\langle P_{2}\rangle_{\text{d}}\) (\(|\boldsymbol{d}|=\sqrt{2}\)). Figure 4 shows their temperature dependence. \(\langle P_{2}\rangle_{\text{e}}\) points 2/5 on cooling. This magnitude is expected for ground states with the optimal spin configuration for all pairs of spins connected by edges. On the other hand, as for the diagonal magnitude at low temperatures, although a naïve average over the degenerating ground states (ca. 0.78) is within error bars, we will need to consider some particular spin configurations from the whole set of ground states for complete understanding. The uniaxiality (nematic order with a finite σ) may choose some particular spin configurations from the whole set of ground-state configurations. Indeed, the simulation result is close to one (≈ 0.725), which is expected for the subset involving the disorder of one-third spins.26)

Change in “averaged” symmetry

Both quantities, \(\langle P_{2}\rangle_{\text{e}}\) and \(\langle P_{2}\rangle_{\text{d}}\), vary starting from the limiting magnitudes with increasing temperature (Fig. 4). The former is roughly parallel with the nematic order parameter. On the other hand, the latter starts to decrease appreciably around \(T\approx 0.1V_{0}/k_{\text{B}}\), which corresponds to the significant maximum in heat capacity Fig. 3(a). Since their temperature dependences differ, they cross around \(T_{\text{c}}\approx 0.285V_{0}/k_{\text{B}}\). Thus, the cluster changes its character from ferroic to antiferroic only upon temperature variation. Considering the case of the 3-state Potts models in the previous section, we can convincingly claim the distinct symmetries at sufficiently high and low temperatures. Although significant fluctuation originating in a small number of spins in a cluster makes it challenging to locate the transition temperature, it seems reasonable to assign it the temperature of the crossing. Note that this choice needs no outside references.

The essential character of our interaction [Eq. (1)] is that it hates the complete uniaxial order but still likes it in part. Indeed, systems on the simple cubic lattice undergo a phase transition from the isotropic (i.e., disordered) state to the uniaxial (nematic) state on cooling with \(0\leq r\lesssim 0.38\).29) Even in the two-dimensional cases, they possess potential instability to the order.22) The lattice geometry, full of triangles with an even-membered circumstance, is essential to establish the order without frustration.

Icositetrachoron
Geometry and ground state

Despite a small number of spins, the octahedral cluster exhibits a qualitative change, i.e., symmetry breaking upon temperature variation, if the interaction between spins prefers mutual twist. The change accompanies enhanced fluctuation in energy and order, yet without singularity in thermodynamic quantities. Here, we proceed to the case of another cluster. Its positive result reinforces our claim of the crucial importance of lattice geometry.

A regular icositetrachoron is a regular polytope, i.e., a hyper polyhedron in dimension four.28) It bears the Schläfli symbol of \(\{3,4,3\}\). In the four-dimensional Cartesian coordinate \((x_{1},x_{2},x_{3},x_{4})\), 24 equivalent vertices are represented by all permutations of \((\pm\sqrt{2}/2,\pm\sqrt{2}/2,0,0)\) with any combination of double signs. Figure 5(a) illustrates the adjacency of vertices. Each vertex is connected to 8 adjacent vertices through equivalent edges (of a unit length), resulting in \(24\times 8\div 2 = 96\) edges. Figure 5(b) is projection into a three-dimensional space with constant \(x_{4}\). The icositetrachoron is projected on a cuboctahedron, of which the vertices are images from the \(x_{4}=0\) sector. Each square face carries an additional vertex, which comes from either of two sectors, \(x_{4}=\pm\sqrt{2}/2\). The additional vertices form a regular octahedron in each sector. Since the distance between two points, \(\boldsymbol{p}\) and \(\boldsymbol{q}\), is defined as \(d=[\sum_{i=1,4} (p_{i}-q_{i})^{2}]^{1/2}\), a variety of distances between vertices is limited to \(\{1,\sqrt{2},\sqrt{3}, 2\}\), corresponding to the kth nearest-neighbor vertices. We can find these distances in Fig. 5(b) while taking only large spheres (vertices in the \(x_{4}=0\) sector) into account in the projection.


Figure 5. (Color online) Representations of a regular icositetrachoron. (a) A graph representing connections between adjacent vertices. (b) The projection into a three-dimensional space (\(x_{4}=\text{constant}\)). Large spheres on vertices of a cuboctahedron belong to the sector \(x_{4}=0\), while small spheres at centers of square faces represent vertices belonging to either of the sectors with \(x_{4}=\pm\sqrt{2}/2\). The latter is doubly degenerate in this projection. Only the edges (i.e., bonds) from small to large spheres are indicated by thin bonds. The bonds between small spheres are invisible in this projection because they are inside the cuboctahedron.

Although it is not easy to recognize from the projection, 12 regular triangles and 8 regular octahedrons gather around each vertex. The “lattice” geometry of the icositetrachoron is compatible with the ground states mimicking that of the antiferromagnetic 3-state Potts model, as confirmed by the successful coloring with R, G, and B in Fig. 5. Besides, the spin systems on these “lattices” have the degeneracy of the ground states without energetic frustration, similarly to the triangular lattice.22) For example, each spin can flip without an energy penalty, as in the octahedron case (Fig. 1). With fixed R and G, the eight (\(= 24/3\)) spins can independently take either B or B′, resulting in the degeneracy of \(2^{8}\) for a particular pair of R and G. However, some of R or G spins can flip distinctly from the octahedron, resulting in additional degeneracy. The characterization of the ground states is challenging and currently incomplete.

Simulation results

The simulation results of the icositetrachoron cluster are generally similar to the octahedron cluster, as seen in Fig. 3. However, the anomalies in heat capacity and susceptibility are more significant in the icositetrachoron than in the octahedron. The origin of more evident behavior may originate in the increased spin number and/or enhanced connectivity among spins. Remarkably, despite only 24 spins involved, the results for the icositetrachoron cluster enable similar analyses to the triangular lattice based on similar temperature dependence.22)

Figure 3 compares the heat capacity C and susceptibility χ of the icositetrachoron cluster (the coordination number of a vertex being \(z=8\)) with others. The general trends are similar to, yet more significant than, the octahedron cluster (\(z=4\)). Namely, the heat capacity exhibits peaks around \(k_{\text{B}}T/zV_{0}\approx 0.03\) and 0.1, whereas the susceptibility only has a pronounced peak at the latter. The two peaks in heat capacity are similar to the triangular lattice, which undergoes a phase transition at a lower temperature and a crossover between the isotropic and sensitive (short-ranged nematic) states at a higher temperature. Its susceptibility shows a remarkable peak only at the higher temperature.

Figure 6 shows the nematic order parameter σ [Eq. (2)] and two-point correlations \(\langle P_{2}(\boldsymbol{s}_{i}\cdot\boldsymbol{s}_{j})\rangle\) at all possible distances as functions of temperature for the icositetrachoron cluster. Since the assumed interaction is partially ferroic (i.e., nematic), the σ is positive irrespective of temperature because of the finite-size correlation. However, upon decreasing temperature, it significantly grows around the temperature of the heat capacity peak at a higher temperature. Thus, the anomalies at the higher temperature originate in the crossover between the essentially free (well disordered) and the locally ordered nematic states. On the other hand, anomalous behavior is insignificant at the lower temperature. In contrast, two of the two-point correlations exhibit notable increases there. It is noteworthy that the two distances bear the same color as the starting (\(|\boldsymbol{d}|=0\)) spin in the coloring of Fig. 5. The contrasting temperature dependences followed by two distance groups resemble the triangular lattice.22)


Figure 6. (Color online) Nematic order parameter σ (circle) and orientational correlation \(\langle P_{2}[\boldsymbol{s}(0)\cdot\boldsymbol{s}(\boldsymbol{d})]\rangle\) (cross or plus-sign) at different distances \(|\boldsymbol{d}|\) in a cluster consisting of 24 spins on vertices of a regular icositetrachoron.

In the case of the triangular lattice, the analysis of the decay behavior of the two-point correlation offered a handy way to see the occurrence of a phase transition.26) It is doubtful whether this applies to the present case or whether the dependence is meaningful because we have only four distances. However, surprisingly, the distance dependence of the icositetrachoron cluster resembles the triangular lattice.22) Figure 7 illustrates those at selected temperatures. The decay at high temperatures is more substantial than a single power law, and the exponential law may apply. The power law describes the decay quite well at lower temperatures, as evident from the figure. Finally, the dependence exhibits a zigzag on further cooling, reflecting the splitting into the two distance groups, \(\{1,\sqrt{3}\}\) and \(\{\sqrt{2}, 2\}\). The former and the latter, respectively, bear different and same colors in the coloring in Fig. 5. Thus, the zigzag suggests the order of the antiferromagnetic 3-state Potts type consisting of three sublattices.


Figure 7. (Color online) Two-point correlation of spin orientations in a cluster consisting of 24 spins on vertices of a regular icositetrachoron as a function of the logarithmic distance in dimension four at selected temperatures, \(k_{\text{B}}T/V_{0}=(0.81)^{j-10}\) (\(j=5{\text{–}}15\) from the bottom to the top).

Some attempts to analyze the decay behaviors indicated that a fit assuming two separate power-laws, \(\sigma_{1j}d^{-\nu}\) (\(j=0,1\)), with a shared power ν and different coefficients \(\sigma_{1j}\) reasonably describe the decay at a single temperature for the two distance groups. We choose \(j=0\) to the \(\{\sqrt{2}, 2\}\) group based on the same color as the central spin (\(d=0\)) in the coloring in Fig. 5. In the spirit of the analyses for the triangular lattice under external fields,26) the difference in fit parameters, such as \(\Delta\sigma_{1}=\sigma_{10}-\sigma_{11}\), may serve as an effective order parameter. Figure 8 indicates its temperature dependence. It is effectively null through high-temperature anomalies but becomes finite below the low-temperature anomaly in heat capacity.


Figure 8. (Color online) \(\Delta\sigma_{1}\) (left axis, circle with error bar) and locally-defined apparent order parameters, \(A_{2\phi}^{k}\) defined by Eq. (4) (right axis, cross for \(k=1\) and plus-sign for \(k=2\)), for a cluster consisting of 24 spins on vertices of a regular icositetrachoron.

Practical order parameter

Alternately, we may define the quantity that senses the formation of the Potts-like order. In the case of the triangular lattice,22) the quantities were defined for hexagons around an arbitrarily chosen site, and they successfully served as locally-defined order parameters. Because the minimum path around a site is a square, we can take the following form in the present case: \begin{equation} A_{2\phi}^{k} = \frac{1}{4}[1-2P^{*}(d_{k})+P^{*}(\sqrt{2}d_{k})], \end{equation} (4) where \begin{equation*} P^{*}(d) = \frac{\langle P_{2}[\boldsymbol{s}(0)\cdot\boldsymbol{s}(d)]\rangle}{g^{\circ}(d)} \end{equation*} with \(g^{\circ} (d)\) being an “averaged” decay function of the correlation function necessary to correct the distance dependence. The index k represents the square consisting of the kth nearest neighbors from the central vertex.

Since the variety of d is limited to \(\{1,\sqrt{2},\sqrt{3}, 2\}\), only \(k=1, 2\), corresponding to \(d=1,\sqrt{2}\), are useful: \(k=4\) (\(d=2\)) points to a single vertex for each vertex and the \(k=3\), which means to be the nearest to the \(k=4\) vertex, results in a square with \(d=1\), i.e., \(A_{2\phi}^{3}=A_{2\phi}^{1}\). By definition, \(A_{2\phi}^{k}\) vanishes if the spin alignment on the square has the fourfold symmetry while assuming the adequate correction about the distance dependence of the correlation function.

The definition of \(g^{\circ} (d)\) depends on k of \(A_{2\phi}^{k}\). For \(k=1\), \(g^{\circ} (d)\) is a weighted average of two decay functions, i.e., \(g^{\circ} (d)=\frac{1}{3}(\sigma_{10}+2\sigma_{11})d^{-\nu}\). On the other hand, \(g^{\circ} (d)\) is given by \(\sigma_{10}d^{-\nu}\) for \(k=2\) because both the d's belong to the same distance group. Thus, definite \(A_{2\phi}^{1}\) indicates the violation of the fourfold symmetry, while \(A_{2\phi}^{2}\) should be null irrespective of such a phenomenon. Conversely, the deviation of the latter from zero is an indicator of the manipulation of the former.

Figure 8 compares the temperature dependence of \(A_{2\phi}^{k}\) with \(\Delta\sigma_{1}\). As expected, \(A_{2\phi}^{2}\) remains effectively null at all temperatures while \(A_{2\phi}^{1}\) deviates from zero around \(k_{\text{B}}T\approx 0.3V_{0}\). Thus, the temperature dependence indicates that the fourfold symmetry of the square formed by the nearest-neighbor spins is violated at low temperatures. Definite yet gradual temperature dependences of \(A_{2\phi}^{1}\) with \(\Delta\sigma_{1}\) remind us of the behavior of quantities similar to the former at short distances for the triangular lattice.22) Note that their growth becomes sharper at longer distances (around several to tens lattice spacings) and the blurred dependences originate in the uncertainty in the averaged decay function employed.

4. Concluding Remarks

We computationally examined properties of spin clusters that would capture the necessary condition for the phase transition, which the authors22) recently identified as accompanying only a short-range spin order in dimension two. MC simulations of the octahedron and icositetrachoron clusters indicated a “transition” between the states where the spins on the square consisting of nearest-neighbor vertices are all similar and alternate upon temperature variation. The transition of the icositetrahedron cluster accompanies fluctuation more pronounced than the octahedral cluster yet without singularity. Surprisingly, the decay of its two-point correlation function against distance is qualitatively similar to the large triangular lattice: Exponential at high temperatures, zigzag (splitting into two distance groups) due to the formation of sublattices at low temperatures, and power-law in between. Thus, the behavior of the icositetrachoron cluster resembles that of the supposedly infinite triangular lattice for the same spin model despite its small size. Since the sublattice order in the present cases spreads over the whole system, their resemblance to the behavior of the triangular lattice22) indicates its essential role in the phase transition accompanying only a short-range order on the triangular lattice. This conclusion is consistent with the behavior of the triangular lattice under external fields.26) Even if one cannot judge to what sublattice a spin belongs from the spin orientation under the fixed spin order at a far distant place, the successive determinability of the belonging is a matter for the same spin model on the triangular lattice.

One may claim that the chosen lattice geometry and the interaction between spins subject to the present study are somewhat artificial and difficult to find in nature. Although the claim is partly valid, the present results demonstrate a possibility that fascinating phenomena may emerge for nanomaterials by suitable design of interaction and lattice geometry. For example, the non-uniqueness of the spin alignment originating in the interaction [Eq. (1)] offers an exciting possibility for ordered states. When the lattice is bipartite, all spins on one sublattice are parallel, and the rest on the other are disordered on cones satisfying the optimum conditions because of the entropy gain.29) If the numbers of spins (sites) on two sublattices are different, the minorities should be parallel for the same reason. Under an external field, a partition into a spin group parallel along the field and others disordered on cones happen at finite temperatures.26) The so-called order-by-disorder mechanism31) is behind such uniaxial and superlattice orders. In the present case, the “lattice” is not bipartite. However, a similar situation will occur under fields. It is another possibility to tune cluster properties.

Finally, we comment on the headless spin model involving an interaction preferring mutual twists. Although models with Eq. (1) give, by its nature, fruits such as a phase transition involving the order-by-disorder mechanism31) on the simple cubic lattice,29) an unexpected phase transition on the triangular lattice,22) and a “transition” of clusters in the present work, points of interest are not limited to such. We can imagine a twist range of minimum energy beyond its limit \(\theta_{\text{min}}\leq\arccos\sqrt{3/7}\) by Eq. (1). Such \(\theta_{\text{min}}\) opens the following possibilities: Spin groups different from the trio find the minimum energy with \(\theta_{\text{min}}\) for all possible pairs in the group for particular angles. They originate from apex angles of triangles formed by the center of regular polyhedra and their vertices. Namely, \(\theta_{\text{min}}=\arccos(1/3)\) (tetrahedron and cube) and \(\theta_{\text{min}}=\arccos\sqrt{1/5}\) (icosahedron). The octahedron does not contribute because of \(\theta_{\text{min}}=\pi/2\), resulting in the spin trio. The corresponding groups consist of four and six spins, respectively. They are interesting by the ground state with the absolute minimum in energy without frustration on any complex graphs embedded on a spherical or flat surface under the four-color theorem.32) It means that the latter interaction separates graph complexity, frustration, and vast degeneracy of ground states, which are often simultaneously discussed as if they are inseparable. Note that the stability of the sextet requires spins to be headless. Usual directional spins can enjoy the stability of either the trio or quartet.

Acknowledgments

This work was supported in part by JSPS Grant-in-Aid for Scientific Research on Innovative Areas “Discrete Geometric Analysis for Materials Design” (JP20H04629) and Grant-in-Aid for Scientific Research (B) (JP21H01046).


References

  • 1 H. E. Stanley, Introduction to Phase Transitions and Critical Phenomena (Oxford University Press, New York, 1971). Google Scholar
  • 2 J. J. Binney, N. J. Dowrick, A. J. Fisher, and M. E. J. Newman, The Theory of Critical Phenomena (Oxford University Press, New York, 1992). CrossrefGoogle Scholar
  • 3 H. Nishimori and G. Ortiz, Elements of Phase Transitions and Critical Phenomena (Oxford University Press, New York, 2010). CrossrefGoogle Scholar
  • 4 Y. Nambu, Phys. Rev. 117, 648 (1960). 10.1103/PhysRev.117.648 CrossrefGoogle Scholar
  • 5 H. Kamerlingh Onnes, Comm. Phys. Lab. Univ. Leiden 122 & 124 (1911). Google Scholar
  • 6 S. Honda, K. Yamazaki, S. Honda, and H. Morii, Structure 12, 1507 (2004). 10.1016/j.str.2004.05.022 CrossrefGoogle Scholar
  • 7 S. Honda, T. Akiba, Y. S. Kato, S. Honda, M. Sekijima, M. Ishimura, A. Ooishi, H. Watanabe, T. Odahara, and K. Harata, J. Am. Chem. Soc. 130, 15327 (2008). 10.1021/ja8030533 CrossrefGoogle Scholar
  • 8 K. Binder, H. Rauch, and V. Wildpaner, J. Phys. Chem. Solids 31, 391 (1970). 10.1016/0022-3697(70)90119-8 CrossrefGoogle Scholar
  • 9 K. Binder, Physica 62, 508 (1972). 10.1016/0031-8914(72)90237-6 CrossrefGoogle Scholar
  • 10 J. Merikoski, J. Timonen, M. Manninen, and P. Jena, Phys. Rev. Lett. 66, 938 (1991). 10.1103/PhysRevLett.66.938 CrossrefGoogle Scholar
  • 11 P. V. Hendriksen, S. Linderoth, and P.-A. Lindgård, Phys. Rev. B 48, 7259 (1993). 10.1103/PhysRevB.48.7259 CrossrefGoogle Scholar
  • 12 J. H. Luscombe, M. Luban, and F. Borsa, J. Chem. Phys. 108, 7266 (1998). 10.1063/1.476144 CrossrefGoogle Scholar
  • 13 A. M. Läuchli, J. Sudan, and E. S. Sørensen, Phys. Rev. B 83, 212401 (2011). 10.1103/PhysRevB.83.212401 CrossrefGoogle Scholar
  • 14 M. Martins and W. Wurth, J. Phys.: Condens. Matter 28, 503002 (2016). 10.1088/0953-8984/28/50/503002 CrossrefGoogle Scholar
  • 15 I. M. L. Billas, A. Châtelain, and W. A. de Heer, Science 265, 1682 (1994). 10.1126/science.265.5179.1682 CrossrefGoogle Scholar
  • 16 Y. Miyazaki, A. Bhattacharjee, M. Nakano, K. Saito, S. M. J. Aubin, H. J. Eppley, G. Christou, D. N. Hendrickson, and M. Sorai, Inorg. Chem. 40, 6632 (2001). 10.1021/ic010567w CrossrefGoogle Scholar
  • 17 K. Mitsumoto, E. Oshiro, H. Nishikawa, T. Shiga, Y. Yamamura, K. Saito, and H. Oshio, Chem.—Eur. J. 17, 9612 (2011). 10.1002/chem.201101404 CrossrefGoogle Scholar
  • 18 S. A. Wolf, A. Y. Chtchelkanova, and D. M. Treger, IBM J. Res. Dev. 50, 101 (2006). 10.1147/rd.501.0101 CrossrefGoogle Scholar
  • 19 P. M. Chaikin and T. C. Lubensky, Principles of Condensed Matter Physics (Cambridge University Press, Cambridge, U.K., 1995). CrossrefGoogle Scholar
  • 20 P. C. Hohenberg, Phys. Rev. 158, 383 (1967). 10.1103/PhysRev.158.383 CrossrefGoogle Scholar
  • 21 N. D. Mermin and H. Wagner, Phys. Rev. Lett. 17, 1133 (1966). 10.1103/PhysRevLett.17.1133 CrossrefGoogle Scholar
  • 22 K. Saito, M. Hishida, and Y. Yamamura, J. Phys. Soc. Jpn. 90, 124003 (2021). 10.7566/JPSJ.90.124003 LinkGoogle Scholar
  • 23 J. M. Kosterlitz and D. J. Thouless, J. Phys. C 6, 1181 (1973). 10.1088/0022-3719/6/7/010 CrossrefGoogle Scholar
  • 24 J. M. Kosterlitz, J. Phys. C 7, 1046 (1974). 10.1088/0022-3719/7/6/005 CrossrefGoogle Scholar
  • 25 Y. Tomita, Phys. Rev. E 90, 032109 (2014). 10.1103/PhysRevE.90.032109 CrossrefGoogle Scholar
  • 26 K. Saito and Y. Yamamura, J. Phys. Soc. Jpn. 91, 074007 (2022). 10.7566/JPSJ.91.074007 LinkGoogle Scholar
  • 27 H. Park, Phys. Rev. B 49, 12881 (1994). 10.1103/PhysRevB.49.12881 CrossrefGoogle Scholar
  • 28 H. S. M. Coxeter, Regular Polytopes (Dover, New York, 1973). Google Scholar
  • 29 K. Saito, M. Hishida, and Y. Yamamura, J. Phys. Soc. Jpn. 86, 084602 (2017). 10.7566/JPSJ.86.084602 LinkGoogle Scholar
  • 30 N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller, J. Chem. Phys. 21, 1087 (1953). 10.1063/1.1699114 CrossrefGoogle Scholar
  • 31 J. Villain, R. Bidaux, J.-P. Carton, and R. Conte, J. Phys. (Paris) 41, 1263 (1980). 10.1051/jphys:0198000410110126300 CrossrefGoogle Scholar
  • 32 K. Appel and W. Haken, Bull. Am. Math. Soc. 82, 711 (1976). 10.1090/S0002-9904-1976-14122-5 CrossrefGoogle Scholar

Cited by

View all 1 citing articles

full accessThree-Step Ordering of Classical Headless Spins Preferring Twists on Icositetrachoric Honeycomb, a Four-Dimensional Analog of the Triangular Lattice

054003, 10.7566/JPSJ.92.054003