betaFeedback

WorksPhysics and Astronomy

Algorithm for spin symmetry operation search

Abstract

A spin space group provides a suitable way to fully exploit the symmetry of a spin arrangement with a negligible spin-orbit coupling. There has been a growing interest in applying spin symmetry analysis with the spin space group in the field of magnetism. However, there is no established algorithm to search for spin symmetry operations of the spin space group. This paper presents an exhaustive algorithm for determining spin symmetry operations of commensurate spin arrangements. The present algorithm searches for spin symmetry operations from the symmetry operations of a corresponding nonmagnetic crystal structure and determines their spin-rotation parts by solving a Procrustes problem. An implementation is distributed under a permissive free software license in spinspg v0.1.1: https://github.com/spglib/spinspg.

Introduction

When the spin–orbit coupling (SOC) is negligible, a spin space group is an appropriate concept to fully exploit the symmetry of a corresponding spin arrangement [11, 17, 15, 27]. The spin arrangement comprises a crystal structure and magnetic moments. A spin symmetry operation of the spin space group is assumed to act on the spatial and spin coordinates simultaneously, generalizing a magnetic symmetry operation of a magnetic space group. The spin space group was first introduced to analyze an extra symmetry of a spin Hamiltonian for neutron scattering experiments [1, 2] and has recently been applied to the field of magnetism [23]: for example, an analysis of a symmetry-adapted tensor of transport properties with negligible SOC [24, 29], symmetries of a spin Hamiltonian [28], magnon band structure [5], and classification of antiferromagnetism [22].

When we consider spin space groups of given spin arrangements, we have to first exhaustively search for their spin symmetry operations. To the best of our knowledge, there is no rigorous algorithm to search for spin symmetry operations. Therefore, the development of an algorithm and its implementation would benefit the spin symmetry analysis.

There are a few differences between spin space groups and magnetic space groups in terms of symmetry search algorithms. First, a spin space group may contain nontrivial operations acting on only spin coordinates, called a spin-only group, which complicates the group structure of the spin space group. Second, although we only need to consider at most double enlarged cells for magnetic space groups, the unit cell size may arbitrarily change between a spin arrangement and its nonmagnetic correspondence. Lastly, spin rotations, which simultaneously act on magnetic moments, do not have to belong to a crystallographic point group.

Here, we present a rigorous and robust algorithm for determining spin symmetry operations of a given commensurate spin arrangement, extending our magnetic symmetry operation search algorithm Shinohara et al., to be published [19]. The present algorithm fully exploits the group structure of the spin space groups and outputs spin symmetry operations as a coset decomposition of the spin space group, based on the seminal works by Litvin and Opechowski [12, 11, 13]. We explicitly denote basis vectors of space groups and spin space groups and employ a lattice algorithm to deal with the case when the unit cell size varies with and without magnetic moments. We search for the spin-rotation parts from three-dimensional orthogonal groups by solving a well-known optimization problem called a Procrustes problem [7]. Note that we restrict the present algorithm to commensurate spin arrangements in order to use similar inputs and outputs to an existing space group search implementation [21]. The implementation is distributed under the BSD 3-clause license in SPINSPG v0.1.1 on top of a crystal symmetry search algorithm [21]. For magnetic crystal structures tabulated in MAGNDATA [6], the present algorithm and implementation have been used to identify physical properties free from SOC [25].

This paper is organized as follows. In § 2, we give definitions of spin space groups and their derived groups. In § 3, we provide an algorithm for determining a spin-only group, spin translation group, and spin space group of a given spin arrangement. In § 4, we demonstrate the present spin symmetry operation search to a spin arrangement of a NiAs-type CrSe. The notations and terminology in this paper are summarized in Table 1.

SymbolMeaning
A⊔BA \sqcup BDisjoint union of set AA and BB with A∩B=∅A \cap B = \varnothing
E(3)E(3)Three-dimensional Euclidean group
O(3)O(3)Three-dimensional orthogonal group
E\mathbf{E}Identity matrix
SN\mathcal{S}_NSymmetric group of degree NN
∥y∥2=yTy\lVert y \rVert_2 = \sqrt{y^{\mathsf{T}}y}l2l^2 norm of vector yy
∥B∥F=Tr⁡BTB\lVert B \rVert_F = \sqrt{\operatorname{Tr} B^{\mathsf{T}}B}Frobenius norm of matrix BB
11Trivial group
r\boldsymbol{r}Position in Cartesian coordinates
m\boldsymbol{m}Magnetic moments
Spin symmetry operation (g,W)(g,W)Pair of spatial operation gg and spin rotation WW
(R,v)A=(ARA−1,Av)(R,v)_A = (ARA^{-1},Av)Spatial operation with basis vectors AA
A=(a1,a2,a3)A=(a_1,a_2,a_3)Basis vectors
X=(x1,⋯ ,xN)X=(x_1,\cdots,x_N)Array of point coordinates
T=(t1,⋯ ,tN)T=(t_1,\cdots,t_N)Array of atomic types
M=(m1,⋯ ,mN)M=(m_1,\cdots,m_N)Array of magnetic moments
Spin arrangement (A,X,T,M)(A,X,T,M)Pair of crystal structure and magnetic moments
(A,X,T)(A,X,T)Crystal structure ignoring magnetic moments
ϵmag\epsilon_{\mathrm{mag}}Absolute tolerance to compare magnetic moments
S,R\mathcal{S},\mathcal{R}Space group
Translation subgroup T(R)\mathcal{T}(\mathcal{R})Translation parts of R\mathcal{R} with identity rotations
Point group P(R)\mathcal{P}(\mathcal{R})Group obtained from the rotation parts of R/T(R)\mathcal{R}/\mathcal{T}(\mathcal{R})
TA\mathcal{T}_ATranslation group spanned by AA
σg\sigma_gPermutation of sites induced by gg
AS=AU−1A_S=AU^{-1}Primitive basis vectors of T(S)\mathcal{T}(S)
Spin-only group Pso=Pso(G)\mathcal{P}_{\mathrm{so}}=\mathcal{P}_{\mathrm{so}}(\mathcal{G})Spin-rotation parts with identity spatial operations
σi,n^i\sigma_i,\hat{n}_iEigenvalue and eigenvector of MMTMM^{\mathsf{T}}
n^∥\hat{n}_{\parallel}Parallel direction for collinear spin arrangement
n^⊥\hat{n}_{\perp}Perpendicular direction for coplanar spin arrangement
AD=ASVA_D=A_SVBasis vectors of T(D(G))\mathcal{T}(\mathcal{D}(\mathcal{G}))
Spin translation group Gst=Gst(G)\mathcal{G}_{\mathrm{st}}=\mathcal{G}_{\mathrm{st}}(\mathcal{G})Subgroup of G\mathcal{G} with identity rotations
Mg=(mσg(1)⋯mσg(N))M_g=(m_{\sigma_g(1)}\cdots m_{\sigma_g(N)})Array of magnetic moments permuted by σg\sigma_g
Spin space group G\mathcal{G}See Sec. 2.2
Family space group F(G)\mathcal{F}(\mathcal{G})Space group composed of spatial parts of G\mathcal{G}
Maximal space subgroup D=D(G)\mathcal{D}=\mathcal{D}(\mathcal{G})Subgroup of G\mathcal{G} with identity spin-rotation parts
Spin-space-group typeSee Sec. 2.2
(P,p)(P,p)Transformation on spatial coordinates
QQTransformation matrix on spin coordinates
Family spin point group B(G)\mathcal{B}(\mathcal{G})Spin-rotation parts of G\mathcal{G}
Spin point group U(G)\mathcal{U}(\mathcal{G})Pairs of spatial and spin rotations of G\mathcal{G}
M\mathcal{M}Magnetic space group

Table 1. Notation and terminology in this paper.

Group structure of spin space group

We provide the definitions of spin symmetry operations and spin arrangements in Sec. 2.1. We define the spin space group in Sec. 2.2. Then, we introduce its derived groups, the spin-only group (Sec. 2.3) and spin translation group (Sec. 2.4). Although these groups were already discussed in Litvin [12], Litvin & Opechowski [11], and Litvin [13], we consider it beneficial to summarize these results because we fully exploit the group structure of the spin space group in searching for spin symmetry operations.

We note that spin-only groups and spin translation groups complicate the group structure of the spin space groups. For example, a spin point group, which ignores translation parts of spin symmetry operations of a spin space group, cannot be computed without going through the spin space group due to the existence of a nontrivial spin translation group in general. Although the analysis of spin point groups is not required to determine spin symmetry operations, we discuss the group structure of spin point groups in Appendix B for completeness.

Spin symmetry operation and spin arrangement

A spin symmetry operation comprises a spatial operation in the three-dimensional Euclidean group E(3)\mathrm{E}(3) and a spin rotation in the three-dimensional orthogonal group O(3)\mathrm{O}(3). The spin symmetry operation (g,W)∈E(3)×O(3)(g,\mathbf{W}) \in\mathrm{E}(3) \times\mathrm{O}(3) acts on a pair of position r\mathbf{r} and magnetic moments m\mathbf{m} as

(g,W)(r,m)=(gr,Wm).(1)(g,\mathbf{W})(\mathbf{r},\mathbf{m})=(g\mathbf{r},\mathbf{W}\mathbf{m}). \tag*{(1)}

The product of two spin symmetry operations (g,W)(g,\mathbf{W}) and (g′,W′)(g',\mathbf{W}') is defined as

(g,W)(g′,W′)=(gg′,WW′).(2)(g,\mathbf{W})(g',\mathbf{W}')=(gg',\mathbf{W}\mathbf{W}'). \tag*{(2)}

Although uncommon in crystallography, we suppose both gg and WW are represented with Cartesian coordinates for later convenience. We denote basis vectors

A=(a1,a2,a3)=(a1xa2xa3xa1ya2ya3ya1za2za3z).(3)\mathbf{A}=(\mathbf{a}_1,\mathbf{a}_2,\mathbf{a}_3)=\begin{pmatrix}a_{1x} & a_{2x} & a_{3x}\\a_{1y} & a_{2y} & a_{3y}\\a_{1z} & a_{2z} & a_{3z}\end{pmatrix}. \tag*{(3)}

When gg with a matrix part R\mathbf{R} and a translation part v\mathbf{v} are represented with A\mathbf{A}, we explicitly write g=(R‾,v‾)Ag=(\overline{\mathbf{R}},\overline{\mathbf{v}})_{\mathbf{A}}, where R=AR‾A−1\mathbf{R}=\mathbf{A}\overline{\mathbf{R}}\mathbf{A}^{-1} and v=Av‾\mathbf{v}=\mathbf{A}\overline{\mathbf{v}}. A spin arrangement is a set of pairs of a crystal structure and magnetic moments.

Spin space group

Let G\mathcal{G} be a subgroup of E(3)×O(3)\mathrm{E}(3)\times\mathrm{O}(3). When the following F(G)\mathcal{F}(\mathcal{G}) and D(G)\mathcal{D}(\mathcal{G}) are space groups, G\mathcal{G} is called a spin space group [11],

F(G)={g∈E(3)∣∃W∈O(3) s.t. (g,W)∈G}(4)\mathcal{F}(\mathcal{G})=\{g\in\mathrm{E}(3)\mid\exists W\in\mathrm{O}(3)\ \text{s.t.}\ (g,W)\in\mathcal{G}\} \tag*{(4)}
D(G)={g∈E(3)∣(g,E)∈G},(5)\mathcal{D}(\mathcal{G})=\{g\in\mathrm{E}(3)\mid(g,E)\in\mathcal{G}\}, \tag*{(5)}

where EE stands for the identity matrix. For a spin space group G\mathcal{G}, we call F(G)\mathcal{F}(\mathcal{G}) a family space group and D(G)\mathcal{D}(\mathcal{G}) a maximal space subgroup. The maximal space subgroup D(G)\mathcal{D}(\mathcal{G}) is a normal subgroup of G\mathcal{G}. On the contrary, the family space group F(G)\mathcal{F}(\mathcal{G}) is not a subgroup of G×O(3)\mathcal{G}\times\mathrm{O}(3) in general. Although Litvin & Opechowski [11] did not impose the condition that D(G)\mathcal{D}(\mathcal{G}) is crystallographic, we impose this condition to guarantee that a given spin arrangement is commensurate. We confine our discussion and algorithms to commensurate spin arrangements.

A spin symmetry operation (g,W)(g,W) is transformed by a transformation (P,p)(P,p) on the spatial coordinates and a transformation matrix QQ on the spin coordinates as

(g,W)↦((P,p)−1g(P,p),Q−1WQ).(6)(g,W)\mapsto\left((P,p)^{-1}g(P,p),Q^{-1}WQ\right). \tag*{(6)}

Two spin space groups G1\mathcal{G}_1 and G2\mathcal{G}_2 belong to the same spin-space-group type if they are transformed to the other by a pair of an orientation-preserving transformation (P,p)(P,p)1 on the spatial coordinates and a transformation matrix Q\mathbf{Q} on the spin coordinates. In the spin space group, we can consider rotating magnetic moments independently with spatial coordinates. On the other hand, we consider rotating magnetic moments only in association with spatial rotations and time-reversal operations in the magnetic space group [14]. Thus, the spin space group can be regarded as a supergroup of the magnetic space group under an appropriate correspondence between spin symmetry operations and magnetic symmetry operations as discussed in Appendix § 6.

Spin-only group

A spin-only group of a spin space group G\mathcal{G} is a set of spin symmetry operations of G\mathcal{G} with identity spatial operations,

Pso(G)={W∈O(3)∣((E,0),W)∈G}.(7)\mathcal{P}_{\mathrm{so}}(\mathcal{G}) = \{\mathbf{W} \in O(3) \mid((\mathbf{E},0),\mathbf{W}) \in\mathcal{G}\}. \tag*{(7)}

A direct product of an identity in spatial coordinates and Pso(G)\mathcal{P}_{\mathrm{so}}(\mathcal{G}), 1×Pso(G)1 \times\mathcal{P}_{\mathrm{so}}(\mathcal{G}), is a subgroup of G2\mathcal{G}^{2}.

Spin translation group

A spin translation group of a spin space group G\mathcal{G} is a set of spin symmetry operations with identity rotations in spatial coordinates,

Gst(G)={((E,v),W)∣((E,v),W)∈G}.(8)\mathcal{G}_{\mathrm{st}}(\mathcal{G}) = \{((\mathbf{E},\mathbf{v}),\mathbf{W}) \mid((\mathbf{E},\mathbf{v}),\mathbf{W}) \in\mathcal{G}\}. \tag*{(8)}

Litvin [12] classified the spin translation groups under the transformation in Eq. (6). We denote a translation subgroup and a point group of space group R\mathcal{R} as T(R)\mathcal{T}(\mathcal{R}) and P(R)\mathcal{P}(\mathcal{R}), respectively,

T(R)={(E,t)∣(E,t)∈R}(9)\mathcal{T}(\mathcal{R}) = \{(\mathbf{E},\mathbf{t}) \mid(\mathbf{E},\mathbf{t}) \in\mathcal{R}\} \tag*{(9)}
P(R)={W∣∃v s.t. (W,v)∈R}.(10)\mathcal{P}(\mathcal{R}) = \{\mathbf{W} \mid\exists\mathbf{v}\ \mathrm{s.t.}\ (\mathbf{W},\mathbf{v}) \in\mathcal{R}\}. \tag*{(10)}

Then, the spin-only group Pso(G)\mathcal{P}_{\mathrm{so}}(\mathcal{G}) and translation subgroup of D(G)\mathcal{D}(\mathcal{G}) are a normal subgroup of Gst(G)\mathcal{G}_{\mathrm{st}}(\mathcal{G}). Because (T(D(G))×1)∩(1×Pso(G))=1(\mathcal{T}(\mathcal{D}(\mathcal{G})) \times1) \cap(1 \times\mathcal{P}_{\mathrm{so}}(\mathcal{G})) = 1, their direct product T(D(G))×Pso(G)\mathcal{T}(\mathcal{D}(\mathcal{G})) \times\mathcal{P}_{\mathrm{so}}(\mathcal{G}) is also a normal subgroup of Gst(G)\mathcal{G}_{\mathrm{st}}(\mathcal{G}). Thus, we can consider a factor group Gst(G)/(T(D(G))×Pso(G))\mathcal{G}_{\mathrm{st}}(\mathcal{G})/(\mathcal{T}(\mathcal{D}(\mathcal{G})) \times\mathcal{P}_{\mathrm{so}}(\mathcal{G})), which is finite for commensurate spin arrangements. Finally, Gst(G)\mathcal{G}_{\mathrm{st}}(\mathcal{G}) is a normal subgroup of G\mathcal{G}. The factor group G/Gst(G)\mathcal{G}/\mathcal{G}_{\mathrm{st}}(\mathcal{G}) is isomorphic to P(F(G))\mathcal{P}(\mathcal{F}(\mathcal{G})) and thus finite.

Spin symmetry operation search

We provide an algorithm to search for spin symmetry operations from a given spin arrangement represented by the following four objects: (1) basis vectors of its lattice A=(a1,a2,a3)\boldsymbol{A} = (\boldsymbol{a}_1,\boldsymbol{a}_2,\boldsymbol{a}_3), (2) an array of point coordinates of sites in its unit cell X=(x1,⋯ ,xN)\boldsymbol{X} = (\boldsymbol{x}_1,\cdots,\boldsymbol{x}_N), (3) an array of atomic types of sites in its unit cell T=(t1,⋯ ,tN)\boldsymbol{T} = (t_1,\cdots,t_N), and (4) an array of magnetic moments of sites in its unit cell M=(m1,⋯ ,mN)\boldsymbol{M} = (\boldsymbol{m}_1,\cdots,\boldsymbol{m}_N), where NN is the number of sites in the unit cell.

To search for spin symmetry operations robustly, comparisons of point coordinates and magnetic moments should be performed within tolerances in practice. We adopt the same absolute tolerance parameter for point coordinates as Togo & Tanaka [21]. For magnetic moments, we use another tolerance parameter ϵmag\epsilon_{\mathrm{mag}} to identify that two magnetic moments Wmi\boldsymbol{W}\boldsymbol{m}_i and mσg(i)\boldsymbol{m}_{\sigma_g(i)} are equal to the other if

∥Wmi−mσg(i)∥2<ϵmag.(11)\left\lVert\boldsymbol{W}\boldsymbol{m}_i-\boldsymbol{m}_{\sigma_g(i)}\right\rVert_2 < \epsilon_{\mathrm{mag}}. \tag*{(11)}

The present algorithm extends the authors’ previous work on detecting magnetic symmetry operations ([19]). We first consider a space group of nonmagnetic crystal structure (A,X,T)(\boldsymbol{A},\boldsymbol{X},\boldsymbol{T}) in § 3.1. Next, we search for normal subgroups of the spin space group: the spin-only group Pso(G)\mathcal{P}_{\mathrm{so}}(\mathcal{G}) in § 3.2 and the translation subgroup T(D(G))\mathcal{T}(\mathcal{D}(\mathcal{G})) in § 3.3. With these normal subgroups of Gst(G)\mathcal{G}_{\mathrm{st}}(\mathcal{G}), we search for coset representatives of the spin translation group Gst(G)/(T(D(G))×Pso(G))\mathcal{G}_{\mathrm{st}}(\mathcal{G})/(\mathcal{T}(\mathcal{D}(\mathcal{G})) \times\mathcal{P}_{\mathrm{so}}(\mathcal{G})) in Sec. 3.4. We search for coset representatives of the spin space group G/Gst(G)\mathcal{G}/\mathcal{G}_{\mathrm{st}}(\mathcal{G}) in Sec. 3.5. Because the notation in this section is abstract to unambiguously deal with several basis vectors, it may be helpful to read it alongside the examples in Sec. 4.

Space group of nonmagnetic crystal structure

A candidate for spatial operations of G\mathcal{G} can be derived from symmetry operations of a crystal structure (A,X,T)(\mathbf{A},\mathbf{X},\mathbf{T}) ignoring the magnetic moments. We also employ the space group S\mathcal{S} given as a stabilizer of E(3)\mathrm{E}(3) preserving (A,X,T)(\mathbf{A},\mathbf{X},\mathbf{T}):

S={g=(R‾,v‾)A∈E(3) | ∃σg∈SN, ∀i=1,⋯N,R‾xi+v‾≡xσg(i)(mod1)ti=tσg(i)},(12)\mathcal{S} = \left\{ g = (\overline{R},\overline{v})_{\mathbf{A}} \in\mathrm{E}(3) \,\middle|\, \begin{gathered} \exists\sigma_g \in\mathcal{S}_N,\ \forall i = 1,\cdots N, \\ \overline{R}\mathbf{x}_i+\overline{v}\equiv\mathbf{x}_{\sigma_g(i)}\pmod{1} \\ t_i=t_{\sigma_g(i)} \end{gathered} \right\}, \tag*{(12)}

where SN\mathcal{S}_N is a symmetric group of degree NN. Note that g=(R‾,v‾)Ag = (\overline{R},\overline{v})_{\mathbf{A}} maps point coordinates xi\mathbf{x}_i to R‾xi+v‾\overline{R}\mathbf{x}_i+\overline{v}. The mapped point coordinates coincide with point coordinates in X\mathbf{X} up to modulo one, inducing permutation σg\sigma_g.

The existing crystal symmetry search algorithm {(citep}(Togo & Tanaka, 2018)} can find primitive basis vectors AS\mathbf{A}_{\mathcal{S}} of T(S)\mathcal{T}(\mathcal{S}) and coset decomposition of S\mathcal{S} over TA\mathcal{T}_{\mathbf{A}}, where we write a translation subgroup formed by A\mathbf{A} as

TA={(E,n)A∣n∈Z3}.(13)\mathcal{T}_{\mathbf{A}}=\left\{(E,\mathbf{n})_{\mathbf{A}}\mid\mathbf{n}\in\mathbb{Z}^{3}\right\}. \tag*{(13)}

The input basis vector A\mathbf{A} can be represented as an integer linear combination of AS\mathbf{A}_{\mathcal{S}}. Thus, an integer matrix U∈Z3×3U\in\mathbb{Z}^{3\times3} exists such that A=ASU\mathbf{A}=\mathbf{A}_{\mathcal{S}}U. The translation subgroup TAS\mathcal{T}_{\mathbf{A}_{\mathcal{S}}} is decomposed as

TAS=∐t‾S(E,t‾S)ASTA,(14)\mathcal{T}_{\mathbf{A}_{\mathcal{S}}}=\coprod_{\overline{t}^{\mathcal{S}}}\left(E,\overline{t}^{\mathcal{S}}\right)_{\mathbf{A}_{\mathcal{S}}}\mathcal{T}_{\mathbf{A}}, \tag*{(14)}

where t‾S\overline{t}^{\mathcal{S}} is a centering vector in a unit cell spanned by A\mathbf{A}. The coset decomposition of S\mathcal{S} is written as

S=∐R‾S(R‾S,v‾R‾S)ASTAS.(15)\mathcal{S}=\coprod_{\overline{R}^{\mathcal{S}}}\left(\overline{R}^{\mathcal{S}},\overline{v}_{\overline{R}^{\mathcal{S}}}\right)_{\mathbf{A}_{\mathcal{S}}}\mathcal{T}_{\mathbf{A}_{\mathcal{S}}}. \tag*{(15)}

Here, v‾R‾S\overline{v}_{\overline{R}^{S}} is a translation part of a symmetry operation with a matrix part R‾S\overline{R}^{S}.

For the spin arrangement (A,X,T,M)(\mathbf{A},\mathbf{X},\mathbf{T},\mathbf{M}), the spin space group can be expressed as a stabilizer of S×O(3)\mathcal{S} \times\mathrm{O}(3) that preserves (A,X,T,M)(\mathbf{A},\mathbf{X},\mathbf{T},\mathbf{M}),

G={(g,W)∈S×O(3)∣Wmi=mσg(i) (i=1,…,N)}.(16)\mathcal{G} = \left\{(g,W) \in\mathcal{S} \times\mathrm{O}(3) \mid W\mathbf{m}_i = \mathbf{m}_{\sigma_g(i)}\ (i=1,\ldots,N)\right\}. \tag*{(16)}

This expression serves as the starting point for the spin symmetry operation search.

For notation simplicity, we denote the maximal space subgroup of G\mathcal{G} as D=D(G)\mathcal{D} = \mathcal{D}(\mathcal{G}).

Spin-only group search

Because a symmetry operation in the spin-only group of G\mathcal{G} does not change the order of point coordinates, the spin-only group of (A,X,T,M)(\mathbf{A},\mathbf{X},\mathbf{T},\mathbf{M}) is expressed as

Pso={W∈O(3)∣Wmi=mi (i=1,…,N)}.(17)\mathcal{P}_{\mathrm{so}} = \left\{W \in\mathrm{O}(3) \mid W\mathbf{m}_i = \mathbf{m}_i\ (i=1,\ldots,N)\right\}. \tag*{(17)}

As shown in Table 2, when a spin space group G\mathcal{G} is a stabilizer of a spin arrangement, spin-only groups are classified into four types up to transformations [11, 15]: nonmagnetic, collinear, coplanar, and noncoplanar spin arrangements.

Spin arrangementSpin-only groupEigenvalues of MMTMM^{\mathsf{T}}
Nonmagnetic∞∞m≅O(3)\infty\infty m \cong O(3)σ1=σ2=σ3=0\sigma_1=\sigma_2=\sigma_3=0
Collinear∞m≅SO⁡(2)⋊Z2\infty m \cong\operatorname{SO}(2)\rtimes\mathbb{Z}_2σ1>σ2=σ3=0\sigma_1>\sigma_2=\sigma_3=0
Coplanarmmσ1≥σ2>σ3=0\sigma_1\ge\sigma_2>\sigma_3=0
Noncoplanar11σ1≥σ2≥σ3>0\sigma_1\ge\sigma_2\ge\sigma_3>0

Table 2. Classification of spin-only groups up to transformations. The spin-only groups are represented in Hellmann–Mauguin symbols [8].

In § 3.2.1, we provide an algorithm for detecting Pso\mathcal{P}_{\mathrm{so}} using the eigenvalue decomposition of MMTMM^{\mathsf{T}}. The spin-only group search should be performed within tolerances for magnetic moments in practice. In fact, selecting appropriate tolerances is challenging. In § 3.2.2, we propose a robust algorithm, building on the previous section, to alleviate the difficulty.

Spin-only group search by eigenvalue decomposition

We consider a moment tensor of M∈R3×NM \in\mathbb{R}^{3 \times N},

MMT=∑i=1Nmi⊗mi.(18)MM^{\mathsf{T}} = \sum_{i=1}^{N} \mathbf{m}_i \otimes\mathbf{m}_i. \tag*{(18)}

Because MMTM M^{\mathsf{T}} is a symmetric semi-definite matrix, we can consider its eigenvalue decomposition,

MMT=∑r=13σrn^r⊗n^r,(19)M M^{\mathsf{T}} = \sum_{r=1}^{3} \sigma_r \hat{\mathbf{n}}_r \otimes\hat{\mathbf{n}}_r, \tag*{(19)}

where σ1≥σ2≥σ3≥0\sigma_1 \ge\sigma_2 \ge\sigma_3 \ge0 and {n^r}r=13\{\hat{\mathbf{n}}_r\}_{r=1}^{3} are orthonormal.

The spin-only group is classified based on the eigenvalues {σi}i=13\{\sigma_i\}_{i=1}^{3}, as summarized in Table 2. In a nonmagnetic spin arrangement, all magnetic moments are zero. In this case, all eigenvalues are zero, σ1=σ2=σ3=0\sigma_1 = \sigma_2 = \sigma_3 = 0. In a collinear spin arrangement, all magnetic moments align parallel or antiparallel to a direction n^∥\hat{\mathbf{n}}_{\parallel}. The eigenvector n^1\hat{\mathbf{n}}_1 with the largest eigenvalue should be parallel or antiparallel to n^∥\hat{\mathbf{n}}_{\parallel} and the other eigenvalues σ2\sigma_2 and σ3\sigma_3 should be zero. In a coplanar spin arrangement, all magnetic moments align perpendicular to a direction n^⊥\hat{\mathbf{n}}_{\perp}. In this case, the eigenvector n^3\hat{\mathbf{n}}_3 with the smallest eigenvalue should be parallel or antiparallel to n^⊥\hat{\mathbf{n}}_{\perp} and the smallest eigenvalue σ3\sigma_3 should be zero. In other cases, the spin arrangement is noncoplanar. Note that the site order of magnetic moments MM does not affect the classification because MMTM M^{\mathsf{T}} in Eq. (18) remains invariant under a permutation of the site order.

Numerically robust spin-only group search

The spin-group search algorithm in the previous section requires judging whether the eigenvalues are zero or positive, which needs an additional tolerance parameter. To reduce the number of tolerance parameters for usability, we modify the spin-group search algorithm solely with the tolerance ϵmag\epsilon_{\mathrm{mag}} in Eq. (11) as follows.

  1. We compute the eigenvalues σi\sigma_i and eigenvectors n^i\hat{\mathbf{n}}_i in Eq. (19).

  1. If all magnetic moments are close to zero within ϵmag\epsilon_{\mathrm{mag}},

∥mi∥2<ϵmag(i=1,…,N),(20)\lVert\mathbf{m}_i \rVert_2 < \epsilon_{\mathrm{mag}} \qquad(i = 1,\ldots,N), \tag*{(20)}

the spin arrangement is nonmagnetic.

  1. If not, we check if the eigenvector n^1\hat{n}_1 is a parallel or antiparallel direction for all magnetic moments,

2∥mi−(mi⋅n^1)n^1∥2<ϵmag(∀i=1,…,N).(21)2 \left\lVert\mathbf{m}_i - (\mathbf{m}_i \cdot\hat{n}_1)\hat{n}_1 \right\rVert_2 < \epsilon_{\mathrm{mag}} \quad(\forall i = 1,\ldots,N). \tag*{(21)}

When spin symmetry operations in a collinear spin-only group prescribed by a direction n^1\hat{n}_1 act on mi\mathbf{m}_i, the acted magnetic moments draw a cone as shown in Figure 1 (a). The left-hand side of the above inequality is the largest displacement between the acted magnetic moments, which corresponds to the diameter of the cone with direction n^1\hat{n}_1. If the inequality holds for all magnetic moments, the spin arrangement is collinear.

Image not extracted

Figure 1. Magnetic moments acted by (a) collinear and (b) coplanar spin-only groups. (a) A collinear spin-only group along axis n^1\hat{\mathbf{n}}_1 is generated from rotations along n^1\hat{\mathbf{n}}_1 and mirror operations preserving n^1\hat{\mathbf{n}}_1. The red dotted line indicates the largest displacement between magnetic moments acted the collinear spin-only group. (b) A coplanar spin-only group along axis n^3\hat{\mathbf{n}}_3 is generated from a mirror operation perpendicular to n^3\hat{\mathbf{n}}_3. The red dotted line indicates the largest displacement between magnetic moments acted the coplanar spin-only group.

  1. If not, we check if the eigenvector n^3\hat{n}_3 is a perpendicular direction for all magnetic moments,

2∣(mi⋅n^3)∣<ϵmag(∀i=1,…,N).(22)2 \left|(\mathbf{m}_i \cdot\hat{n}_3)\right| < \epsilon_{\mathrm{mag}} \quad(\forall i = 1,\ldots,N). \tag*{(22)}

The left-hand side of the above inequality is the largest displacement between magnetic moments acted by a coplanar spin-only group along n^3\hat{n}_3 as shown in Figure 1 (b). If the inequality holds for all magnetic moments, the spin arrangement is coplanar.

  1. Otherwise, the spin arrangement is noncoplanar.

Translation subgroup search

We search for primitive basis vectors AD\mathbf{A}_D of the translation subgroup T(D(G))={(E,v)∣((E,v),E)∈G}\mathcal{T}(D(\mathcal{G})) = \{(\mathbf{E},\mathbf{v}) \mid((\mathbf{E},\mathbf{v}),\mathbf{E}) \in\mathcal{G}\}. The group–subgroup relationships of translation subgroups are shown in Figure 2. Because TAD\mathcal{T}_{A_D} is between TAS\mathcal{T}_{A_S} and TA\mathcal{T}_A, TA⊴TAD⊴TAS\mathcal{T}_A \trianglelefteq\mathcal{T}_{A_D} \trianglelefteq\mathcal{T}_{A_S}, we only need to examine finite coset representatives of TAS/TA\mathcal{T}_{A_S}/\mathcal{T}_A for candidates of TAS/TAD\mathcal{T}_{A_S}/\mathcal{T}_{A_D}.

Image not extracted

Figure 2. Group–subgroup relationship of translation subgroups and space groups derived from spin space group. The nodes represent translation subgroups and space groups. Each edge indicates that a lower group is a subgroup of an upper group in a diagram. Note that SDTAs\mathcal{S}_{D}\mathcal{T}_{A_s} could be a proper subgroup of S\mathcal{S}.

For every coset (E,tS)ASTA\left(\mathbf{E},\mathbf{t}^{S}\right)_{A_S}\mathcal{T}_A in Eq. (14), we check if (E,tS)AS\left(\mathbf{E},\mathbf{t}^{S}\right)_{A_S} preserves magnetic moments M\mathbf{M},

TAD={g=(E,tˉS)AS | gTA∈TAS/TA, mi=mσg(i) (i=1,…,N)}TA.(23)\mathcal{T}_{AD}=\left\{g=\left(E,\bar{\mathbf{t}}^{S}\right)_{AS}\ \middle|\ g^{T}A\in\mathcal{T}_{AS}/\mathcal{T}_{A},\ \mathbf{m}_{i}=\mathbf{m}_{\sigma_{g}(i)}\ (i=1,\ldots,N)\right\}\mathcal{T}_{A}. \tag*{(23)}

An integer matrix V∈Z3×3\mathbf{V}\in\mathbb{Z}^{3\times3} exists such that AD=ASVA_{D}=A_{S}\mathbf{V}. A lattice algorithm to generate V\mathbf{V} from centering vectors tˉS\bar{\mathbf{t}}^{S} in Eq. (23) is presented in Appendix C.

Spin translation group search

We can find a spin rotation W\mathbf{W} for a given symmetry operation g∈Sg\in S by solving a Procrustes problem [7] as follows. We write magnetic moments permuted by gg as

Mg=(mσg(1),⋯ ,mσg(N)),(24)\mathbf{M}_{g}=\left(\mathbf{m}_{\sigma_{g}(1)},\cdots,\mathbf{m}_{\sigma_{g}(N)}\right), \tag*{(24)}

where σg\sigma_{g} is a permutation of NN sites induced by gg. Rather than directly searching for W∈O(3)\mathbf{W}\in O(3) such that Wmi=mσg(i)\mathbf{W}\mathbf{m}_{i}=\mathbf{m}_{\sigma_{g}(i)}, we choose a candidate W~\widetilde{\mathbf{W}} by solving the following Procrustes problem:

W~=arg⁡min⁡W∈O(3)∥Mg−WM∥F.(25)\widetilde{\mathbf{W}}=\underset{\mathbf{W}\in O(3)}{\arg\min}\left\lVert\mathbf{M}_{g}-\mathbf{W}\mathbf{M}\right\rVert_{F}. \tag*{(25)}

Here the displacement between the magnetic moments and operated ones by (g,W)(g,\mathbf{W}) is measured by the Frobenius norm ∥⋅∥F\lVert\cdot\rVert_{F}. The solution to Eq. (25) can be explicitly written as

W~=YZT,(26)\widetilde{\mathbf{W}}=\mathbf{Y}\mathbf{Z}^{\mathsf{T}}, \tag*{(26)}

where Y\mathbf{Y} and Z\mathbf{Z} are orthogonal matrices of the singular value decomposition of MgMT\mathbf{M}_{g}\mathbf{M}^{\mathsf{T}},

MgMT=YΣZT.(27)\mathbf{M}_{g}\mathbf{M}^{\mathsf{T}}=\mathbf{Y}\boldsymbol{\Sigma}\mathbf{Z}^{\mathsf{T}}. \tag*{(27)}

After we obtain W~\widetilde{\mathbf{W}}, (g,W~)(g,\widetilde{\mathbf{W}}) is taken as a spin symmetry operation if the condition

∥mi−W~mσg(i)∥2<ϵmag(28)\left\lVert\mathbf{m}_{i}-\widetilde{\mathbf{W}}\mathbf{m}_{\sigma_{g}(i)}\right\rVert_{2}<\epsilon_{\mathrm{mag}} \tag*{(28)}

holds for every site ii. If Eq. (28) does not hold for some site ii, we reject (g,W~)(g,\widetilde{W}) as a spin symmetry operation.

For coset representatives of (E,t‾D)ADTAD∈TAS/TAD(E,\overline{t}^{D})_{A_D}\mathcal{T}_{A_D}\in\mathcal{T}_{A_S}/\mathcal{T}_{A_D}, we search for a corresponding spin rotation Wt‾DW_{\overline{t}^{D}} using the above algorithm if it exists. The coset decomposition of Gst\mathcal{G}_{\mathrm{st}} can be written as

Gst={g=((E,t‾D)AD,Wt‾D) | (E,t‾D)ADTAD∈TAS/TADWt‾Dmi=mσg(i)  (∀i=1,…,N)}(TAD×Pso),(29)\mathcal{G}_{\mathrm{st}}=\left\{g=\left((E,\overline{t}^{D})_{A_D},W_{\overline{t}^{D}}\right)\ \middle|\ \begin{array}{l}(E,\overline{t}^{D})_{A_D}\mathcal{T}_{A_D}\in\mathcal{T}_{A_S}/\mathcal{T}_{A_D}\\ W_{\overline{t}^{D}}m_i=m_{\sigma_g(i)}\;(\forall i=1,\ldots,N)\end{array}\right\}\left(\mathcal{T}_{A_D}\times\mathcal{P}_{\mathrm{so}}\right), \tag*{(29)}

where t‾D\overline{t}^{D} takes a distinct translation up to TAD\mathcal{T}_{A_D} and we choose Wt‾D=0=EW_{\overline{t}^{D}=0}=E.

Because the translation subgroup of F(G)\mathcal{F}(\mathcal{G}) is TAD\mathcal{T}_{A_D}, spatial operation parts of G\mathcal{G} should belong to a maximal subgroup SD\mathcal{S}_D of S\mathcal{S} with its translation subgroup TAD\mathcal{T}_{A_D}. Figure 2 shows the group–subgroup relationship between S\mathcal{S} and SD\mathcal{S}_D. A rotation ASR‾SAS−1∈P(S)A_S\overline{R}^{S}A_S^{-1}\in\mathcal{P}(\mathcal{S}) is compatible with TAD\mathcal{T}_{A_D} if V−1R‾SVV^{-1}\overline{R}^{S}V is an integer matrix [9]. Thus, the compatible subgroup SD\mathcal{S}_D is written as

SD={(R‾S,v‾R‾S+t‾R‾S)AS | (R‾S,v‾R‾S)ASTAS∈S/TASV−1R‾SV∈Z3×3(E,t‾R‾S)ASTAS∈TAS/TA}TAD,(30)\mathcal{S}_D=\left\{\left(\overline{R}^{S},\overline{v}_{\overline{R}^{S}}+\overline{t}_{\overline{R}^{S}}\right)_{A_S}\ \middle|\ \begin{array}{l}\left(\overline{R}^{S},\overline{v}_{\overline{R}^{S}}\right)_{A_S}\mathcal{T}_{A_S}\in\mathcal{S}/\mathcal{T}_{A_S}\\ V^{-1}\overline{R}^{S}V\in\mathbb{Z}^{3\times3}\\ (E,\overline{t}_{\overline{R}^{S}})_{A_S}\mathcal{T}_{A_S}\in\mathcal{T}_{A_S}/\mathcal{T}_A\end{array}\right\}\mathcal{T}_{A_D}, \tag*{(30)}

where t‾R‾S\overline{t}_{\overline{R}^{S}} is a centering vector in TAS/TA\mathcal{T}_{A_S}/\mathcal{T}_A. The additional centering vector t‾R‾S\overline{t}_{\overline{R}^{S}} is necessary for SD\mathcal{S}_D to form a group [16, 20].

For a coset representative of SD/TAD\mathcal{S}_D/\mathcal{T}_{A_D}, a corresponding spin-rotation part WR‾SW_{\overline{R}^{S}} can also be determined by solving the Procrustes problem as presented in Sec. 3.4. Finally, we obtain all spin symmetry operations in a coset decomposition

G={((R‾S,v‾R‾S+t‾R‾S)AS,WR‾S) | (R‾S,v‾R‾S+t‾R‾S)ASTAD∈SD/TADWR‾Smi=mσg(i)  (∀i=1,…,N)}Gst.(31)\mathcal{G}=\left\{\left(\left(\overline{R}^{S},\overline{v}_{\overline{R}^{S}}+\overline{t}_{\overline{R}^{S}}\right)_{A_S},W_{\overline{R}^{S}}\right)\ \middle|\ \begin{array}{l}\left(\overline{R}^{S},\overline{v}_{\overline{R}^{S}}+\overline{t}_{\overline{R}^{S}}\right)_{A_S}\mathcal{T}_{A_D}\in\mathcal{S}_D/\mathcal{T}_{A_D}\\ W_{\overline{R}^{S}}m_i=m_{\sigma_g(i)}\;(\forall i=1,\ldots,N)\end{array}\right\}\mathcal{G}_{\mathrm{st}}. \tag*{(31)}

Examples of spin symmetry operation search

We consider a spin arrangement of the NiAs-type CrSe ([4, 12]) (#2.35 of MAGNDATA ([6])) as an example of spin symmetry operation search, illustrated in Figure 3. The hexagonal lattice of CrSe has the following basis vectors

Image not extracted

Figure 3. Spin arrangement example for NiAs-type CrSe. The gray and orange balls represent Cr and Se atoms, respectively. The red arrows denote magnetic moments of Cr atoms with equal magnitudes.

A=(a−12a0032a000c).A = \begin{pmatrix} a & -\frac{1}{2}a & 0 \\ 0 & \frac{\sqrt{3}}{2}a & 0 \\ 0 & 0 & c \end{pmatrix}.

The fractional coordinates and magnetic moments of atoms are as follows.

Cr:x1=(0,0,0)T, m1=(−12mx,−32mx,−mz)T\mathrm{Cr}: \boldsymbol{x}_{1} = (0,0,0)^{\mathrm{T}},\ \boldsymbol{m}_{1} = \left(-\frac{1}{2}m_x,-\frac{\sqrt{3}}{2}m_x,-m_z\right)^{\mathrm{T}}
Cr:x2=(13,23,0)T, m2=(mx,0,−mz)T\mathrm{Cr}: \boldsymbol{x}_{2} = \left(\frac{1}{3},\frac{2}{3},0\right)^{\mathrm{T}},\ \boldsymbol{m}_{2} = (m_x,0,-m_z)^{\mathrm{T}}
Cr:x3=(23,13,0)T, m3=(−12mx,32mx,−mz)T\mathrm{Cr}: \boldsymbol{x}_{3} = \left(\frac{2}{3},\frac{1}{3},0\right)^{\mathrm{T}},\ \boldsymbol{m}_{3} = \left(-\frac{1}{2}m_x,\frac{\sqrt{3}}{2}m_x,-m_z\right)^{\mathrm{T}}
Cr:x4=(0,0,12)T, m4=(12mx,32mx,mz)T\mathrm{Cr}: \boldsymbol{x}_{4} = \left(0,0,\frac{1}{2}\right)^{\mathrm{T}},\ \boldsymbol{m}_{4} = \left(\frac{1}{2}m_x,\frac{\sqrt{3}}{2}m_x,m_z\right)^{\mathrm{T}}
Cr:x5=(13,23,12)T, m5=(−mx,0,mz)T\mathrm{Cr}: \boldsymbol{x}_{5} = \left(\frac{1}{3},\frac{2}{3},\frac{1}{2}\right)^{\mathrm{T}},\ \boldsymbol{m}_{5} = (-m_x,0,m_z)^{\mathrm{T}}
Cr:x6=(23,13,12)T, m6=(12mx,−32mx,mz)T\mathrm{Cr}: \boldsymbol{x}_{6} = \left(\frac{2}{3},\frac{1}{3},\frac{1}{2}\right)^{\mathrm{T}},\ \boldsymbol{m}_{6} = \left(\frac{1}{2}m_x,-\frac{\sqrt{3}}{2}m_x,m_z\right)^{\mathrm{T}}
Se:x7=(0,13,14)T, m7=(0,0,0)T\mathrm{Se}: \boldsymbol{x}_{7} = \left(0,\frac{1}{3},\frac{1}{4}\right)^{\mathrm{T}},\ \boldsymbol{m}_{7} = (0,0,0)^{\mathrm{T}}
Se:x8=(13,0,14)T, m8=(0,0,0)T\mathrm{Se}: \boldsymbol{x}_{8} = \left(\frac{1}{3},0,\frac{1}{4}\right)^{\mathrm{T}},\ \boldsymbol{m}_{8} = (0,0,0)^{\mathrm{T}}
Se:x9=(13,13,34)T, m9=(0,0,0)T\mathrm{Se}: \boldsymbol{x}_{9} = \left(\frac{1}{3},\frac{1}{3},\frac{3}{4}\right)^{\mathrm{T}},\ \boldsymbol{m}_{9} = (0,0,0)^{\mathrm{T}}
Se:x10=(0,23,34)T, m10=(0,0,0)T\mathrm{Se}: \boldsymbol{x}_{10} = \left(0,\frac{2}{3},\frac{3}{4}\right)^{\mathrm{T}},\ \boldsymbol{m}_{10} = (0,0,0)^{\mathrm{T}}
Se:x11=(23,0,34)T, m11=(0,0,0)T\mathrm{Se}: \boldsymbol{x}_{11} = \left(\frac{2}{3},0,\frac{3}{4}\right)^{\mathrm{T}},\ \boldsymbol{m}_{11} = (0,0,0)^{\mathrm{T}}
Se:x12=(23,23,14)T, m12=(0,0,0)T\mathrm{Se}: \boldsymbol{x}_{12} = \left(\frac{2}{3},\frac{2}{3},\frac{1}{4}\right)^{\mathrm{T}},\ \boldsymbol{m}_{12} = (0,0,0)^{\mathrm{T}}

Here, the magnetic moments are represented with Cartesian coordinates.

Space group of nonmagnetic crystal structure

The space-group type of S\mathcal{S} for the crystal structure ignoring magnetic moments is P63/mmc\mathrm{P}6_{3}/\mathrm{mmc} (No. 194). One of the primitive basis vectors for S\mathcal{S} and the transformation matrix are

AS=(0−12a0−33a36a000−c)A_{\mathcal{S}} = \begin{pmatrix} 0 & -\frac{1}{2}a & 0 \\ -\frac{\sqrt{3}}{3}a & \frac{\sqrt{3}}{6}a & 0 \\ 0 & 0 & -c \end{pmatrix}
U=(−1−10−21000−1)U = \begin{pmatrix} -1 & -1 & 0 \\ -2 & 1 & 0 \\ 0 & 0 & -1 \end{pmatrix}

with A=ASUA = A_{\mathcal{S}}U as defined in Sec. 3.1. There are three coset representatives for TAS/TA\mathcal{T}_{A_{\mathcal{S}}}/\mathcal{T}_{A},

TAS=g1TA⊔g2TA⊔g3TA(32)\mathcal{T}_{A_{\mathcal{S}}}=g_{1}\mathcal{T}_{A}\mathbin{\sqcup}g_{2}\mathcal{T}_{A}\mathbin{\sqcup}g_{3}\mathcal{T}_{A} \tag*{(32)}
g1=(E,(0,0,0)T)ASg_{1}=\left(E,(0,0,0)^{\mathsf{T}}\right)_{A_{\mathcal{S}}}
g2=(E,(−1,−1,0)T)ASg_{2}=\left(E,(-1,-1,0)^{\mathsf{T}}\right)_{A_{\mathcal{S}}}
g3=(E,(−1,0,0)T)AS.g_{3}=\left(E,(-1,0,0)^{\mathsf{T}}\right)_{A_{\mathcal{S}}}.

Spin-only group

The moment tensor of MM is given by

MMT=(3mx20003mx20006mz2).MM^{\mathsf{T}}= \begin{pmatrix} 3m_{x}^{2} & 0 & 0 \\ 0 & 3m_{x}^{2} & 0 \\ 0 & 0 & 6m_{z}^{2} \end{pmatrix}.

Its eigenvalues 3mx23m_{x}^{2}, 3mx23m_{x}^{2}, and 6mz26m_{z}^{2} are all positive. Consequently, the spin arrangement is noncoplanar with Pso=1\mathcal{P}_{\mathrm{so}}=1.

Translation subgroup of maximal space subgroup

The translation g1g_{1} in Eq. (32) belongs to TAD\mathcal{T}_{A_{\mathrm{D}}}. On the other hand, g2g_{2} does not belong to TAD\mathcal{T}_{A_{\mathrm{D}}} because it maps x1x_{1} to x3x_{3}, whereas m1≠m3m_{1}\ne m_{3}. Similarly, g3g_{3} does not belong to TAD\mathcal{T}_{A_D}. Therefore, TAD\mathcal{T}_{A_D} is identical to TAS\mathcal{T}_{A_S} with AD=ASVA_D = A_S V as defined in Sec. 3.3 and we can choose V=EV = E.

Coset representatives of spin translation group

There are three candidates for the spatial operation parts of coset representatives of Gst\mathcal{G}_{\mathrm{st}}: g1g_1, g2g_2, and g3g_3. For translation g1g_1 in Eq. (32), we obtain

Mg1MT=(3mx20003mx20006mz2).M_{g_1}M^{\mathrm{T}} = \begin{pmatrix} 3m_x^2 & 0 & 0 \\ 0 & 3m_x^2 & 0 \\ 0 & 0 & 6m_z^2 \end{pmatrix}.

From the singular value decomposition of Mg1MTM_{g_1}M^{\mathrm{T}}, we calculate the spin-rotation part W~1\widetilde{W}_1 for g1g_1 as W~1=E\widetilde{W}_1 = E. For translation g2g_2, the singular value decomposition of Mg2MTM_{g_2}M^{\mathrm{T}} is

Mg2MT=(−12320−32−120001)(3mx20003mx20006mx2)(100010001)T.M_{g_2}M^{\mathrm{T}} = \begin{pmatrix} -\frac{1}{2} & \frac{\sqrt{3}}{2} & 0 \\ -\frac{\sqrt{3}}{2} & -\frac{1}{2} & 0 \\ 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} 3m_x^2 & 0 & 0 \\ 0 & 3m_x^2 & 0 \\ 0 & 0 & 6m_x^2 \end{pmatrix} \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{pmatrix}^{\mathrm{T}}.

Hence, a candidate for a spin-rotation part with g2g_2 is

W~2=(−12320−32−120001)(100010001)T=(−12320−32−120001).\widetilde{W}_2 = \begin{pmatrix} -\frac{1}{2} & \frac{\sqrt{3}}{2} & 0 \\ -\frac{\sqrt{3}}{2} & -\frac{1}{2} & 0 \\ 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{pmatrix}^{\mathrm{T}} = \begin{pmatrix} -\frac{1}{2} & \frac{\sqrt{3}}{2} & 0 \\ -\frac{\sqrt{3}}{2} & -\frac{1}{2} & 0 \\ 0 & 0 & 1 \end{pmatrix}.

Similarly, a candidate for a spin-rotation part with g3g_3 is

W~3=(−12−32032−120001).\widetilde{W}_3 = \begin{pmatrix} -\frac{1}{2} & -\frac{\sqrt{3}}{2} & 0 \\ \frac{\sqrt{3}}{2} & -\frac{1}{2} & 0 \\ 0 & 0 & 1 \end{pmatrix}.

The spin symmetry operations (gi,W~i)(g_i,\widetilde{W}_i) (i=1,2,3i = 1, 2, 3) all preserve magnetic moments.

Consequently, the spin translation group is obtained as

Gst=(g1,W~1)(TAD×1)∪(g2,W~2)(TAD×1)∪(g3,W~3)(TAD×1).\mathcal{G}_{\mathrm{st}} = (g_1,\widetilde{W}_1)(\mathcal{T}_{A_D} \times1) \cup(g_2,\widetilde{W}_2)(\mathcal{T}_{A_D} \times1) \cup(g_3,\widetilde{W}_3)(\mathcal{T}_{A_D} \times1).

Coset representatives of spin space group

Because TAD\mathcal{T}_{A_D} and TAS\mathcal{T}_{A_S} coincide in this example, S\mathcal{S} is fully compatible with TAD\mathcal{T}_{A_D}, SD=S\mathcal{S}_D = \mathcal{S}. For 24 coset representatives of SD/TAD\mathcal{S}_D/\mathcal{T}_{A_D}, we search for their spin-rotation parts by solving the Procrustes problems. For example, a six-fold spatial rotoinversion

g=(((−110−10000−1),0)AD)∈SDg=\left(\left(\begin{pmatrix}-1&1&0\\-1&0&0\\0&0&-1\end{pmatrix},0\right)_{A_D}\right)\in\mathcal{S}_D

gives a permutation of sites σg=(4,5,6,1,2,3,12,7,11,9,10,8)\sigma_g=(4,5,6,1,2,3,12,7,11,9,10,8) and the singular value decomposition of MgMM_gM

MgM=(−1000−1000−1)(3mx20003mx20006mx2)(100010001)TM_gM=\begin{pmatrix}-1&0&0\\0&-1&0\\0&0&-1\end{pmatrix}\begin{pmatrix}3m_x^2&0&0\\0&3m_x^2&0\\0&0&6m_x^2\end{pmatrix}\begin{pmatrix}1&0&0\\0&1&0\\0&0&1\end{pmatrix}^{\mathrm{T}}

which results in W~=−E\widetilde{W}=-E. Similarly, all of the 24 coset representatives of SD/TAD\mathcal{S}_D/\mathcal{T}_{A_D} give spin symmetry operations preserving MM.

A magnetic space group M\mathcal{M} of the NiAs-type CrSe is P31m′P31m^\prime (BNS number 157.55), which has ∣M/TAD∣=6|\mathcal{M}/\mathcal{T}_{A_D}|=6 coset representatives. Because the spin translation group Gst\mathcal{G}_{\mathrm{st}} contains three-fold spin-rotation parts, these spin symmetry operations ((g2,W~2g_2,\widetilde{W}_2) and (g3,W~3g_3,\widetilde{W}_3) in Sec. 4.4) are not captured in M\mathcal{M}. Similarly, the spin symmetry operations with six-fold spatial rotations or rotoinversions are not preserved in M\mathcal{M}.

Conclusion

We have presented the algorithm for determining spin symmetry operations of a given spin arrangement. The spin-only group is robustly determined from the eigenvalue decomposition of the moment tensor of magnetic moments, MMTMM^{\mathrm{T}}. We have explicitly considered the three translation subgroups to address the enlargement of the unit cell due to the spin translation group: the translation subgroup spanned by the input basis vectors TA\mathcal{T}_A, one by the primitive basis vector TAS\mathcal{T}_{A_S}, and one by the primitive basis vectors for the spin space group TAD\mathcal{T}_{A_D}. Spin-rotation parts of the coset representatives of the spin translation group and spin space group are found by solving the Procrustes problem to match the original magnetic moments and permuted ones. The present algorithm is implemented in SPINSPG under a permissive license. In future work, it will be beneficial to identify the spin-space-group type and a suitable transformation from the spin symmetry operations. Our presented algorithm and implementation will advance spin symmetry analysis in crystallography and condensed matter physics.

Note added After this work was completed, we became aware of recent preprints on the classification and enumeration of spin space-group types ([26, 18, 10]). Two of these works, Xiao et al. [26] and Jiang et al. [10], appear to use a workflow similar to ours to determine spin symmetry operations although they only provide a brief description and their implementations are not available to the public.

Our presented algorithm uses the Procrustes problem to determine a spin-rotation part W\mathbf{W}. On the other hand, Xiao et al. [26] directly determines W\mathbf{W} for a symmetry operation gg such that selected three magnetic moments mi\mathbf{m}_i (i=1,2,3i = 1,2,3) are transformed into mσg(i)\mathbf{m}_{\sigma_g(i)}, which may not be robust against numerical noises and experimental uncertainty of magnetic moments.

Jiang et al. [10] identifies spin symmetry operations on top of SPGLIB (Togo & Tanaka, [21]) similar to ours. However, they do not incorporate translation subgroups TAD\mathcal{T}_{AD} and TA\mathcal{T}_{A}. This omission makes their algorithm complicated and non-exhaustive.

Correspondence between spin symmetry operation and magnetic symmetry operation

When we consider a spin symmetry operation ((R,v),W)((R,\mathbf{v}),\mathbf{W}) for a Hamiltonian without the spin–orbit coupling (SOC), we derive the condition that ((R,v),W)((R,\mathbf{v}),\mathbf{W}) has a corresponding magnetic symmetry operation for the Hamiltonian with SOC. The SOC introduces an additional term to the Hamiltonian proportional to L^⋅σ^\hat{L}\cdot\hat{\sigma}, with the angular momentum operator L^\hat{L} and the Pauli matrices σ^=(σ^x,σ^y,σ^z)\hat{\boldsymbol{\sigma}} = (\hat{\sigma}_x,\hat{\sigma}_y,\hat{\sigma}_z). A spin symmetry operation ((R,v),W)((R,v),W) acts on L^\hat{L} and σ^\hat{\boldsymbol{\sigma}} as

L^↦(det⁡R)(det⁡W)(∑ν=x,y,zRμνL^ν)μ=x,y,z(33)\hat{L} \mapsto(\det R)(\det W)\left(\sum_{\nu=x,y,z}R_{\mu\nu}\hat{L}_{\nu}\right)_{\mu=x,y,z} \tag*{(33)}
σ^↦(∑ν=x,y,zWμνσ^ν)μ=x,y,z,(34)\hat{\boldsymbol{\sigma}} \mapsto\left(\sum_{\nu=x,y,z}W_{\mu\nu}\hat{\sigma}_{\nu}\right)_{\mu=x,y,z}, \tag*{(34)}

where RR and WW are represented with Cartesian coordinates. Note that det⁡W\det W in Eq. (33) reflects a time reversal operation. Thus, ((R,v),W)((R,v),W) acts on the SOC term as

L^⋅σ^↦(det⁡R)(det⁡W)∑μ,ν=x,y,z[RTW]μνL^μσ^ν.(35)\hat{L}\cdot\hat{\boldsymbol{\sigma}} \mapsto(\det R)(\det W)\sum_{\mu,\nu=x,y,z}\left[R^{\mathsf{T}}W\right]_{\mu\nu}\hat{L}_{\mu}\hat{\sigma}_{\nu}. \tag*{(35)}

To preserve the SOC term, (det⁡R)(det⁡W)RTW(\det R)(\det W)R^{\mathsf{T}}W should be identity. Because (det⁡R)R(\det R)R and (det⁡W)W(\det W)W belong to SO(3)\mathrm{SO}(3), this condition is equivalent to

(det⁡R)R=(det⁡W)W.(36)(\det R)R=(\det W)W. \tag*{(36)}

Therefore, if ((R,v),W)((R,v),W) satisfies Eq. (36), it has a corresponding magnetic symmetry operation ((R,v),det⁡W)((R,v),\det W) with symmetry operation part (R,v)(R,v) and time-reversal part det⁡W\det W. We identify det⁡W=1\det W=1 with an identity operation and det⁡W=−1\det W=-1 with a time-reversal operation. Here, we assume that a magnetic symmetry operation ((R,v),θ)((R,v),\theta) acts on a magnetic moments mm as θ(det⁡R)Rm\theta(\det R)Rm.

Conversely, a magnetic symmetry operation ((R,v),θ)((R,v),\theta) is mapped to a spin symmetry operation ((R,v),θR)((R,v),\theta R). We can confirm this mapping satisfies Eq. (36) by (det⁡θR)θR=(det⁡R)R(\det\theta R)\theta R=(\det R)R. Because this mapping is injective, the index of a magnetic space group M\mathcal{M} in its translation subgroup TAD\mathcal{T}_{AD} is a divisor of the index of a corresponding spin space group G\mathcal{G} in TAD\mathcal{T}_{AD}, ∣G/TAD∣≡0(mod∣M/TAD∣)|\mathcal{G}/\mathcal{T}_{AD}|\equiv0\pmod{|\mathcal{M}/\mathcal{T}_{AD}|}.

B Spin point group

We define a family spin point group of a spin space group G\mathcal{G} as

B(G)={W∈O(3)∣∃g s.t. (g,W)∈G}.(37)\mathcal{B}(\mathcal{G}) = \{W \in\mathrm{O}(3) \mid\exists g\ \text{s.t.}\ (g,W) \in\mathcal{G}\}. \tag*{(37)}

When we write a coset decomposition of G\mathcal{G} by its spin translation group Gst(G)\mathcal{G}_{\mathrm{st}}(\mathcal{G}) as

G=⨆R((R,vR),WR)Gst(G),(38)\mathcal{G} = \bigsqcup_{R} ((R,\boldsymbol{v}_{R}),W_{R})\mathcal{G}_{\mathrm{st}}(\mathcal{G}), \tag*{(38)}

A spin point group of G\mathcal{G} is

U(G)={(R,WRB(Gst(G)))}R∈P(F(G)).(39)\mathcal{U}(\mathcal{G}) = \{(R,W_{R}\mathcal{B}(\mathcal{G}_{\mathrm{st}}(\mathcal{G})))\}_{R\in\mathcal{P}(\mathcal{F}(\mathcal{G}))}. \tag*{(39)}

The spin point group U(G)\mathcal{U}(\mathcal{G}) is a subgroup of O(3)×(O(3)/B(Gst(G)))\mathrm{O}(3)\times(\mathrm{O}(3)/\mathcal{B}(\mathcal{G}_{\mathrm{st}}(\mathcal{G}))) and is isomorphic to G/Gst(G)\mathcal{G}/\mathcal{G}_{\mathrm{st}}(\mathcal{G}).

In general, we cannot choose WRW_{R} so that {(R,WR)}\{(R,W_{R})\} is a group under the multiplication in O(3)×O(3)\mathrm{O}(3)\times\mathrm{O}(3). We show one of the counterexamples in which {(R,WR)}\{(R,W_{R})\} is not closed as a group, with a spin arrangement

A=(400060008)X=(0.40.60.40.40.40.60.40.400.250.50.75)M=(10−10010−10.40.40.40.4),A = \begin{pmatrix}4 & 0 & 0 \\ 0 & 6 & 0 \\ 0 & 0 & 8\end{pmatrix} \qquad X = \begin{pmatrix}0.4 & 0.6 & 0.4 & 0.4 \\ 0.4 & 0.6 & 0.4 & 0.4 \\ 0 & 0.25 & 0.5 & 0.75\end{pmatrix} \qquad M = \begin{pmatrix}1 & 0 & -1 & 0 \\ 0 & 1 & 0 & -1 \\ 0.4 & 0.4 & 0.4 & 0.4\end{pmatrix},

illustrated in Figure 4. This spin arrangement is noncoplanar with PSO(G)=1\mathcal{P}_{\mathrm{SO}}(\mathcal{G}) = 1. The spin space group G\mathcal{G} is obtained as

Image not extracted

Figure 4. Spin arrangement example in Appendix B. Its spin point group is not closed under O(3)×O(3)O(3) \times O(3) due to the spin translation group.

G={g1,g2,g3,g4,g5,g6,g7,g8}(T(D(G))×PSO(G))\mathcal{G}=\{g_1,g_2,g_3,g_4,g_5,g_6,g_7,g_8\}\left(\mathcal{T}(\mathcal{D}(\mathcal{G}))\times\mathcal{P}_{\mathrm{SO}}(\mathcal{G})\right)
g1=((1,0)A,1)g_1=((1,0)_{\mathcal{A}},1)
g2=((1,12c)A,2z)g_2=\left(\left(1,\frac{1}{2}c\right)_{\mathcal{A}},2_z\right)
g3=((1,34c)A,mxy)g_3=\left(\left(1,\frac{3}{4}c\right)_{\mathcal{A}},m_{xy}\right)
g4=((1,14c)A,mxˉy)g_4=\left(\left(1,\frac{1}{4}c\right)_{\mathcal{A}},m_{\bar{x}y}\right)
g5=((2,34c)A,4z−)g_5=\left(\left(2,\frac{3}{4}c\right)_{\mathcal{A}},4_z^{-}\right)
g6=((2,14c)A,4z+)g_6=\left(\left(2,\frac{1}{4}c\right)_{\mathcal{A}},4_z^{+}\right)
g7=((m,12c)A,mx)g_7=\left(\left(m,\frac{1}{2}c\right)_{\mathcal{A}},m_x\right)
g8=((m,0)A,my),g_8=((m,0)_{\mathcal{A}},m_y),

where we denote a mirror operation along xyxy axis as mxym_{xy}. The spin point group of G\mathcal{G} is

U(G)={(1,1),(1,m‾xy),(2,4z−),(m,mx)}(1×B(Gst(G)))(40)\mathcal{U}(\mathcal{G})=\{(1,1),(1,\overline{m}_{xy}),(2,4_z^{-}),(m,m_x)\}\left(1\times\mathcal{B}(\mathcal{G}_{\mathrm{st}}(\mathcal{G}))\right) \tag*{(40)}
B(Gst(G))={1,2z}\mathcal{B}(\mathcal{G}_{\mathrm{st}}(\mathcal{G}))=\{1,2_z\}

The coset representatives in Eq. (40) are not closed as a subgroup of O(3)×O(3)O(3)\times O(3) because (2,4z−)−1=(2,4z+)(2,4_z^{-})^{-1}=(2,4_z^{+}) does not belong to the coset representatives. In fact, we cannot choose coset representatives to form a subgroup of O(3)×O(3)O(3)\times O(3) in this example because U(G)≅Z4\mathcal{U}(\mathcal{G})\cong\mathbb{Z}_4, B(Gst(G))≅Z2\mathcal{B}(\mathcal{G}_{\mathrm{st}}(\mathcal{G}))\cong\mathbb{Z}_2, and Z4\mathbb{Z}_4 cannot be written as an internal semidirect product Z2⋊Z2\mathbb{Z}_2\rtimes\mathbb{Z}_2.

Generating transformation matrix from centering vectors

To demonstrate how to generate a transformation matrix V\boldsymbol{V} in § 3.3, we consider the following coplanar fcc structure with the conventional basis,

A=(a000a000a)A = \begin{pmatrix} a & 0 & 0 \\ 0 & a & 0 \\ 0 & 0 & a \end{pmatrix}
X=(001212012012012120)X = \begin{pmatrix} 0 & 0 & \frac{1}{2} & \frac{1}{2} \\ 0 & \frac{1}{2} & 0 & \frac{1}{2} \\ 0 & \frac{1}{2} & \frac{1}{2} & 0 \end{pmatrix}
M=(00mm0000mm00).M = \begin{pmatrix} 0 & 0 & m & m \\ 0 & 0 & 0 & 0 \\ m & m & 0 & 0 \end{pmatrix}.

One of the primitive basis vectors for the nonmagnetic crystal structure is

AS=AU−1=a2(011101110)A_S = AU^{-1} = \frac{a}{2}\begin{pmatrix} 0 & 1 & 1 \\ 1 & 0 & 1 \\ 1 & 1 & 0 \end{pmatrix}
U=(−1111−1111−1).U = \begin{pmatrix} -1 & 1 & 1 \\ 1 & -1 & 1 \\ 1 & 1 & -1 \end{pmatrix}.

There are four centering vectors in TAS/TA\mathcal{T}_{A_S}/\mathcal{T}_A,

TAS=g1TA⊔g2TA⊔g3TA⊔g4TA\mathcal{T}_{A_S} = g_1\mathcal{T}_A \sqcup g_2\mathcal{T}_A \sqcup g_3\mathcal{T}_A \sqcup g_4\mathcal{T}_A
g1=(E,(0,0,0)T)ASg_1 = \left(E,(0,0,0)^{\mathsf{T}}\right)_{A_S}
g2=(E,(1,0,0)T)ASg_2 = \left(E,(1,0,0)^{\mathsf{T}}\right)_{A_S}
g3=(E,(0,1,0)T)ASg_3 = \left(E,(0,1,0)^{\mathsf{T}}\right)_{A_S}
g4=(E,(0,0,1)T)AS.g_4 = \left(E,(0,0,1)^{\mathsf{T}}\right)_{A_S}.

The half of centering vectors form the translation subgroup of D(G)\mathcal{D}(\mathcal{G}),

T(D(G))=g1TA⊔g2TA.\mathcal{T}(\mathcal{D}(\mathcal{G})) = g_1\mathcal{T}_A \sqcup g_2\mathcal{T}_A.

Because T(D(G))\mathcal{T}(\mathcal{D}(\mathcal{G})) is spanned by column vectors of UU and centering vectors of g1g_1 and g2g_2, we can rewrite T(D(G))\mathcal{T}(\mathcal{D}(\mathcal{G})) as

T(D(G))=ASU~Z5\mathcal{T}(\mathcal{D}(\mathcal{G})) = A_S\widetilde{U}\mathbb{Z}^{5}
U~=(−111011−110011−100).\widetilde{U} = \begin{pmatrix} -1 & 1 & 1 & 0 & 1 \\ 1 & -1 & 1 & 0 & 0 \\ 1 & 1 & -1 & 0 & 0 \end{pmatrix}.

We need to find a transformation matrix V∈Z3×3V \in\mathbb{Z}^{3 \times3} such that

T(D(G))=ASVZ3,(41)\mathcal{T}(\mathcal{D}(\mathcal{G})) = A_S V \mathbb{Z}^{3}, \tag*{(41)}

and it is known to be achieved by the Hermite normal form of U~\widetilde{U} [3],

U~=(100000100001200)(−111011−110001−1000001001000),\widetilde{U} = \begin{pmatrix} 1 & 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 & 0 \\ 0 & 1 & 2 & 0 & 0 \end{pmatrix} \begin{pmatrix} -1 & 1 & 1 & 0 & 1 \\ 1 & -1 & 1 & 0 & 0 \\ 0 & 1 & -1 & 0 & 0 \\ 0 & 0 & 0 & 1 & 0 \\ 0 & 1 & 0 & 0 & 0 \end{pmatrix},

where U~\widetilde{U} is decomposed into a product of a lower triangular integer matrix and a unimodular matrix. Because the above 5×55 \times5 integer matrix is unimodular, we obtain

T(D(G))=AS(100000100001200)Z5=AS(100010012)Z3.\begin{aligned} \mathcal{T}(\mathcal{D}(\mathcal{G})) = A_S \begin{pmatrix} 1 & 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 & 0 \\ 0 & 1 & 2 & 0 & 0 \end{pmatrix} \mathbb{Z}^{5} \\ &= A_S \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 1 & 2 \end{pmatrix} \mathbb{Z}^{3}. \end{aligned}

Consequently, we generate the integer matrix

V=(100010012).V = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 1 & 2 \end{pmatrix}.

Synopsis

This paper presents an algorithm for determining the spin symmetry operations of a given spin arrangement. Spin symmetry operations of a spin space group act simultaneously on both the spatial and spin coordinates of the spin arrangement.

References

  1. [1]Brinkman, W. & Elliott, R. J. (1966). J. Appl. Phys. 37(3), 1457–1459.
  2. [2]Brinkman, W. F., Elliott, R. J. & Peierls, R. E. (1966). Proc. Math. Phys. Eng. Sci. 294(1438), 343–358.
  3. [3]Cohen, H. (1993). A Course in Computational Algebraic Number Theory. Springer Berlin Heidelberg.DOI
  4. [4]Corliss, L. M., Elliott, N., Hastings, J. M. & Sass, R. L. (1961). Phys. Rev. 122, 1402–1406.
  5. [5]Corticelli, A., Moessner, R. & McClarty, P. A. (2022). Phys. Rev. B, 105, 064430.
  6. [6]Gallego, S. V., Perez-Mato, J. M., Elcoro, L., Tasci, E. S., Hanson, R. M., Momma, K., Aroyo, M. I. & Madariaga, G. (2016). J. Appl. Cryst. 49(5), 1750–1776.
  7. [7]Gower, J. C. & Dijksterhuis, G. B. (2004). Procrustes Problems. Oxford University Press.DOI
  8. [8]Hahn, T., Klapper, H., Müller, U. & Aroyo, M. (2016). In International Tables for Crystallography, edited by M. I. Aroyo, vol. A, chap. 3.2, pp. 720–776. Wiley, 6th ed.
  9. [9]Hart, G. L. W. & Forcade, R. W. (2008). Phys. Rev. B, 77, 224115. URL: https://link.aps.org/doi/10.1103/PhysRevB.77.224115DOI
  10. [10]Jiang, Y., Song, Z., Zhu, T., Fang, Z., Weng, H., Liu, Z.-X., Yang, J. & Fang, C., (2023). Enumeration of spin-space groups: Towards a complete description of symmetries of magnetic orders. https://arxiv.org/abs/2307.10371.
  11. [11]Litvin, D. & Opechowski, W. (1974). Physica, 76(3), 538–554.
  12. [12]Litvin, D. B. (1973). Acta Cryst. A, 29(6), 651–660.
  13. [13]Litvin, D. B. (1977). Acta Cryst. A, 33(2), 279–287.
  14. [14]Litvin, D. B. (2014). Magnetic group tables. IUCr. https://www.iucr.org/publ/978-0-9553602-2-0.DOI
  15. [15]Liu, P., Li, J., Han, J., Wan, X. & Liu, Q. (2022). Phys. Rev. X, 12, 021016.
  16. [16]Nebe, G. (2011). In International Tables for Crystallography, edited by H. Wondratschek & U. Müller, vol. A1, chap. 1.4, pp. 27–40. Springer, 2nd ed.
  17. [17]Opechowski, W. (1986). Crystallographic and Metacrystallographic Groups. North-Holland.DOI
  18. [18]Ren, J., Chen, X., Zhu, Y., Yu, Y., Zhang, A., Li, J., Li, C. & Liu, Q., (2023). Enumeration and representation of spin space groups. https://arxiv.org/abs/2307.10369.
  19. [19]Shinohara, K., Togo, A. & Tanaka, I. (to be published 2023). Acta Cryst. A.
  20. [20]Stokes, H. T. & Campbell, B. J. (2017). Acta Cryst. A, 73(1), 4–13.
  21. [21]Togo, A. & Tanaka, I., (2018). Spglib: a software library for crystal symmetry search. https://github.com/spglib/spglib.
  22. [22]Šmejkal, L., Sinova, J. & Jungwirth, T. (2022a). Phys. Rev. X, 12, 031042.
  23. [23]Šmejkal, L., Sinova, J. & Jungwirth, T. (2022b). Phys. Rev. X, 12, 040501.
  24. [24]Železný, J., Zhang, Y., Felser, C. & Yan, B. (2017). Phys. Rev. Lett. 119, 187204.
  25. [25]Watanabe, H., Shinohara, K., Nomoto, T., Togo, A. & Ryotaro, A., (2023). Spin-group symmetry analysis of emergent electromagnetic properties. submitted.
  26. [26]Xiao, Z., Zhao, J., Li, Y., Shindou, R. & Song, Z.-D., (2023). Spin space groups: Full classification and applications. https://arxiv.org/abs/2307.10364.
  27. [27]Yang, J., Liu, Z.-X. & Fang, C., (2021). Symmetry invariants in magnetically ordered systems having weak spin-orbit coupling. https://arxiv.org/abs/2105.12738.
  28. [28]Zelenskiy, A., Monchesky, T. L., Plumer, M. L. & Southern, B. W. (2022). Phys. Rev. B, 106, 144433.
  29. [29]Zhang, Y., Železný, J., Sun, Y., van den Brink, J. & Yan, B. (2018). New J. Phys. 20(7), 073028.

Paper details

vlnc:W_dabeb3th
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