betaFeedback

WorksPhysics and Astronomy

Exact magnetic ground states and competing interlayer exchanges in layered kagome magnets

Abstract

Magnetic ground states in layered kagome magnets are investigated using a Heisenberg model with six interlayer exchange interactions. For isotropic two-sublattice models with identical diagonal Fourier exchange functions, a coplanar single-wave-vector ground state is shown to exist for arbitrary exchange ranges. Extending the three-exchange model of Rosenfeld and Mushnikov reveals that longer-range interactions can change the preferred magnetic order from spiral to ferromagnetic and back to spiral, rather than merely modifying the spiral period. The analysis includes ground-state phase diagrams, harmonic stability, magnetic-field response, and sensitivity to individual exchange interactions. The propagation vector, internal layer angle, saturation field, and planar susceptibility are also found to be insufficient to determine all six exchange constants uniquely. These results provide a basis for predicting noncollinear magnetic order from first-principles exchange interactions.

Introduction

Structural stability and a sizable calculated magnetic moment do not by themselves determine the magnetic ground state of a material. This distinction is particularly important for layered kagome compounds, where magnetic order affects the symmetry of electronic states and can be changed by composition or applied field. Recent reviews describe the diversity of magnetic configurations and their relationship to topological and transport properties in the kagome families [1, 2, 3]. Experiments on RMn6Sn6RMn_{6}Sn_{6} materials establish that competition between interlayer interactions can produce incommensurate order and complex field-induced structures [4, 5, 6, 7, 8, 9]. The magnetic stacking along the crystallographic cc axis is therefore central to understanding these compounds.

The practical need has grown with computational materials discovery. High-throughput density-functional calculations have examined 312 substitutional AT6X6AT_{6}X_{6} compositions and identified additional stable magnetic candidates [10]. Searches of related AT6X4AT_{6}X_{4} and AT6X5AT_{6}X_{5} structure types broaden the range of candidate compounds [11]. Such studies typically examine a limited number of candidate magnetic structures, frequently including collinear arrangements. A lowest-energy collinear solution, however, need not be the lowest-energy state of a frustrated interlayer exchange model. Competing exchanges and nearly degenerate collinear states make noncollinear order a plausible alternative, although not every predicted compound is expected to be a spiral magnet.

The three-exchange double-layer model of Rosenfeld and Mushnikov (RM) provides a reference for RMn6X6RMn_{6}X_{6} magnets [12]. It describes the competition between inequivalent adjacent-layer couplings and a further interlayer interaction; subsequent work examined nested spirals and their stability [13, 14]. First-principles calculations can now resolve more distant exchange paths, whose effects need not reduce to a change of spiral period. We therefore retain six interlayer exchanges and examine how they determine the magnetic stacking, stability, and field response. Both the turn angle between repeats and the relative orientation of the two layers are allowed to relax.

We use the Luttinger–Tisza (LT) method [15, 16, 17]. For the isotropic classical model, the equivalent Mn planes of RMn6Sn6RMn_{6}Sn_{6} have identical diagonal Fourier exchange functions, so the LT eigenmode satisfies the fixed-length constraint. This gives the exact zero-field ground state within the rigid-layer model. The exact exchange spectrum determines the phase diagrams and provides the starting point for the stability and magnetic-response calculations. The model assumes equal-length layer moments and omits rare-earth magnetism, longitudinal fluctuations, and additional anisotropic or non-Heisenberg interactions.

Six-exchange two-sublattice Heisenberg model

Real-space exchange geometry and conventions

The model consists of two magnetic layers, AA and BB, in each repeat nn. The unit vectors eAn\mathbf{e}_{An} and eBn\mathbf{e}_{Bn} give their spin directions. Counting each exchange bond once, the energy is

H=−∑n[J1eAn⋅eBn+J2eBn⋅eA,n+1+J4eAn⋅eB,n+1+J5eBn⋅eA,n+2+J3(eAn⋅eA,n+1+eBn⋅eB,n+1)+J6(eAn⋅eA,n+2+eBn⋅eB,n+2)].(1)\mathcal{H}=-\sum_{n}\left[J_{1}\mathbf{e}_{An}\cdot\mathbf{e}_{Bn}+J_{2}\mathbf{e}_{Bn}\cdot\mathbf{e}_{A,n+1} +J_{4}\mathbf{e}_{An}\cdot\mathbf{e}_{B,n+1}+J_{5}\mathbf{e}_{Bn}\cdot\mathbf{e}_{A,n+2} +J_{3}\left(\mathbf{e}_{An}\cdot\mathbf{e}_{A,n+1}+\mathbf{e}_{Bn}\cdot\mathbf{e}_{B,n+1}\right) +J_{6}\left(\mathbf{e}_{An}\cdot\mathbf{e}_{A,n+2}+\mathbf{e}_{Bn}\cdot\mathbf{e}_{B,n+2}\right)\right]. \tag*{(1)}

The first three bonds reproduce the established three-exchange layered model [12]. The J4J_{4}, J5J_{5}, and J6J_{6} paths describe additional interlayer interactions. Positive JaJ_{a} favors parallel alignment on the corresponding bond. The exchange labels denote the particular paths shown in Figure 1, not a universal shell ordering. All energies below are per two-layer crystallographic repeat.

Six effective interlayer exchanges in the two-Mn-plane kagome model, illustrated on a YMn₆Sn₆-inspired layer stack

Figure 1. Six effective interlayer exchanges in the two-Mn-plane kagome model, illustrated on a YMn₆Sn₆-inspired layer stack. The arrows indicate the six effective interlayer exchange paths, not individual microscopic Mn–Mn bonds. The same-sublattice terms J3J_3 and J6J_6 also contain the corresponding BB–BB interactions, omitted for clarity. The nonmagnetic atoms are illustrative.

Exact homogeneous spiral energy

A planar spiral with a constant rotation angle between successive two-layer repeats is described by

eAn=(cos⁡nΦ,sin⁡nΦ,0),\mathbf{e}_{A n} = (\cos n\Phi, \sin n\Phi, 0),
eBn=(cos⁡(nΦ+δ),sin⁡(nΦ+δ),0).(2)\mathbf{e}_{B n} = (\cos(n\Phi+ \delta), \sin(n\Phi+ \delta), 0). \tag*{(2)}

Here Φ=2πq\Phi= 2\pi q is the rotation between consecutive crystallographic repeats and δ\delta is the internal AA–BB phase. Direct substitution into Eq. (1) yields

E(Φ,δ)=−J1cos⁡δ−J2cos⁡(Φ−δ)−J4cos⁡(Φ+δ)−J5cos⁡(2Φ−δ)−2J3cos⁡Φ−2J6cos⁡2Φ.(3)E(\Phi, \delta) = -J_1\cos\delta- J_2\cos(\Phi-\delta) - J_4\cos(\Phi+\delta) - J_5\cos(2\Phi-\delta) - 2J_3\cos\Phi- 2J_6\cos2\Phi. \tag*{(3)}

The four bonds connecting AA and BB layers can be collected into C(Φ)C(\Phi), while D(Φ)D(\Phi) contains the bonds between like layers:

C(Φ)=J1+J2e−iΦ+J4eiΦ+J5e−2iΦ,(4)C(\Phi) = J_1 + J_2e^{-i\Phi} + J_4e^{i\Phi} + J_5e^{-2i\Phi}, \tag*{(4)}
D(Φ)=2J3cos⁡Φ+2J6cos⁡2Φ.(5)D(\Phi) = 2J_3\cos\Phi+ 2J_6\cos2\Phi. \tag*{(5)}

The energy is E=−D−Re⁡[eiδC]E = -D - \operatorname{Re}[e^{i\delta}C]. If C≠0C \ne0, the minimizing phase satisfies

eiδ∗(Φ)=C(Φ)∗∣C(Φ)∣,(6)e^{i\delta^*(\Phi)} = \frac{C(\Phi)^*}{\lvert C(\Phi)\rvert}, \tag*{(6)}

and hence the lowest homogeneous-spiral energy is

E0(Φ)=−2J3cos⁡Φ−2J6cos⁡2Φ−∣C(Φ)∣.(7)E_0(\Phi) = -2J_3\cos\Phi- 2J_6\cos2\Phi- \lvert C(\Phi)\rvert. \tag*{(7)}

If C=0C=0, the intersublattice exchange does not determine δ\delta. The dependence of the energy on the spiral wave vector may then develop a cusp, where the derivative formulas cannot be used.

Momentum-space exchange matrix

With the Fourier convention consistent with Eq. (4), the Hermitian exchange matrix is

J(Φ)=(D(Φ)C(Φ)C(Φ)∗D(Φ)).(8)J(\Phi)=\begin{pmatrix} D(\Phi) & C(\Phi) \\ C(\Phi)^* & D(\Phi) \end{pmatrix}. \tag*{(8)}

Its eigenvalues are

λ±(Φ)=D(Φ)±∣C(Φ)∣.(9)\lambda_{\pm}(\Phi)=D(\Phi)\pm\lvert C(\Phi)\rvert. \tag*{(9)}

The upper exchange eigenvalue reproduces the optimized spiral energy, E0=−λ+E_0=-\lambda_+. Its relation to the unrestricted classical ground state follows from the LT argument in Sec. III.

Exact ground states for arbitrary-range exchange networks

The six exchanges in Eq. (1) describe the selected interlayer bonds. The ground-state result below also applies when more distant bonds are included. Consider an isotropic classical Heisenberg model with two unit spins per primitive cell and translationally invariant bilinear exchanges Jαβ(R)J_{\alpha\beta}(\mathbf{R}), where α,β=A,B\alpha,\beta=A,B and R\mathbf{R} connects primitive cells. Each physical bond is counted once, with Jβα(−R)=Jαβ(R)J_{\beta\alpha}(-\mathbf{R})=J_{\alpha\beta}(\mathbf{R}). The exchange matrix is

J(q)=(DA(q)C(q)C(q)∗DB(q)),Dα(q)=∑RJαα(R)eiq⋅R,C(q)=∑RJAB(R)eiq⋅R.(10)J(\mathbf{q})=\begin{pmatrix} D_A(\mathbf{q}) & C(\mathbf{q}) \\ C(\mathbf{q})^* & D_B(\mathbf{q}) \end{pmatrix}, \qquad D_\alpha(\mathbf{q})=\sum_{\mathbf{R}}J_{\alpha\alpha}(\mathbf{R})e^{i\mathbf{q}\cdot\mathbf{R}}, \qquad C(\mathbf{q})=\sum_{\mathbf{R}}J_{AB}(\mathbf{R})e^{i\mathbf{q}\cdot\mathbf{R}}. \tag*{(10)}

Here the exchange-matrix convention includes both directions of each like-sublattice bond; the Hamiltonian carries the corresponding factor 1/21/2. Sublattice positions may instead be included in the Fourier phases by a change of basis, without changing the eigenvalues. For equivalent sublattices, in the precise sense that DA(q)=DB(q)≡D(q)D_A(\mathbf{q})=D_B(\mathbf{q})\equiv D(\mathbf{q}) for every q\mathbf{q}, the upper eigenvalue and an associated normalized eigenvector are

λ+(q)=D(q)+∣C(q)∣,v+(q)=12(1C(q)∗/∣C(q)∣).(11)\lambda_+(\mathbf{q})=D(\mathbf{q})+\lvert C(\mathbf{q})\rvert, \qquad v_+(\mathbf{q})=\frac{1}{\sqrt{2}}\begin{pmatrix}1 \\ C(\mathbf{q})^*/\lvert C(\mathbf{q})\rvert\end{pmatrix}. \tag*{(11)}

The expression for the eigenvector assumes C≠0C\ne0; at C=0C=0 the two sublattice phases are degenerate. In particular, equality of the diagonal functions does not require equality of the different AA–BB bonds.

Exact zero-field ground state. The following statement is an application of the LT method [15, 16, 17]. The Fourier bound below is the LT “weak constraint”, and the equal sublattice amplitudes of the optimal eigenvector ensure that the “strong constraint” of unit spin length is satisfied. Single-wave-vector helical ground states of Bravais-lattice Heisenberg models were obtained in the same way by Yoshimori [18]. We state the two-sublattice form explicitly because it is used throughout. Consider the isotropic bilinear Hamiltonian

H=−12∑R,R′∑α,β=A,BJαβ(R−R′) eRα⋅eR′β.(12)\mathcal{H}=-\frac{1}{2}\sum_{\mathbf{R},\mathbf{R}'}\sum_{\alpha,\beta=A,B}J_{\alpha\beta}(\mathbf{R}-\mathbf{R}')\,\mathbf{e}_{\mathbf{R}\alpha}\cdot\mathbf{e}_{\mathbf{R}'\beta}. \tag*{(12)}

Every spin satisfies ∣eRα∣=1\lvert\mathbf{e}_{\mathbf{R}\alpha}\rvert=1. Assume translational invariance, a well-defined exchange Fourier transform, and DA(q)=DB(q)D_A(\mathbf{q})=D_B(\mathbf{q}) at every wave vector. Then at least one exact global zero-field ground state is a coplanar single-Q\mathbf{Q} spiral, with

EGSNc=−max⁡q[D(q)+∣C(q)∣].(13)\frac{E_{\mathrm{GS}}}{N_c}=-\max_{\mathbf{q}}\left[D(\mathbf{q})+\lvert C(\mathbf{q})\rvert\right]. \tag*{(13)}

Here NcN_c is the number of primitive cells, and Q\mathbf{Q} maximizes the expression in brackets.

Proof. We use the Fourier convention eqα=Nc−1/2∑ReRαe−iq⋅Re_{\mathbf{q}\alpha}=N_c^{-1/2}\sum_{\mathbf{R}}e_{\mathbf{R}\alpha}e^{-i\mathbf{q}\cdot\mathbf{R}}. The largest exchange eigenvalue limits the exchange energy of every Fourier component: eq†J(q)eq≤Λeq†eq\mathbf{e}_{\mathbf{q}}^\dagger J(\mathbf{q})\mathbf{e}_{\mathbf{q}}\leq\Lambda\mathbf{e}_{\mathbf{q}}^\dagger\mathbf{e}_{\mathbf{q}}, where Λ=max⁡qλ+(q)\Lambda=\max_{\mathbf{q}}\lambda_+(\mathbf{q}). Summing over wave vectors and using the fixed spin lengths gives

H=−12∑qeq†J(q)eq≥−Λ2∑qeq†eq=−NcΛ.(14)\mathcal{H}=-\frac{1}{2}\sum_{\mathbf{q}}\mathbf{e}_{\mathbf{q}}^\dagger J(\mathbf{q})\mathbf{e}_{\mathbf{q}} \geq-\frac{\Lambda}{2}\sum_{\mathbf{q}}\mathbf{e}_{\mathbf{q}}^\dagger\mathbf{e}_{\mathbf{q}} =-N_c\Lambda. \tag*{(14)}

No spin configuration can have an exchange energy below this value, including nonuniform, noncoplanar, and multi-q\mathbf{q} states. A planar spiral reaches this energy. At the preferred wave vector Q\mathbf{Q}, we take eiδ=C(Q)∗/∣C(Q)∣e^{i\delta}=C(\mathbf{Q})^*/\lvert C(\mathbf{Q})\rvert and write

eAR=(cos⁡Q⋅R,sin⁡Q⋅R,0),eBR=(cos⁡(Q⋅R+δ),sin⁡(Q⋅R+δ),0).(15)\mathbf{e}_{A\mathbf{R}}=(\cos\mathbf{Q}\cdot\mathbf{R},\sin\mathbf{Q}\cdot\mathbf{R},0), \qquad \mathbf{e}_{B\mathbf{R}}=(\cos(\mathbf{Q}\cdot\mathbf{R}+\delta),\sin(\mathbf{Q}\cdot\mathbf{R}+\delta),0). \tag*{(15)}

Each spin has unit length. The sublattice phase vector (1,eiδ)T(1,e^{i\delta})^T is an eigenvector of J(Q)J(\mathbf{Q}) with eigenvalue Λ\Lambda, and its complex conjugate gives the corresponding mode at −Q-\mathbf{Q}. The energy of this planar spiral therefore equals the lowest energy allowed by Eq. (14), establishing Eq. (13). When C(Q)=0C(\mathbf{Q})=0, either relative phase is allowed and the same conclusion follows. Q.E.D.

The planar spiral is an exact ground state, although other magnetic structures may have the same energy: distinct maximizing wave vectors may yield additional degenerate states. Finite periodic samples require a compatible Q\mathbf{Q}; the unrestricted continuous-wave-vector statement holds in the thermodynamic limit. Arbitrarily distant AAAA, BBBB, and ABAB exchanges only add Fourier harmonics to DD and CC, without enlarging the 2×22\times2 eigenvalue problem. Longer exchange range can nevertheless generate additional competing ordering vectors.

General number of sublattices. For ss unit spins per cell, let λmax⁡(q)\lambda_{\max}(\boldsymbol{q}) denote the largest eigenvalue of the s×ss \times s Hermitian exchange matrix. The same argument gives

EGSNc≥−s2max⁡qλmax⁡(q).(16)\frac{E_{\mathrm{GS}}}{N_c} \geq-\frac{s}{2}\max_{\boldsymbol{q}}\lambda_{\max}(\boldsymbol{q}). \tag*{(16)}

A single-QQ spiral reaches the lowest possible exchange energy if the exchange eigenvector at the preferred wave vector has ∣vα∣=1/s|v_\alpha| = 1/\sqrt{s} for every sublattice, which is the generalized LT condition [16, 17]. The relative phases of its components then describe a coplanar spiral with unit spins on every sublattice and the lowest possible exchange energy. Other ground states are also possible: combinations of degenerate eigenmodes may also satisfy the local constraints. For two inequivalent sublattices with DA≠DBD_A \ne D_B, the maximizing eigenvector generally has unequal amplitudes, so the Fourier maximum alone need not give the ground state.

The result is exact for the stated classical Heisenberg model, but does not establish the validity of a rigid-layer reduction for a material. Additional magnetic sites within a kagome plane, unequal moment lengths, non-Heisenberg interactions, anisotropy, applied fields, and longitudinal fluctuations require separate treatment. In particular, a two-layer ground state can still be unstable to intralayer degrees of freedom omitted when each plane is represented by one spin.

Ground-state determination and exchange dependence

A useful form of the spiral energy

Writing x=cos⁡Φx = \cos\Phi, the intersublattice exchange contribution takes the form

P3(x)=∣C(Φ)∣2=a0+a1x+a2x2+a3x3,(17)P_3(x) = |C(\Phi)|^2 = a_0 + a_1 x + a_2 x^2 + a_3 x^3, \tag*{(17)}
a0=J12+J22+J42+J52−2(J1J5+J2J4),a1=2[J1(J2+J4)+J2J5]−6J4J5,a2=4(J1J5+J2J4),a3=8J4J5.\begin{aligned} a_0 &= J_1^2 + J_2^2 + J_4^2 + J_5^2 - 2(J_1J_5 + J_2J_4),\\ a_1 &= 2[J_1(J_2 + J_4) + J_2J_5] - 6J_4J_5,\\ a_2 &= 4(J_1J_5 + J_2J_4),\\ a_3 &= 8J_4J_5. \end{aligned}

Thus, for P3(x)>0P_3(x) > 0,

E0(x)=−2J3x−2J6(2x2−1)−P3(x).(18)E_0(x) = -2J_3x - 2J_6(2x^2 - 1) - \sqrt{P_3(x)}. \tag*{(18)}

This form is convenient for determining the preferred spiral period.

Finding the preferred spiral period

In the interior interval 0<Φ<π0 < \Phi< \pi, homogeneous spirals satisfy dE0/dΦ=0dE_0/d\Phi= 0. Since dx/dΦ=−sin⁡Φ≠0dx/d\Phi= -\sin\Phi\ne0, this is equivalent to

−2J3−8J6x−P3′(x)2P3(x)=0.(19)-2J_3 - 8J_6x - \frac{P_3'(x)}{2\sqrt{P_3(x)}} = 0. \tag*{(19)}

The possible equilibrium spiral periods can also be found from the polynomial equation

4(2J3+8J6x)2P3(x)−[P3′(x)]2=0.(20)4(2J_3 + 8J_6x)^2P_3(x) - [P_3'(x)]^2 = 0. \tag*{(20)}

The equation is generally fifth order in xx. Squaring the equilibrium condition introduces additional roots that need not satisfy Eq. (19). Only solutions with −1<x<1-1 < x < 1 and P3(x)>0P_3(x) > 0 that satisfy the original equilibrium condition are retained, Eq. (19). Their energies are compared with those of the collinear states Φ=0\Phi= 0 and Φ=π\Phi= \pi. At P3(x)=0P_3(x) = 0, direct energy comparison is necessary because the derivative may be discontinuous. Introducing Q(x)=[P3′(x)]2−2P3(x)P3′′(x)Q(x) = [P_3'(x)]^2 - 2P_3(x)P_3''(x), the second derivative of the spiral energy is

d2E0dx2=−8J6+Q(x)4P3(x)3/2.(21)\frac{d^2E_0}{dx^2} = -8J_6 + \frac{Q(x)}{4P_3(x)^{3/2}}. \tag*{(21)}

At an equilibrium spiral wave vector, d2E0/dΦ2=sin⁡2Φ d2E0/dx2d^2E_0/d\Phi^2 = \sin^2\Phi\,d^2E_0/dx^2. A positive second derivative indicates that the spiral resists a small change in its period. The sign of this curvature distinguishes a local pitch minimum from a maximum.

Ground-state selection

The ground state has the lowest energy among the stationary spirals, collinear endpoints, and cusp configurations P3(x)=0P_3(x) = 0. Its internal layer angle follows from C(Φ)C(\Phi) when C≠0C \ne0.

Ground-state phase diagrams

Exact three-exchange reference and phase nomenclature

The RM phase diagrams provide a reference [12]. Their Figure 2 shows separate results for J1>0J_1 > 0 and J1<0J_1 < 0. The same axes x=J2/J1x = J_2/J_1 and y=J3/J1y = J_3/J_1 are retained, with −3≤x≤3-3 \leq x \leq3 and −2≤y≤2-2 \leq y \leq2, and their labels F, AF1, AF2, AF3, H1, and H2. The first four denote distinct collinear stacking patterns; H1 and H2 denote spiral structures for which the two consecutive turn angles have the same and opposite signs, respectively. In particular, H1 and H2 do not denote two distinct minima of the spiral energy at fixed exchange parameters. In RM’s Table 1 the collinear assignments are: F: (δ,Φ)=(0,0)(\delta,\Phi) = (0,0), AF1: (π,0)(\pi,0), AF2: (0,π)(0,\pi), and AF3: (π,π)(\pi,\pi). These assignments are used throughout. In the corresponding layer sequences, AF1 alternates within each repeat, whereas AF2 alternates between repeats.

Figure 2. Ground-state phase diagrams in the RM coordinates [12]. Left column: numerically evaluated three-exchange model, J4=J5=J6=0J_4 = J_5 = J_6 = 0; right column: calculated six-exchange sections at J4/J1=0.15J_4/J_1 = 0.15, J5/J1=0.30J_5/J_1 = 0.30, and J6/J1=−0.18J_6/J_1 = -0.18. Upper row: J1>0J_1 > 0; lower row: J1<0J_1 < 0. F and AF1–AF3 denote the four collinear stacking patterns, while H1 and H2 distinguish the sign relation of consecutive spiral turn angles, as in the original paper. Phase boundaries are resolved on a finite numerical grid.

With J4=J5=J6=0J_4 = J_5 = J_6 = 0, the present Hamiltonian reduces exactly to their zero-field exchange energy,

ERM(Φ,δ)=−J1cos⁡δ−J2cos⁡(Φ−δ)−2J3cos⁡Φ.(22)E_{\mathrm{RM}}(\Phi,\delta) = -J_1\cos\delta- J_2\cos(\Phi-\delta) - 2J_3\cos\Phi. \tag*{(22)}

The optimized energy is ERM,0(Φ)=−2J3cos⁡Φ−∣J1+J2e−iΦ∣E_{\mathrm{RM},0}(\Phi)=-2J_3\cos\Phi-\lvert J_1+J_2e^{-i\Phi}\rvert. For an interior stationary pitch, writing u=cos⁡Φu=\cos\Phi, the stationary equation implies

u=J12J22/(4J32)−J12−J222J1J2,−1<u<1(23)u=\frac{J_1^2J_2^2/(4J_3^2)-J_1^2-J_2^2}{2J_1J_2},\qquad-1<u<1 \tag*{(23)}

provided J1J2J3≠0J_1J_2J_3\ne0. The unsquared stationary equation additionally requires J1J2/J3<0J_1J_2/J_3<0; a value of uu in (−1,1)(-1,1) alone is not sufficient. The spiral solution is compared with the two collinear endpoints, Φ=0\Phi=0 and π\pi, and the lowest energy is selected. The continuous spiral–collinear boundaries, in RM notation, lie on four curves

k:x+2y(x+1)=0,l:x−2y(x+1)=0,m:x−2y(x−1)=0,n:x+2y(x−1)=0.(24)\begin{aligned} k &: x+2y(x+1)=0,\qquad l:x-2y(x+1)=0,\\ m &: x-2y(x-1)=0,\qquad n:x+2y(x-1)=0. \tag*{(24)} \end{aligned}

Only the parts of these curves separating actual lowest-energy states are phase boundaries. Their assignments to adjacent phases differ between the two signs of J1J_1, as specified in Table 2 of Ref. [12].

Representative phase diagrams

Figure 2 places the original three-exchange diagram beside a six-exchange example. Both use the same axes and phase labels. The upper and lower rows correspond to J1>0J_1>0 and J1<0J_1<0, respectively. For the extended panels, the parameters are fixed at J4/J1=0.15J_4/J_1=0.15, J5/J1=0.30J_5/J_1=0.30, and J6/J1=−0.18J_6/J_1=-0.18, and xx and yy are varied. These values illustrate the model; they are not fitted to experiment. For the three-exchange panels, the analytical equilibrium-pitch expression was checked by directly minimizing the exchange energy. For the six-exchange panels, the exchange energy was minimized numerically with respect to the spiral period. For the extended model, the upper exchange eigenvalue is compared at the two collinear wave vectors and at the interior maxima. Phase boundaries separate states with equal ground-state energies within the numerical resolution. The additional exchanges reshape the collinear and spiral regions. The phase labels identify the globally minimizing single-qq states, as guaranteed by the equivalent-layer Luttinger–Tisza construction. The dark lines mark changes in the minimizing state. Boundaries between collinear states are straight lines that follow from the endpoint eigenvalues λ+(0)=2J3+2J6+∣C(0)∣\lambda_+(0)=2J_3+2J_6+\lvert C(0)\rvert and λ+(π)=−2J3+2J6+∣C(π)∣\lambda_+(\pi)=-2J_3+2J_6+\lvert C(\pi)\rvert. In panel (d), the collinear equal-energy conditions give C(π)=0C(\pi)=0, or J2/J1=1.15J_2/J_1=1.15, for AF2 and AF3, and J3/J1=[∣C(π)∣−∣C(0)∣]/4J1=0.65J_3/J_1=[\lvert C(\pi)\rvert-\lvert C(0)\rvert]/4J_1=0.65 for AF1 and AF2 in the relevant parameter sector. These straight lines represent global phase boundaries only along segments where no spiral has lower energy. In particular, the narrow H1 region near J2/J1=1.15J_2/J_1=1.15 interrupts the simple collinear assignment: the optimized spiral energy is lower there.

Distinct roles of the longer-range exchanges

The three added interactions modify the spiral in physically different ways. The J4J_4 path contributes −J4cos⁡(Φ+δ)-J_4\cos(\Phi+\delta) and changes the complex intersublattice exchange amplitude; J5J_5 contributes −J5cos⁡(2Φ−δ)-J_5\cos(2\Phi-\delta) and adds a longer-range intersublattice harmonic. By contrast, the equivalent-layer J6J_6 coupling contributes −2J6cos⁡2Φ-2J_6\cos2\Phi to the diagonal exchange spectrum. Thus J6J_6 changes the optimal pitch without directly changing the minimizing internal angle at a fixed pitch. The resulting pitch and internal angle must therefore be determined together. Separate sweeps of each coupling for both signs of J1J_1, including the numerical phase-atlas conventions and limitations, are given in the Supplemental Material [19]. They demonstrate changes of the collinear–spiral regions, but a change of phase label alone does not establish the number of competing local spiral minima.

Controlled progression from three to six exchanges

Their combined effect is examined by holding (J1,J2,J3)/J1=(1,0.25,−0.12)(J_1,J_2,J_3)/J_1=(1,0.25,-0.12) fixed and successively adding J4/J1=0.025J_4/J_1=0.025, J5/J1=0.045J_5/J_1=0.045, and J6/J1=−0.035J_6/J_1=-0.035. Each step adds one nonzero interaction without readjusting the preceding ones. The exchange constants are illustrative, not fitted to a particular compound.

For every parameter set, the exact function E0(Φ)E_0(\Phi) in Eq. (7) is minimized over 0≤Φ≤π0\le\Phi\le\pi, including both endpoints. The internal angle follows from Eq. (6). Table 1 lists the resulting ground-state quantities. For a perpendicular field, the reported saturation scale is BsatMs=Λ−λ+(0)+2KB_{\mathrm{sat}}M_s=\Lambda-\lambda_+(0)+2K, using the same K/J1=0.04K/J_1=0.04 for all four cases and a uniform ferromagnetic k=0k=0 mode. This field scale is derived in Sec. IX C. The added J4J_4 reduces the spiral wave vector, while J5J_5 makes the ferromagnetic state energetically preferred. The negative J6J_6 then restores a nonzero spiral pitch. Table 1 also shows the accompanying changes in the internal angle and hard-axis saturation field. These are comparisons between distinct Hamiltonians, not phase transitions induced by continuously varying one parameter. The example demonstrates

Modelqqδ\delta (rad)E0/J1E_0/J_1BsatMs/J1B_{\mathrm{sat}}M_s/J_1
J1J_1–J3J_30.24280.2421−1.03080.1008
J1J_1–J4J_40.09900.1068−1.03540.0804
J1J_1–J5J_500−1.08000.0800
J1J_1–J6J_60.20330.2266−1.05830.1283

Table 1. Controlled three-to-six-exchange sequence, with J1=1J_1=1, J2=0.25J_2=0.25, J3=−0.12J_3=-0.12, and successively added (J4,J5,J6)=(0.025,0.045,−0.035)(J_4,J_5,J_6)=(0.025,0.045,-0.035). Energy is per two-layer repeat. The saturation field uses K=0.04K=0.04. Values are obtained by global minimization of Eq. (7).

ends_mid=1 strates that longer-range exchange paths can change the magnetic order, not only the spiral period.

Quantitative sensitivity to exchange interactions

The equilibrium pitch and internal phase respond together when an exchange parameter changes. For a spiral whose equilibrium angles vary continuously with exchange, define u=(Φ,δ)T\mathbf{u} = (\Phi, \delta)^{T} and the homogeneous-angle Hessian Hij=∂2E/∂ui∂ujH_{ij} = \partial^{2}E/\partial u_i\partial u_j. Differentiating the two equilibrium equations with respect to JaJ_a yields

∂u∂Ja=−H−1∂∂Ja(∂ΦE∂δE).(25)\frac{\partial\mathbf{u}}{\partial J_a} = -H^{-1}\frac{\partial}{\partial J_a}\begin{pmatrix}\partial_{\Phi}E \\ \partial_{\delta}E\end{pmatrix}. \tag*{(25)}

For the four intersublattice bonds let r=(0,−1,1,−2)\mathbf{r} = (0,-1,1,-2) correspond to a=(1,2,4,5)a = (1,2,4,5), and put tr=δ+rΦt_r = \delta+ r\Phi and wr=JrABcos⁡trw_r = J_r^{AB}\cos t_r, where JrAB=(J1,J2,J4,J5)J_r^{AB} = (J_1,J_2,J_4,J_5) in the stated order. The explicit Hessian is

H=(∑rr2wr+2J3cos⁡Φ+8J6cos⁡2Φ∑rrwr∑rrwr∑rwr).(26)H = \begin{pmatrix} \sum_r r^2w_r + 2J_3\cos\Phi+ 8J_6\cos2\Phi& \sum_r rw_r \\ \sum_r rw_r & \sum_r w_r \end{pmatrix}. \tag*{(26)}

The mixed-derivative vector in Eq. (25) is (rsin⁡tr,sin⁡tr)T(r\sin t_r,\sin t_r)^T for a=1,2,4,5a = 1,2,4,5, (2sin⁡Φ,0)T(2\sin\Phi,0)^T for a=3a = 3, and (4sin⁡2Φ,0)T(4\sin2\Phi,0)^T for a=6a = 6. These expressions determine the local changes of both angles directly from the equilibrium Hessian. The derivative of the reduced wave vector is ∂q/∂Ja=(2π)−1∂Φ/∂Ja\partial q/\partial J_a = (2\pi)^{-1}\partial\Phi/\partial J_a.

For a perpendicular field, when the uniform exchange eigenmode is ferromagnetic and KK is fixed, hsat=BsatMs=λ+(Φ∗)−λ+(0)+2Kh_{\mathrm{sat}} = B_{\mathrm{sat}}M_s = \lambda_{+}(\Phi_*)-\lambda_{+}(0)+2K. Because the equilibrium pitch makes the first derivative of the energy vanish,

∂hsat∂Ja=∂λ+(Φ)∂Ja∣Φ∗−∂λ+(0)∂Ja.(27)\frac{\partial h_{\mathrm{sat}}}{\partial J_a} = \left.\frac{\partial\lambda_{+}(\Phi)}{\partial J_a}\right|_{\Phi_*} - \frac{\partial\lambda_{+}(0)}{\partial J_a}. \tag*{(27)}

For the present parameter set C(0)>0C(0) > 0, so these derivatives are cos⁡tr−1\cos t_r - 1 for a=1,2,4,5a = 1,2,4,5, 2(cos⁡Φ−1)2(\cos\Phi- 1) for a=3a = 3, and 2(cos⁡2Φ−1)2(\cos2\Phi- 1) for a=6a = 6.

The derivatives are evaluated for the six-exchange endpoint of Table I. The parameters are J1=1J_1 = 1, J2=0.25J_2 = 0.25, J3=−0.12J_3 = -0.12, J4=0.025J_4 = 0.025, J5=0.045J_5 = 0.045, J6=−0.035J_6 = -0.035, and K=0.04K = 0.04. The minimum occurs at Φ=1.2772\Phi= 1.2772 and δ=0.2266\delta= 0.2266, with hsat=0.1283h_{\mathrm{sat}} = 0.1283. The eigenvalues of the homogeneous-angle Hessian are 0.1619 and 1.0735 in the chosen exchange-energy units, showing that small uniform changes of the two angles cost exchange energy. For this parameter set, J6J_6 and J3J_3 have the largest local effects on the pitch, whereas J6J_6 has the largest effect on the saturation field. These derivatives describe a smooth spiral branch; they do not apply at a transition between competing minima. Numerical values and finite-difference checks are provided in the Supplemental Material.

Harmonic stability of spiral states

Changes of spin direction within the spiral plane

A minimum of E0(Φ)E_0(\Phi) is stable against a uniform change of pitch, but the individual layers may also undergo nonuniform rotations. Writing

θA,n=nΦ+uA,n,(28)\theta_{A,n} = n\Phi+ u_{A,n}, \tag*{(28)}
θB,n=nΦ+δ∗+uB,n,δ∗=−arg⁡C(Φ).(29)\theta_{B,n} = n\Phi+ \delta_* + u_{B,n}, \qquad\delta_* = -\arg C(\Phi). \tag*{(29)}

The intersublattice bonds have relative displacements r=0,−1,+1,−2r = 0,-1,+1,-2, respectively, and strengths JrAB=(J1,J2,J4,J5)J_r^{AB} = (J_1,J_2,J_4,J_5). The corresponding exchange contributions to the in-plane spin-rotation energy are

wr=JrABcos⁡(δ∗+rΦ),W(p)=∑r=0,−1,1,−2wreirp.(30)w_r = J_r^{AB}\cos(\delta_* + r\Phi), \qquad W(p) = \sum_{r=0,-1,1,-2} w_r e^{irp}. \tag*{(30)}

For Fourier amplitudes uα,n=N−1/2∑puα(p)eipnu_{\alpha,n} = N^{-1/2}\sum_p u_\alpha(p)e^{ipn}, the energy through second order is

ΔE∥=12∑pu(p)†H∥(p)u(p),(31)\Delta E_{\parallel} = \frac{1}{2}\sum_p u(p)^\dagger H_{\parallel}(p)u(p), \tag*{(31)}

with

H∥(p)=(A(p)−W(p)−W(p)∗A(p)).(32)H_{\parallel}(p) = \begin{pmatrix} A(p) & -W(p) \\ -W(p)^* & A(p) \end{pmatrix}. \tag*{(32)}
A(p)=∑rwr+2J3cos⁡Φ(1−cos⁡p)+2J6cos⁡2Φ(1−cos⁡2p).(33)A(p) = \sum_r w_r + 2J_3\cos\Phi(1-\cos p) + 2J_6\cos2\Phi(1-\cos2p). \tag*{(33)}

For each wave vector pp, the AA and BB layers can rotate mainly together or against one another. The corresponding coefficients of the quadratic spin-rotation energy are

λ∥,±(p)=A(p)±∣W(p)∣.(34)\lambda_{\parallel,\pm}(p) = A(p) \pm|W(p)|. \tag*{(34)}

At p=0p = 0, phase relaxation implies ∑rwr=∣C(Φ)∣≥0\sum_r w_r = |C(\Phi)| \ge0, so the acoustic branch obeys λ∥,−(0)=0\lambda_{\parallel,-}(0) = 0 and the optical branch has λ∥,+(0)=2∣C(Φ)∣\lambda_{\parallel,+}(0) = 2|C(\Phi)|. The zero-energy mode represents a common rotation of every spin in the plane. A spiral resists all small in-plane distortions only if λ∥,−(p)≥0\lambda_{\parallel,-}(p) \ge0 for every pp in the Brillouin zone, not merely E0′′(Φ)>0E_0''(\Phi) > 0.

Tilting spins out of the spiral plane

Small out-of-plane rotations are described by

eαn=1−zαn2(cos⁡θαn,sin⁡θαn,0)+zαnz^,(35)\mathbf e_{\alpha n} = \sqrt{1-z_{\alpha n}^2}(\cos\theta_{\alpha n},\sin\theta_{\alpha n},0) + z_{\alpha n}\hat{\mathbf z}, \tag*{(35)}

and retain terms quadratic in the deviations from the planar spiral. At an equilibrium spiral, the exchange field is parallel to each reference spin, with magnitude

Λ(Φ)=D(Φ)+∣C(Φ)∣=−E0(Φ).(36)\Lambda(\Phi) = D(\Phi) + |C(\Phi)| = -E_0(\Phi). \tag*{(36)}

The exchange-energy change for out-of-plane spin rotations is described by

H⊥(p)=Λ(Φ)I−J(p),(37)H_{\perp}(p) = \Lambda(\Phi)I - J(p), \tag*{(37)}

where J(p)J(p) is the exchange matrix in Eq. (8). Its eigenvalues are

λ⊥,−(p)=Λ(Φ)−[D(p)+∣C(p)∣]=E0(p)−E0(Φ),(38)\lambda_{\perp,-}(p) = \Lambda(\Phi) - [D(p) + |C(p)|] = E_0(p) - E_0(\Phi), \tag*{(38)}
λ⊥,+(p)=Λ(Φ)−[D(p)−∣C(p)∣].\lambda_{\perp,+}(p) = \Lambda(\Phi) - [D(p) - |C(p)|].

Here E0(p)E_0(p) denotes the same spiral energy as Eq. (7), evaluated at wave vector pp. The upper transverse eigenvalue is not independent of the lower one: λ⊥,+(p)=λ⊥,−(p)+2∣C(p)∣\lambda_{\perp,+}(p) = \lambda_{\perp,-}(p) + 2|C(p)|. Equation (38) relates transverse stability directly to the exchange spectrum. A global exchange-energy minimum has a nonnegative lower transverse branch. The in-plane fluctuation energies provide the complementary stability test.

Effect of easy-plane anisotropy

An easy-plane anisotropy is included as

HK=K∑n,α=A,B(eαnz)2,K>0.(39)\mathcal{H}_K = K\sum_{n,\alpha=A,B}(e^z_{\alpha n})^2, \qquad K > 0. \tag*{(39)}

Easy-plane anisotropy penalizes a spin component perpendicular to the spiral plane. It therefore increases the restoring energy for out-of-plane deviations without changing the in-plane result. In the normalization of Eq. (31), the transverse matrix gains 2KI2K I. For a planar spiral ii, all small out-of-plane rotations have nonnegative energy when

K≥Kc,i=12[E0(Φi)−min⁡p∈[−π,π]E0(p)].(40)K \ge K_{c,i} = \frac{1}{2}\left[E_0(\Phi_i) - \min_{p\in[-\pi,\pi]} E_0(p)\right]. \tag*{(40)}

The critical anisotropy is nonnegative, and it vanishes for the global ground state, for which E0(Φi)=min⁡pE0(p)E_0(\Phi_i)=\min_p E_0(p); Eq. (40) is therefore relevant for metastable spirals. At the threshold, one out-of-plane distortion still costs no energy to quadratic order. A positive restoring energy requires K>Kc,iK>K_{c,i}. Small in-plane rotations are stable when the eigenvalues in Eq. (34) are nonnegative throughout the Brillouin zone. This condition tests stability against small spin rotations; it does not determine the energy barriers between distinct magnetic states.

The harmonic spectra above refer to rotations about zero-field stationary spirals. An applied field changes the stationary spin configuration, so its stability must be examined using the Hessian evaluated at that field-dependent configuration. In particular, local stability of a planar solution is distinct from its equilibrium selection among competing planar and nonplanar states.

Response to a magnetic field

Unrestricted isotropic ground state

The field response depends on whether the spiral plane is free to rotate. We first consider the equilibrium response of the isotropic Hamiltonian, without constraining the spins to their zero-field plane. We denote the largest exchange eigenvalue by Λ=max⁡pλ+(p)\Lambda=\max_p\lambda_+(p) and its ordering wave vector by Q\mathbf{Q}. For the exchange parameters considered here, C(0)>0C(0)>0, so the uniform ferromagnetic mode has eigenvalue λ+(0)=D(0)+∣C(0)∣\lambda_+(0)=D(0)+\lvert C(0)\rvert. A field hh along zz couples to both layers through −h∑n,αenαz-h\sum_{n,\alpha}e_{n\alpha}^z.

A conical spiral combines the optimal transverse exchange mode at Q\mathbf{Q} with a uniform magnetization mm along the field:

eAn=(scos⁡nQ,ssin⁡nQ,m),eBn=(scos⁡(nQ+δQ),ssin⁡(nQ+δQ),m),(41)\begin{aligned} \mathbf{e}_{An}&=(s\cos nQ,s\sin nQ,m),\\ \mathbf{e}_{Bn}&=(s\cos(nQ+\delta_Q),s\sin(nQ+\delta_Q),m), \tag*{(41)} \end{aligned}

where s=1−m2s=\sqrt{1-m^2} and δQ=−arg⁡C(Q)\delta_Q=-\arg C(Q). Its energy per two-layer repeat is

Econe(m,h)=−Λ+[Λ−λ+(0)]m2−2hm.(42)E_{\mathrm{cone}}(m,h)=-\Lambda+[\Lambda-\lambda_+(0)]m^2-2hm. \tag*{(42)}

The conical spiral is the classical magnetization process of an isotropic helimagnet [20], and it is an exact ground state under the conditions stated here, for fields along the cone axis. At fixed uniform magnetization mm, the Fourier-space exchange bound puts the uniform magnetization in the p=0p=0 ferromagnetic eigenmode and bounds every remaining Fourier component by Λ\Lambda. The conical spiral in Eq. (41) has exactly this lowest exchange energy while preserving the length of each spin. Thus the cone is a global classical ground state for the present isotropic equal-diagonal exchange matrix when the uniform p=0p=0 mode is ferromagnetic (C(0)>0C(0)>0). For other exchange parameters, the uniform mode may have a different sublattice character, and the conical result then requires a separate examination.

Minimizing Eq. (42) yields

hsat=Λ−λ+(0),(43)h_{\mathrm{sat}}=\Lambda-\lambda_+(0), \tag*{(43)}
m(h)=min⁡(1,h/hsat),h≥0,m(h)=\min(1,h/h_{\mathrm{sat}}),\qquad h\geq0,
Egs(h)={−Λ−h2/hsat,0<h<hsat,−λ+(0)−2h,h≥hsat.E_{\mathrm{gs}}(h)= \begin{cases} -\Lambda-h^2/h_{\mathrm{sat}}, & 0<h<h_{\mathrm{sat}},\\ -\lambda_+(0)-2h, & h\geq h_{\mathrm{sat}}. \end{cases}

For a nonferromagnetic ground state with hsat>0h_{\mathrm{sat}}>0, the equilibrium wave vector Q\mathbf{Q} is independent of field below saturation. Rotational invariance makes this result independent of the direction of the laboratory field. The magnetization per repeat is 2m2m, so the corresponding susceptibility is 2/hsat2/h_{\mathrm{sat}} below saturation. If the exchange ground state is already ferromagnetic, Λ=λ+(0)\Lambda=\lambda_+(0) and hsat=0h_{\mathrm{sat}}=0: the zero-temperature isotropic ferromagnet is fully polarized for any h>0h>0, so the finite-saturation-field expressions h/hsath/h_{\mathrm{sat}} and 2/hsat2/h_{\mathrm{sat}} do not apply. A positive easy-plane anisotropy gives a separate, finite hard-axis saturation scale, as discussed below.

Constrained in-plane response

When the spiral plane is fixed and contains the field, individual layer angles can relax instead of forming the unrestricted cone. For a noncollinear stationary spiral, the harmonic Hessian in Eq. (32) determines its initial response. The magnetic field induces spin rotations with Fourier components at p=±Φip=\pm\Phi_i; for a generic incommensurate pitch, the source vector at p=Φip=\Phi_i is vi=(1,eiδi)T\mathbf{v}_i=(1,e^{i\delta_i})^T. The susceptibility follows from the quadratic change in exchange energy per two-layer repeat,

Eiplanar(h)=E0(Φi)−12χiplanarh2+O(h4),(44)E_i^{\mathrm{planar}}(h)=E_0(\Phi_i)-\frac{1}{2}\chi_i^{\mathrm{planar}}h^2+O(h^4), \tag*{(44)}
χiplanar=12vi†H(Φi)+vi.\chi_i^{\mathrm{planar}}=\frac{1}{2}\mathbf{v}_i^\dagger\mathbf{H}(\Phi_i)^{+}\mathbf{v}_i.

Here ++ denotes the inverse on the field-coupled subspace. A zero-energy mode coupled to the field produces a divergent harmonic response; a pseudoinverse does not remove that physical divergence. The factor 1/21/2 follows from the two conjugate Fourier sources and the energy normalization per repeat; it is not the susceptibility per spin. Commensurate cases where the two source wave vectors coincide require separate treatment. The susceptibility obtained in this way does not establish whether the distorted planar state is stable against all spin rotations.

Perpendicular field with easy-plane anisotropy

RM define their magnetic energy per pair of layers. Their Zeeman contribution is −BMs(cos⁡θA+cos⁡θB)-BM_s(\cos\theta_A+\cos\theta_B), and their easy-plane anisotropy is K(cos⁡2θA+cos⁡2θB)K(\cos^2\theta_A+\cos^2\theta_B), with MsM_s and KK defined for one layer [12]. Our unit-spin field variable is therefore h=BMsh = BM_s; in particular, the reduced field BMs/J1BM_s/J_1 used in RM’s Fig. 4 is exactly h/J1h/J_1, without an additional factor of two.

Adding the easy-plane term in Eq. (39) to the isotropic cone of Eq. (42) gives the perpendicular-field response. The spins cant uniformly while retaining their zero-field pitch and relative layer angle. Writing m=M/Ms=cos⁡θm = M/M_s = \cos\theta, the energy per two-layer repeat is

E⊥(m,B)=−Λ+[Λ−λ+(0)+2K]m2−2BMsm,(45)E_{\perp}(m,B)=-\Lambda+\left[\Lambda-\lambda_{+}(0)+2K\right]m^2-2BM_sm, \tag*{(45)}

where Λ=max⁡kλ+(k)\Lambda=\max_k\lambda_{+}(k). Minimization gives

Keff=K+12[Λ−λ+(0)],(46)K_{\mathrm{eff}}=K+\frac{1}{2}\left[\Lambda-\lambda_{+}(0)\right], \tag*{(46)}
Bsat=2KeffMs,m(B)=min⁡(1,BMs2Keff).B_{\mathrm{sat}}=\frac{2K_{\mathrm{eff}}}{M_s},\qquad m(B)=\min\left(1,\frac{BM_s}{2K_{\mathrm{eff}}}\right).

Equations (45) and (46) reduce directly to RM’s Eqs. (8) and (9) when J4=J5=J6=0J_4=J_5=J_6=0. The additional exchanges change the effective anisotropy and saturation field, not the form of the canting response. The uniform cone is the exact classical ground state also in the presence of the easy-plane term. At fixed mm the cone attains the exchange bound, and ∑n,α(en,αz)2≥(∑n,αen,αz)2/2N\sum_{n,\alpha}(e^z_{n,\alpha})^2 \geq(\sum_{n,\alpha}e^z_{n,\alpha})^2/2N, with equality only for uniform canting. The result assumes rigid equal-length layer moments, a common easy plane, and a uniform ferromagnetic k=0k=0 mode.

For the three-exchange and six-exchange endpoints of Table I, Eq. (46) gives Keff/J1=0.0504K_{\mathrm{eff}}/J_1=0.0504 and 0.06410.0641, respectively. The corresponding saturation fields, already listed in that table, differ by about 27%. The increase reflects the change in the exchange spectrum.

In-plane field: six-exchange energy and limits

For an in-plane field, let ana_n and bnb_n be the independent planar angles of the two layers in repeat nn. The direct six-exchange generalization of RM’s (10) is

E∥=−1N∑n=0N−1{BMs[cos⁡an+cos⁡bn]+J1cos⁡(bn−an)+J2cos⁡(an+1−bn)+J3[cos⁡(an+1−an)+cos⁡(bn+1−bn)]+J4cos⁡(bn+1−an)+J5cos⁡(an+2−bn)+J6[cos⁡(an+2−an)+cos⁡(bn+2−bn)]}.(47)E_{\parallel}=-\frac{1}{N}\sum_{n=0}^{N-1}\left\{BM_s[\cos a_n+\cos b_n]+J_1\cos(b_n-a_n)+J_2\cos(a_{n+1}-b_n) +J_3[\cos(a_{n+1}-a_n)+\cos(b_{n+1}-b_n)] +J_4\cos(b_{n+1}-a_n)+J_5\cos(a_{n+2}-b_n) +J_6[\cos(a_{n+2}-a_n)+\cos(b_{n+2}-b_n)]\right\}. \tag*{(47)}

The sum is over two-layer repeats with periodic or suitably commensurate boundary conditions. Our expression reduces to RM’s planar energy for J4=J5=J6=0J_4=J_5=J_6=0. Here the spins are restricted to the plane, so their possible rotation out of that plane is excluded. A positive easy-plane anisotropy makes such configurations physically relevant, but only a sufficiently strong anisotropy stabilizes them against out-of-plane spin rotations. A minimum within the plane need not be stable in three dimensions.

Recovery of the RM in-plane-field comparison

Instead of assuming that the extended model reproduces the RM field response, we calculated the constrained-planar response of RM for J1=1J_1=1, J3=−0.12J_3=-0.12, and J2=0.20,0.22,0.25,0.30J_2=0.20,0.22,0.25,0.30 with the six-exchange extension at J2=0.25J_2=0.25 and (J4,J5,J6)=(0.025,0.045,−0.035)(J_4,J_5,J_6)=(0.025,0.045,-0.035). For each parameter set, we minimized Eq. (47) on periodic chains chosen to approximate the infinite-chain pitch closely: N=101,67,70N=101,67,70 and 104104 for J2=0.20,0.22,0.25J_2=0.20,0.22,0.25 and 0.300.30, and N=59N=59 for the six-exchange model. The residual mismatch between the optimal pitch and an allowed k/Nk/N is below 2×10−42\times10^{-4}; exact commensurability is not implied. A fixed chain length pins the zero-field pitch to q=k/Nq=k/N; for N=40N=40 this changes the low-field slope by up to 7% (Supplemental Material [19]). At each of 101 fields in 0≤h/J1≤0.100\leq h/J_1\leq0.10, the lowest energy was taken from relaxations started from the three best commensurate spirals, the ferromagnet, a random configuration and the solution at the preceding field, using analytical gradients. The RM curves recover the qualitative low-field distortion and differing approaches to saturation reported in Ref. [12]. The triangles in Figure 3 mark h=Λ−λ+(0)h=\Lambda-\lambda_{+}(0), the local-stability threshold of the in-plane ferromagnet. This threshold need not coincide with the equilibrium transition field. The six-exchange example changes the preferred zero-field pitch from q≃0.24281q\simeq0.24281 to q≃0.20327q\simeq0.20327 and raises this field from 0.0298J10.0298J_1 to 0.048J10.048J_1. These results demonstrate the sensitivity of constrained-planar field response to longer-range exchange, but do not isolate individual contributions of J4J_4, J5J_5, and J6J_6. The plotted energy is the smallest found among the numerical starting configurations at each field, not a proof of the absolute planar minimum. Apparent jumps can reflect crossings of competing planar branches or finite-chain commensuration; assigning an equilibrium first-order transition requires converged branch energies, a system-size study, and stability against nonplanar rotations.

Figure 3. Constrained in-plane magnetization. (a) RM limit for J1=1J_1 = 1, J3=−0.12J_3 = -0.12, and four values of J2J_2. (b) Comparison at J2=0.25J_2 = 0.25 with the six-exchange extension (J4,J5,J6)=(0.025,0.045,−0.035)(J_4, J_5, J_6) = (0.025, 0.045, -0.035). Each point is the lowest energy found among six initial configurations on periodic chains selected to approximate the zero-field pitch (the allowed values are q=k/Nq = k/N). Triangles mark h=Λ−λ+(0)h = \Lambda- \lambda_{+}(0), where the in-plane ferromagnet becomes locally stable. The planar curves are sampled at intervals of 0.001J10.001J_1; triangles indicate local stability rather than an equilibrium transition.

a. Transverse stability of field-distorted planar states. The in-plane relaxation fixes a stationary set of angles {φαn(h)}\{\varphi_{\alpha n}(h)\}, but does not establish stability against rotations out of the easy plane. Let H⊥(h,0)H_{\perp}(h,0) be the transverse Hessian of the exchange and Zeeman energy, evaluated at that stationary planar configuration with K=0K=0, using the same quadratic-energy normalization as Eq. (31). Since the easy-plane anisotropy vanishes for every planar spin configuration, the stationary angles are independent of KK, and the transverse Hessian at finite anisotropy is H⊥(h,K)=H⊥(h,0)+2KIH_{\perp}(h,K)=H_{\perp}(h,0)+2KI. Consequently, the minimum anisotropy required for local trans-verse nonnegativity of this particular branch is

Kc(h)=max⁡{0,−12λmin⁡[H⊥(h,0)]}.(48)K_{\mathrm{c}}(h)=\max\left\{0,-\frac{1}{2}\lambda_{\min}\left[H_{\perp}(h,0)\right]\right\}. \tag*{(48)}

This field-dependent condition extends Eq. (40) to nonuniform planar configurations. It is not an equilibrium BB–KK phase boundary: in-plane stability and comparison with other planar and nonplanar states remain necessary. This transverse criterion is separate from the planar energy comparison and from the exact hard-axis conical response in Eqs. (45) and (46).

b. Field response along the exchange sequence. Adding the exchange interactions in sequence also changes the spin-wave stability spectra and hard-axis magnetization. In particular, the five-exchange ferromagnet has Λ=λ+(0)\Lambda=\lambda_{+}(0) and hence hsat=2Kh_{\mathrm{sat}}=2K, whereas the sixth exchange restores a spiral and a finite exchange contribution to saturation. The full spectral and magnetization comparison is provided in the Supplemental Material [19]; these hard-axis curves follow the exact conical solution and should not be confused with nonlinear in-plane field relaxation.

Information obtainable from magnetic measurements

A measured spiral pitch qq, internal phase δ\delta, hard-axis saturation scale hsath_{\mathrm{sat}}, and constrained-planar susceptibility χ∥\chi_{\parallel} provide four quantities for six exchange constants. At most four combinations of the exchanges can therefore be determined, and at least two remain free. For the representative six-exchange parameters, the 4×64\times6 Jacobian of these observables has the maximal rank four, so exactly four combinations are constrained to first order. The sensitivity differs strongly between these combinations. With exchanges expressed in the units and parameterization used in the Supplemental Material, the singular values of the relative Jacobian ∂ln⁡Oμ/∂Ja\partial\ln O_{\mu}/\partial J_a are 72.9, 9.3, 6.0 and 1.0. Their magnitudes and condition number depend on this choice of exchange coordinates and observable normalization; they do not by themselves

quantify experimental uncertainties. The weakest combination changes all four observables by only about 1% for a change of 0.01J10.01J_1, so it is poorly determined by data of realistic precision. The Jacobian and its assumptions are given in the Supplemental Material [19]. Wave-vector-resolved measurements, such as inelastic neutron scattering, or independent first-principles exchange information are therefore necessary.

Discussion

The influence of an interlayer exchange on the ordering wave vector depends on its contribution to the exchange spectrum, not simply on its distance or magnitude. Even relatively small additional couplings can change the preferred stacking. Consequently, collinear energy comparisons alone may miss the magnetic ground state of a layered kagome magnet.

Different measurements probe different aspects of this competition. The hard-axis saturation field measures the separation between the ordering and uniform exchange eigenvalues, together with the anisotropy contribution, whereas the in-plane response probes the relaxation of individual layer directions. The latter can involve competing stationary configurations, so the finite-chain curves cannot alone establish equilibrium field-induced transitions. The weakly constrained exchange combination found in § 8.5 further motivates momentum-resolved excitation measurements or independent first-principles exchange calculations.

The conclusions apply to the classical rigid-layer model. For specific AT6X6\mathrm{AT}_{6}\mathrm{X}_{6} compounds [10, 11], the validity of intralayer rigidity and the possible importance of other magnetic sublattices, anisotropic interactions, and longitudinal fluctuations must be assessed separately.

Conclusions

For two equivalent magnetic layers per repeat, the Luttinger–Tisza construction gives an exact classical single-QQ ground state. In the six-exchange model, additional interlayer couplings can change the preferred order from spiral to ferromagnetic and back to spiral. The finite-chain planar field response illustrates competing configurations but does not, by itself, establish equilibrium first-order transitions.

The four magnetic observables considered here constrain at most four combinations of the six exchange constants, with one combination weakly determined. Determining the magnetic stacking therefore requires a sufficiently complete exchange description and comparison with noncollinear states; momentum-resolved measurements can supply further constraints.

Acknowledgments

Work at Ames National Laboratory was supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division. Ames National Laboratory is operated for the U.S. DOE by Iowa State University under Contract No. DE-AC02-07CH11358.

References

References

  1. [1]X. Xu, J.-X. Yin, Z. Qu, and S. Jia, “Quantum interactions in topological R166 kagome magnet,” Rep. Prog. Phys. 86, 114502 (2023), doi:10.1088/1361-6633/acf3d.DOI
  2. [2]P. Negi, K. Medhi, A. Pancholi, and S. Roychowdhury, “Magnetic Kagome materials: bridging fundamental properties and topological quantum applications,” Mater. Horiz. 12, 4510–4544 (2025), doi:10.1039/D5MH00120J.DOI
  3. [3]A. K. Sharma, S. Chatterjee, P. Yanda, C. Felser, and C. Shekhar, “Topological quantum materials: kagome, chiral, and square-net frameworks,” npj Quantum Mater. 11, 55 (2026), doi:10.1038/s41535-026-00902-7.DOI
  4. [4]G. Venturini, D. Fruchart, and B. Malaman, “Incommensurate magnetic structures of RMn₆Sn₆ (R = Sc, Y, Lu) compounds from neutron diffraction study,” J. Alloys Compd. 236, 102–110 (1996), doi:10.1016/0925-8388(95)01998-7.DOI
  5. [5]N. J. Ghimire et al., “Competing magnetic phases and fluctuation-driven scalar spin chirality in the kagome metal YMn₆Sn₆,” Science Advances 6, eabe2680 (2020), doi:10.1126/sciadv.abe2680.DOI
  6. [6]Q. Wang et al., “Field-induced topological Hall effect and double-fan spin structure with a c-axis component in the metallic kagome antiferromagnetic compound YMn₆Sn₆,” Phys. Rev. B 103, 014416 (2021), doi:10.1103/PhysRevB.103.014416.DOI
  7. [7]R. L. Dally et al., “Chiral properties of the zero-field spiral state and field-induced magnetic phases of the inter-ant kagome metal YMn₆Sn₆,” Phys. Rev. B 103, 094413 (2021), doi:10.1103/PhysRevB.103.094413.DOI
  8. [8]S. X. M. Riberolles et al., “New insight into tuning magnetic phases of RMn₆Sn₆ kagome metals,” npj Quantum Mater. 9, 42 (2024), doi:10.1038/s41535-024-00656-0.DOI
  9. [9]N. L. T. V. Trevisan, and R. J. M. McQueeney, “High-field magnetic phase diagrams of the RMn₆Sn₆ (R = Gd–Tm) kagome metals,” Phys. Rev. B 111, 054410 (2025), doi:10.1103/PhysRevB.111.054410.
  10. [10]S. Chen, Z. Zhang, V. Antropov, and Y. Sun, “Computational prediction of ferromagnetic AT₆X₆ kagome compounds,” Phys. Rev. B 113, 224412 (2026), doi:10.1103/t61c-xmn3.DOI
  11. [11]S. Chen, V. Antropov, and Y. Sun, “High-throughput identification of ferromagnetic kagome candidates in the AT₆X₄ and AT₆X₅ families,” arXiv:2608.22666 (2026).arxiv.org/abs/2608.22666
  12. [12]E. V. Rosenfeld and N. V. Mushnikov, “Double-flat-spiral magnetic structures: Theory and application to the RMn₆X₆ compounds,” Physica B: Condens. Matter 403, 1898–1906 (2008), doi:10.1016/j.physb.2007.10.220.
  13. [13]E. V. Rosenfeld, N. V. Mushnikov, and V. V. Dyakin, “Nested spirals in layered magnets,” Phys. Met. Metallogr. 107, 557–568 (2009), doi:10.1134/S0031918X09060052.
  14. [14]E. V. Rosenfeld, N. V. Mushnikov, and V. V. Dyakin, “Spiral structures in layered magnets,” Phys. Status Solidi B 246, 2187–2193 (2009), doi:10.1002/pssb.200945143.DOI
  15. [15]J. M. Luttinger and L. Tisza, “Theory of dipole interaction in crystals,” Phys. Rev. 70, 954 (1946).DOI
  16. [16]D. H. Lyons and T. A. Kaplan, “Method for determining ground-state spin configurations,” Phys. Rev. 120, 1580 (1960).DOI
  17. [17]T. A. Kaplan and N. Menyuk, “Spin ordering in three- dimensional crystals with strong competing exchange interactions,” Philos. Mag. 87, 3711 (2007).
  18. [18]A. Yoshimori, “A new type of antiferromagnetic structure in the rutile type crystal,” J. Phys. Soc. Jpn. 14, 807 (1959).DOI
  19. [19]See Supplemental Material at [URL will be inserted by publisher] for the phase atlases of the individual longer-range exchanges, the sensitivity data, the harmonic spectra along the exchange sequence, the finite-chain calculations, and the identifiability analysis.
  20. [20]Y. Kitano and T. Nagamiya, “Magnetization process of a screw spin system,” Prog. Theor. Phys. 31, 1 (1964).

Paper details

vlnc:W_tej48z32
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 13

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