Resolving the Magnetic Ground State and Field-Induced Transitions in Magnetic Dirac Semimetal Candidate EuMnSb$_2$

Abstract: The magnetic structure of a magnetic topological semimetal EuMnSb$_2$ is investigated in fields up to 30 T using polarized and unpolarized neutron diffraction, pulsed-field x-ray magnetic circular dichroism and pulsed-field magnetometry. We determine the zero-field magnetic structures of the Eu and Mn sublattices, and find that magnetic transitions induced by applied fields below 2 T correspond to changes in the magnetic order of the Eu spins alone without detectable perturbation to the order of the Mn spins. An additional magnetic transition is observed at fields close to the saturation field for the Eu spins. We present a mean-field model which describes key features of the magnetic behavior and allows us to estimate the dominant Eu--Eu and Eu--Mn exchange interactions responsible for the coupling between magnetism and electronic topology.

• The paper resolves EuMnSb₂’s magnetic structure as C-type antiferromagnetic Mn moments locked along the a axis and canted A-type antiferromagnetic Eu moments, using spherical neutron polarimetry on a single crystal.
• Element-specific XMCD shows that fields up to 30 T reorient only the Eu spins, while the Mn sublattice remains rigid; magnetometry identifies spin flops at 1.5 T for H ∥ a and 0.5 T for H ∥ c, plus a higher-field reorientation near saturation.
• The mean-field model finds dominant Eu interlayer antiferromagnetic exchange (J₁ = −0.149 meV) and reproduces key spin-flop behavior, while the microscopic Eu–Mn coupling and the Hc₂ transition for H ∥ c remain unresolved.

EuMnSb$_2$ is a layered antimonide in which Dirac-like electronic bands coexist with two magnetically ordered sublattices, making the precise magnetic structure a prerequisite for interpreting its topological properties. Electronic structure calculations predict that the material is either a gapped Dirac semimetal or a Weyl semimetal depending on the assumed magnetic configuration, and magnetotransport measurements show that Eu ordering strongly modulates the conductivity. Yet four prior neutron diffraction studies disagreed on the Eu spin arrangement, owing to severe neutron self-absorption by natural Eu ($\sigma_a = 4530$ b at 1.8 Å), partial $bc$-plane twinning, and sample-to-sample variation between growth methods. The work reported here resolves these ambiguities by applying spherical neutron polarimetry (SNP)—which is immune to absorption corrections because it relies on intensity ratios for opposite neutron polarizations—together with unpolarized diffraction, pulsed-field XMCD, and steady- and pulsed-field magnetometry, all performed on a single Sn-flux-grown crystal (2608.16724).

Sample and experimental approach

The crystal studied was grown by the Sn flux method and belongs to the semiconducting class of EuMnSb$_2$: the zero-field $c$-axis resistivity rises steeply on cooling, with a shoulder near $T_{\mathrm{Eu1}} \simeq 21$ K reflecting coupling between transport and Eu magnetic order. X-ray characterization established a 2:1 twin ratio in the $bc$ plane, high crystalline quality, and no detectable Sn inclusions. SNP measurements were carried out on the D3 diffractometer at the ILL using CryoPad at $\lambda = 0.83$ Å, with polarization matrices $P_{\alpha\beta}$ collected at 2, 7.5, and 30 K. Pulsed-field magnetization up to 35 T (pulse length ≈7 ms) was measured at Oxford's Nicholas Kurti laboratory down to 500 mK, and pulsed-field XMCD at the Eu $M_5$ (1130 eV) and Mn $\sigma_a = 4530$0 (640.5 eV) edges was recorded at BESSY II in fields up to 30 T.

Zero-field magnetic structures

At 30 K, where only the Mn sublattice is ordered, the polarization matrix for the (210) reflection confirms the previously reported C-type collinear antiferromagnetic structure with Mn moments of approximately 4.2 $\sigma_a = 4530$1 along the $\sigma_a = 4530$2 axis: vanishing $\sigma_a = 4530$3 excludes a $\sigma_a = 4530$4-axis component, while non-zero $\sigma_a = 4530$5 indicates nuclear–magnetic interference from an $\sigma_a = 4530$6-plane spin component. A notable finding is that the magnetic phase is almost entirely a single time-reversal domain of the Mn structure; the authors leave the origin of this large domain imbalance as an open question.

Below both Eu transitions, refinement of the SNP data at 7.5 K and 2 K yields a canted A-type antiferromagnetic Eu structure consistent in outline with Wilde et al., but quantitatively distinct: the canting angle in the $\sigma_a = 4530$7 plane refines to about 35° from $\sigma_a = 4530$8 at 7.5 K (38° at 2 K), close to earlier values of 31° and 41°, whereas the $\sigma_a = 4530$9-plane canting is only ~15° from $bc$0 (16° at 2 K)—substantially smaller than the ~50° reported previously at 5 K. This discrepancy matters because the size of the transverse Eu moment directly affects which topological band structure calculations should target.

Field-induced transitions and the rigidity of the Mn sublattice

Pulsed-field XMCD provides the key evidence that all field-induced transitions below 30 T involve the Eu spins alone. The Eu $bc$1 edge signal shows canting beginning at very low field and saturating around 15 T for both $bc$2 and $bc$3, while the Mn $bc$4 edge XMCD remains absent up to 30 T, demonstrating that the strong Mn–Mn exchange keeps the Mn sublattice collinear and effectively rigid throughout the investigated field range. This justifies treating the Mn moments as a fixed staggered background in any model of the Eu physics.

Magnetization data reveal spin-flop transitions at $bc$5 T for $bc$6 and 0.5 T for $bc$7, with no corresponding anomaly for $bc$8, whose magnetization instead shows negative curvature. In pulsed fields the magnetization saturates at approximately 19 T at the lowest temperature, and for $bc$9 an additional step $_2$0 appears a few tesla below saturation, indicating a further field-induced spin reorientation unique to that orientation. The resulting field–temperature phase diagrams map $_2$1, $_2$2, and the field-suppressed $_2$3 for both orientations.

Mean-field model

The authors construct a minimal Hamiltonian comprising nearest-neighbor AFM exchange $_2$4 between Eu layers within a Mn-layer spacing, FM exchange $_2$5 across Mn layers, Zeeman coupling, and two phenomenological staggered fields $_2$6 and $_2$7 representing the effective field exerted by the ordered Mn sublattice on the Eu moments. Exchange alone cannot stabilize the observed canting—single-ion anisotropy terms are symmetry-forbidden—and the $_2$8-axis staggered component in particular requires an anisotropic Eu–Mn exchange tensor coupling an $_2$9-axis Mn moment to a $c$0-axis Eu moment. The fitted parameters are:

Parameter	Value (meV)
$c$1	−0.149
$c$2	0.002
$c$3	0.081
$c$4	0.021

The dominance of $c$5 over $c$6 by nearly two orders of magnitude identifies the interlayer AFM coupling as the primary interaction governing the Eu spin dynamics. The model reproduces the spin-flop transitions for $c$7 and $c$8 reasonably well, including the 0.5 T flop for $c$9. For $T_{\mathrm{Eu1}} \simeq 21$0, the simulation predicts a step-like feature at 15.5 T associated with pairwise alignment of the four Eu spins just before saturation, closely resembling the experimentally observed $T_{\mathrm{Eu1}} \simeq 21$1 for $T_{\mathrm{Eu1}} \simeq 21$2; the authors therefore suggest that $T_{\mathrm{Eu1}} \simeq 21$3 corresponds to an analogous pre-saturation spin alignment. The model nonetheless fails in two respects: it predicts linear rather than negatively curved magnetization for $T_{\mathrm{Eu1}} \simeq 21$4, and it produces no $T_{\mathrm{Eu1}} \simeq 21$5 anomaly for $T_{\mathrm{Eu1}} \simeq 21$6. The staggered fields remain phenomenological parameters, and the authors note that establishing their microscopic origin would require a detailed calculation of the Eu–Mn exchange tensor, with antisymmetric Dzyaloshinskii–Moriya interactions (allowed by the absence of inversion at relevant Eu–Mn paths) and dipolar couplings identified as candidate mechanisms.

Limitations and open questions

Three issues remain unresolved. First, the origin of the pronounced single time-reversal-domain population, inferred from non-zero off-diagonal polarization matrix elements at (210) and (220), is unexplained despite being energetically unexpected. Second, the microscopic identity of the $T_{\mathrm{Eu1}} \simeq 21$7 transition for $T_{\mathrm{Eu1}} \simeq 21$8—and whether it is related to the pre-saturation pair-alignment transition predicted for $T_{\mathrm{Eu1}} \simeq 21$9—is inferred but not demonstrated. Third, the microscopic mechanism coupling Eu and Mn spins, which underpins both the canted ground state and the coupling between magnetism and electronic topology, is represented only phenomenologically. Additionally, the conclusions apply specifically to the semiconducting Sn-flux-grown composition; given the documented sensitivity of structure and transport to growth route and stoichiometry, extension to metallic tetragonal or Mn-rich samples is not established here.

Conclusion

By combining absorption-free spherical neutron polarimetry, element-specific pulsed-field XMCD, and high-field magnetometry on a single crystal, this study settles the long-standing disagreement over the EuMnSb$bc$0 magnetic ground state: a C-type AFM Mn sublattice locked along $bc$1, and a canted A-type AFM Eu sublattice with a smaller $bc$2-axis component than previously reported, undergoing low-field spin flops and a pre-saturation reorientation driven entirely by the Eu spins. The extracted exchange parameters establish $bc$3 as the dominant Eu interaction and provide a quantitative framework for modeling magnetism–topology coupling in $bc$4Mn$bc$5 compounds, while leaving the microscopic Eu–Mn coupling mechanism and the high-field $bc$6 transition as clearly posed open problems.