Axion electrodynamics in a topologically trivial antiferromagnet Authors: Abhilash Mishra, Robin Karothiya, Syed Qamar Abbas Shah, Arti Kashyap, Christian Binek Published: 2026 ABSTRACT Axion electrodynamics, originally proposed by Frank Wilczek in the context of high energy physics, can manifest in condensed matter systems as an isotropic linear magnetoelectric (ME) response. Yet, experimental realization of a continuous and purely monopolar ME response has remained elusive. Here, we demonstrate such a response by transforming the topologically trivial antiferromagnet chromia into an isotropic ME medium. Using low-frequency AC ME susceptometry, supported by Monte Carlo simulations, we reveal a monopolar ME response \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{{ij}}=\theta {\delta }_{{ij}}$$\end{document}αij=θδij resulting from suppressed quadrupolar contributions. The zero crossing of the temperature-dependent θ at 168 K provides direct evidence that the measured isotropic ME response arises from the axion contribution in chromia. These results establish a condensed-matter platform for axion electrodynamics and open pathways toward ME Hall effects, magnetophotovoltaic responses, and monopolar-based devices. Introduction The linear magnetoelectric (ME) effect is a tensorial property that relates the electrically induced magnetization \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${M}_{{{\rm{j}}}}$$\end{document}Mj to an applied electric field \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${E}_{{{\rm{i}}}}$$\end{document}Ei such that \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{{\mu }}}_{0}{M}}_{{{\rm{j}}}}={{{\alpha }}}_{{{\rm{ij}}}}{E}_{i}$$\end{document}μ0Mj=αijEi1 where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\mu }}}}_{0}$$\end{document}μ0 is the vacuum permeability and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}$$\end{document}αij is ME susceptibility tensor. The transposed tensor, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$({{\rm{\alpha }}}_{{{\rm{ji}}}})$$\end{document}(αji), governs the converse ME effect, in which a magnetic field induces polarization. In the archetypal ME antiferromagnet Cr2O3 (chromia), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}$$\end{document}αij is diagonal in the basis aligned with the crystallographic c-axis ([0001]) and its perpendicular directions. Symmetry constraints imposed by chromia’s magnetic point group 3̅'m’ break spatial inversion (P) and time-reversal (T) symmetries while preserving their combination (PT). These constraints dictate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}={\alpha }_{\perp }$$\end{document}αij=α⊥ for i=j=1i=j=1 and 2, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}={\alpha }_{\parallel }$$\end{document}αij=α∥ for i=j=3i=j=3, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}=0$$\end{document}αij=0 otherwise2–4. Such a susceptibility tensor can be decomposed into a trace free and a pseudoscalar component as1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}=\frac{1}{3}\left({\alpha }_{\perp }-{\alpha }_{\parallel }\right){Q}_{{ij}}+\theta {\delta }_{{ij}}.$$\end{document}αij=13α⊥−α∥Qij+θδij. Here, Qij = diag(1, 1, −2) is a traceless tensor and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta=\frac{1}{3}\left(2{\alpha }_{\perp }+{\alpha }_{\parallel }\right)$$\end{document}θ=132α⊥+α∥. The Qij-term represents the ME quadrupole, while θ captures the pseudoscalar, or axion, contribution. In contemporary descriptions of magnetoelectricity, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}$$\end{document}αij is often interpreted within a multipolization framework2, where the traceless component corresponds to the quadrupole term and the pseudoscalar component to the ME monopole. The connection between the decomposition in Eq. (1) and axion electrodynamics hinges on the ME monopole response \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{{ij}}=\theta {\delta }_{{ij}}$$\end{document}αij=θδij. In macroscopic electrodynamics, the constitutive relations of a material can be derived from an effective Lagrangian density \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L({{\bf{E}}},{{\bf{B}}})$$\end{document}L(E,B) via \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{D}}}=\frac{\partial L}{\partial {{\bf{E}}}},{{\bf{H}}}=-\frac{\partial L}{\partial {{\bf{B}}}}.$$\end{document}D=∂L∂E,H=−∂L∂B. Gauge invariance permits adding the topological axion term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{\theta }\propto \theta \,$$\end{document}Lθ∝θE · B to the standard Lagrangian3. For a constant axion field θ, the Euler–Lagrange equations reproduce the standard Maxwell equations in the bulk, while the constitutive relations include the pseudoscalar ME coupling. When θ varies in space or time, as occurs at surfaces of topological insulators, the θ-term produces the modified field equations of axion electrodynamics. In topologically trivial materials such as engineered Cr₂O₃ powder, θ is constant throughout the bulk and up to the surface, so the bulk Maxwell equations remain unchanged, and the effect of θ appears entirely through the constitutive relations. A medium whose constitutive tensor contains only the monopole component \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{{ij}}=\theta {\delta }_{{ij}}$$\end{document}αij=θδij was implicitly envisioned in Wilczek’s 1987 proposal3, where he speculated that “it is not beyond the realm of possibility that fields whose properties partially mimic those of axion fields can be realized in condensed-matter systems.” Our engineered material realizes the macroscopic response envisioned by Wilczek more directly than topological insulators, because it exhibits a bulk pseudoscalar ME response corresponding to a constant θ throughout the sample, without relying on surface-induced θ variations. The existence of a non-zero axion component in materials was long debated because the Post constraint was widely assumed to enforce \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\alpha }}}}_{{{\rm{ij}}}}={{{\rm{\alpha }}}}_{{{\rm{ji}}}}$$\end{document}αij=αji and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop{\sum }_{{\rm{i}}}{{{\alpha }}}_{{{\rm{ii}}}}=0$$\end{document}∑iαii=04,5. This debate was resolved only after measurements in Cr₂O₃ demonstrated a finite trace with the unambiguous experimental confirmation of a non-zero axion piece6,7. Here, the temperature-dependent parallel and perpendicular ME susceptibilities sum up to a finite trace of (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{\alpha }}}_{{{\rm{ij}}}}$$\end{document}αij) such that \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta=\frac{1}{3}\left(2{\alpha }_{\perp }+{\alpha }_{\parallel }\right)\ne 0$$\end{document}θ=132α⊥+α∥≠0. However, θ ≠ 0 does not preclude the presence of the ME quadrupole response. Hence, the ME properties of single-crystalline chromia are not fully captured by axion electrodynamics. Nevertheless, the observation of a non-zero axion component has motivated the search for ME materials with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{\perp }={\alpha }_{\parallel }$$\end{document}α⊥=α∥, where the quadrupolar ME response would be zero and only the pure monopolar ME response remains. Although single-crystalline chromia inevitably exhibits a quadrupolar response, it has been proposed that its ME susceptibility can be exploited to generate a magnetic field of a magnetic monopole8,9. This effect arises from the ME response to a mirror charge induced by a real point charge in proximity of the surface8. This prediction should not be conflated with the realization of a material exhibiting a purely monopolar ME response. Although related, the monopolar ME response and the monopole magnetic stray field are two distinct phenomena. We demonstrate a long theorized continuous, purely isotropic ME monopole response near room-temperature, distinct from the quantized low-temperature responses in topological materials6. Previous efforts to realize ME isotropic response have included effective medium approaches using powder samples. Some of us recently demonstrated that ME annealing of powders derived from materials with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop{{\mathrm{lim}}}_{T\to {T}_{N}^{-}}\frac{{\alpha }_{\parallel }(T)}{{\alpha }_{\perp }(T)} < 0$$\end{document}limT→TN−α∥(T)α⊥(T)<0, such as chromia (Néel temperature TN = 307 K), cannot produce isotropic ME response10. Intuitively, this is because ME annealing requires the application of electric and magnetic fields during cooling below TN. This protocol imposes a directional bias, thereby breaking isotropy. Decades earlier, Shtrikman and Treves noted that powders composed of diagonal ME materials with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop{{\mathrm{lim}}}_{T\to {T}_{N}^{-}}\frac{{\alpha }_{\parallel }(T)}{{\alpha }_{\perp }(T)} > 0$$\end{document}limT→TN−α∥(T)α⊥(T)>0 could, in principle, exhibit isotropic ME behaviour in response to ME annealing11. However, no such material has been experimentally identified to date. Recently, metamaterials with broken parity and time inversion symmetry leading to axion response have been proposed but not experimentally realized as of today12. Furthermore, a quantized isotropic ME response has been predicted for ideal topological insulator (TI) materials13. However, in realistic magnetic TI materials, this response rapidly decays due to disorder. In addition, magnetic TIs such as Cr-doped (Bi,Sb)₂Te₃ exhibit very low magnetic ordering temperatures (~15 K)14,15. This limits the practicality of TIs for realizing a monopolar ME response, particularly for applications near room temperature. Here, we present the realization of an engineered, topologically trivial material that exhibits a continuous, pure, and robust ME monopole response up to the Néel temperature, making it promising for fundamental studies and practical applications. In contrast to the previously observed quantized, low-temperature topological ME effect16, the monopolar response reported here is non-quantized and is measured by quasi-static susceptometry up to 307 K. Figure 1 is a schematic illustration that contrasts two distinct scenarios. It shows why ME annealing fails to generate a monopolar ME response in ordinary chromia powder (Fig. 1a), and how ME annealing a chromia single crystal followed by powdering at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T < {T}_{N}$$\end{document}T TN, zero-crossing at T = 168 K is fundamentally inaccessible under any field-cooling protocol (see ref. 10). This limitation applies to all powders with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathop{{\mathrm{lim}}}_{T\to {T}_{N}^{-}}\frac{{\alpha }_{\parallel }(T)}{{\alpha }_{\perp }(T)} < 0$$\end{document}limT→TN−α∥(T)α⊥(T)<0, encompassing all experimentally investigated ME crystalline materials. Conversely, observing a zero-crossing at T = 168 K in a chromia-based engineered material indicates that pure isotropic ME response has been realized. The present study identifies an unexplored pathway for engineering axion response of ME samples. The emergence of a pure ME monopole response in chromia-based samples leads to the realization of the predicted zero-crossing at T = 168 K, as confirmed by both experiment and simulation. Crucially, the origin of this monopole response can be traced to a subtle but decisive symmetry consideration. As illustrated in Fig. 1c, a π-rotation of an antiferromagnetic single-domain grain about an axis perpendicular to its threefold axis leaves the Néel vector, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{L}}}={{{\bf{M}}}}_{{{\bf{1}}}}-{{{\bf{M}}}}_{{{\bf{2}}}}+{{{\bf{M}}}}_{{{\bf{3}}}}-{{{\bf{M}}}}_{{{\bf{4}}}}$$\end{document}L=M1−M2+M3−M4, invariant. During powdering, grains may undergo random C₂ rotations that alter their crystallographic orientation but preserve the sign of their ME response. As a result, the ensemble retains the single-domain polarity encoded during ME annealing, enabling a non-zero isotropic ME response. The result of powdering a ME-annealed single crystal is sketched in Fig. 1b. As illustrated, Néel vectors are oriented such that, upon applying an electric field in any radial direction, all grains aligned along the field lines contribute equally to the magnetization response. Strikingly, once the sample is heated above TN, its pure axion response is lost irreversibly, with no field cooling or known protocol capable of restoring it. Results and discussion Next, we present experimental data together with quantitative simulations that incorporate the known temperature dependencies of α∥ and α⊥. Figure 2a shows the result of our low-frequency ME AC susceptometry measurements for (i,j) = (3,3) squares; (3,1) down triangles; (1,1) up triangles; (2,2) diamonds. Strictly speaking, the susceptometry does not directly measure \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{| }}}{\alpha }_{{ij}}(T){{\rm{| }}}$$\end{document}∣αij(T)∣, but rather the induced magnetic moment \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${m}_{i}$$\end{document}mi, which is related to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{{ij}}$$\end{document}αij via \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{{ij}}={\mu }_{0}{m}_{i}/(A{V}_{j})$$\end{document}αij=μ0mi/(AVj), where Vj is the amplitude of the applied sinusoidal voltage across electrodes oriented normal to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\hat{e}}_{j}$$\end{document}e^j, and A is the electrode area. Consequently, Fig. 2a presents \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|{m}_{i}/{V}_{j}|\propto {{\rm{| }}}{\alpha }_{{ij}}\left(T\right)A{{\rm{| }}}$$\end{document}∣mi/Vj∣∝∣αijTA∣, while Fig. 2b shows the specific susceptibility \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|{m}_{i}/(A{V}_{j})|$$\end{document}∣mi/(AVj)∣ vs. T.Fig. 2Isotropic magnetoelectric response.a Temperature dependence of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|{m}_{i}/{V}_{j}\right|$$\end{document}mi/Vj for (i,j) = (3,3) squares, (3,1) down triangles, (1,1) up triangles, and (2,2) diamonds measured via AC ME susceptometry using electrode pairs of different sizes. The inset shows the cuboid sample with matrix resin (green) and embedded Cr2O3 grains (blue ellipsoids). Red arrows depict antiferromagnetic order. A representative pair of top and bottom electrodes (golden layers) is connected to a voltage source. Vertical line marks T = 168 K where tensor elements have a zero crossing. b Data normalized with respect to the respective electrode area, resulting in data collapse. The red line shows the result of a simulation of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|{\alpha }_{33}\right|$$\end{document}α33 vs. T as representation of diagonal elements. Orange line is the simulated null result of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|{\alpha }_{31}\right|$$\end{document}α31 vs. T as representation of vanishes off-diagonal elements. Isotropic magnetoelectric response. a Temperature dependence of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|{m}_{i}/{V}_{j}\right|$$\end{document}mi/Vj for (i,j) = (3,3) squares, (3,1) down triangles, (1,1) up triangles, and (2,2) diamonds measured via AC ME susceptometry using electrode pairs of different sizes. The inset shows the cuboid sample with matrix resin (green) and embedded Cr2O3 grains (blue ellipsoids). Red arrows depict antiferromagnetic order. A representative pair of top and bottom electrodes (golden layers) is connected to a voltage source. Vertical line marks T = 168 K where tensor elements have a zero crossing. b Data normalized with respect to the respective electrode area, resulting in data collapse. The red line shows the result of a simulation of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|{\alpha }_{33}\right|$$\end{document}α33 vs. T as representation of diagonal elements. Orange line is the simulated null result of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|{\alpha }_{31}\right|$$\end{document}α31 vs. T as representation of vanishes off-diagonal elements. The temperature dependence of the tensor elements exhibits several notable features. Most strikingly, apart from the off-diagonal element \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\it{{\alpha }}}_{31}$$\end{document}α31, which remains essentially zero across the entire temperature range, all measured diagonal components touch the zero line at T ≈ 168 K, the temperature at which the axion piece of chromia is expected to vanish (see SI1). Furthermore, the diagonal elements follow a common functional form within the level of noise. For isotropic behavior, i.e., \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{{ii}}=\theta$$\end{document}αii=θ, one would expect all nonzero tensor elements to quantitatively conform to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left|\theta (T)\right|$$\end{document}θ(T). The apparent lack of data collapse among the diagonal elements is an experimental artifact due to the cuboidal shape of the sample, which has six faces of three differing areas. The sample volume contributing to the ME response depends on the size of the pair of opposing electrodes selected to measure a particular susceptibility component. The inset of Fig. 2a shows a schematic of the sample, including a particular pair of electrodes (golden layers at the top and bottom). Cr₂O₃ granules, represented as blue ellipsoids with opposing red arrows indicating Néel vector orientation, are embedded in the solid resin matrix (green). Voltage applied across the sample generates an electric field between the electrodes. The electric field inside each chromia grain depends primarily on the ratio of permittivities between the grain and the surrounding epoxy and is largely independent of grain size, with only minor corrections due to particle shape. Neglecting strong shape effects is justified by the moderate dielectric contrast (εchromia ≈ 11–12, εresin ≈ 3–5). The schematic in Fig. 2a is illustrative; the actual microstructure is densely packed. Normalizing each dataset by its corresponding electrode area leads to a collapse of all diagonal elements within experimental uncertainty, as illustrated in Fig. 2b. The lines in Fig. 2b show simulations of the ME responses |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{33}$$\end{document}α33| vs. T (red) and |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{31}$$\end{document}α31| vs. T (orange), both in good agreement with the experimental data. The simulation data in Fig. 2b represent the specific case in which a single crystal in its ME single-domain state was milled at T < TN. The fact that the simulation data follow the functional form of the measured diagonal elements, together with the vanishing off-diagonal elements, confirms that a purely isotropic ME response has been achieved (See SI4 for absolute values of the simulated ME response of annealed chromia powder, and SI5 for an estimate and simulation result of the response after ME annealing the single crystal and powdering it at T < TN). Next, we outline the simulation approaches. Two separate scenarios were simulated to clarify how the sequence of ME annealing and milling influences the resulting ME response in Cr2O3. In the first scenario, the milling of a single crystal into powder at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T > {T}_{N}$$\end{document}T>TN (i.e., in the paramagnetic phase) is simulated by populating positions in a cuboid with grains whose c-axes are randomly oriented. Figure 3a shows the distribution of these crystallographic c-axis orientations and the absence of magnetic order (gray dots). During subsequent field cooling through the Néel temperature, the applied collinear electric and magnetic fields generate an orientation-dependent ME energy bias.Fig. 3Monte–Carlo simulations.a Single crystal milled into randomly oriented grains above Néel temperature (paramagnetic state). b Field cooling below TN, each grain independently enters the antiferromagnetic phase and selects one of the two Néel vector states. Grains in certain orientation ranges preferentially adopt the opposite state, producing the characteristic ring pattern. c Single crystal, simulated as collection of grains with c-axes aligned along the z-axis of a global coordinate system, after ME annealing to below TN. The Néel vector in each grain points up along the c-axis represented by the dot at 0°. d Single crystal milled into powder at T < TN, grain is in a + 1 orientation defined in such a way that the grain’s Néel vector has a positive z-component in the global coordinate system. Monte–Carlo simulations. a Single crystal milled into randomly oriented grains above Néel temperature (paramagnetic state). b Field cooling below TN, each grain independently enters the antiferromagnetic phase and selects one of the two Néel vector states. Grains in certain orientation ranges preferentially adopt the opposite state, producing the characteristic ring pattern. c Single crystal, simulated as collection of grains with c-axes aligned along the z-axis of a global coordinate system, after ME annealing to below TN. The Néel vector in each grain points up along the c-axis represented by the dot at 0°. d Single crystal milled into powder at T < TN, grain is in a + 1 orientation defined in such a way that the grain’s Néel vector has a positive z-component in the global coordinate system. The Néel vector is constrained to each grain’s local c-axis but may adopt either polarity dictated by the ME free energy contribution. As the system crosses TN, every grain independently selects one of the two Néel states with probabilities determined by this ME energy bias (Fig. 3b, see “Methods” for the free energy expression). An analytic condition10.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tan }^{2}{{\rm{\vartheta }}}=-\frac{{{{\rm{\alpha }}}}_{\parallel }\left({T}_{N}^{-}\right)}{{{{\rm{\alpha }}}}_{\perp }\left({T}_{N}^{-}\right)}$$\end{document}tan2ϑ=−α∥TN−α⊥TN−also yields the critical angle ϑϑ separating two Néel states (see SI3). This prediction is in excellent agreement with the Monte–Carlo results shown in Fig. 3b, where blue (red) dots indicate grains whose Néel vectors have a positive (negative) z-component in the sample’s global coordinate system. This two-domain configuration results in an anisotropic ME response where different tensor components exhibit distinct temperature dependencies and zero-crossing behaviors. Representative diagonal (α33) and off-diagonal (α31-like) orientations are shown in Supplementary Fig. S4a, b and agree with the behavior reported in ref. 10. In the second simulated scenario, a single crystal is first cooled through TN under collinear electric and magnetic fields, forming a uniform single-domain state. In our simulation, this is implemented by treating the crystal as an ensemble of grains whose c-axes are perfectly aligned with the sample’s global z-axis. The single-domain state requires every grain’s Néel vector to align with the global +z direction (Fig. 3c). Milling the crystal into a powder at \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T < {T}_{N}$$\end{document}T TN = 307 K, a voltage of V = 500 V is applied, generating an electric field of E = 5 × 105 V m−1 along the crystallographic c-axis. Simultaneously, a magnetic field B = μ₀H = 7 T is applied along the same axis. The sample is cooled under these collinear fields to T = 200 K < TN preparing the crystal into an antiferromagnetic single domain state. Following this ME annealing protocol, the crystal is maintained at T ≈ 280 K (