Observing strong optical nonreciprocity from magnetically doped Dirac semimetals Authors: Bardia Nabavi, Sina Jafari Ghalekohneh, Hervé Ness, Dimitar Pashov, Mark van Schilfgaarde, Ian Leahy, Anthony Rice, Andrew Norman, Kirstin Alberi, Bo Zhao Published: 2026 ABSTRACT Many contemporary photonic and optoelectronic technologies rely on strong electromagnetic nonreciprocity at infrared frequencies. However, achieving a strong nonreciprocal effect in the infrared remains extremely challenging due to the lack of materials with inherently large nonreciprocal optical responses. Magnetic Weyl semimetals offer a promising solution, but the choice of materials has thus far been limited. Here, we present a different approach of accessing strong nonreciprocity by doping the Dirac semimetal Cadmium Arsenide (Cd3As2) with a magnetic dopant, Manganese (Mn). We attribute the strong optical nonreciprocity to the collective magnetic ordering of the Mn dopant that creates magnetic Weyl cones by lifting the degeneracy of Dirac cones via time-reversal symmetry breaking. The enabled nonreciprocity exhibits strong tunability through doping concentration and external magnetic fields, providing a new material platform and design rational for accessing strong electromagnetic nonreciprocity. Introduction Magneto-optical nonreciprocity is the foundation of many modern photonics and optoelectronics technologies1–5. Recent studies on radiative energy systems further underpin the critical role of such nonreciprocity to advanced energy harvesting and heat flow architecture3,6,7. Despite the profound implications of nonreciprocity in controlling light propagation and light-matter interactions, achieving a strong nonreciprocal effect, especially in the infrared region, is extremely challenging, primarily due to the lack of materials exhibiting strong nonreciprocal responses8. The nonreciprocity is related to the degree of asymmetry of the permittivity tensor \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{\varepsilon }}$$\end{document}ε¯¯, which can be characterized by the parameter \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g=||\overline{\overline{{\varepsilon }_{{{\rm{a}}}}}}||/||\overline{\overline{{\varepsilon }_{{{\rm{s}}}}}}||$$\end{document}g=∣∣εa¯¯∣∣/∣∣εs¯¯∣∣, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\parallel \cdot \parallel$$\end{document}∥⋅∥ is the matrix norm, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{{\varepsilon }_{{{\rm{s}}}}}}$$\end{document}εs¯¯ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{{\varepsilon }_{{{\rm{a}}}}}}$$\end{document}εa¯¯ are the symmetric and antisymmetric part of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{\varepsilon }}$$\end{document}ε¯¯, respectively8. Traditionally, nonreciprocal effects rely on the magneto-optical effect in semiconductors or ferromagnetic materials. g is on the order of 10−2 and 10−4, respectively, under a practical magnetic field B ∼~ 1 T (Fig. 1a), making achieving strong nonreciprocal light-matter interactions challenging, especially at the subwavelength scale9. The necessity of the external magnetic field also imposes significant challenges for miniaturizing nonreciprocal photonic components and systems such as on-chip optical isolators10,11. It is highly desirable to achieve strong nonreciprocity without the need for external magnetic fields10,12.Fig. 1Nonreciprocity originated from transforming Dirac cones to Weyl cones.a Average degree of nonreciprocity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{g}=\int_{{\lambda}_{1}}^{{{\lambda }_{2}}}gd\lambda /({\lambda }_{2}-{\lambda }_{1})$$\end{document}ḡ=∫λ1λ2gdλ/(λ2−λ1) where λ1 = 2 μm and λ2 =20 μm for magnetic Weyl semimetals (MWSM)8, doped semiconductors (SC)60, and ferromagnetic metals (FM)23. SC and FM are considered under an external magnetic field of B = 1 T. The data for (MnxCd1-x)3As2 is from the theoretical model of this work for the 15% Mn doping level. b With time-reversal symmetry (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathcal{T}}}$$\end{document}T) breaking, the degeneracy of the electronic band structure of a Dirac semimetal can be lifted and converted to that of a magnetic Weyl semimetal with two Weyl nodes separated by 2b in momentum space. E and k show momentum and wavevector, respectively. c Lattice structure of Mn-doped Cd3As2. Manganese, Arsenic, and Cadmium atoms are shown in red, black, and white, respectively. d Magnetic moment M versus in-plane external magnetic field B measured at room temperature which magnetic dopants exhibit a hysteresis response to applied field. Magnetometry is performed at room temperature on a sample with 12% Mn impurities. e Schematic of the forward and backward reflection measurement setup to detect the nonreciprocity. For such planar systems, the contrast of the two reflectivity directly shows whether the system is reciprocal. θi and θr are angles of incidence and reflection, respectively. f Schematic (left) and SEM image (right) of the Mn-doped Cd3As2 samples structure fabricated through MBE. The layers from the topmost to bottom are (MnxCd1-x)3As2, ZnxCd1-xTe, and ZnTe. The substrate is GaAs. An internal Infrared photon source is directed into the sample with incident angle θ and magnetization direction Φ and the reflected signal is collected by a Fourie Transform Infrared (FTIR) spectrometer, where a Zinc Selenide (ZnSe) polarizer can be added. Nonreciprocity originated from transforming Dirac cones to Weyl cones. a Average degree of nonreciprocity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{g}=\int_{{\lambda}_{1}}^{{{\lambda }_{2}}}gd\lambda /({\lambda }_{2}-{\lambda }_{1})$$\end{document}ḡ=∫λ1λ2gdλ/(λ2−λ1) where λ1 = 2 μm and λ2 =20 μm for magnetic Weyl semimetals (MWSM)8, doped semiconductors (SC)60, and ferromagnetic metals (FM)23. SC and FM are considered under an external magnetic field of B = 1 T. The data for (MnxCd1-x)3As2 is from the theoretical model of this work for the 15% Mn doping level. b With time-reversal symmetry (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathcal{T}}}$$\end{document}T) breaking, the degeneracy of the electronic band structure of a Dirac semimetal can be lifted and converted to that of a magnetic Weyl semimetal with two Weyl nodes separated by 2b in momentum space. E and k show momentum and wavevector, respectively. c Lattice structure of Mn-doped Cd3As2. Manganese, Arsenic, and Cadmium atoms are shown in red, black, and white, respectively. d Magnetic moment M versus in-plane external magnetic field B measured at room temperature which magnetic dopants exhibit a hysteresis response to applied field. Magnetometry is performed at room temperature on a sample with 12% Mn impurities. e Schematic of the forward and backward reflection measurement setup to detect the nonreciprocity. For such planar systems, the contrast of the two reflectivity directly shows whether the system is reciprocal. θi and θr are angles of incidence and reflection, respectively. f Schematic (left) and SEM image (right) of the Mn-doped Cd3As2 samples structure fabricated through MBE. The layers from the topmost to bottom are (MnxCd1-x)3As2, ZnxCd1-xTe, and ZnTe. The substrate is GaAs. An internal Infrared photon source is directed into the sample with incident angle θ and magnetization direction Φ and the reflected signal is collected by a Fourie Transform Infrared (FTIR) spectrometer, where a Zinc Selenide (ZnSe) polarizer can be added. Magnetic Weyl semimetals exhibit unique characteristics because of their exceptional electronic and optical properties, which arise from a topological band structure featuring Weyl nodes that act as magnetic monopoles in momentum space7,13. The momentum-separation 2b of the Weyl nodes leads to remarkable phenomena including a giant anomalous Hall effect13,14 and intrinsic magneto-optical15,16 responses without the need of external fields that far exceed those of conventional ferromagnetic materials17. These properties, emerging from the interplay of topology and magnetism, make magnetic Weyl semimetals promising candidates for future optical and electronic applications18,19. For example, magnetic Weyl semimetals can provide nonreciprocity reaching g ~ 1, about a hundred times stronger than of traditional semiconductors6,20–22 and a thousand times stronger than ferromagnetic materials23 under magnetic field B ∼~ 1 T (Fig. 1a). Despite these promising material candidates, accessing the unique properties of magnetic Weyl fermions remains a daunting task due to the limited availability of magnetic Weyl semimetals and challenges in material synthesis24. An alternative promising pathway to access the magnetic Weyl cones could be through magnetically doped Dirac semimetals. Magnetic doping lifts the degeneracy of a three-dimensional Dirac cone by breaking time-reversal symmetry, forming a pair of Weyl cones (Fig. 1b), as suggested in several theoretical studies25–29. Such possibility represents a potential new paradigm to access Weyl fermions and its associated strong nonreciprocity. However, there has been no experimental works so far to demonstrate the nonreciprocity obtained through this approach. Here, we report observing strong optical nonreciprocal properties from magnetically doped Dirac semimetals. We synthesize manganese (Mn) doped Cd3As2, a Dirac semimetal, through molecular beam epitaxy (MBE) (Fig. 1c). Mn atoms induce a local magnetic field that gives rise to the hysteresis response to the external magnetic field, as revealed by magnetometry (Fig. 1d). This intrinsic magnetic ordering leads to a substantially enhanced degree of nonreciprocity, significantly larger than traditional materials30,31, as depicted in Fig. 1a. We propose a reflectivity-based nonreciprocity detection methodology and measure the reflectivity of the samples in the infrared range (Fig. 1e,f). We observe strong contrast between the forward and backward reflectivity for the doped sample (Fig. 2a-d) under no external field, indicating a strong intrinsic nonreciprocal behavior. Our findings could have a far-reaching impact on generating magnetic Weyl semimetals and promote the development of next-generation nonreciprocal photonics technologies.Fig. 2Spectral directional reflectivity contrast of Mn-doped Cd3As2.a–d Incident angles θ are varied by adjusting the optical path within the reflectometer enabling comprehensive angular mapping of optical nonreciprocal response. Theoretical and measured p-polarized contrast ηp, between the forward and backward reflectivity at room temperature, shown at different angles and wavelengths for the (a, b) undoped (c, d) sample with 15% Mn doping, respectively. Azimuthal angle φ is fixed at 0°. Spectral directional reflectivity contrast of Mn-doped Cd3As2. a–d Incident angles θ are varied by adjusting the optical path within the reflectometer enabling comprehensive angular mapping of optical nonreciprocal response. Theoretical and measured p-polarized contrast ηp, between the forward and backward reflectivity at room temperature, shown at different angles and wavelengths for the (a, b) undoped (c, d) sample with 15% Mn doping, respectively. Azimuthal angle φ is fixed at 0°. Results In a recent paper29, the electronic structures of bulk Cd3As2 with different doping levels of Mn impurities is calculated from first-principles Quasi-particle self-consistent GW (QSGW) method and model Hamiltonians29. The magnetic Mn impurities (low 2% doping) are predicted to generate a large local magnetic field which leads to the splitting of the Dirac points of the pristine bulk Cd3As2 into pairs of Weyl points. The electronic structure of Mn-doped Cd3As2 can be emulated by the electronic structure of pristine Cd3As2 with an appropriate external magnetic field29, indicating that the Mn-induced local field act as an external applied magnetic field. The study reveals that the Mn spin orientation axis has non-zero components in the x, y and z directions when the doping level is around 2%, where the x-y plane is the (001) surface plane of the Cd3As2 samples (Fig. 1c). Generically, the internal magnetic field has non-zero components in all three directions. Such an internal magnetic field will result in a wavevector connecting the Weyl points that can also have three components. Similar to magneto-optical effects, 2b along the γ direction (vector connecting the Weyl nodes) gives rise to the non-zero off-diagonal components εαβ = −εβα (α ≠≠ β), α, β, γ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\in\,$$\end{document}∈{x, y, z}32. For this reason, the induced anisotropy give rise to a full permittivity tensor with nine components:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{\varepsilon }}=\left[\begin{array}{ccc}{\varepsilon }_{xx} & {\varepsilon }_{xy} & {\varepsilon }_{xz}\\ -{\varepsilon }_{xy} & {\varepsilon }_{xx} & {\varepsilon }_{yz}\\ -{\varepsilon }_{xz} & -{\varepsilon }_{yz} & {\varepsilon }_{zz}\end{array}\right]$$\end{document}ε¯¯=εxxεxyεxz−εxyεxxεyz−εxz−εyzεzz In this study, we obtain the components of the permittivity from the complex conductivity σαβ(ω) calculated from the modeled electronic structure33,34:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{\alpha \beta }\left(\omega \right)={\delta }_{\alpha \beta }+\frac{1}{{\varepsilon }_{0}}\frac{i}{\omega }{\sigma }_{\alpha \beta }(\omega )$$\end{document}εαβω=δαβ+1ε0iωσαβ(ω)where δαβ is Kronecker delta, ε0 is the vacuum permittivity, ω is the angular frequency. Electromagnetic nonreciprocity is manifested as the asymmetric permittivity tensor, i.e., \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{\varepsilon }}\ne {\overline{\overline{\varepsilon }}}^{{{\rm{T}}}}$$\end{document}ε¯¯≠ε¯¯T3,35,36. We demonstrate this asymmetricity using a robust reflection spectroscopy approach. For reciprocal systems, i.e., systems with materials with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline{\overline{\varepsilon }}={\overline{\overline{\varepsilon }}}^{{{\rm{T}}}}$$\end{document}ε¯¯=ε¯¯T, one can show that the scattering matrix S is always symmetric5. The reflectivity matrix that is derived from the scattering matrix is also symmetric5. For the planar systems we consider in this work, the reflectivity matrix can be expressed as37:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{R}}}=\left[\begin{array}{cc}{\left|{r}_{{ss}}\right|}^{2} & {\left|{r}_{{sp}}\right|}^{2}\\ {\left|{r}_{{ps}}\right|}^{2} & {\left|{r}_{{pp}}\right|}^{2}\end{array}\right]$$\end{document}R=rss2rsp2rps2rpp2 Here, rij with i, j ∈∈ {s, p} is the reflection coefficients where the first subscript denotes the polarization of the reflected wave, and the second subscript indicates the polarization of the incident wave. The p (s) polarization is defined as when the magnetic (electric) field is perpendicular to the plane of incidence. The diagonal (i = j) and off-diagonal components (i ≠≠ j) of R are the co-polarization and cross-polarization reflectivity, respectively. Similar to traditional magneto-optical Kerr effect (MOKE), the orientation of 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b in Mn doped-Cd3As2 can introduce polarization rotation and conversion to the incident light. 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b governs the shape of the permittivity tensor corresponding to various MOKE configurations. More detailed discussion on MOKE is provided in the Supplementary Information Section I. We define the plane of incidence (POI) as the plane formed by the z axis and the incident wave. Depending on direction of 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b relative to the POI, three conventional MOKE configurations are identified38: longitudinal, Voigt or transverse, and polar MOKE configurations39 where 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b is in x, y, and z directions, respectively, as detailed in the Supplementary Information Section I. Since 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b in Mn-doped Cd3As2 is generally not aligned along the coordinate axis, one could expect that the samples are typically in a hybrid MOKE configuration. In this work, we propose an approach based on reflection spectroscopy to reveal the nonreciprocity for such a general material system. As detailed in the Supplementary Information Fig. S1, the symmetry of the reflection coefficient of opaque planar structure is only sensitive to the magnetic field in the x-y plane and is in fact not affected by the magnetic field in the z direction. This property helps us isolate the effect of the magnetic field in the z direction. The in-plane magnetic fields in general can induce polarization changes to the incident waves, as indicated by the non-zero cross-polarization terms (Supplementary Information Fig. S1). However, by rotating the sample about the z axis, one could find a configuration that allows the combined magnetic field in the x-y plane to be perpendicular to the POI. In this case, we effectively reach a pure Voigt configuration that helps to clearly reveal the nonreciprocal response since this configuration does not produce polarization conversion to the incident wave. If the sample is nonreciprocal, reflection contrast can be induced only for p-polarized waves, where the magnetic field of the electromagnetic wave aligns with the internal magnetic field and is most sensitive to its effect. The s-polarized waves (electric field perpendicular to POI), on the other hand, do not experience any nonreciprocal effect and consistently show no reflection asymmetry. Thus, they can be used as a control experiment (Supplementary Information Section II). The Mn-doped Cd3As2 (001) thin films are synthesized with MBE on a ZnₓCd1-xTe/ZnTe/GaAs (001) substrate structure (Fig. 1f). The ZnₓCd1-xTe composition is chosen to be lattice matched to Cd3As2. The films are doped up to several atomic % using elemental Mn sourced from an effusion cell40. The concentration is controlled through the Mn//Cd flux ratio. The thickness of the Mn-doped Cd3As2, ZnₓCd1-xTe, and ZnTe layers are 500, 125, and 20 nm, respectively. All materials besides the Cd3As2 are expected to be reciprocal. We conduct magnetization measurements at 300 K on Mn-doped Cd3As2 with ~15% Mn substitution in the (001) surface plane of the Cd3As2 sample, which reveal a clear hysteresis loop, as shown in Fig. 1d. The presence of non-zero remanent magnetization and coercivity at room temperature indicates robust collective magnetic behavior, far beyond the dilute paramagnetic limit, among Mn spins, which is not achieved in conventional Weyl semimetals at temperatures higher than Curie temperature16. The presence of robust magnetic hysteresis at the room temperature indicates that the Mn magnetic ordering transition temperature, if there is one, should be above the room temperature. This result confirms that Mn atoms in this doping regime are not oriented completely randomly. Instead, the Mn ions align along a preferred direction, generating a net magnetic field oriented to minimize the magnetostatic energy, a behavior likely driven by shape anisotropy. Such order generates an effective internal magnetic field that could be sufficient to break time-reversal symmetry and stabilize a Weyl semimetal phase. The experimental observation supports the theoretical picture in which Mn-induced magnetism drives the splitting of Dirac nodes into Weyl points, enabling topological band structure modification via magnetic doping. We construct our experiments, accordingly, as indicated in Fig. 1e. We define a reflection contrast as the difference between the reflectivity of the forward and backward scenarios41:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\eta }_{i}(\lambda,\theta,\varphi )={r}_{i}(\lambda,\theta,\varphi )-{r}_{i}(\lambda,\theta,\varphi+\uppi )$$\end{document}ηi(λ,θ,φ)=ri(λ,θ,φ)−ri(λ,θ,φ+π)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\eta }_{i}$$\end{document}ηi defines the reflection contrast for a certain polarization, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${r}_{i}={|{r}_{ii}|}^{2}+{|{r}_{ji}|}^{2}$$\end{document}ri=∣rii∣2+∣rji∣2 represents the reflectivity for incident waves with i-polarization. θ, φ, and λ are angle of incident, azimuthal angle, and wavelength respectively. ri(λ, θ, φ) corresponds to the forward configuration, whereas ri(λ, θ, φ + π) is the reflectivity of the corresponding backward setup. We note that the polarizations of the forward and backward incident waves are the same. We measure the reflectivity of the undoped sample first for a range of λ and θ and the results for p-polarized waves is shown in Fig. 2a,b, which essentially exhibit no contrast because of the reciprocal behavior. However, in sharp contrast, for the doped sample, we measure a strong reflectivity contrast for the p-polarization waves as shown in Fig. 2c,d. The optical nonreciprocity in Mn-doped Cd3As2 shows a non-monotonic spectral behavior across the infrared range. Strong nonreciprocal reflectivity is observed for a wide wavelength and angular range except some regions indicated by the dark color in the contour plots in which the reflectivity contrast vanishes. This trend is reproducible across field cycles and absent in undoped Cd3As2, confirming its intrinsic origin. The dispersive behavior of nonreciprocity arises from the intrinsic optical properties of the material. Here, we highlight the agreement between the theoretical predictions and experimental observations, which provides strong supports for our model and understanding of the material. Note that the φ = 0° corresponds to the Voigt configuration where the local magnetic field is aligned along the y axis. The measured reflectivity of the s-polarized waves for both the undoped and doped samples exhibits diminishing contrasts, as shown in the Supplementary Information Fig. S2, indicating a reciprocal behavior. They also serve as a control and sanity check for the reflectivity contrast measurements. All measured results agree with our simulation based on a 4××4 scattering matrix method (Supplementary Information Section III) with the dielectric function obtained from the calculated electronic band structures. The results clearly demonstrate the nonreciprocal behavior of the Mn-doped Cd3As2 as compared to the undoped ones. We note that, although the Mn doping introduces local exchange fields, band structure calculations for 2% Mn-doped Cd3As2 sample demonstrating the transformation of the Dirac cone into a Weyl cone indicate that the nonreciprocity we observe is most plausibly linked to magnetic Weyl topology29, rather than to local exchange band splitting as in dilute ferromagnetic semiconductors such as Mn-doped GaAs42,43. The close agreement between simulations and experimental measurements of optical nonreciprocity further validates the modeling approach. The nonreciprocal response spans a broad wavelength range in the infrared, marking a unique feature of magnetic Weyl semimetals. The in-plane magnetic field can be further understood from the measured η while rotating the POI by changing φ (angle between 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{b}}}$$\end{document}b and the y axis). The Voigt configuration corresponds to φ = 0° (Fig. 2a-d) where only the p-polarized waves have a non-zero contrast η. When φ = 90°, the in-plane magnetic field becomes completely parallel to the POI (longitudinal MOKE). Such configuration produces no contrast for neither the s- or the p-polarized waves (Supplementary Information Figs. S1 and S2). For azimuthal angles in between, one triggers hybrid longitudinal and transverse MOKE effects. In hybrid MOKE configurations, cross-polarized components rps and rsp are non-zero. They contribute alongside rpp to the overall reflection asymmetry, resulting in non-zero contrasts for both the s- and p- polarized waves. The s-polarized reflectivity contrast is maximized at φ= 45° and 135°, where polarization conversion is maximized, consistent with theoretical studies44. The theoretical calculations (Fig. 3a,b) and the experimental measurements (Fig. 3c,d) agree well. The dark band at φ = 90° shows diminishing contrasts for both polarizations (Supplementary Information Section I) and represents a symmetry axis for all contours plots.Fig. 3Optical nonreciprocity in Mn-doped Cd3As2.a–d Reflection contrasts η for different wavelength λ and azimuthal angle φφ for (a, c) p- and (b, d) s-polarized waves. Theoretical calculations are shown in the left panels, whereas the measurements are on the right panels. The incident angle is fixed at θ = 40°. All data are for Mn-doped Cd3As2 with 15% Mn content measured at room temperature. e Computed real and imaginary part of diagonal and off-diagonal permittivity ε from the k·p model with Zeeman term γB = 50 meV that mimics external magnetic field and Fermi level EF = 30 meV which fits best with the 15% Mn-doped Cd3As2 sample. f Experimental and theoretical forward (φ) and backward (φ+ π) reflectivity at room temperature for p-polarized waves. The angle of incidence is fixed at θ = 40°. The difference shows the nonreciprocity. The highlighted pink region most likely corresponds to the ENZ region. Optical nonreciprocity in Mn-doped Cd3As2. a–d Reflection contrasts η for different wavelength λ and azimuthal angle φφ for (a, c) p- and (b, d) s-polarized waves. Theoretical calculations are shown in the left panels, whereas the measurements are on the right panels. The incident angle is fixed at θ = 40°. All data are for Mn-doped Cd3As2 with 15% Mn content measured at room temperature. e Computed real and imaginary part of diagonal and off-diagonal permittivity ε from the k·p model with Zeeman term γB = 50 meV that mimics external magnetic field and Fermi level EF = 30 meV which fits best with the 15% Mn-doped Cd3As2 sample. f Experimental and theoretical forward (φ) and backward (φ+ π) reflectivity at room temperature for p-polarized waves. The angle of incidence is fixed at θ = 40°. The difference shows the nonreciprocity. The highlighted pink region most likely corresponds to the ENZ region. Moreover, as shown in Fig. 3b,d, maximum s-polarized nonreciprocity is concentrated around λ = 5 μm, which could be related to the oscillatory behavior of the permittivity (Fig. 3e). A detailed discussion on the origin of the s-polarized nonreciprocity is presented in the Supplementary Information Section II. The approach presented above provides a way to not only detect nonreciprocity but also quantify the direction of the in-plane internal magnetic field direction, which is critical to ensure the performance and operation of photonic and optoelectronic Weyl devices. By adjusting the magnetic moments within the sample, one can also tune both the co- and cross-polarized reflection coefficients, allowing polarization conversion and enabling control over nonreciprocal behavior through the azimuthal angle45. A full characterization of polarization conversion and nonreciprocal reflectivity across different configurations is provided in the Supplementary Information Fig. S1 and Supplementary Information Section I. Figure 3e shows the calculated real and imaginary parts of the diagonal and off-diagonal permittivity from the simulated electronic band structure. It is clear to see that the off-diagonal components are of the same order of magnitude with the diagonal components, a unique property related to the magnetic Weyl system. The diagonal permittivity approaches zero at around 3 μm, marking the onset of probable epsilon-near-zero (ENZ) region46,47, which could correspond to the valley in the reflectivity spectra in that region (Fig. 3f). One unique advantage of the nonreciprocity enabled by Mn-doped Cd3As2 is its sensitivity to the magnetic field strength produced by the dopant, making it potentially tunable by adjusting the doping level. This contrasts with traditional magnetic Weyl semimetals, where the level of nonreciprocity is difficult to tune. To showcase the tunability through doping concentration, we fabricate three more samples with different doping levels. For each sample, we repeat the previously discussed process and identify the configuration corresponding to the effective Voigt configuration. We then measure the forward and backward reflectivity spectra for the p-polarized waves at oblique angles (θ = 40°) as shown in Fig. 4a. For the undoped Cd3As2, the two reflectivity spectra overlap because of reciprocity. However, as the Mn concentration increases, a clear contrast between the spectra emerges and grows with doping. The rate of increase in non-reciprocity with Mn concentration, however, is not constant (Supplementary Information Section V) and indirectly reflects the complex magnetic ordering of the Mn ions in the samples25.Fig. 4Effect of Mn concentration on Cd3As2 nonreciprocity.a, b Experimental and theoretical reflection spectra in the Voigt configuration at 300 K for different Mn% in the absence of an applied magnetic field. γB is the value of Zeeman term used in the k·p model which mimics the effective magnetic field of the magnetic Mn impurities to capture nonreciprocity (1 meV is equivalent to a 17.3 Tesla B-field strength). The Fermi levels for all samples are 30 meV in all calculations. 0% corresponds to the case of undoped reciprocal Cd3As2 sample. For each sample, we conduct measurement for p-polarized waves without applying external magnetic field under the Voigt configuration. For better visualization, the curves are vertically offset by 0.4. Effect of Mn concentration on Cd3As2 nonreciprocity. a, b Experimental and theoretical reflection spectra in the Voigt configuration at 300 K for different Mn% in the absence of an applied magnetic field. γB is the value of Zeeman term used in the k·p model which mimics the effective magnetic field of the magnetic Mn impurities to capture nonreciprocity (1 meV is equivalent to a 17.3 Tesla B-field strength). The Fermi levels for all samples are 30 meV in all calculations. 0% corresponds to the case of undoped reciprocal Cd3As2 sample. For each sample, we conduct measurement for p-polarized waves without applying external magnetic field under the Voigt configuration. For better visualization, the curves are vertically offset by 0.4. As discussed in ref. 29, a k·p model in a magnetic field can emulate the effects of Mn doping in Cd3As2. Such a magnetic field is an input parameter in the model and represents, in average, the effects of the Mn magnetic moments. To compare our experimental data to theory, we calculate p-polarized reflectivity (rp) using our model to tune the value of the magnetic field in the k·p model to match, in the best possible way, the experiment. Figure 4b shows the corresponding theoretical simulation results, for increasing fields, which agree well with the experiment (especially at low fields). Besides the Mn doping level, another unique feature of Mn-doped Cd3As2 is that the nonreciprocity can also be controlled and reprogrammed by an external magnetic field. This is conceivable since an external magnetic field can enhance the alignment of the spins of Mn atoms and therefore increase (or decrease) the nonreciprocity. To directly probe the magnetic effect on the nonreciprocity, we measure the reflectivity contrast as a function of externally applied magnetic field for the 15% Mn-doped Cd3As2 sample (Fig. 5). The external magnetic field is applied along the φ = 0° direction so that it can be aligned with the internal magnetic field. A magnetic assembly using permanent magnets (Fig. 5a and Supplementary Information Fig. S7) is employed to apply discrete magnetic fields up to 0.36 T, during which reflectivity is measured at θ = 40° for φ = 0° and φ = 180°. We first gradually increase the applied magnetic field from 0 T to ++0.36 T, then decrease it back to 0 T, followed by sweeping it to -−0.36 T and finally returning to 0 T. It is seen that magnetization saturates at 0.36 T, as indicated by the vanishing reflectivity contrast and the absence of further response to stronger external fields. To better quantify the broadband nonreciprocal effect, we use an average contrast defined as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{\eta }={\int }_{{\lambda }_{1}}^{{\lambda }_{2}}\eta d\lambda /({\lambda }_{2}-{\lambda }_{1})$$\end{document}η¯=∫λ1λ2ηdλ/(λ2−λ1), where λ1 = 3 μm and λ2 = 15 μm, a band with strong nonreciprocity.Fig. 5Field-dependent nonreciprocal optical response.a Magnetic reflectometry setup. b P-polarized and s-polarized nonreciprocal contrast (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{\eta }={\int }_{{\lambda }_{1}}^{{\lambda }_{2}}\eta d\lambda /({\lambda }_{2}-{\lambda }_{1})$$\end{document}η¯=∫λ1λ2ηdλ/(λ2−λ1), where λ1 = 3 μm and λ2 = 15 μm) as a function of the external magnetic field for the 15% Mn-doped Cd3As2 sample at room temperature. Since the applied magnetic field is aligned parallel or antiparallel to the Weyl momentum-space separation vector 2b, the magnitude of 2b increases or decreases, while its direction remains unchanged as the magnetic domains become polarized. Azimuthal φ and incident θ angle are fixed at 0°, and 40°. Field-dependent nonreciprocal optical response. a Magnetic reflectometry setup. b P-polarized and s-polarized nonreciprocal contrast (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{\eta }={\int }_{{\lambda }_{1}}^{{\lambda }_{2}}\eta d\lambda /({\lambda }_{2}-{\lambda }_{1})$$\end{document}η¯=∫λ1λ2ηdλ/(λ2−λ1), where λ1 = 3 μm and λ2 = 15 μm) as a function of the external magnetic field for the 15% Mn-doped Cd3As2 sample at room temperature. Since the applied magnetic field is aligned parallel or antiparallel to the Weyl momentum-space separation vector 2b, the magnitude of 2b increases or decreases, while its direction remains unchanged as the magnetic domains become polarized. Azimuthal φ and incident θ angle are fixed at 0°, and 40°. We probe the reflectivity using the same way as in Fig. 1e for both polarizations. We plot the s- and p-polarized reflectivity contrast \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{\eta }$$\end{document}η¯ in Fig. 5b, and a clear hysteresis loop is observed for p-polarized wave, indicating that the applied field assists in aligning the Mn spins and the resulting collective internal magnetization. This behavior confirms that external fields enhance the effective magnetization of the system, thereby influencing the nonreciprocal optical response. In contrast, s-polarized light, which does not exhibit a nonreciprocity and can be used as a control, shows negligible field dependence. Similar reciprocal behavior is observed in the undoped Cd3As2 samples, where both s- and p-polarized waves show no sensitivity to the applied magnetic field, serving as a control experiment (Supplementary Information Fig. S6). These results provide direct optical evidence for field-induced spin alignment and its role in modulating topological and magneto-optical properties in Mn-doped Cd3As2. The hysteresis loop exhibits clear asymmetry. Recent studies on magnetic Weyl semimetals48, also reveal similar behavior. Therefore, we believe that the asymmetry of the hysteresis loop in Fig. 5 may also arise from local regions whose magnetic polarization is difficult to be flipped under the external magnetic field. Similar measurements performed on samples with lower Mn concentrations reveal a weaker magnetic response, characterized by smaller remanence, reduced coercivity, and a narrower hysteresis loop with lower energy loss, thereby indicating that the degree of the nonreciprocal response increases with doping. The hysteretic behavior indicates a memory effect, suggesting that the optical state of the system retains information about its magnetic history. Such magnetically reconfigurable nonreciprocity could also serve as a building block for topological photonic devices and hybrid spin-photon interfaces. Overall, our work establishes magnetically doped Dirac semimetals as viable and tunable platform for realizing strong optical nonreciprocity at infrared frequencies. Through precise molecular beam epitaxy growth and systematic characterization, we demonstrate that Mn-doped Cd3As2 exhibits robust, broadband nonreciprocal optical response at room temperature, originating from collective Mn spin alignment. It is likely that the combination of magnetization hysteresis, doping-level dependence, and external-field tunability reflects the time-reversal symmetry breaking which is signature of the magnetic Weyl phase. More direct evidence such as observance of the Weyl cone or electric transport could be still necessary, which will be reported in future work. These features culminate in an unprecedented level of control over nonreciprocal light-matter interactions2, with g approaching unity, far exceeding conventional semiconductors and ferromagnetic materials. The reconfigurable and hysteretic nature of the nonreciprocal effect introduces new possibilities for photonics48. Recent advances in nonreciprocal nanophotonics could be used to further enhance the nonreciprocal response49. This work lays the foundation for a new class of nonreciprocal materials with engineered topological and magnetic functionalities, paving the way toward compact, tunable, and ultrafast photonic components for next-generation quantum and classical information technologies. It also paves the way for high-performance nonreciprocal thermal emitters8 for thermal50 and energy51 applications. Methods First-principle QSGW and Hamiltonian model Electronic structure calculations for both undoped and Mn-doped Cd₃As₂ are performed using the first-principles quasiparticle self-consistent GW (QSGW) approach29. In these calculations, a single Mn atom substitutes for one Cd atom in the unit cell, corresponding to a ∼2% doping level. From these calculations emerged that in the 2% doping case, the original Dirac point of Cd3As2 split into pairs of Weyl points, which are displaced in the reciprocal k-space. It is found that such a splitting can be emulated by the application of an appropriate external magnetic field; suggesting that the magnetic moment of the substitutional Mn impurity acts indeed as a local magnetic field. In general, the connecting vector (2b) between a pair of Weyl points is parallel to the applied field. The perturbation under a field in the x direction causes no energy splitting of the Weyl points, unlike the case of a field parallel to z direction. In practice, doping level in Cd3As2 samples is significantly higher, and the precise magnetic ordering of the dopants remains an open question. Furthermore, the presence of spatially varying local magnetic moments suggests a complicated magnetic ordering, making it challenging to model realistic systems with first-principles calculations. To account for this discrepancy, we employ a model Hamiltonian for undoped Cd3As2 that incorporates a Zeeman term to stimulate the effect of local magnetic moments introduced by the Mn impurities in doped Cd3As229. Thus, in our calculations, we treat the internal magnetic field as a tunable parameter, alongside temperature and Fermi level to reflect the theoretically observed nonreciprocity. In the present paper, we employ the k·p Hamiltonian developed and described in detail in ref. 29. The model is specifically designed to reproduce, with high fidelity, the electronic structure obtained from first principle many-body (QSGW) calculations for undoped and Mn-doped bulk Cd3As229. The model includes a Zeeman (i.e. an applied external magnetic field emulating the magnetic Mn atom). A detailed analysis of the electronic structures from the k·p model with an applied magnetic field B and from the QSGW electronic structures of Cd3As2 (with one Cd atom replaced by one Mn, i.e. approx. 2% impurity doping) is provided in ref. 29. The use of this model allows us to systematically investigate the electronic structure and transport properties of Mn-doped Cd3As2 under variations in key critical parameters such as temperature, Fermi level, and applied field (we note that the applied magnetic field emulates the Mn impurity concentration and its effective global magnetic order). The k·p Hamiltonian comprises four orbitals (s, px, py, pz) and two spin states per orbital, with a spin-orbit coupling term, HSO, as described in ref. 52. The Zeeman term HB = -−γBσ, couples the magnetic field B = (Bx, By, Bz) to the spin Pauli matrices σ = (σx, σy, σz)29. The presence of this external magnetic field is needed to emulate the local fields induced by the Mn impurities. The full Hamiltonian H8(k) depends on the three components of the crystal momentum k = (kx, ky, kz)29. Upon diagonalization at each k-point, the resulting eigenvalues are used to construct the electronic band structure and additional relevant physical quantities. The electrical conductivity tensor is then calculated using linear response theory, following the formalism outlined in refs. 52–55:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{\alpha \beta }\left(\omega \right)={\sigma }_{\alpha \beta }^{g}\left(\omega \right)+\frac{{e}^{2}}{i\omega }{S}_{\alpha \beta }\left(\omega \right)$$\end{document}σαβω=σαβgω+e2iωSαβωwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{\alpha \beta }^{g}\left(\omega \right)$$\end{document}σαβgω is gauge term52,54,55 and α, β ∈∈{x, y, z}. The gauge term is given by:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{\alpha \beta }^{g}\left(\omega \right)=-\frac{{iN}{e}^{2}}{m^{*}\omega}{\delta }_{\alpha \beta }$$\end{document}σαβgω=−iNe2m*ωδαβwhere N is number of electrons (determined by Fermi energy and temperature), and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${m^{*}}$$\end{document}m* is the effective mass. This gauge term is retained in all calculations to ensure that, in the DC limit ω→→ 0, the longitudinal conductivity σαα(0) has a vanishing imaginary part contribution52. The current-current correlation function Sαβ(ω) is defined as below:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${S}_{\alpha \beta }\left(\omega \right)=-2\int \frac{d\varepsilon }{2\pi }\,{f}_{\varepsilon }{{\rm{Tr}}}[\;{j}_{\alpha }{{\rm{Im}}}G\left(\varepsilon \right)\,{j}_{\beta }{G}_{+}\left(\varepsilon+\hbar\omega \right)+{j}_{\alpha }{G}_{-}\left(\varepsilon -\hbar\omega \right)\,{j}_{\beta }{{\rm{Im}}}G\left(\varepsilon \right)]$$\end{document}Sαβω=−2∫dε2πfεTr[jαImGεjβG+ε+ħω+jαG−ε−ħωjβImGε]where fε is the Fermi-Dirac distribution function, jα,β is the current operator, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${G}_{\pm }$$\end{document}G± denotes the Green’s function as defined:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${G}_{\pm }={\left(\varepsilon -{{{\bf{H}}}}_{8}\left({{\bf{k}}}\right)\pm {{\rm{i}}}\vartheta \right)}^{-1}$$\end{document}G±=ε−H8k±iϑ−1with (ϑ →→ 0+), and the imaginary part of the Green’s function is given by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathfrak{I}}G\left(\varepsilon \right)=\frac{1}{2}({G}_{+}-{G}_{-})$$\end{document}IGε=12(G+−G−). Using a compact notation for the eigenstates of H8(k), En ≡≡ En(k) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${|n}\rangle \equiv |{u}_{n}({{\bf{k}}})\rangle$$\end{document}∣n⟩≡∣un(k)⟩, one recovers the standard expression for the dynamical conductivity σαβ(ω) (for α ≠≠ β) as detailed in refs. 52–57:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma }_{\alpha \beta }(\omega )=-{{\rm{i}}}{e}^{2}\hslash {\sum}_{n,m\ne n}\frac{(\frac{{f}_{{E}_{m}}-{f}_{{E}_{n}}}{\hslash \omega })(\langle n|\,{j}_{\alpha }|m\rangle \langle m|\,{j}_{\beta }|n\rangle )}{({E}_{m}-{E}_{n})+\hslash \omega+{{\rm{i}}}\vartheta } \\=\frac{{e}^{2}}{{{\rm{i}}}\omega }{\sum}_{n,m\ne n}\frac{(\,\,{f}_{{E}_{n}}-{f}_{{E}_{m}})(\langle n|\,{j}_{\alpha }|m\rangle \langle m|\,{j}_{\beta }|n\rangle )}{({E}_{n}-{E}_{m})-\hslash \omega -{{\rm{i}}}\vartheta }$$\end{document}σαβ(ω)=−ie2ℏ∑n,m≠n(fEm−fEnℏω)(⟨n∣jα∣m⟩⟨m∣jβ∣n⟩)(Em−En)+ℏω+iϑ=e2iω∑n,m≠n(fEn−fEm)(⟨n∣jα∣m⟩⟨m∣jβ∣n⟩)(En−Em)−ℏω−iϑ The matrix elements 〈n|jα|m〉 of the α component of the current operator is obtained in the k-space from the matrix elements \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\langle {u}_{n}\left(k\right)\left|{\nabla }_{{k}_{\alpha }}{H}_{8}\left({{\bf{k}}}\right)\right|{u}_{m}(k)\rangle$$\end{document}⟨unk∇kαH8kum(k)⟩. Analytical expressions for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\nabla }_{{k}_{\alpha }}$$\end{document}∇kαH8(k) can be directly derived from the Hamiltonian H8(k) as shown in ref. 29. In the numerical calculations, the sums over states account for the k-dependence of the eigenstates:10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sum}_{n,m\ne n}\ldots \to \int \frac{{dk}}{{\left(2\pi \right)}^{3}}{\sum }_{n=1,m\ne n}^{8}\ldots$$\end{document}∑n,m≠n…→∫dk2π3∑n=1,m≠n8… Numerical calculations are performed on a regular cubic k-point grid centered at the Γ point, spanning from -−0.155 to 0.155 Å-1 along each of the three Cartesian directions. A dense mesh of (455××455××455) points is used to accurately capture the contributions from the Dirac/Weyl crossing points57. The dielectric tensor εαβ(ω) is then obtained from the conductivity tensor as described in Eq. (2). Our approach is based on a model Hamiltonian that fully captures the electronic structure of both pristine and doped Cd3As2. Using such a model Hamiltonian enables systematic tuning of key physical parameters, including temperature, Fermi level and applied magnetic field, to explore their influence on the optical conductivity of Cd3As229. This model demonstrates significantly better agreement with experimental observations compared to simplified two-band model33,34,46. Optical reciprocity and reflection contrast To rigorously investigate the emergence of optical nonreciprocity in our system, we employ a reciprocity test based on reflection spectroscopy. The theoretical foundation of this approach arises from the reciprocity condition derived from Maxwell’s equations in linear, time invariant, and nonmagnetic media. In a reciprocal system, the electromagnetic response for a wave propagating in each direction must be identical to that of the time-reversed counterpart propagating in the opposite direction5. Expressing the fields in terms of modal amplitudes via the scattering matrix S, simplifies the reciprocity condition to S = ST, implying that the power transmitted or reflected from forward direction from port m to n should be equal to transferred power in backward direction from port n to m \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$({\left|{{{\bf{S}}}}_{{nm}}\right|}^{2}={\left|{{{\bf{S}}}}_{{mn}}\right|}^{2})$$\end{document}(Snm2=Smn2) as depicted in Fig. 1e. Since the scattering matrix relates amplitude of incoming (aa) and outcoming (bb) field via Snmam = bn it follows that in planar photonic systems, reciprocity under in-plane rotation implies:11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${r}_{i}(\lambda,\theta,\varphi )={r}_{i}(\lambda,\theta,\varphi+\uppi )$$\end{document}ri(λ,θ,φ)=ri(λ,θ,φ+π) This angular transformation effectively swaps the incident and reflected in-plane momentum vectors, which, under reciprocity, must yield identical reflectivity. Hence, any asymmetry in ri(λ, θ, φ) versus ri(λ, θ, φ+ π) directly signals a breakdown of reciprocity. In Mn-doped Cd3As2, time reversal symmetry is explicitly broken by magnetic impurities, leading to electromagnetic nonreciprocity. We implement nonreciprocity test by measuring the reflectivity spectra while rotating the sample azimuthally in-plane from an initial angle φ to φ+ π, keeping the polarization and angle of incidence fixed. In reciprocal systems (e.g. undoped Cd3As2), the reflectivity remains invariant under this transformation. However, in Mn-doped Cd3As2, we observe clear deviations from this symmetry, confirming the emergence of nonreciprocal optical response due to time-reversal symmetry breaking. This discussion holds in the absence of polarization conversion, i.e. when the sample is aligned in the Voigt configuration. Sample fabrication Sample fabrication begins with the epitaxial growth of thin films from elemental sources using MBE on GaAs (001) substrates with a 4° miscut (Supplementary Information Fig. S8). To improve film quality, a lattice-matched ZnₓCd1-xTe/ZnTe buffer structure is employed as an intermediate layer, following the procedure described in ref. 40. (MnₓCd1-x)3As2 films are subsequently grown under similar conditions to those used for unalloyed Cd3As240 with the addition of Mn supplied from a dedicated effusion cell. Reflection high-energy electron diffraction (RHEED) patterns during growth are comparable to those of undoped films, though they become progressively hazier with increasing Mn concentration. After growth an electron transparent cross-section sample was prepared using a focused ion beam (FIB) lift out method and examined in a Thermo Fisher Spectra 200 aberration corrected scanning transmission electron microscope (STEM) operated at 200 kV. The layer structure (Supplementary Information Fig. S8b) was revealed by STEM high angle annular dark field (HAADF) images and STEM X-ray energy dispersive spectroscopy (EDS) elemental maps. High-resolution X-ray diffraction (XRD) confirms a single-set peaks, shifting toward smaller lattice constants as Mn incorporation increases, consistent with lattice contraction due to Mn substitution (Supplementary Information Fig. S9). The Mn composition is estimated by combining effusion cell flux calibration with compositional analysis via Rutherford backscattering spectrometry (RBS). Film thicknesses are independently measured using ellipsometry. Film surface roughness is characterized by scanning electron microscopy (SEM) to confirm a flat, planar structure, ensuring that the measured reflectivity contrast arises solely from nonreciprocity (Supplementary Information Fig. S10). Experimental measurements Magnetization measurements are carried out using the ACMS II option in a Quantum Design Physical Property Measurement System (PPMS). Samples are mounted on a quartz rod using a thin layer of General Electric (GE) varnish to minimize background signals. Magnetization is measured within the plane of the film. A significant diamagnetic background is observed originating from the GaAs substrate, while the Mn-doped Cd3As2 film exhibits a distinct low-field hysteresis loop. The GaAs diamagnetism is field linear and is removed by subtracting a linear fit to the raw magnetization for fields greater than 1 T. Optical measurements are performed using Nicolet iS50 FTIR controlled via OMNIC Spectra software. To perform the polarization-resolved measurements, an internal ZnSe polarizer is employed, allowing us to distinguish between s-and p-polarized waves. Internal infrared source of the spectrometer is directed onto the sample, and the reflected beam is collected using VeeMAX III reflectometer, which enables angle-resolved measurements, as depicted in Fig. 1e and Supplementary Information Fig. S7b. To ensure that the observed reflectivity contrast originates solely from nonreciprocity, the laser spot is fixed to probe the same sample location throughout all measurements. The control experiments with s-polarized waves and undoped samples verify our experimental approach as illustrated in Fig. 2b and Supplementary Information Fig. S2. Finally, the reflected wave is detected by a DTGS detector. Due to high noise to signal ratio at small oblique angles, measurements are done at angle θ = 40°. The captured reflectivity data are post-processed in MATLAB using locally weighted average linear regression. To investigate the magneto-optical response of the material we design a home-built magnetic setup to apply external field to the sample during the experiment. The setup consists of a linear array of permanent Neodymium magnets (Supplementary Information Fig. S7a), enabling the application of magnetic fields up to 0.36 T, sufficient to reach magnetic saturation. This gap-tunable configuration allows continuous variation of the applied field by adjusting the magnet spacing (l), making it suitable for hysteresis studies. We have characterized the spatial profile of the applied magnetic field using the gaussmeter (SISCO-EFD-WT10D) by mapping the field strength across the laser spot region. Within the laser spot size, the field variation is found to be negligible ( < 5% of the applied field), confirming that the sample experiences a nearly uniform magnetic field during the measurement. We emphasize that in our measurements, the input polarizer selects the incident polarization state (s or p), while the detector collects the total reflected intensity regardless of its polarization state. Therefore, ri as defined in Eq. (4) represents the reflectivity for incident i-polarization, which includes both co-polarized (rii) and cross-polarized (rji) contributions. For the hysteresis measurements, the magnetic field is gradually increased from 0 T to ++0.36 T in four steps by adjusting the spacing between the magnets. Upon reaching saturation, the field is decreased back to 0 T. The magnet array is then rotated to reverse the field direction, and the same field ramp is applied from 0 T to -−0.36 T and subsequently reduced to 0 T. The full measurement loop allows for the observation of hysteretic optical behavior. We perform this measurement for both Mn-doped and undoped Cd3As2 samples, with the undoped sample serving as a control. The doped sample exhibits a clear optical hysteresis response. In contrast, the undoped sample shows no such behavior as illustrated in Supplementary Information Fig. S6. Optical modeling and simulations We employ a 4××4 scattering matrix method to directly compute the reflection coefficients of the sample58,59. The 4××4 scattering matrix method provides a robust framework for modeling the propagation of electromagnetic waves through multilayered structures with a general dielectric tensor. This method enables simultaneous treatment of both s- and p-polarizations under arbitrary angles of incidence. The electric and magnetic field components tangential to each interface are expressed as four-component vectors. For each layer, Maxwell’s equations are recast into a first-order differential system:12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{d}{{dz}}{{\mathbf{\psi }}}\left(z\right)=i{{\bf{K}}}(z){{\mathbf{\psi }}}\left(z\right)$$\end{document}ddzψz=iK(z)ψzwhere ψ(z) = [Ex, Ey, Hx, Hy]T is the state vector and K(z) is the 4××4 propagation matrix determined by the material’s permittivity and permeability tensors. E and H are electric and magnetic field respectively. The general solution in each layer involves the eigenmodes of K, which are propagated and matched across interfaces using boundary conditions that enforce continuity of tangential fields. By recursively applying the transfer matrices across all layers, we relate the input and output field amplitudes at the outermost boundaries of the multilayer stack. The resulting scattering matrix S encodes reflection and transmission coefficients for all polarization channels, including off-diagonal terms rsp and rps which are critical for capturing nonreciprocal effects. This formalism is particularly suited for gyrotropic or magneto-optical materials, where standard scalar methods fail to capture the correct polarization dynamics. A detailed derivation and numerical implementation of the solver are provided in the Supplementary Information Section III. Supplementary information Supplementary Information Transparent Peer Review file Supplementary Information Transparent Peer Review file