Magnon-Exciton Coupling

• Magnon–exciton coupling is the interaction between optical excitations—ranging from crystal-field d–d transitions to Coulomb-bound electron–hole pairs—and collective spin waves, activating weak or forbidden processes.
• It manifests in various regimes including sideband absorption in antiferromagnets, coherent modulation of excitonic resonances by driven spin precession, and hybrid spin–orbit exciton–magnon modes.
• Studies in systems like KCuF₃, LiNiPO₄, and CrSBr show that this coupling not only probes magnetic phase structures but also mediates exciton interactions and enables microwave–optical transduction.

Magnon–exciton coupling denotes the coupling of an optical electronic excitation to a collective spin excitation. In insulating transition-metal compounds, the “exciton” may be a local crystal-field $d$–$d$ excitation; in semiconductors and van der Waals magnets it is more commonly a bound electron–hole pair. The magnon supplies spin angular momentum, crystal momentum, or both, thereby activating otherwise weak or forbidden optical processes, and in other settings it dynamically modulates exciton energies, optical susceptibilities, and transport. Across the literature, the same label therefore covers several distinct regimes: exciton–magnon sideband absorption in ordered antiferromagnets, coherent modulation of excitonic resonances by driven spin precession, hybrid spin–orbit exciton–magnon modes, and magnon-mediated effective interactions between excitons or between electrons and holes (Eremin et al., 2011, Sun et al., 8 Apr 2026, Gloppe et al., 2020, Dhakal et al., 2024).

1. Conceptual scope and elementary phenomenology

The basic quasiparticles are defined differently in different material classes. In localized transition-metal antiferromagnets, “excitons” refers to local crystal-field excitations such as spin-forbidden or magnetic-dipole-allowed $d$–$d$ transitions of ions like Cu$^{2+}$ or Ni$^{2+}$. In semiconductors and van der Waals magnets, excitons are Coulomb-bound electron–hole pairs with finite oscillator strength. Magnons are quantized spin-wave excitations of an ordered magnetic lattice; in antiferromagnets they are shaped by exchange, anisotropy, and sometimes Dzyaloshinskii–Moriya interactions (Sun et al., 8 Apr 2026, Bae et al., 2022).

A standard optical manifestation is the exciton–magnon transition, or magnon sideband, in which an exciton is created while a magnon is either annihilated or created. In the idealized picture, such sidebands appear at

$$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$

with “hot” sidebands requiring thermally occupied magnons and “cold” sidebands surviving as magnon-emission processes at low temperature (Sun et al., 8 Apr 2026). This sideband language is natural for KCuF$_3$ and LiNiPO$_4$, but it is not exhaustive.

A second regime is coherent modulation: a driven magnon mode modifies an excitonic resonance in time. In CrSBr, for example, the exciton energy follows the spin canting angle as

$E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$

so a microwave-driven or optically launched magnon that modulates $d$0 imprints sidebands or time-domain oscillations on reflectivity (Adak et al., 3 Apr 2026, Datta et al., 2024). A third regime occurs when local spin–orbit excitons and magnons overlap spectrally and hybridize into mixed modes, as in FePS$d$1, where the low-lying excitations are most accurately viewed as hybrid spin–orbit exciton–magnon modes rather than pure spin waves or pure local excitons (Dhakal et al., 2024).

These categories are related but not interchangeable. A plausible implication is that the term “magnon–exciton coupling” is best understood as a family of exchange-, symmetry-, and susceptibility-mediated processes, rather than a single universal Hamiltonian.

2. Exchange-activated optical coupling in insulating antiferromagnets

A canonical microscopic treatment was developed for KCuF$d$2, where the coupling is described by the exchange-induced magnetic dipole mechanism. The background magnetism is modeled by the nearest-neighbor Heisenberg Hamiltonian

$d$3

with antiferromagnetic $d$4 along the $d$5 axis and weak ferromagnetic $d$6 in plane, while light couples through the magnetic dipole operator

$d$7

Using a canonical transformation, the effective absorption operator contains a two-site term

$d$8

which couples an orbital excitation on one site to spin-wave creation on the antiferromagnetically coupled neighbor (Eremin et al., 2011).

The physical picture is explicitly two-site. A virtually hopping carrier traverses an antiferromagnetic bond, the magnetic field of light drives a magnetic-dipole transition on one site, and the superexchange loop supplies the phase structure needed to emit a finite-$d$9 transverse magnon while maintaining the photon’s $d$0 constraint. In the low-temperature limit this leads to an interaction vertex

$d$1

with exchange-renormalized matrix element $d$2 (Eremin et al., 2011).

For KCuF$d$3, the resulting selection rules are those of magnetic dipole transitions, not electric dipole transitions. In $d$4 site symmetry, the observed A2 excitation to $d$5 and A3 excitation to $d$6 are M1-active, and the zero-magnon lines plus sidebands follow a $d$7 polarization dependence. The sidebands are strong for $d$8 in the geometry discussed. By contrast, the canonical exchange-induced electric dipole mechanism yields different selection rules and, for KCuF$d$9, predicts a polarization dependence not borne out experimentally (Eremin et al., 2011).

The absorption spectrum also carries a structure-dependent momentum filter. The derived coefficient

$d$0

contains the factor $d$1, which retains the low-energy van Hove singularities at the magnetic Brillouin-zone points $d$2 and $d$3 while filtering out the large-density-of-states $d$4 and $d$5 points. With $d$6 K and $d$7 K from inelastic neutron scattering, the magnon density of states has singularities at $d$8, $d$9, $^{2+}$0, and $^{2+}$1 meV, but the optical sidebands appear only near $^{2+}$2–$^{2+}$3 meV. Experimentally, KCuF$^{2+}$4 exhibits zero-magnon lines at $^{2+}$5 (A2) and $^{2+}$6 (A3), with sidebands at offsets of about $^{2+}$7 meV and $^{2+}$8 meV, consistent with the $^{2+}$9 and $^{2+}$0 singularities of the transverse magnon density of states (Eremin et al., 2011).

This comparison established an enduring distinction between mechanism and lineshape. In KCuF$^{2+}$1, the absence of high-energy $^{2+}$2 sidebands and the M1 polarization rules identify the exchange-induced magnetic dipole channel rather than the exchange-induced electric dipole channel (Eremin et al., 2011).

3. Sidebands as probes of magnetic phase structure

LiNiPO$^{2+}$3 provides a complementary case in which exciton–magnon coupling is not only symmetry-resolved but magnetically switchable. The material crystallizes in orthorhombic Pnma with lattice parameters $^{2+}$4 Å, $^{2+}$5 Å, and $^{2+}$6 Å; Ni$^{2+}$7 sites experience a low-symmetry $^{2+}$8 crystal field, and below $^{2+}$9 K the ordered moments are predominantly aligned along the $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$0 axis with a small canting toward the $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$1 axis. Optical absorption is predominantly observed for polarization $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$2, and magneto-absorption was measured in pulsed fields up to $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$3 T with $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$4 (Sun et al., 8 Apr 2026).

In this system the relevant optical features are broad hot magnon sidebands centered around $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$5 nm ($$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$6 eV) and near $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$7 nm ($$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$8 eV). Their width is $$E_{\mathrm{sideband}} = E_{\mathrm{exciton}} \pm \hbar \omega_m,$$9 meV, much larger than typical magnon bandwidths, so the spectral width is set by the intrinsic linewidth of the electronic excitation, for example by Franck–Condon effects, rather than by the magnon dispersion. The temperature dependence at zero field follows an activated form,

$_3$0

A naive fit yields $_3$1 meV, whereas including a temperature-dependent prefactor $_3$2 gives the best agreement for $_3$3 meV, consistent with a quasi-two-dimensional magnon gap in the $_3$4 planes and prior neutron reports of $_3$5 meV (Sun et al., 8 Apr 2026).

The intensity is controlled by both a magnon factor and a spin-dependent optical matrix element. A commonly used heuristic is

$_3$6

where $_3$7 is the Bose occupation. This dependence becomes experimentally visible across the field-induced phase sequence. The sideband switches sharply near the I$_3$8II transition around $_3$9 T, is strongly suppressed in plateau-like phases I and V, and reappears in phases II and VII. Representative spectra were reported at $_4$0 T for phase I, $_4$1 T for phase II, $_4$2 T for phase V, and $_4$3 T for phase VII (Sun et al., 8 Apr 2026).

The phase selectivity is tied to magnetic structure. Plateau-like phases I and V show strong suppression, consistent with finite magnon gaps and near-collinear spin structures with reduced transverse spin components. By contrast, canted phases IV and VII enhance the coupling because large $_4$4-axis spin components maximize transverse spin fluctuations and antisymmetric exchange pathways. LiNiPO$_4$5 also has strong Dzyaloshinskii–Moriya interactions, with $_4$6, which provide an efficient antisymmetric exchange channel for exciton–magnon coupling in noncollinear and canted states (Sun et al., 8 Apr 2026).

This field dependence has an important methodological consequence: magnon sideband spectroscopy can function as a phase-sensitive optical probe of spin order. In LiNiPO$_4$7, the intensity map at $_4$8 nm follows the magnetic phase diagram, showing strong suppression in phases I and V and high intensity in canted phases, notably IV, despite its narrow field window (Sun et al., 8 Apr 2026).

4. Layered van der Waals magnets and hybrid spin–orbital regimes

The layered antiferromagnet CrSBr has become a central platform because it hosts strong excitons, GHz antiferromagnetic resonances, and unusually large sensitivity of exciton energies to interlayer spin alignment. CrSBr is an orthorhombic van der Waals semiconductor with A-type antiferromagnetic order: intralayer exchange aligns Cr spins ferromagnetically along the $_4$9 axis, and interlayer exchange aligns successive layers antiparallel below the Néel temperature. In reflectivity, one study resolves a high-energy exciton $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$0 near $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$1 eV and a low-energy exciton $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$2 near $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$3 eV, with shifts

$E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$4

where $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$5 meV for $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$6 and $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$7 meV for $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$8 (Adak et al., 3 Apr 2026). A related magneto-optical study on thick flakes identifies an A exciton near $E_X - E_{X,0} = -\Delta_B \cos^2(\theta/2),$9 eV and a B exciton near $d$00 eV, with field-induced shifts $d$01 meV and $d$02 meV for in-plane switching at $d$03 T, and $d$04 meV and $d$05 meV under $d$06 at $d$07 T (Shi et al., 2024).

Time-domain measurements further show that above-gap optical excitation launches coherent magnons whose time-dependent interlayer spin texture modulates exciton energies. At $d$08 K and $d$09 T along $d$10, transient reflectance reveals two modes at $d$11 GHz and $d$12 GHz. Their coherence times are $d$13–$d$14 ns in bulk and exceed $d$15 ns in a $d$16-layer sample, with propagation lengths of about $d$17 and $d$18 and dynamic exciton modulation $d$19 meV in bulk (Bae et al., 2022). The optical signature is dispersive: the oscillatory signal flips phase by $d$20 across excitonic resonances, indicating dynamic energy modulation rather than a simple intensity change (Bae et al., 2022).

Symmetry control is unusually explicit in CrSBr. With $d$21, the canted antiferromagnetic state is invariant under a $d$22 rotation about $d$23. Under this symmetry, the optical magnon is bright because out-of-phase precession changes the interlayer angle $d$24, whereas the acoustic magnon is dark because in-phase precession leaves $d$25 nearly unchanged. Tilting the in-plane field by $d$26 breaks the $d$27 symmetry and hybridizes bright and dark modes. Experiments using $d$28, $d$29, $d$30, and $d$31 show avoided crossings near $d$32–$d$33 T, with a splitting that grows linearly with $d$34 and reaches as much as half the low-energy magnon frequency; the linewidth narrows by a factor of $d$35 at the avoided crossing (Diederich et al., 2022). Under uniaxial tensile strain along $d$36, the interlayer exchange decreases, the saturation field is reduced, and at a critical strain $d$37 the dark branch develops an approximately field-independent segment, described as a dispersionless dark magnon band (Diederich et al., 2022).

CrSBr also illustrates that magneto-exciton coupling can be large without reducing to bare Zeeman splitting. In a thick $d$38 nm flake, a $d$39 T in-plane field is sufficient to induce AFM-to-FM switching, while out-of-plane fields to $d$40 T drive continuous canting. The B exciton redshift of $d$41 meV at $d$42 K under $d$43 is too large to be explained by a reasonable exciton $d$44 factor alone; the study therefore attributes the shift to exchange coupling to the evolving magnetic order and interlayer hybridization, with a microscopic emphasis on Cr–S orbital hybridization and the larger Bohr radius of the B exciton (Shi et al., 2024).

FePS$d$45 represents a different limit, in which the elementary excitations themselves are already mixed. Fe$d$46 carries unquenched orbital degrees of freedom with $d$47, and the single-ion problem is governed by

$d$48

with $d$49 meV. Because spin–orbit coupling, trigonal crystal field, and intersite exchange are comparable, the ground state and excitations admix $d$50, $d$51, and $d$52 sectors. Linear Flavor Wave Theory then yields low-energy modes at $d$53 meV, $d$54 meV, and $d$55 meV, in quantitative agreement with recent experiments, and the modes are explicitly hybrid spin–orbit exciton–magnon excitations rather than pure magnons (Dhakal et al., 2024). This case is significant because it shows that an $d$56-only description can be inadequate when orbital moments are not quenched.

5. Hybrid interfaces, resonant photonics, and microwave–optical transduction

A distinct realization of magnon–exciton coupling occurs at heterointerfaces, where magnons do not necessarily create optical sidebands by exchange-assisted absorption but instead generate effective fields that act directly on excitonic degrees of freedom. In a MoSe$d$57/YIG van der Waals heterostructure, coherent ferromagnetic resonance of a $d$58-thick YIG film at $d$59 GHz modulates the valley excitons in an adjacent MoSe$d$60 flake through a dynamical valley Zeeman effect. The effective Hamiltonian is written as

$d$61

where $d$62 labels the $d$63 valleys, $d$64 is the magnon operator, and $d$65 is the exciton operator. Experimentally, the reflectance signals satisfy $d$66, showing opposite modulation for $d$67 and $d$68 helicities. Calibrated single-magnon couplings are

$d$69

well above the dipolar upper bound $d$70, which identifies short-range interfacial exchange as the dominant channel (Gloppe et al., 2020).

In bulk CrSBr, microwave-to-optical transduction exploits the strong dependence of exciton energies on antiferromagnetic canting. The coupled system is modeled by

$d$71

which linearizes to

$d$72

The estimated single-particle coupling is $d$73 kHz for $d$74 in the bulk crystal, about three orders of magnitude larger than off-resonant magneto-optical coupling typically reported in YIG. Without cavity enhancement, coherent sidebands are observed over an intrinsically broadband window of $d$75 MHz, set by the magnon linewidth, and the measured lower bound on overall conversion efficiency is $d$76 (Adak et al., 3 Apr 2026). The experiment further shows that multiple self-hybridized exciton-polariton resonances inherit the magnon-coupled response, which broadens the usable optical detuning range and mitigates optical dissipation (Adak et al., 3 Apr 2026).

Photonic structure strongly modifies how this coupling is observed. A transfer-matrix analysis for CrSBr emphasizes that coherent magnons may primarily shift the exciton resonance energy,

$d$77

with simulation parameters $d$78, $d$79, and $d$80, but the optical readout can nevertheless appear as a red-shift, a nearly vanishing response, or a blue-shift depending on photonic interference and whether the exciton–photon sector is in weak or strong coupling. The same study defines the reflectance contrast by

$d$81

and shows that a Bayesian Optimization Structure Search can raise $d$82 from $d$83 for $d$84 nm CrSBr on SiO$d$85(285 nm)/Si to amplitudes exceeding $d$86 with a metallic mirror, while reducing the exciton linewidth to $d$87 meV would allow $d$88 to approach the theoretical maximum $d$89 (Budak et al., 9 Mar 2026).

These developments establish an important interpretive caution. Optical spectra near exciton resonances do not provide a one-to-one measure of microscopic exciton–magnon coupling unless interference, cavity effects, and polariton composition are modeled quantitatively (Budak et al., 9 Mar 2026).

6. Mediated interactions, transport, and unresolved issues

Beyond spectroscopy and transduction, magnons can mediate effective interactions among excitons or between spatially separated electrons and holes. In CrSBr, a field-tunable, density-dependent redshift of the B exciton was interpreted as an attractive magnon-mediated exciton–exciton interaction. The exciton energy is written as

$d$90

with $d$91 a small cant-angle adjustment of a magnon mode. Minimization of the total energy gives the absorbed-photon energy

$d$92

so the density-dependent redshift is maximal near $d$93 and vanishes in strictly AFM or FM states. Experimentally, the additional redshift is largest at intermediate fields, negligible for the A exciton, and not attributable to heating by repetition-rate controls (Datta et al., 2024).

A kinetic theory for bilayer CrSBr develops a different many-body consequence: magnon–exciton drag. In that treatment, effective coupling arises from an orbital mechanism in which magnons tilt the layer magnetizations, enabling charge-carrier tunneling that mixes intra- and interlayer excitons and modulates the exciton energy. Integrating out magnons yields efficient two-magnon scattering, two-magnon absorption, and two-magnon emission. Despite rather small renormalization of the exciton energy and effective mass, the exciton–magnon scattering time is in the sub-ps range and strongly decreases with increasing magnon population. Solving the Boltzmann kinetic equation then predicts that magnons can efficiently drag excitons, producing large and nearly isotropic propagation that can significantly exceed intrinsic anisotropic diffusion. The same theory identifies negative group velocity of magnons at small wavevectors due to long-range dipole–dipole interactions, which supplies a microscopic route toward negative drag in far-from-equilibrium regimes (Iakovlev et al., 5 Dec 2025).

Magnon mediation can also act across an insulating barrier. In a trilayer consisting of an electron layer, an antiferromagnetic insulator barrier, and a hole layer, uncompensated interfaces generate a magnon-mediated interlayer interaction whose sign depends on whether the two interfaces contact the same or opposite antiferromagnetic sublattices. Changing the barrier thickness by a single atomic layer switches the interaction between attractive and repulsive. The paper derives an analytical expression for the critical temperature of indirect exciton condensation and estimates $d$94 to be around $d$95 K for realistic parameters, with anisotropy playing a crucial role because it opens the magnon gap that defines the attractive frequency window (Johansen et al., 2019).

Several recurring interpretive issues are now clear. One is mechanism discrimination: in KCuF$d$96, the decisive distinction is between magnetic-dipole and electric-dipole exchange channels (Eremin et al., 2011). Another is linewidth interpretation: in LiNiPO$d$97, the broad sideband width reflects the intrinsic electronic excitation rather than the magnon bandwidth (Sun et al., 8 Apr 2026). A third is optical readout nonuniqueness: in CrSBr, photonic interference can invert or suppress the apparent spectral signature of an unchanged microscopic coupling (Budak et al., 9 Mar 2026). Finally, some spectra remain beyond single-magnon descriptions; for KCuF$d$98, additional higher-energy sidebands likely involve multi-magnon processes or fractionalized spin excitations such as spinons (Eremin et al., 2011).

Taken together, these results define magnon–exciton coupling as a broad research domain spanning exchange-assisted optical absorption, coherent susceptibility modulation, hybrid spin–orbital quasiparticles, and magnon-mediated many-body forces. The common thread is that excitonic observables become direct probes, and in some cases control knobs, of spin-wave physics.