Magnetic Excitons Overview

• Magnetic excitons are electron–hole excitation modes whose energies and optical properties are directly governed by magnetic order.
• They exhibit strong coupling with magnetic phases in materials like van der Waals magnets and quantum Hall systems, revealing diverse many-body interactions.
• Their tunable dynamics are key to developing advanced spintronic and quantum optoelectronic applications through engineered magnetic and electronic band structures.

Magnetic excitons are excitonic optical excitations or neutral collective electron–hole modes whose energies, oscillator strengths, line shapes, selection rules, transport, or lifetimes are governed by magnetism. In contemporary condensed-matter usage, the term does not denote a single universal microscopic object. In van der Waals magnetic semiconductors it commonly refers to bound electron–hole excitons whose properties are strongly imprinted by magnetic order because the same transition-metal–ligand orbitals generate both the local moments and the optical transitions (Adak et al., 11 Feb 2026). In other settings it can denote a local many-body excitation between quantum-entangled Zhang–Rice states in a multiferroic antiferromagnet (Son et al., 2021), an interlayer electron–hole pair in a bilayer quantum Hall system (Doretto et al., 2012), or a spin-selected exciton that condenses in an excitonic-insulator phase (Liu et al., 2024). The common element is a direct coupling between excitonic structure and magnetic degrees of freedom rather than a merely perturbative Zeeman shift.

1. Conceptual scope and definitions

A central feature of magnetic-exciton physics in magnetic semiconductors is that excitons and magnetic moments arise from the same orbital manifold. In van der Waals magnets, partially filled transition-metal $d$ orbitals, metal–ligand hybridization, crystal-field splitting, and Coulomb correlations define both the semiconducting gap and the local magnetic moments, so exchange, spin–orbit coupling, and lattice symmetry directly affect the exciton (Adak et al., 11 Feb 2026). This produces a spectrum spanning localized Frenkel-like multiplet excitons, Wannier–Mott excitons, charge-transfer or Zhang–Rice-type excitons, and mixed-character hybrid excitons, rather than a single canonical form (Adak et al., 11 Feb 2026).

The term also covers markedly different microscopic limits. In NiI$_2$, the magnetic exciton is a local, strongly bound optical excitation between a Zhang–Rice triplet ground state and a Zhang–Rice singlet excited state of the NiI$_6$ cluster; it appears as an ultrasharp peak at $1.384\ \mathrm{eV}$ with a $5.34\ \mathrm{meV}$ linewidth at $2\ \mathrm{K}$, and its optical visibility is enabled only in the multiferroic spiral phase below $T_{N2}=59.5\ \mathrm{K}$ (Son et al., 2021). In bilayer quantum Hall physics at total filling factor $\nu_T=1$, a magnetic exciton is instead a neutral interlayer electron–hole pair created by transferring an electron between lowest-Landau-level branches associated with different layers, so that the excitation is naturally described as a pseudospin flip and can be approximately bosonized (Doretto et al., 2012).

A common misconception is to identify magnetic excitons with magnons or with any optically active state in a magnetic material. The literature distinguishes these cases sharply. Magnetic excitons are often ordinary excitons whose properties are magnetically tunable, whereas magnons are spin-wave excitations; in several systems the relevant phenomenon is not identity but coupling, including coherent and incoherent exciton–magnon scattering, magnon-assisted emission, or magnon-dressed propagation (Dirnberger et al., 9 Jul 2025).

2. Microscopic mechanisms of magneto-exciton coupling

The strongest documented microscopic mechanism is magnetic-order-driven reconstruction of the band-edge electronic structure. In thick layered CrSBr, two bright excitons near the $I$ point of the Brillouin zone, $A$ and $_2$0, are associated with the $_2$1 and $_2$2 transitions, respectively. CrSBr is an A-type van der Waals antiferromagnet at low temperature, with ferromagnetic alignment within each layer and antiferromagnetic alignment between layers. As temperature crosses the Néel temperature $_2$3, both excitons show an inflection in peak position. An in-plane field along the easy axis $_2$4 drives a direct AFM-to-FM transition in a $_2$5 crystal at $_2$6, shifting the $_2$7 exciton by about $_2$8 and the $_2$9 exciton by about $_6$0, with enhanced photoluminescence and reduced linewidth. Under a hard-axis $_6$1-field, magnetic circular dichroism tracks spin canting up to a saturation field of about $_6$2, and the $_6$3 exciton redshifts by about $_6$4 at $_6$5 (Shi et al., 2024). The reported origin is increased interlayer hybridization during canting and FM alignment, with particular sensitivity of the $_6$6 state because $_6$7 carries S $_6$8 weight and the $_6$9-exciton Bohr radius is $1.384\ \mathrm{eV}$0–$1.384\ \mathrm{eV}$1 times larger than that of $1.384\ \mathrm{eV}$2 (Shi et al., 2024).

In NiI$1.384\ \mathrm{eV}$3, the decisive mechanism is symmetry breaking by multiferroicity. The ZRT$1.384\ \mathrm{eV}$4ZRS excitation is described as a Frenkel-type exciton with even parity that would be optically dark in one-photon absorption unless both inversion and time-reversal symmetries are broken. Below $1.384\ \mathrm{eV}$5, a spiral proper-screw magnetic structure breaks inversion symmetry and generates electric polarization. Second harmonic generation appears only in this low-temperature phase, and the $1.384\ \mathrm{eV}$6 exciton appears only there, with polarization anisotropy locked to multiferroic domains rather than to a crystallographic axis (Son et al., 2021). In this case, magnetism controls not only the energy but the very optical activity of the exciton.

Magnetic proximity produces another mechanism. In monolayer TMDs on magnetic substrates, the exchange field modifies the full Bethe–Salpeter exciton problem rather than only the band edges. Rotating the substrate magnetization away from the out-of-plane direction mixes spin character in the conduction and valence states, so optically dark excitons acquire finite oscillator strength and bright excitons can darken. This “conversion” is explicitly presented as a many-body excitonic effect that cannot be captured by a single-particle band-shift picture (Scharf et al., 2017).

Spatially nonuniform magnetic fields can also generate effective excitonic potentials. For monolayer TMDs near a magnetized ferromagnetic disk or wire, the effective center-of-mass potential

$1.384\ \mathrm{eV}$7

is dominated by the Zeeman term rather than the diamagnetic term for realistic fields, producing valley-selective confinement: one valley is trapped just inside the magnetic border and the other just outside it (Chaves et al., 2024).

3. Representative material platforms

CrSBr has emerged as a model platform because it combines robust antiferromagnetic order, strong optical excitons, and unusually large field-tunable exciton shifts. A later high-field study reported that CrSBr hosts both localized Frenkel-like and delocalized Wannier–Mott-like excitons, described as a duality rare among other magnetic or nonmagnetic 2D materials. In that account, the high-energy exciton $1.384\ \mathrm{eV}$8 is an order of magnitude more sensitive to magnetic-order changes than $1.384\ \mathrm{eV}$9, and measurements up to $5.34\ \mathrm{meV}$0 were used to infer different spatial extents and distinct coupling to lattice vibrations, including a strong temperature-dependent redshift for $5.34\ \mathrm{meV}$1 between antiferromagnetic and ferromagnetic phases that is almost temperature-invariant for $5.34\ \mathrm{meV}$2 (Smiertka et al., 19 Jun 2025).

CrCl$5.34\ \mathrm{meV}$3 illustrates two distinct regimes. In monolayer CrCl$5.34\ \mathrm{meV}$4, biaxial compressive strain drives a FM-to-AFM transition at about $5.34\ \mathrm{meV}$5. Solving an effective Bethe–Salpeter equation, the low-energy bright peaks are found at about $5.34\ \mathrm{meV}$6, $5.34\ \mathrm{meV}$7, $5.34\ \mathrm{meV}$8, and $5.34\ \mathrm{meV}$9 in the FM phase and at about $2\ \mathrm{K}$0, $2\ \mathrm{K}$1, $2\ \mathrm{K}$2, and $2\ \mathrm{K}$3 in the AFM phase; the lowest bright excitons shift by about $2\ \mathrm{K}$4, larger than the underlying band-gap change, and the lowest exciton changes abruptly from a highly localized real-space state in FM to a more Wannier-like state in AFM (Ebrahimian et al., 25 Sep 2025). In bulk CrCl$2\ \mathrm{K}$5 at room temperature, by contrast, the relevant claim is that strong short-range ferromagnetic correlations and quasi-2D confinement stabilize exceptionally bound excitons even though the crystal is not long-range magnetically ordered. Optical spectroscopy identifies an A-exciton at $2\ \mathrm{K}$6, a B-exciton at $2\ \mathrm{K}$7, and a quasiparticle gap of $2\ \mathrm{K}$8, implying binding energies of $2\ \mathrm{K}$9 and $T_{N2}=59.5\ \mathrm{K}$0, respectively (Ermolaev et al., 1 May 2025).

Ni-based antiferromagnets realize more localized and symmetry-sensitive forms. In NiI$T_{N2}=59.5\ \mathrm{K}$1, the magnetic exciton is multiferroic-enabled and tied to a quantum-entangled ZRT ground state with about $T_{N2}=59.5\ \mathrm{K}$2 ligand-hole contribution and a ZRS excited state with about $T_{N2}=59.5\ \mathrm{K}$3 ligand-hole contribution (Son et al., 2021). In suspended few-layer NiPS$T_{N2}=59.5\ \mathrm{K}$4, narrow optical lines near $T_{N2}=59.5\ \mathrm{K}$5 are interpreted as localized magnetic excitons tied to zigzag antiferromagnetic order; the main line $T_{N2}=59.5\ \mathrm{K}$6 appears at about $T_{N2}=59.5\ \mathrm{K}$7, a nearby $T_{N2}=59.5\ \mathrm{K}$8 line lies roughly $T_{N2}=59.5\ \mathrm{K}$9 higher, $\nu_T=1$0 is about $\nu_T=1$1 above $\nu_T=1$2, and a broader $\nu_T=1$3 feature is roughly $\nu_T=1$4 above the main line. These are assigned to bare and magnon-dressed excitons, with strain tuning used to infer increasing delocalization with the number of magnon-mediated hops (Wolff et al., 23 Dec 2025).

MnPS$\nu_T=1$5 represents a different limit, dominated by self-trapped or Frenkel-like emission with strong spin–lattice division of labor. Its broad photoluminescence, large $\nu_T=1$6 Stokes shift, and exceptionally long exciton lifetime of order $\nu_T=1$7 below $\nu_T=1$8 identify a regime in which magnetic order affects radiative recombination while phonons dominate nonradiative decay (Peng et al., 9 Oct 2025).

4. Dynamics, transport, and recombination

Magnetic excitons are distinguished from ordinary excitons not only spectrally but dynamically. In CrSBr, exciton transport has been reported to be driven by spin excitations rather than by conventional diffusion. Time-resolved photoluminescence imaging defines the mean-squared displacement

$\nu_T=1$9

and an effective transport coefficient

$I$0

Near the Néel temperature $I$1, $I$2 reaches approximately $I$3, the propagation becomes nearly isotropic in the $I$4-plane despite the highly anisotropic excitonic band structure, and the behavior is attributed to drag from incoherent magnon currents. At $I$5, the exciton cloud can transiently contract rather than expand, with $I$6 and an inward velocity of about $I$7. In bilayers, the propagation follows $I$8 with $I$9 to $A$0, consistent with superdiffusion (Dirnberger et al., 9 Jul 2025).

Exciton lifetimes can be switched by magnetic phase transitions. In thin CrSBr layers, time-resolved photoluminescence around $A$1 shows a step-like reduction of the exciton lifetime from $A$2 in the AFM phase at $A$3 to $A$4 in the FM phase at $A$5, with the abrupt change occurring at about $A$6 for $A$7. The effect persists below $A$8 and nearly disappears in the paramagnetic phase. Ab initio BSE predicts a larger free-exciton oscillator strength in AFM by a factor of about $A$9, which would imply the opposite lifetime trend; the reported resolution is that the experimentally relevant low-temperature states are strongly localized excitons, whose oscillator strength increases in the FM phase because the localization volume becomes larger (Kalitukha et al., 8 Jan 2026).

MnPS$_2$00 separates radiative and nonradiative channels particularly clearly. Its exciton lifetime satisfies

$_2$01

and the extracted radiative rate obeys

$_2$02

Below $_2$03, the radiative channel is interpreted as magnon-assisted, with a fitted boson energy of $_2$04 matching the magnon energy at the Brillouin-zone boundary, while the long lifetime is controlled by phonon-mediated nonradiative recombination with a fitted phonon energy of $_2$05 (Peng et al., 9 Oct 2025). This division between magnon-assisted emission and phonon-limited quenching is one of the clearest demonstrations that magnetic excitons need not couple identically to all bosonic reservoirs.

5. Engineered and field-driven magnetic excitons

Diluted magnetic semiconductors provide a classical magnetic-exciton setting in which exchange with localized magnetic ions dominates the dynamics. In a Zn$_2$06Mn$_2$07Se quantum well of width $_2$08, the giant Zeeman effect tunes the bright $_2$09 heavy-hole exciton toward excited $_2$10 and $_2$11 states. The quoted bound-state energies are $_2$12 for $_2$13, $_2$14 for $_2$15, and $_2$16 for $_2$17, with nominal resonance fields $_2$18 for $_2$19 and $_2$20 for $_2$21. Quantum kinetic theory beyond the Markov approximation predicts that exciton–impurity correlations bridge several-meV detunings and populate excited states already below the nominal resonance field, with substantial spin transfer into the optically dark $_2$22 exciton parabola protected against radiative decay (Ungar et al., 2019).

Colloidal nanoplatelets realize a related exchange-coupled limit. In CdSe/(Cd,Mn)S core/shell nanoplatelets, magnetic excitons are excitonic states exchange-coupled to Mn$_2$23 ions in the shell. Polarized photoluminescence shows sign reversal of the circular polarization relative to nonmagnetic reference samples, optically detected magnetic resonance at $_2$24 yields $_2$25, and Mn spin-lattice relaxation times of $_2$26, $_2$27, and $_2$28 are used to infer Mn concentrations of about $_2$29, $_2$30, and $_2$31 in different samples (Shornikova et al., 2020). Here the magnetic exciton functions as a spectroscopic probe of localized-spin polarization and carrier leakage into the magnetic region.

Magnetic fields can also engineer confinement directly. For monolayer TMDs near ferromagnetic disks or wires, the Zeeman contribution is predicted to dominate over the diamagnetic term unless fields approach $_2$32. One valley is then confined inside the disk border and the opposite valley outside it, with wavefunction peaks separated by about $_2$33 for $_2$34, making the effect accessible to circularly polarized photoluminescence (Chaves et al., 2024). Magnetic proximity furnishes a complementary route: in TMD/magnetic-substrate heterostructures, rotation of the substrate magnetization converts dark and bright excitons by spin mixing, again requiring a full excitonic treatment rather than a single-particle description (Scharf et al., 2017).

6. Collective phases, magnetic moments, and theoretical formalisms

Magnetic excitons also appear as collective or condensed phases. In one-dimensional systems, spontaneous excitonic instability is defined by a lowest exciton energy $_2$35, and a unified critical-temperature relation is derived as

$_2$36

Within this framework, $_2$37. First-principles $_2$38-BSE calculations further predict single-chain staircase Scandocene wires to be antiferromagnetic spin-triplet excitonic insulators with $_2$39, and Chromocene wires to be ferromagnetic half-excitonic insulators with a spin-down instability at $_2$40 but no instability in the spin-up channel, where $_2$41 (Liu et al., 2024).

In bilayer quantum Hall systems at $_2$42, the magnetic exciton becomes a bosonized interlayer electron–hole pair. The effective Hamiltonian

$_2$43

supports a zero-momentum condensate corresponding to the Halperin $_2$44 state at small $_2$45, and a finite-momentum condensate at intermediate layer separation $_2$46. The latter has a gapped neutral spectrum, and the crossing of ground-state energies near $_2$47–$_2$48 in the one-mode treatment, and around $_2$49 for $_2$50 in the two-mode treatment, is interpreted as strong evidence for a first-order quantum phase transition (Doretto et al., 2012).

The magnetic response of excitons themselves now has a distinct quantum-geometric formulation. A fully quantum theory of the exciton orbital magnetic moment expresses

$_2$51

namely a single-particle term, a relative-motion term, and a center-of-mass term associated with exciton-band geometry. In biased bilayer graphene, this resolves the long-standing discrepancy in the valley $_2$52-factor of $_2$53-excitons: the free electron–hole approximation gives $_2$54, an envelope-weighted approximation gives $_2$55 and $_2$56, while the full theory yields $_2$57 and $_2$58, in close agreement with measured values $_2$59 and $_2$60 (Sethi et al., 8 Sep 2025).

The diversity of theoretical descriptions reflects the absence of a single microscopic archetype. Current work employs $_2$61-BSE for excitonic spectra and orbital moments, DFT/HSE06 and DFT+$_2$62+$_2$63 for magnetically ordered band structures, quantum kinetic theory for exciton–impurity correlations, bosonization and Bogoliubov theory for collective condensates, and spin-wave or Boltzmann-type approaches for exciton–magnon dynamics (Adak et al., 11 Feb 2026). A plausible implication is that “magnetic exciton” is best regarded as a unifying phenomenological category defined by strong coupling between excitons and magnetic order, rather than by a unique internal composition.