Can an exciton-polariton condensate in a van der Waals magnet be controlled magnetically?

A condensate that inherits the magnet's spin order

The paper asks whether a condensate of exciton-polaritons — hybrid light-matter quasiparticles formed when cavity photons strongly couple to excitons — can be tuned by magnetism rather than by redesigning the structure. The authors answer with CrSBr, an air-stable layered A-type antiferromagnet whose tightly bound excitons are polarized along the crystallographic b axis. Exfoliated CrSBr flakes are tens of micrometres long along a but only about 1 µm wide along b, so they act as microwires that confine light laterally.

The device is a simple sandwich: a 200-nm gold mirror, a 70-nm PMMA spacer, the CrSBr flake, a second 70-nm spacer and a 40-nm semi-transparent gold mirror. The bare cavity has a Q factor of 62, and the lower polariton mode studied has a lifetime broadening of 14 meV, corresponding to a lifetime of 47 fs. Because the wire is narrow along b, polaritons are quantized into discrete lateral states labelled Q0, Q1 and Q2; along the long a axis they remain essentially parabolic.

This geometry matters for the central claim. The condensate is not a generic photon fluid but a state built from excitons whose energy depends on how the CrSBr spins are ordered. That coupling is the handle the magnetic field later turns.

Intensity jump, line narrowing and coherence

The authors identify condensation through several independent signatures rather than a single measurement. In a cavity 1.25 µm wide and 40 µm long, pumped at 1.325 eV with 240 fs pulses at 85 kHz and held at 5 K, the emission from the excited lateral state Q1 grows nonlinearly with pump fluence. Above a threshold of about 30 µJ cm⁻², raising the fluence by only a factor of three increases the Q1 peak at 1.276 eV by more than two orders of magnitude.

The comparison that makes this convincing is the control excitation. Pumping off-resonantly at 1.771 eV produces linear scaling instead, so the nonlinear jump is tied to the near-resonant pumping condition rather than to fluence alone. The Q1 line also narrows from 19 meV to 8.7 meV and shifts upward by 5 meV across the threshold.

Coherence is measured separately in a Michelson interferometer, where one arm inverts the real-space image and an optical delay line sets the temporal overlap. Below threshold there are no interference fringes; above it, fringes appear and their visibility decays with delay. A Gaussian fit gives a coherence time of 427 ± 7 fs, roughly an order of magnitude longer than the spontaneous polariton lifetime, which the authors treat as direct evidence of condensation rather than ordinary lasing-like emission.

Condensation appears in an excited lateral state

One of the paper's less intuitive results is where the condensate forms. In two samples, condensation occurs in the excited state Q1 rather than the ground state Q0, and Q1 dominates the emission once the threshold is crossed even though Q0 and Q1 emit comparably below it. The pump photons sit 46 ± 3 meV above the Q1 emission.

The authors connect that energy gap to relaxation channels. The difference closely matches an optical magnon mode of CrSBr reported at about 45 meV, and multi-phonon or single A3g phonon emission could also bridge it. Density functional theory calculations show electrons localizing on Cr atoms and holes on S atoms, producing a structural distortion that increases interlayer separation and couples efficiently to the interlayer breathing A1g phonon mode.

The paper's interpretation is that surface exciton states, seen as dispersionless resonances at 1.312 eV and 1.325 eV, provide an efficient entry point, and that phonon- and magnon-assisted scattering then moves excitation into the polariton states. This is presented as a consistent picture rather than a proven causal chain, and the authors note that wire-to-wire variation in confinement can change the details.

A 2 T field shifts the condensate by 10.5 meV

The magnetic-control experiment uses a separate microwire, cavity 5, in a magneto-optical cryostat with an out-of-plane field up to 5 T. Under continuous-wave excitation at 1.907 eV and 10 K, the quantized polariton modes and the surface exciton resonance redshift by more than 10 meV as the field rises from 0 T to 2 T, while a defect emission peak at 1.265 eV stays nearly unchanged. The shift saturates above 2 T, consistent with CrSBr's antiferromagnetic-to-ferromagnetic transition under out-of-plane bias.

The authors argue this is not simple Zeeman splitting or a magnetic polaron effect, because the shape and size of the shift track a reconfiguration of collective spin order. Each confined mode inherits a shift set by its exciton fraction and by the roughly 18 meV shift of bulk excitons, while surface excitons shift less because of their different magnetic environment.

Condensation survives in both phases. Q1 shows the same superlinear intensity increase at 0 T and at 2 T, and the condensed state itself shifts by up to 10.5 meV between the antiferromagnetic and ferromagnetic configurations. The authors contrast this with electric-field tuning, which perturbs the potential felt by polaritons, whereas the magnetic field changes the magnetic order, the symmetry and ultimately the exciton wavefunction from quasi-one-dimensional to extended three-dimensional.

The blueshift may not be only exciton repulsion

The paper also reports a monotonic increase of polariton energy with fluence, which differs from earlier reports of decreasing polariton energies or competing opposite shifts. The authors attribute part of this to repulsive Coulomb interactions between excitons or to saturable absorption, the usual explanations in polariton condensates.

They then add a CrSBr-specific possibility. Magnon-induced magnetic disorder can reduce the oscillator strength of excitons, and when excitons are strongly coupled to photons a smaller Rabi gap raises the lower polariton energy. Because the microwire polaritons have a sizable photon fraction, this effect could be pronounced. The authors present this as a suggested contribution alongside exciton-exciton repulsion, not as an isolated measurement.

This distinction matters for readers comparing the 5 meV threshold blueshift with the 10.5 meV magnetic shift. They are different quantities obtained under different conditions: one is a fluence-driven change near threshold under pulsed excitation, the other a field-driven change between magnetic phases.

Where the magnetic handle is demonstrated and where it is not

The magnetic tuning is shown on one microwire, cavity 5, under continuous-wave excitation at 10 K, while the condensation fingerprints come mainly from cavity 1 at 5 K under pulsed excitation. The authors report similar signatures in other comparable microcavities, including at temperatures above 20 K, but the 10.5 meV figure belongs to the specific field-sweep sample and geometry.

The relaxation picture is explicitly tentative. Surface exciton assignment is described as tentative, resting on the absence of dispersion, the different dielectric environment and strain, and the smaller magnetic shift relative to bulk excitons. The phonon and magnon scattering pathways are supported by energy matching and calculations rather than resolved in time.

The proposed applications — ultrafast modulation via coherent magnons, magnetic-memory integration and microwave-to-optical transduction — are framed as prospects. Nothing in the reported data demonstrates those functions, so they should be read as motivation for the platform rather than as results.

Spin order becomes a tuning knob for coherent emission

The paper's distinctive contribution is not that polariton condensation occurs in a van der Waals magnet, but that the condensate's energy can be moved by 10.5 meV through a 2 T field that reconfigures spin order, while condensation persists in both the antiferromagnetic and ferromagnetic phases. That combination — a macroscopic coherent state whose energy follows the magnetic configuration — is the result worth tracking.

For readers judging follow-up work, the relevant threshold is the 2 T saturation: shifts should be compared at fields below and above the antiferromagnetic-to-ferromagnetic transition, and the exciton fraction of the probed mode should be stated, since each quantized mode inherits a different share of the bulk exciton shift. Comparisons that ignore which mode, which phase and which excitation scheme produced a shift will mix quantities the paper keeps separate.

Sources used in this answer