Imaging exciton–polariton transport in MoSe2 waveguides

For decades, the dream of optical computing has faced a single stubborn wall. Light travels fast, but it doesn't slow down easily, and it doesn't interact with itself. Electrons carry information more flexibly, but they lose energy as heat and reach speeds that physics is increasingly reluctant to improve. What we need is something in between — a carrier that moves like light but interacts with matter the way electrons do. A group led by Zhe Fei at Iowa State recently demonstrated that a crystal just three atoms thick can produce exactly that. The key idea is the exciton-polariton. An exciton forms when light is absorbed by a semiconductor, promoting an electron and leaving behind a positively charged hole. The two stay bound together by electrostatic attraction, orbiting each other like a miniature hydrogen atom within the crystal. When those excitons couple strongly enough to photons, they hybridize into a new quasiparticle: part light and part matter, inheriting long-range propagation from the photon and strong material interaction from the exciton. Hu and colleagues describe this as a "half-light and half-matter quasiparticle." The challenge is to achieve that hybridization in a form that's actually usable — not cryogenically cooled, not buried inside a bulky mirror cavity, but open to the air at room temperature. The material they chose is molybdenum diselenide, or MoSe2, which is a transition-metal dichalcogenide — a class of atomically thin van der Waals semiconductors. What makes these materials special is that their excitons are unusually tightly bound. The A exciton of MoSe2 sits near 1.55 electronvolts, and that large binding energy keeps the exciton intact at room temperature rather than dissolving into thermal noise. Large binding energy also means strong oscillator strength — the exciton couples powerfully to light. Previous exciton-polariton experiments with two-dimensional materials embedded them inside optical microcavities, sandwiching them between mirrors to trap photons long enough for coupling to occur. Hu and colleagues stripped that scaffolding away entirely. They placed exfoliated MoSe2 flakes on silicon dioxide and silicon, letting the flakes themselves act as planar waveguides — no mirrors, no cavities, just the material and the modes it naturally supports. To observe what was happening inside those flakes, the team used scattering-type scanning near-field optical microscopy, or s-SNOM. Here's what that means concretely. A metal-coated atomic force microscope tip, sharpened to a point with a radius of curvature around 25 nanometers, is brought very close to the sample surface. A tunable titanium-sapphire laser, covering a range of 1.3 to 1.8 electronvolts, illuminates the tip. The tip concentrates the light far below the diffraction limit — think of it as a nanoscale antenna — and launches in-plane propagating modes directly into the MoSe2 flake. A fraction of those modes travel to the flake edge, scatter back into free-space photons, and interfere with light that reflects directly off the tip. As the tip scans toward the edge, that interference produces standing-wave fringes in the near-field amplitude signal. Those fringes are what the images show: oscillating stripes whose spacing encodes the polariton wavelength. The mode identified by Hu and colleagues is the TM0 mode — the fundamental transverse-magnetic waveguide mode, excited because the electric field under the tip is oriented perpendicular to the sample surface. The period of the fringes translates to polariton wavelength through a geometric phase relation that accounts for the incident angle of the laser. At an excitation energy of 1.38 electronvolts — corresponding to a free-space wavelength of 900 nanometers — the measured polariton wavelength comes out to roughly 380 nanometers, which is less than half the free-space value. That compression is one of the central results. And it's tunable. By changing the thickness of the MoSe2 flake, the polariton wavelength can be adjusted from about 600 nanometers down to around 300 nanometers. By scanning the laser frequency toward the A exciton at 1.55 electronvolts, the dispersion — the relationship between polariton energy and momentum — curves back on itself in a characteristic curve. That back-bending is not a glitch; it is the hallmark signature of strong coupling. When two modes hybridize, their combined dispersion avoids crossing and instead curves away from each other, producing upper and lower polariton branches separated by the Rabi splitting. From the extent of that back-bending, Hu and colleagues estimate a Rabi splitting of approximately 100 millielectronvolts. That number deserves attention. The Rabi splitting is the energy gap that opens between the upper and lower polariton branches during the anti-crossing. For strong coupling to be genuine — not just a minor adjustment — that splitting must exceed the linewidths of both the photonic mode and the exciton. At room temperature, the A exciton of MoSe2 has a Lorentzian linewidth of about 100 millielectronvolts. A Rabi splitting of the same magnitude indicates that the system is right at the boundary of strong coupling. The team's supplementary analysis shows that cooling the material to low temperatures narrows the exciton linewidth to around 50 millielectronvolts, which sharpens the back-bending in the calculated dispersion and pushes the system more firmly into the strong coupling regime. The room-temperature result is impressive precisely because it works at all. To validate their findings, the team conducted a multilayer transfer-matrix calculation — a standard method for computing how light propagates through a stack of layers with different dielectric properties. They used published dielectric constants for MoSe2 in the plane and set the out-of-plane permittivity to 8.3. The imaginary part of the computed reflection coefficient, plotted as a function of energy and in-plane momentum, reveals a bright feature where the TM0 mode resides — and that feature aligns closely with the experimental data points extracted from the fringe measurements. The back-bending in the calculated dispersion matches what the images show. Mode positions, degree of bending, and the energy at which the polariton wavelength compresses most strongly all line up. The propagation lengths are where operating under ambient conditions becomes particularly significant. Hu and colleagues extract the propagation length from the linewidth of the Fourier peak in the fringe pattern. The in-plane polariton momentum is treated as a complex number: the real part is two pi divided by the polariton wavelength, setting the oscillation, and the imaginary part is one over twice the propagation length, setting the decay. At low excitation energies — far from the exciton resonance — the propagation length exceeds 12 micrometers. In one image at 1.35 electronvolts, fringes remain visible thirty micrometers from the edge. Near the A exciton, where absorption is stronger, the propagation length falls to around 2 micrometers. That energy dependence is the expected trade-off: closer to resonance, coupling is stronger and wavelength compression is greater, but the exciton's absorption also takes a larger bite out of the propagation. Twelve micrometers at room temperature, in a flake of material resting on glass, with no enclosing cavity. Hu and colleagues note that these propagation lengths compare favorably with graphene surface plasmons and hexagonal boron nitride phonon-polaritons under similar conditions. The key advantage MoSe2 offers is the large exciton binding energy, which maintains the strongly coupled state stable without refrigeration. What this package of results opens up is a pathway toward nanophotonic circuitry that does not require cryogenic cooling or precision-fabricated mirror stacks. Polariton wavelengths down to 300 nanometers mean you could route optical signals in features far smaller than conventional photonics allows. Propagation lengths exceeding 12 micrometers mean signals can travel between functional elements on a chip. The planar waveguide geometry — a flake on a substrate — is compatible with standard device fabrication in a way that optical cavities are not. The paper suggests that patterning the flakes, operating at low temperatures to sharpen the coupling, or exploring alternative polaritonic modes could further enhance performance. What Hu and colleagues created is not a finished device. It's a demonstration that the key ingredients coexist: strong coupling, long propagation, wavelength compression, and room-temperature stability, all within a material that can be grown in a lab and transferred to a chip. Light, compressed into a wave a fraction of its original wavelength, traveling over ten micrometers through a crystal just three atoms thick, at room temperature. That's the result. The implications for what comes next are just beginning to be explored. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.