Deep inside a cryostat, beneath layers of superconducting coils, lies a stack of hundreds of semiconductor layers. Each layer is just a few nanometers thick: gallium arsenide, then aluminum gallium arsenide. The electrons within these layers are confined to a plane, like ants on a sheet of paper. When the magnetic field reaches 36 tesla, the stack begins to behave like a radio receiver tuned to a frequency that no ordinary detector can hear. This is not a metaphor: the physics is exactly that. An axion passing through a magnetic field converts into a photon. The detector's job is to hear that whisper amid the noise.
The axion is a hypothetical particle, a dark matter candidate, the invisible substance that holds galaxies together. Its mass, according to calculations, lies in the range from fractions to several millielectronvolts. One millielectronvolt is roughly one two-hundred-billionth of the electron's mass. To catch such a particle, you need a resonator tuned precisely to its frequency. Conventional microwave cavities, like those in the ADMX experiment, work well but only up to frequencies of a few gigahertz. An axion with a mass of 1 meV corresponds to a frequency of about 242 GHz, which is already in the terahertz range, where cavities lose their sharpness.
This is where the stack of quantum wells comes into play. In a strong magnetic field, electrons in the wells begin to move in circles at the cyclotron frequency. This frequency depends on the magnetic field and the electron's effective mass. If you tilt the sample, the normal component of the magnetic field changes, and the cyclotron frequency shifts. Thus, by simply rotating by fractions of a degree, you can tune the detector's resonant frequency. The axion field, interacting with the magnetic field through the inverse Primakoff effect, induces an electric field in the medium. At the epsilon-near-zero point, where the dielectric permittivity is nearly zero, this field is amplified by tens to hundreds of times. The signal radiates from the resonator's surface and is focused onto a quantum-dot photodetector, an ultrasensitive single-photon sensor.
Calculations for three benchmark configurations show that the signal amplification factor reaches 70–390. With an integration time of 300 days, a broad scan can achieve sensitivity to the axion-photon coupling constant of about 5×10⁻¹⁴ GeV⁻¹ for masses 0.6–5 meV. This surpasses the current limits of the CAST experiment. An optimistic configuration comes close to the band of theoretically preferred QCD axions, the DFSZ and KSVZ models. The same detector, without modification, searches for dark photons with kinetic mixing χ > 6×10⁻¹⁶ in the same mass range.
The future of this approach lies in scaling. Semiconductor technology is mature: molecular beam epitaxy allows growing structures with record electron mobility. Magnets at 36 tesla already exist, and 50 tesla magnets are under development. The immediate next step is to fabricate a prototype multilayer structure a few millimeters thick and test the resonant response in a real magnet. In parallel, quantum-dot photodetectors can be improved, reducing the dark count to microhertz. Machine learning can help optimize the design of multilayer structures with thousands of layers for maximum signal amplification.
A radio receiver tuned to the whisper of an invisible field is already on the threshold. Its antenna is a stack of semiconductor layers; its tuning knob is a simple tilt. And if the axion truly exists, this receiver will be the first to hear it.
The detector's resonant frequency is tuned by simply tilting the sample in the magnetic field: changing the angle by hundredths of a degree shifts the resonance by a linewidth, allowing the axion mass to be scanned without a single moving mechanical part inside the cryostat.