Stability, degeneracy, and scalability of a 600-site cavity array microscope

Optical cavities are a foundational technology for controlling light-matter interactions. While interfacing a single cavity to either an atom or ensemble has become a standard tool, the advent of single atom control in large atomic arrays has spurred interest in a new frontier of ``many-cavity QED,’’ featuring many independent resonators capable of separately addressing individual quantum emitters. In this fast-evolving landscape, the cavity array microscope was recently introduced – employing free space intra-cavity optics to engineer a two-dimensional array of tightly spaced cavity TEM$_{00}$ modes with wavelength-scale waists, ideally suited for interfacing with atom arrays. Here we realize the next-generation of this architecture, achieving hundreds of degenerate cavity modes with improved, uniform finesse, and explore the technical features of the system which will enable further scalability. In particular, we study imperfections, including optical aberrations, field of view constraints, array non-degeneracies, and losses from optical elements. We identify the sensitivity to these various vectors and exposit the control knobs and techniques necessary to align and operate the system in a stable manner. Ultimately, we lay out a pathway towards operation with tens of thousands of independent cavities while maintaining compatibility with existing atom arrays, paving the way to myriad applications including highly parallelized remote entanglement generation, fast and non-destructive mid-circuit readout, and the implementation of hybrid atom-photon Hamiltonians.

💡 Research Summary

This paper presents a comprehensive development of the Cavity Array Microscope (CAM), a many‑cavity quantum electrodynamics platform that leverages free‑space intra‑cavity optics to generate a two‑dimensional lattice of tightly spaced TEM₀₀ modes. Building on the first‑generation device that interfaced 43 cavities with a neutral‑atom array, the authors redesign the system to support more than 600 independent cavities while simultaneously improving three key performance metrics: finesse, field‑of‑view, and mutual degeneracy.

The architecture consists of a microlens array (MLA) that creates an array of optical potential wells, a primary 30× demagnifying 4f telescope (300 mm spherical lens + 10 mm high‑NA aspheric lens) that projects the MLA plane onto the central image plane where atoms are trapped, and a secondary 20× retro‑reflecting telescope that returns the light to the MLA without placing a mirror in the atom plane. With an MLA focal length of 4.8 mm and a pitch of 300 µm, the intrinsic waist at the MLA is ~35 µm; after demagnification the waist at the atom plane is 1.15 µm and the inter‑cavity spacing is 10 µm. The effective cavity length is 1.042 m, giving a free‑spectral range of 143.8 MHz.

A paraxial ABCD‑matrix model reduces the whole system to an effective half‑planar two‑mirror cavity formed by the planar input mirror and a curved “micro‑mirror” with radius equal to the MLA focal length. This model predicts the waist dependence on the distance between the input mirror and the MLA (Δd_Mirror). When Δd_Mirror = f_MLA/2 the waist matches the analytical expression w₀ = (λ · f_MLA)/(π · M), which the experiment confirms. Sensitivity analysis shows that most optics tolerate millimeter‑scale misalignments, but the high‑NA aspheric lens is extremely sensitive: a longitudinal shift of only ~10 µm changes the waist appreciably, demanding active feedback for long‑term stability.

Experimental characterization uses a spatial light modulator (SLM) to address individual cavities sequentially and a shared avalanche photodiode for readout. With a narrow‑linewidth 780 nm laser, the authors sweep the input‑mirror piezo to scan across multiple free‑spectral ranges, extracting the cavity linewidth κ ≈ 1.26 MHz and finesse. Across 603 cavities the average finesse is 114 ± 17, limited primarily by coating absorption and surface roughness (RMS < 15 nm). The corresponding single‑atom cooperativity exceeds 10, thanks to the small waist and a 5.7 mm atom‑surface separation provided by the aspheric lenses.

Degeneracy is quantified by measuring the resonance frequencies of all cavities relative to a reference cavity shifted by 80 MHz. 537 cavities are found to be degenerate within a single linewidth, indicating a residual non‑degeneracy caused by nanometer‑scale optical‑path‑length variations from stress and surface irregularities on the optics. The authors demonstrate that these variations can be compensated by fine‑tuning the SLM phase and by a low‑bandwidth feedback loop that stabilizes the aspheric lens position.

Loss analysis identifies three dominant contributions: (i) mirror and anti‑reflection coating losses (~0.5 % per surface), (ii) scattering from the MLA and aspheric lenses (~0.7 %), and (iii) residual mis‑alignment or asymmetry in the aspheric lens (~0.3 %). The authors argue that replacing the current asphere with a higher‑quality, stress‑relieved element would raise the average finesse toward 150 and improve degeneracy.

Scalability is addressed by proposing two straightforward upgrades: (1) a denser MLA (pitch reduced to 150 µm) and (2) a larger‑NA objective that expands the usable field of view beyond the current 140 µm radius limited by field curvature. Simulations suggest that with these modifications the CAM could support tens of thousands of cavities while preserving the same cooperativity and atom‑surface distance, making it competitive with the largest neutral‑atom arrays (≈10⁴ atoms). The authors also compare the CAM’s information bandwidth (number of cavities × minimum cavity linewidth) against other resonator platforms (nanophotonic cavities, fiber‑based micro‑mirrors, macroscopic mirrors), showing that the CAM offers a unique combination of high bandwidth, low surface‑charge susceptibility, and long working distance.

In conclusion, the paper demonstrates a scalable, high‑performance many‑cavity QED platform that can be integrated with existing neutral‑atom arrays. By achieving >600 degenerate cavities with uniform high finesse and identifying clear pathways to reach >10⁴ cavities, the work opens avenues for massively parallel quantum networking, fast non‑destructive mid‑circuit readout, and the exploration of hybrid atom‑photon Hamiltonians at GHz‑scale entanglement rates.