arXiv:2609.22002
Microwave dielectric properties of LiNbO$_{\mathbf{3}}$ and AlN at millikelvin temperatures and single-photon power
Alessandro Reineri, Francesco Crisa, Maithlee Shinde, Daniel Bafia, Akshay Murthy, Changqing Wang, Tanay Roy, Alexander Romanenko, John Zasadzinski, Anna Grassellino, Silvia Zorzetti
Abstract
Efficient bidirectional microwave-optical photon conversion is a key capability for scaling superconducting quantum processors into distributed networks. However, achieving the necessary conversion efficiency requires filling a critical knowledge gap in understanding the loss mechanisms of electro-optic materials. Here, we characterize the microwave properties of single-crystal bulk LiNbO3 and AlN over a broad range of powers, down to single-photon levels, and spanning from millikelvin temperatures to above 1 K. We demonstrate that both materials exhibit two-level systems (TLS) behavior, while piezoelectric-related losses are excluded. We show that TLS-induced dissipation is predominantly localized on the surface rather than being an intrinsic bulk property, a result further corroborated by room-temperature 3D XPS and time-of-flight SIMS analyses. These findings provide useful insights to engineer hybrid architectures that integrate bulk electro-optic crystals within superconducting cavities, proving that microwave quality factors compatible with high-efficiency microwave-optical transduction are within reach.
AI Summary
The paper investigates whether microwave loss channels observed in thin-film electro-optic devices persist in bulk LiNbO$_3$ and AlN at millikelvin temperatures and single-photon powers, relevant to microwave–optical transduction. Measurements use 6-mm, $z$-cut single-crystal cubes inside superconducting cavities: aluminum for AlN and niobium for LiNbO$_3$. Reflection spectroscopy of TM and TE resonances probes electric fields parallel and perpendicular to the crystal axis. Power sweeps reach approximately one intracavity photon at 10 mK; temperature sweeps extend to 700 mK for AlN and 1.5 K for LiNbO$_3$. Finite-element participation ratios convert measured internal quality factors into effective dielectric loss tangents. Single-photon internal quality factors are $7.0\times10^4$–$8.29\times10^4$ for AlN and $4.68\times10^4$–$1.11\times10^5$ for LiNbO$_3$, corresponding to loss tangents of approximately $1.5\times10^{-5}$–$2.5\times10^{-5}$. Power and temperature dependence support TLS absorption alongside a larger power-independent loss contribution. Relatively early TLS saturation is interpreted as evidence for surface-localized defects; room-temperature XPS and ToF-SIMS identify surface contamination and stoichiometric deviations consistent with that interpretation. No separate piezoelectric-loss contribution is identified. At high power, TM-mode quality factors increase logarithmically, whereas TE responses saturate over the accessible range. Strongly interacting TLS and clamp-induced phononic confinement are proposed explanations, but their origin remains unresolved. The standard TLS fit becomes inadequate at high power, and limited delivered power prevents testing the analogous AlN TE regime. Destructive depth profiling uses separate crystals from the same lot; no surface-treatment intervention or transduction measurement is reported.
Central claim may not hold: Low TLS saturation power and surface contamination do not establish the paper's central claim that microwave-active TLS are predominantly surface-localized rather than bulk-distributed. The categorical exclusion of piezoelectric-related dissipation is unsupported by a quantitative bound or a measurement distinguishing that channel from other losses. High-efficiency transduction compatibility is asserted without a quantitative efficiency criterion or calculation, while approximately 77–90% of measured single-photon dissipation is assigned to the unidentified Q_other channel. The stated high-power logarithmic fit branch is singular at its included threshold and cannot join the finite low-power quality factor. The printed photon-number formula overestimates occupation by 2 pi, and the printed BCS gap formula underestimates the gap by pi squared, with consequences for quantitative interpretation if implemented.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.