Reversing Annealing-Induced Optical Loss in Diamond Microcavities

TL;DR

This work addresses the challenge of annealing-induced optical loss in diamond microcavities used for spin-photon interfaces. It systematically analyzes the effect of annealing up to $1200^\circ$C in high vacuum on the optical quality factor, before and after a tri-acid cleaning step, and uses Raman spectroscopy to identify a non-diamond amorphous carbon layer formed during annealing. The key finding is that this amorphous layer increases surface scattering and degrades $Q_{\text{opt}}$, but its removal by cleaning fully restores pre-anneal performance, indicating that color centers can be created in pre-fabricated high-$Q$ cavities without permanent damage. These results support fabrication strategies where diamond resonators are patterned before implantation, with implications for quantum transduction and networking applications that require strong spin-photon coupling.

Abstract

A key challenge for quantum photonic technologies based on spin qubits is the creation of optically active defects in photonic resonators. Several of the most promising defects for quantum applications are hosted in diamond, and are commonly created through ion implantation and annealing at high temperatures and high vacuum. However, the impact of annealing on photonic resonator quality factor, a critical parameter governing their coupling to defects, has not been reported. In this work, we characterize the effect of annealing at temperatures >1200°C in high vacuum on the quality factors of diamond microdisk resonators. We investigate the optical losses associated with a non-diamond layer formed during annealing, and use Raman spectroscopy to analyze the resonator surface morphology and demonstrate that tri-acid cleaning can restore their optical quality factors. These results show the viability of creating defects in pre-fabricated diamond resonators without degrading their optical properties.

Figures (4)

• Figure 1: (a) Illustration of the protocol used to study the effect of annealing and surface treatment on diamond microdisk cavities. Optical $Q$-factors of devices were measured in pre-annealing acid-cleaned (C), annealed (C-A), and acid-cleaned stages (C-A-C). Subjecting the sample to high-temperature annealing creates a non-diamond layer represented by the darker color in (C-A). (b) Schematic of the high-temperature and high-vacuum annealing setup. P1: turbo pump, P2: ion pump, P3: heater, and P4: pressure gauge. (c) Typical time-series of sample temperature (red) and chamber pressure (blue) during the annealing procedure.
• Figure 2: (a) Typical optical spectra of a measured diamond microdisk showing the optical modes of the annealed device (A) superimposed to the same modes after acid-cleaning (A-C). The left inset contains a schematic diagram of the characterization setup (PC: polarization controller, MD: microdisk and PD: photodetector). (b) Scanning electron microscope image of a typical sample (scale bar: 10 $\mu$m). (c) Zoom in on one of the optical modes shown in (a). (d) Representative data and Lorentzian fits of optical modes measured on the clean sample.
• Figure 3: (a) Comparison of the average intrinsic quality factors for various microdisk diameters on the same diamond chip at different stages: clean pre-annealing (C), annealed (C-A) and clean post-annealing (C-A-C). (b) Impact of the pedestal on optical losses: larger (smaller) disks have larger (smaller) pedestals and lower (higher) geometry-limited quality factors. Finite element method simulation shows the electric field of the microdisks. (c) Predictions of optical loss from absorption and scattering extracted from fitting the modified Lorentzian lineshape to resonances exhibiting photothermal effects. Each bar corresponds to measurements from a single resonance of a unique microdisk of the indicated diameter. The absorption component is multiplied by 10.
• Figure 4: Raman spectroscopy of diamond surface. (a) Measurement on the freshly annealed surface (red data, A) showing a native diamond signal around 1350 $\text{cm}^{-1}$ and a broad non-diamond peak $\sim$1600 $\text{cm}^{-1}$. The subsequent tri-acid cleaning removes this non-diamond layer and the spectrum shows just the native diamond signal (blue data, A-C). Only the data collected for 120 seconds is shown for both cases. Inset: Schematics of the Raman spectroscopy setup. (b) Decomposition of Raman spectrum for the data A from (a) to separate different specious contributions to the signal after correcting for the background (black dotted line). The data was fit to three Lorentzian functions and the area under each peak indicates the relative contribution to the signal.