AlGaAs Crystalline Coatings

• AlGaAs crystalline coatings are monocrystalline, epitaxially grown multilayer mirrors using alternating GaAs and AlGaAs layers that offer high reflectivity and reduced mechanical loss.
• They enable ultrastable cavities and interferometric systems to achieve superior frequency stability and lower thermal noise, with demonstrated finesses up to 470,000.
• The substrate-transfer fabrication process introduces unique challenges such as photo-induced birefringence and wavelength-dependent absorption, which are being actively addressed.

Searching arXiv for recent and foundational work on AlGaAs crystalline coatings relevant to precision interferometry, ultrastable cavities, and gravitational-wave detectors. AlGaAs crystalline coatings are monocrystalline, epitaxially grown multilayer mirrors based on alternating GaAs and AlGaAs that are designed to combine high optical reflectivity with low mechanical dissipation. In precision interferometry they are developed as an alternative to conventional amorphous dielectric coatings such as $\mathrm{SiO_2/Ta_2O_5}$, because coating thermal noise in present systems is a major sensitivity limit. Across micromechanical resonators, ultrastable optical cavities, and gravitational-wave detector studies, the central theme is consistent: the crystalline GaAs/AlGaAs materials platform offers substantially lower coating mechanical loss, but its semiconductor band structure, birefringence, substrate-transfer process, and wavelength-dependent absorption impose system-level constraints that are absent in amorphous stacks (0802.0465, Cole et al., 2023, Lee et al., 16 Sep 2025, Liu et al., 23 Jun 2026).

1. Crystalline Bragg mirrors: materials, architecture, and fabrication

In the AlGaAs coating literature, the basic optical element is a distributed Bragg reflector formed from alternating high-index and low-index III–V layers. In the early micromechanical implementation, the structure consisted of 32.5 periods of $\mathrm{Al_{0.12}Ga_{0.88}As}$ and $\mathrm{Al_{0.92}Ga_{0.08}As}$, followed by a 250 nm high-Al etch-protection layer, grown on a 3″ semi-insulating GaAs substrate by molecular beam epitaxy (0802.0465). For a quarter-wave stack centered at $\lambda_0$, the layer design satisfies

$n_H d_H = n_L d_L = \frac{\lambda_0}{4},$

so that reflections add constructively near the stop band (0802.0465). Later ultrastable-cavity implementations used crystalline GaAs/$\mathrm{Al_{0.92}Ga_{0.08}As}$ multilayers with 48.5 periods of quarter-wave layers at 1542 nm and a total coating thickness $d_\text{coat} = 12.0~\mu\text{m}$ (Lee et al., 16 Sep 2025). A room-temperature 1542 nm cavity employed 38.5 pairs of $\mathrm{Al_{0.92}Ga_{0.08}As/GaAs}$, beginning and ending with GaAs, with layer thicknesses 133.2 nm and 115.6 nm, respectively (Ma et al., 4 Feb 2026).

A defining distinction from conventional deposited coatings is that AlGaAs coatings are first grown epitaxially on GaAs and then substrate-transferred to the final optic. The review of gravitational-wave applications describes them as monocrystalline Bragg mirrors made from alternating GaAs and AlGaAs, grown epitaxially on GaAs wafers and then physically bonded onto the final optical substrate such as fused silica, silicon, or sapphire (Cole et al., 2023). Large-area gravitational-wave prototypes used a 35.5-layer quarter-wave $\mathrm{GaAs/Al_{0.92}Ga_{0.08}As}$ stack with target transmission 10 ppm at 1064 nm, substrate-transferred onto 3″ fused-silica disks (Penn et al., 2018). In ultrastable silicon cavities, the mirror stack is substrate-transferred onto superpolished silicon substrates with 1 m radius of curvature, yielding a 6.02 cm cavity finesse of $470{,}000$ at 1542 nm (Lee et al., 16 Sep 2025).

The fabrication logic differs fundamentally from ion-beam-sputtered amorphous coatings. In the micromechanical resonator work, the optical coating and the mechanical resonator were one and the same physical body: beams or cantilevers were etched directly from the crystalline Bragg stack, producing freestanding mirrors with atomically flat optical surfaces (0802.0465). By contrast, large-area interferometric coatings rely on direct bonding of the epitaxial multilayer to a precision-polished substrate, removal of the original GaAs growth wafer, and post-transfer inspection for point defects and unbonded regions (Penn et al., 2018). This substrate-transfer step is technologically enabling, but later birefringence mapping work indicates that it can also imprint spatial non-uniformity that is likely introduced during bonding (Gretarsson et al., 30 Oct 2025).

2. Optical design and reflectivity across application regimes

The optical appeal of AlGaAs coatings derives from the large index contrast between GaAs and AlGaAs, which permits high reflectivity with fewer layers than many oxide systems. In the MHz micromechanical resonators, the DBR was designed for peak reflectivity at 1078 nm at room temperature; calculations predicted $\mathrm{Al_{0.12}Ga_{0.88}As}$0 at 1064 nm below 20 K and $\mathrm{Al_{0.12}Ga_{0.88}As}$1 at 300 K (0802.0465). Cavity-finesse measurements implied $\mathrm{Al_{0.12}Ga_{0.88}As}$2, with measured cavity linewidth $\mathrm{Al_{0.12}Ga_{0.88}As}$3 MHz and $\mathrm{Al_{0.12}Ga_{0.88}As}$4 (0802.0465).

At 1542 nm in a 17 K silicon cavity, the use of 48.5 crystalline quarter-wave periods produced $\mathrm{Al_{0.12}Ga_{0.88}As}$5 and cavity linewidth $\mathrm{Al_{0.12}Ga_{0.88}As}$6, with scattering and absorption of about 5 ppm per mirror (Lee et al., 16 Sep 2025). A room-temperature 48 cm cavity with crystalline GaAs/$\mathrm{Al_{0.12}Ga_{0.88}As}$7 mirrors reached $\mathrm{Al_{0.12}Ga_{0.88}As}$8 on one polarization eigenmode and $\mathrm{Al_{0.12}Ga_{0.88}As}$9 on the other (Ma et al., 1 Jul 2026). Mid-infrared extensions show that the same platform remains viable well beyond the near-IR: at 4.54 $\mathrm{Al_{0.92}Ga_{0.08}As}$0m, substrate-transferred single-crystal $\mathrm{Al_{0.92}Ga_{0.08}As}$1 interference coatings exhibited excess optical loss below 10 ppm (Winkler et al., 2020), and hybrid supermirrors combining a GaAs/AlGaAs crystalline mirror with an ion-beam-sputtered mid-IR mirror achieved total loss of 9.3 ppm and excess loss of 6.8 ppm at 4.45 $\mathrm{Al_{0.92}Ga_{0.08}As}$2m, with cavity finesse up to 396000 (Perner et al., 29 Aug 2025).

Optical absorption is strongly wavelength dependent because the coatings are semiconductors. A key material fact used in the gravitational-wave detector lock-acquisition papers is the GaAs bandgap $\mathrm{Al_{0.92}Ga_{0.08}As}$3, corresponding to $\mathrm{Al_{0.92}Ga_{0.08}As}$4, so light at 532 nm lies deep in the strong absorption regime, while 1064 nm and 1596 nm are below bandgap (Liu et al., 23 Jun 2026). This makes visible auxiliary-light architectures incompatible with AlGaAs/GaAs-coated arm cavities, even though the coatings are attractive at the 1064 nm science wavelength (Tanioka et al., 2024, Liu et al., 23 Jun 2026). The same band-structure logic explains why a 1.5 $\mathrm{Al_{0.92}Ga_{0.08}As}$5m absorption mapping study found that a stack optimized for 1064 nm transmitted about 70% of the laser light at 1530 nm; after correcting for effective penetration depth, the inferred absorption for a high-reflectivity stack at 1530 nm was $\mathrm{Al_{0.92}Ga_{0.08}As}$6 (Steinlechner et al., 2014).

3. Mechanical dissipation, thermal noise, and quality factor

The principal motivation for AlGaAs coatings is reduced mechanical loss and therefore reduced coating Brownian noise. In the micromechanical resonator study, freestanding epitaxial AlGaAs distributed Bragg reflectors produced mechanical quality factors up to 20 000 at 4 K together with optical reflectivity exceeding 99.98% (0802.0465). At room temperature, singly clamped beams reached $\mathrm{Al_{0.92}Ga_{0.08}As}$7 up to 7000 and doubly clamped beams up to 5000; at 4 K, a doubly clamped beam showed $\mathrm{Al_{0.92}Ga_{0.08}As}$8 for the fundamental mode and $\mathrm{Al_{0.92}Ga_{0.08}As}$9 for the second mode (0802.0465). The paper emphasizes that these values are significantly higher than those of comparable freestanding dielectric DBRs and attributes the improvement to the monocrystalline structure, which largely eliminates amorphous defects, dangling bonds, and two-level systems (0802.0465).

For large-area substrate-transferred coatings relevant to interferometers, mechanical ringdown studies decomposed the elastic loss into bulk and shear contributions and found

$\lambda_0$0

after selective defect removal (Penn et al., 2018). Using finite-element energy ratios for a 35 mm cavity with 8 mm diameter crystalline coatings and 250 $\lambda_0$1m beam radius, the effective coating loss was predicted to be

$\lambda_0$2

in agreement with the direct thermal-noise measurement $\lambda_0$3 reported previously (Penn et al., 2018). The paper’s interpretation is explicit: mechanical dissipation in these room-temperature GaAs/AlGaAs coatings arises entirely from bulk loss, with shear loss consistent with zero within uncertainties (Penn et al., 2018).

The review of gravitational-wave detector prospects summarizes the system-level implication by using a coating mechanical loss angle $\lambda_0$4 in upgrade modeling and obtaining a projected strain sensitivity at 100 Hz of $\lambda_0$5, compared with $\lambda_0$6 for Advanced LIGO design (Cole et al., 2023). This is tied to the broader statement, repeated in later gravitational-wave control papers, that crystalline AlGaAs coatings can produce a factor of at least 5 improvement in coating thermal noise compared to current Advanced LIGO coatings (Tanioka et al., 2024, Cole et al., 2023, Liu et al., 23 Jun 2026). A plausible implication is that the optical advantages of AlGaAs are inseparable from their mechanical-loss advantages; in nearly every application, optical performance is pursued as a means of making that thermal-noise reduction usable at the system level.

4. Precision optical references and ultrastable cavities

Ultrastable Fabry–Pérot cavities provide the clearest system-level evidence that the low mechanical loss of AlGaAs coatings translates into improved frequency stability. In a 6 cm cryogenic silicon cavity operated at 17 K with crystalline mirrors, the modified Allan deviation reached

$\lambda_0$7

at $\lambda_0$8, four times lower than the dielectric thermal-noise limit for the same geometry (Lee et al., 16 Sep 2025). From the observed plateau and the coating Brownian-noise model, the inferred cryogenic coating mechanical loss satisfies

$\lambda_0$9

more than tenfold below the dielectric value at the same temperature (Lee et al., 16 Sep 2025). The same work reports long-term drift rates in the tens of $n_H d_H = n_L d_L = \frac{\lambda_0}{4},$0Hz/s range for multiple cryogenic silicon cavities, including AlGaAs-coated systems, indicating that crystalline coatings do not introduce problematic long-term drift (Lee et al., 16 Sep 2025).

At room temperature, a 48 cm cavity employing crystalline AlGaAs coatings reached a fractional frequency instability of $n_H d_H = n_L d_L = \frac{\lambda_0}{4},$1, one of the lowest reported for room-temperature systems, and below the Brownian thermal-noise floor that would be imposed by dielectric coatings (Ma et al., 1 Jul 2026). The cavity operated near a composite coefficient-of-thermal-expansion zero crossing at 297 K and used fused-silica mirrors with bonded GaAs/$n_H d_H = n_L d_L = \frac{\lambda_0}{4},$2 coatings and ULE compensation rings (Ma et al., 1 Jul 2026). The dominant terms in the thermal noise budget were no longer coating Brownian noise but thermo-elastic noise from the fused-silica substrates and Brownian noise from the ULE spacer (Ma et al., 1 Jul 2026). This is a strong system-level marker of success: the coating has been pushed low enough that other subsystems become limiting.

The 2024 multi-temperature cavity study broadened this picture by examining AlGaAs/GaAs optical coatings at 4 K, 16 K, 124 K, and 297 K (Ma et al., 2024). It confirmed that the resonance frequencies respond to optical power in ways that go beyond the photo-thermo-optic effect observed in dielectric coatings, and that external illumination above the GaAs bandgap produces prominent photo-modified birefringence (Ma et al., 2024). The work also showed that closed-cycle cooling can support a 124 K crystalline-mirror cavity with sub-millikelvin outer-shield stability, suggesting that practical cryogenic operation is compatible with the coating technology (Ma et al., 2024).

5. Birefringence, photo-induced effects, and excess noise

Birefringence has emerged as the most distinctive non-ideal feature of AlGaAs coatings in ultrastable optical systems. Static birefringence is now well established. A scanning study of highly reflective substrate-transferred $n_H d_H = n_L d_L = \frac{\lambda_0}{4},$3 coatings at 1064 nm found an average phase difference on reflection between fast and slow axes of

$n_H d_H = n_L d_L = \frac{\lambda_0}{4},$4

consistent with values inferred from high-finesse reference cavities (Gretarsson et al., 30 Oct 2025). Substrate-transferred samples with diameters between 18 and 194 mm showed birefringence non-uniformity at a median level of 0.1 mrad, while an otherwise similar coating left on the growth wafer showed only 0.02 mrad non-uniformity (Gretarsson et al., 30 Oct 2025). Excluding local crystal and bonding defects, the authors conclude that the non-uniformity was imparted during substrate transfer, likely during bonding (Gretarsson et al., 30 Oct 2025). They further find that such non-uniformity is unlikely to have a significant impact on scatter loss in a LIGO-like interferometer (Gretarsson et al., 30 Oct 2025).

Dynamic birefringence is more consequential. In cryogenic silicon cavities, anti-correlated frequency fluctuations between the two polarization eigenmodes were directly measured and identified as birefringent noise (Kedar et al., 2022). A 6 cm cavity at 4.7 K and 16.7 K exhibited birefringent splitting $n_H d_H = n_L d_L = \frac{\lambda_0}{4},$5 kHz, while a 21 cm cavity at 124 K showed $n_H d_H = n_L d_L = \frac{\lambda_0}{4},$6 kHz (Kedar et al., 2022). By simultaneously probing both polarization modes, the authors could cancel this birefringent noise, but the residual noise remained above the expected AlGaAs Brownian limit (Kedar et al., 2022). The same paper reports empirical scalings of the birefringent noise with intracavity power and mode area, while noting that the source of these novel mechanisms remains unclear (Kedar et al., 2022).

At room temperature, the 2026 cavity study identifies spontaneous fluctuations of coating birefringence as a leading contribution to frequency instability in the 1–100 s range (Ma et al., 1 Jul 2026). The 48 cm cavity exhibits a birefringent splitting of 104 kHz and strong sensitivity of that splitting to intracavity power: 25 Hz/$n_H d_H = n_L d_L = \frac{\lambda_0}{4},$7W at $n_H d_H = n_L d_L = \frac{\lambda_0}{4},$8W and 10 Hz/$n_H d_H = n_L d_L = \frac{\lambda_0}{4},$9W at $\mathrm{Al_{0.92}Ga_{0.08}As}$0W (Ma et al., 1 Jul 2026). The paper also shows that the cavity eigenfrequency can be made highly immune to power fluctuations by balancing a photo-birefringent effect against a photo-thermo-optic effect with slope about 4.5 Hz/$\mathrm{Al_{0.92}Ga_{0.08}As}$1W (Ma et al., 1 Jul 2026).

The room-temperature photo-birefringence study develops a unified empirical description of these effects. For a 1542 nm cavity with 38.5 GaAs/$\mathrm{Al_{0.92}Ga_{0.08}As}$2 pairs and penetration depth $\mathrm{Al_{0.92}Ga_{0.08}As}$3, the polarization splitting is $\mathrm{Al_{0.92}Ga_{0.08}As}$4, corresponding to $\mathrm{Al_{0.92}Ga_{0.08}As}$5 (Ma et al., 4 Feb 2026). The authors propose a composite intensity variable

$\mathrm{Al_{0.92}Ga_{0.08}As}$6

and fit the steady-state splitting with a logarithmic law

$\mathrm{Al_{0.92}Ga_{0.08}As}$7

which they interpret as evidence for a primary two-photon process below the GaAs bandgap and a single-photon process above it (Ma et al., 4 Feb 2026). This paper also demonstrates that external LED illumination can be used to reduce noise induced by laser power fluctuations by balancing the photo-thermal-optic response of the mirrors and the photo-birefringent effect at lower intracavity power (Ma et al., 4 Feb 2026).

A related 2025 study goes further by constructing a master-equation model for birefringence under above-band-gap illumination and deriving a photo-optic transfer function with an intensity-dependent pole and DC gain (Wu et al., 29 Nov 2025). It predicts that generation-recombination noise in the coating birefringence should be white below the pole frequency, scale with power the same way laser shot noise does, and be independent of spot size at fixed power (Wu et al., 29 Nov 2025). This suggests that, even if current cavity experiments cannot yet determine the exact GR-noise level, semiconductor carrier dynamics may impose a qualitatively new noise floor distinct from Brownian and thermo-optic noise.

6. Gravitational-wave detectors: thermal-noise gains and wavelength-driven control redesign

The gravitational-wave community’s interest in AlGaAs coatings is explicitly tied to coating thermal noise in the most sensitive band of present interferometers. The review of state of the art states that AlGaAs-based crystalline coatings present the possibility of gravitational-wave observatories having significantly greater range than current systems employing ion-beam-sputtered mirrors, and that given the low thermal noise of AlGaAs at room temperature, these gains may be realized while potentially avoiding cryogenic operation (Cole et al., 2023). Using a coating loss angle $\mathrm{Al_{0.92}Ga_{0.08}As}$8, room-temperature upgrade modeling yields a binary neutron star range of about 600 Mpc, compared with about 175 Mpc for Advanced LIGO design, corresponding to an estimated factor of 40 increase in detection rates (Cole et al., 2023). Later ALS papers restate the system-level claim in slightly different terms: AlGaAs coatings can provide a factor of at least 5 improvement in coating thermal noise compared to current Advanced LIGO coatings (Tanioka et al., 2024, Liu et al., 23 Jun 2026).

Yet the same semiconductor bandgap that makes visible-light absorption severe also forces a redesign of lock acquisition and arm-length stabilization. Current terrestrial detectors employ 532 nm auxiliary lasers generated via second-harmonic generation for arm-length stabilization, but 532 nm is strongly absorbed by AlGaAs/GaAs because it lies well above the GaAs bandgap (Tanioka et al., 2024, Liu et al., 23 Jun 2026). One proposed solution is frequency downconversion to 2128 nm via an optical parametric oscillator; a 2024 tabletop demonstration showed stable arm-length stabilization using 1064 nm science light and 2128 nm auxiliary light, with cavity displacement RMS about 0.7 nm at 0.1 Hz, below the cavity linewidth and therefore sufficient for lock acquisition (Tanioka et al., 2024). The 2026 follow-on paper instead demonstrated a multi-wavelength scheme using 1064 nm, 1596 nm, and off-cavity 532 nm for phase referencing, with the 1596 nm auxiliary beam chosen specifically because it lies outside the absorption bands of AlGaAs/GaAs (Liu et al., 23 Jun 2026).

The 2026 demonstration derives the co-resonance condition between the science and auxiliary wavelengths,

$\mathrm{Al_{0.92}Ga_{0.08}As}$9

and shows that the 1064 nm beam can be brought into resonance by tuning $d_\text{coat} = 12.0~\mu\text{m}$0 by at most one free spectral range while the cavity remains locked to the 1596 nm auxiliary field (Liu et al., 23 Jun 2026). In a 400 mm cavity, the two-color system exhibited linewidths $d_\text{coat} = 12.0~\mu\text{m}$1 and $d_\text{coat} = 12.0~\mu\text{m}$2, with corresponding finesses $d_\text{coat} = 12.0~\mu\text{m}$3 and $d_\text{coat} = 12.0~\mu\text{m}$4 (Liu et al., 23 Jun 2026). The 532 nm beat note linewidth was reduced from about 145 kHz to about 1.4 Hz under PLL control, corresponding to an inferred 1596 nm frequency noise $d_\text{coat} = 12.0~\mu\text{m}$5 Hz and an equivalent 4 km cavity length noise of order $d_\text{coat} = 12.0~\mu\text{m}$6 pm, much smaller than the $d_\text{coat} = 12.0~\mu\text{m}$7 nm arm cavity linewidth (Liu et al., 23 Jun 2026). These demonstrations do not measure coating parameters directly, but they show how control architecture must be redesigned around the semiconductor optical properties of the coating.

7. Limitations, controversies, and open research directions

A recurrent theme in the literature is that AlGaAs coatings solve some of the dominant problems of amorphous mirrors while introducing new system-specific challenges. The most mature concern is excess birefringent noise. Cryogenic cavity studies observed residual noise larger than both cavities’ thermal-noise limits even after birefringent noise cancellation (Kedar et al., 2022). The 2024 multi-temperature study likewise reports excess noise beyond the photo-thermo-optic effect and indicates that photo-modified birefringence is strongly dependent on the intensity of intracavity light at 1.5 $d_\text{coat} = 12.0~\mu\text{m}$8m (Ma et al., 2024). The 2025 cryogenic silicon cavity study reports that earlier generations of crystalline mirrors showed excess global noise with correlation length across the mm-scale beam area, and that the new mirror pair strongly reduced or eliminated this excess sufficiently to reveal the Brownian floor at 17 K (Lee et al., 16 Sep 2025). This suggests that not all excess noise mechanisms are intrinsic to the material; some may depend sensitively on growth, bonding, or process uniformity.

Another recurring issue is bonding and process control. Mechanical ringdown measurements indicate that visible bonding defects contribute only about 5% of the bulk loss in large-area room-temperature samples (Penn et al., 2018), and birefringence mapping indicates that substrate transfer mainly affects non-uniformity rather than average birefringence (Gretarsson et al., 30 Oct 2025). At the same time, several papers imply that bond interfaces, residual stress, or strain relaxation may underlie both static birefringence and illumination-sensitive noise (Gretarsson et al., 30 Oct 2025, Ma et al., 2024, Wu et al., 29 Nov 2025). A plausible implication is that the dominant remaining uncertainties are increasingly in interface and process physics rather than in the optical design of the Bragg stack itself.

Optical absorption at future detector wavelengths remains another threshold issue. The 1.5 $d_\text{coat} = 12.0~\mu\text{m}$9m mapping study found a high-reflectivity-equivalent absorption of $\mathrm{Al_{0.92}Ga_{0.08}As/GaAs}$0 ppm, which the authors describe as very promising but still above the $\mathrm{Al_{0.92}Ga_{0.08}As/GaAs}$1 ppm requirement envisioned for future gravitational-wave detectors (Steinlechner et al., 2014). By contrast, ultrastable-cavity work at 1542 nm reports scattering and absorption of about 5 ppm per mirror in a $\mathrm{Al_{0.92}Ga_{0.08}As/GaAs}$2 cryogenic silicon cavity (Lee et al., 16 Sep 2025). These results are compatible, but they also show that “low absorption” is application-specific: the acceptable optical-loss budget differs sharply between compact reference cavities and megawatt-scale arm cavities.

Several directions for further work are explicit across the literature. Large-area scaling beyond 10 cm and toward 20 cm or more is a major objective for gravitational-wave detectors (Cole et al., 2023). More detailed characterization of bond interfaces, stress management, and process uniformity is repeatedly identified as necessary (Cole et al., 2023, Lee et al., 16 Sep 2025). Dual-polarization interrogation, power-noise balancing using auxiliary illumination, and improved understanding of electro-optic and generation-recombination noise represent a second front, aimed at restoring Brownian-noise-limited performance in systems that are already optically and mechanically competitive (Ma et al., 4 Feb 2026, Wu et al., 29 Nov 2025, Ma et al., 1 Jul 2026, Tanioka et al., 2022). Taken together, these studies indicate that AlGaAs crystalline coatings are no longer limited by proof of principle; the main remaining questions concern manufacturing scale, process reproducibility, and the control of semiconductor-specific noise channels.