{"id":"a96b33cd-7167-41b0-b822-3b3f3b6b799e","arxiv_id":"2502.08012","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A CMOS-compatible acousto-optic modulator reaches 2.1 rad phase modulation at 730 nm and 2.31 GHz with 15 mW drive, corresponding to Vπ·L = 0.26 Vcm.","lead":"A CMOS-fabricated silicon nitride modulator uses high-frequency sound waves to phase-modulate visible light at 2.31 GHz, reaching 2.1 rad modulation depth with 15 mW of microwave power in a 2 mm device. The result matters because compact, low-power visible-light modulators are a missing building block for scalable quantum computer control systems.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"High-power-handling class is inferred from dissimilar passive SiN waveguides, not measured on this released, metallized device; the 15x/100x comparisons hang on that unverified transfer.","rationale":"The reader's weakest_assumption identifies the same load-bearing concern: the paper's advertised significance depends on a high-power-handling class that is inferred from prior SiN waveguide results rather than measured on this device. I agree with that assessment. The experimental core at low optical power appears credible: the modulation depth is extracted from multiple independent sideband ratios, the sideband powers follow the expected Bessel-function behavior, and the reported Vπ is internally consistent with the stated drive conditions. The unexplained saturation above 0.85 V and the absence of error bars are real but secondary; they bound the useful linear range and reduce precision, but they do not threaten the existence of gigahertz phase modulation at the 2.1 rad level. By contrast, if the released, metallized device cannot handle the advertised hundreds of milliwatts, then the 'highest performing resonant phase modulator with high power handling' claim and the 15x/100x comparison lose their basis, even though the low-power measurement stands. The concrete test of ramping optical power on the actual device would settle this directly. Since my concern is the same as the reader's and the reader's CONDITIONAL verdict already encodes it, no verdict change is warranted.","tokens_in":20389,"tokens_out":6165,"duration_ms":60942,"concrete_test":"Measure transmitted optical power and modulation depth versus on-chip input power at 730 nm on the same released device, ramping from about 1 mW to at least 100 mW while monitoring insertion loss and optical spectrum; include an unreleased passive SiN waveguide of identical cross-section as a control. If stable transmission and unchanged modulation depth are not maintained at or above 100 mW, the high-power-handling class and the 15x/100x comparison should be removed or reframed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central significance claim is that this device belongs to the class of GHz modulators handling hundreds of milliwatts of visible light, and that it beats watt-class commercial modulators by 15x in Vπ and 100x in microwave power. That class membership is never demonstrated. The paper cites refs [30,40] for SiN waveguides with high visible power, but those are passive, unreleased structures without the AlN piezoelectric stack and AlCu electrodes used here. Metal absorption, interface quality, and the thermal resistance of the periodic nanopillar supports can all lower the optical damage and thermal threshold relative to a bare SiN guide. No insertion loss, propagation loss, or power-dependent measurement is reported, so the advertised high-power-handling class is an extrapolation rather than a result. The 2.1 rad low-power measurement may stand, but the headline comparison against watt-class bulk LN modulators is not apples-to-apples until the device itself is shown to sustain high optical power. A second related caveat is the saturation near V≈0.85 V: the headline 2.1 rad lies at the edge of a nonlinear regime, so Vπ=0.26 Vcm should be presented as a linear-regime figure of merit, not as a demonstrated operating point for arbitrary drive levels.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a CMOS-fabricated, resonantly enhanced acousto-optic phase modulator for visible light. A 2 mm device with a SiNx waveguide, AlN piezoelectric transducer, and released microstructure produces phase modulation depth 2.1 rad at 2.31 GHz under 15 mW of applied microwave power at 730 nm. The modulation depth is extracted interferometrically from sideband power ratios, cross-checked with five ratios, and the measured sideband conversion efficiencies follow the Bessel-function dependence expected for sinusoidal phase modulation. The authors infer Vπ = 1.32 V, Vπ·L = 0.26 Vcm, and extract mechanical Q = 228 from the resonance lineshape. They claim this is the highest performing visible-light resonant phase modulator in the class that handles hundreds of milliwatts of optical power, with 15x lower Vπ and 100x lower microwave power than bulk lithium niobate modulators, and they outline a path toward integrated Raman-beam control for atomic qubits.","tokens_in":20620,"tokens_out":2716,"duration_ms":26788,"significance":"If the central measurement is taken at face value, this is a significant experimental result: a wafer-scale, CMOS-compatible visible-light phase modulator operating at gigahertz frequencies with low microwave drive power. The measurement methodology is a strength: the modulation depth is determined from multiple independent sideband ratios (Supplement S-IIIb), and the observed sideband conversion efficiency tracks the theoretical Bessel-function curves (Fig. 3b). The demonstrated 2.1 rad modulation depth and the extracted Vπ·L = 0.26 Vcm are credible low-power, linear-regime figures of merit. The paper also provides useful design simulations for photoelastic, moving-boundary, and electromechanical coupling, and it identifies a plausible route to higher-frequency and broader-wavelength operation. However, the headline comparison to 'high-power-handling' modulators and the advertised application class rest on an extrapolation of power handling from other SiN waveguides, not on any measurement of this device's optical power handling, thermal limits, or insertion loss. That comparison needs to be either demonstrated or substantially reframed.","major_comments":[{"comment":"The claim that this device belongs to the class of 'gigahertz-frequency modulators that can handle hundreds of milliwatts of visible-light optical power,' and the associated 15x/100x comparisons, are not supported by measurements on this device. The text cites refs [30,40] for SiNx waveguides operating at hundreds of milliwatts, but those are passive, unreleased waveguides without the AlN/AlCu piezoelectric stack, the released membrane, and the periodically patterned nanopillar supports used here. Metal-induced absorption, interface quality, and the altered thermal path through the nanopillars can all change the optical power-handling threshold. No insertion loss, propagation loss, or power-dependent measurement is reported. The 2.1 rad measurement at low optical power may stand, but the advertised class membership and the comparison to watt-class bulk modulators are not demonstrated. Please either measure power handling (e.g., transmitted power versus incident power, thermal rollover, or damage threshold) on this device, or reframe the claims as 'SiN-based devices are expected to be compatible with high power' and remove the class-based comparison and the 15x/100x headline numbers.","section":"§IV and Abstract"},{"comment":"The headline figure of merit Vπ·L = 0.26 Vcm is extracted from the linear portion of the drive-voltage curve, but the headline 2.1 rad at 15 mW lies in the nonlinear/saturated regime above V ≈ 0.85 V. The text acknowledges saturation and 'increased response' near the top of Fig. 3c, and Fig. 3b shows deviation from Bessel-function behavior above 8 dBm. This means Vπ·L = 0.26 Vcm is a small-signal, linear-regime figure of merit, and it should be presented as such; it should not be used to imply that the device can be driven to 2.1 rad by extrapolating the small-signal Vπ. The paper should state explicitly that the 2.1 rad operating point is in a regime with thermal/mechanical nonlinearities, and that the 0.26 Vcm value describes the linear response only.","section":"§III, Fig. 3c"},{"comment":"The paper does not report the on-chip optical insertion loss or propagation loss of the modulator. For a device advertised as 'efficient' and as a building block for quantum control, the optical loss is a central performance parameter: it determines how much of the input power actually reaches the output and how much is dissipated in the device, which is directly relevant to the power-handling claim. Please report the grating-coupled insertion loss, the waveguide propagation loss, and, if possible, the on-chip loss of the released modulator structure. Without this, the reader cannot assess the practical efficiency of the device or the thermal load that would accompany high optical power.","section":"Experimental Results and Fig. 2a/Fig. 3a"}],"minor_comments":[{"comment":"Fig. 2a is described in the text as measured at a constant microwave power of 15 mW, while Fig. 3a states 12 dBm (≈15.8 mW). Please make the drive powers consistent across figures and text, or explain the difference explicitly.","section":"Fig. 2a vs. Fig. 3a"},{"comment":"Equation (S11) writes 1/Δϵ_i = Σ p_ij S_j, which mixes the permittivity perturbation and the photoelastic tensor in a nonstandard way. The standard form is Δ(1/ε)_i = Σ p_ij S_j, i.e., a perturbation to the inverse permittivity. Please clarify the notation so the sign and magnitude conventions are unambiguous.","section":"Supplement S-II, Eq. (S11)"},{"comment":"The labels 'AOFS' and '/uni0394GND' in Fig. 2b-c are cryptic and appear to be leftovers from a drawing program. Please replace them with clear labels or remove them.","section":"Fig. 2b-c"},{"comment":"In the phase-matching discussion, the estimate Δn ≈ 2 is described as conservative, but the actual refractive-index difference between the optical mode and the microwave mode for this device is not quantified. Please give the simulated or estimated Δn and the resulting phase-matching length for the specific device, so the reader can judge the validity of the sinc-factor approximation for L = 2 mm.","section":"Supplement S-I"}],"recommendation":"major_revision","confidential_remarks":"The core low-power measurement is credible and the sideband-ratio cross-checks are a real strength. The paper's central claim, however, is framed around a 'high-power-handling' class that the device itself is never shown to belong to. That is a load-bearing issue for the headline comparison, but it is fixable by either adding a power-handling measurement or rewriting the claims to separate the demonstrated low-power performance from the expected high-power capability. The saturation issue is also fixable by reporting Vπ as a linear-regime figure of merit. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe core result here is real and worth your time: a CMOS-fabricated, visible-light acousto-optic phase modulator that hits 2.1 rad modulation depth at 2.31 GHz with 15 mW of microwave drive, in a 2 mm device. That's a credible, well-measured number. The measurement is careful: they use an interferometric setup with an acousto-optic frequency shifter to avoid phase-noise problems, extract modulation depth from five independent sideband ratios that agree, and show the conversion efficiency tracks Bessel-function predictions across a drive-power sweep. That's strong evidence of clean sinusoidal phase modulation. The device design—an AlN piezoelectric transducer driving a breathing-mode mechanical resonance in a released SiN/SiO2 waveguide—is sensible, and the work sits in a platform that already has switches and filters, so the integration story is plausible.\n\nWhat I'd flag: the paper's headline significance rests on the device belonging to the class of modulators that handle hundreds of milliwatts of visible light, and that is inferred, not measured. The cites are to passive SiN waveguides without the AlN stack, AlCu electrodes, or nanopillar supports. Metal absorption, interface quality, and thermal resistance could all lower the damage threshold. The comparison to bulk LN modulators (15x lower Vπ, 100x less microwave power) is only apples-to-apples if the device actually sustains high optical power. That's an addressable gap: measure insertion loss, propagation loss, and power-dependent transmission at 730 nm. The saturation above 0.85 V is a real limitation—the 2.1 rad point sits at the edge of a nonlinear regime—but the authors explicitly acknowledge it and frame Vπ as a linear-regime figure. Minor: no error bars or raw data in the main text.\n\nNet: the low-power demonstration is solid and novel—first visible-light GHz AO phase modulator in a CMOS foundry process, to my knowledge. The high-power-handling claim is a reasonable extrapolation but not yet a result. A careful referee should push for that measurement or a re-framing. I'd send it out. It deserves serious review, and the central measurement will stand.","headline":"Solid measured 2.1 rad at 2.31 GHz in a CMOS visible-light AO modulator, but the headline high-power-handling class is inferred, not demonstrated.","tokens_in":21200,"tokens_out":3207,"would_cite":true,"duration_ms":27046,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A CMOS-fabricated 2 mm silicon-nitride device phase-modulates 730 nm light at 2.31 GHz to a 2.1 rad depth using 15 mW of microwave power, which the authors report as the highest-performing visible-light resonant phase modulator in the…","keywords":["acousto-optic phase modulation","visible light","silicon nitride","aluminum nitride","piezo-optomechanics","CMOS photonics","gigahertz modulation","quantum control"],"falsifier":"Launch increasing optical power at 730 nm through the released device, up to hundreds of milliwatts, while monitoring transmitted power, modulation depth, and the mechanical resonance frequency; if the output rolls over, the resonance detunes, or the device damages before reaching that power class, the high-power-handling claim is falsified.","tokens_in":20180,"feed_emoji":"💡","tokens_out":12183,"duration_ms":96414,"temperature":0.7,"pith_summary":"This paper reports a visible-light phase modulator made in a standard CMOS foundry that turns a small microwave drive into a large optical phase shift at gigahertz speed. The device confines a 730 nm optical mode and a mechanically resonating \"breathing\" mode in the same released silicon-nitride microstructure; an aluminum-nitride piezoelectric layer driven by 15 mW of microwave power strains the waveguide at 2.31 GHz, and that strain changes the optical path length through the photoelastic and moving-boundary effects. The measured modulation depth of 2.1 rad in a 2 mm device corresponds to a voltage-length product of $V_\\pi\\cdot L = 0.26$ V cm, which the authors compare favorably with watt-class bulk lithium-niobate modulators for visible-light operation. Because the structure is fabricated on a 200 mm wafer with CMOS-compatible materials, and the same platform already hosts high-speed switches and tunable filters, the paper positions this modulator as a building block for channelized, chip-scale control of the visible laser beams used in atomic and ionic qubit systems.","feed_headline":"15 mW drives 2.1 rad visible phase shift at 2.31 GHz on a CMOS chip","feed_subtitle":"At 0.26 Vcm, it beats visible-light bulk modulators 15x on voltage and 100x on microwave power.","key_machinery":"The load-bearing element is a single released microstructure that confines a fundamental transverse-electric optical mode and a high-$Q$ breathing-mode mechanical resonance in the same cross-section, with an aluminum-nitride piezoelectric film sandwiched between electrode layers. Strain from the driven mechanical resonance changes the waveguide's effective refractive index through the photoelastic effect and the moving-boundary effect, and the accumulated optical phase grows with device length $L$ while the mechanical quality factor $Q$ resonantly amplifies the strain produced by a fixed microwave drive. A Jacobi-Anger expansion, which expresses a sinusoidally phase-modulated wave as sidebands weighted by Bessel functions $J_n(\\alpha)$, connects the measured sideband power ratios back to the modulation depth $\\alpha$.","core_discovery":"The central claim is that a resonantly enhanced breathing-mode mechanical resonance, electrically excited through an embedded piezoelectric aluminum-nitride transducer, can imprint gigahertz-frequency phase modulation on visible light in an integrated silicon-nitride waveguide with $V_\\pi\\cdot L = 0.26$ V cm. At the 2.31 GHz resonance, 15 mW of applied microwave power produces a 2.1 rad modulation depth in a 2 mm long device, enough to push the first-order sideband past its theoretical maximum conversion point ($\\alpha = 1.84$ rad) and to bring the second-order sideband above the remaining carrier. The authors extract a mechanical quality factor of 228, a switching time of 31.4 ns, and close agreement with the Bessel-function sideband weights expected from sinusoidal phase modulation. They also simulate the same geometry to operate across 400-1000 nm optical wavelengths and, by width tuning, across mechanical resonances from 1 to 5 GHz, and on that basis they call it the highest-performing resonant phase modulator with high power handling at visible wavelengths.","pith_inferences":["The paper's \"high power handling\" classification is borrowed from prior silicon-nitride waveguide demonstrations rather than measured on this released, piezoelectrically actuated, metal-contacted structure; direct high-power tests at 730 nm would be needed to confirm the watt-class application target.","If the saturation observed above roughly 0.85 V of drive is thermal in origin, better heat sinking or substrate anchoring could extend modulation depth beyond 2.1 rad, whereas a mechanical nonlinearity would require a different resonator design.","The weak phononic-crystal doublets seen in the resonance spectrum are set by the periodic nanopillar supports, so choosing the support period could select a single high-$Q$ resonance and simplify the drive conditions.","The CMOS-foundry compatibility motivates scaling to many modulators on one die fed by a single laser, an integration path the paper envisions but does not yet demonstrate."],"forward_implications":["A 2.1 rad depth at 2.31 GHz puts first-order sideband conversion near its theoretical maximum of 33.9%, the point where $J_1(\\alpha)$ peaks at $\\alpha = 1.84$ rad.","The $V_\\pi\\cdot L = 0.26$ V cm result implies a half-wave voltage of about 1.32 V for a 2 mm device, with a 31.4 ns switching time set by $Q = 228$.","Because high-speed Mach-Zehnder switches and ring resonators with quality factors above 1.5 million already exist in the same piezo-optomechanical platform, this phase modulator can be combined with them to build a channelized photonic chip that amplitude-modulates, frequency-shifts, and filters visible light for individual qubit control.","Simulations reported in the paper indicate that the geometry supports optical wavelengths from 400 to 1000 nm and, by changing device width, mechanical resonances covering the 1-5 GHz range used by common atomic and ionic qubit species.","The measured on-resonance impedance of 149 $\\Omega$ corresponds to about 50% microwave power coupling, so impedance-matching networks could roughly double the modulation efficiency at the same applied power."],"supporting_citations":[{"why":"Establishes the CMOS piezo-optomechanical platform and the high-speed Mach-Zehnder switches used elsewhere in the same architecture.","marker":"[2]"},{"why":"Demonstrates silicon-nitride waveguides operating with hundreds of milliwatts at 729 nm, the basis for the high-power-handling claim.","marker":"[30]"},{"why":"Provides the commercial resonant phase-modulator baseline that the paper claims to beat by 15x in $V_\\pi$ and 100x in microwave power.","marker":"[36]"},{"why":"Supplies the watt-class bulk lithium-niobate modulator baseline used for the same efficiency comparison.","marker":"[37]"},{"why":"Extends the silicon-nitride high-power evidence to near-visible wavelengths around 780 nm.","marker":"[40]"},{"why":"Demonstrates tunable ring resonators in the same piezo-optomechanical platform, supporting the envisioned qubit control chip.","marker":"[43]"},{"why":"Provides the moving-boundary perturbation theory used to compute the optomechanical coupling integrals.","marker":"[45]"},{"why":"Defines the electromechanical coupling coefficient $k^2$ whose product with optomechanical coupling and quality factor the design optimizes.","marker":"[47]"}],"fun_headline_variants":["CMOS chip achieves 2.1 rad visible phase shift at 2.31 GHz with 15 mW","Low-power GHz visible-light phase modulator on CMOS: 0.26 Vcm","15 mW and 0.26 Vcm: CMOS visible-light phase modulator at GHz","CMOS visible-light phase modulator hits 0.26 Vcm at 2.31 GHz","Gigahertz visible-light phase modulation on CMOS: 15 mW for 2.1 rad"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The device's status as a high-power-handling modulator is inferred from prior silicon-nitride waveguide results rather than measured here; if this released, piezoelectrically actuated structure cannot actually carry hundreds of milliwatts of 730 nm light without damage or thermal detuning, the comparison to watt-class bulk modulators loses its basis even though the 2.1 rad measurement at low optical power may still stand.","fun_headline_variants_meta":{"raw":{"variants":["CMOS chip achieves 2.1 rad visible phase shift at 2.31 GHz with 15 mW","Low-power GHz visible-light phase modulator on CMOS: 0.26 Vcm","15 mW and 0.26 Vcm: CMOS visible-light phase modulator at GHz","CMOS visible-light phase modulator hits 0.26 Vcm at 2.31 GHz","Gigahertz visible-light phase modulation on CMOS: 15 mW for 2.1 rad"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001022,"raw_usage":{"total_tokens":4336,"prompt_tokens":996,"completion_tokens":3340,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":612,"completion_tokens_details":{"reasoning_tokens":3228}},"tokens_in":612,"tokens_out":3340,"duration_ms":21439,"temperature":1.0,"reasoning_tokens":3228,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T11:07:20.942366+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Launch increasing optical power at 730 nm through the released device, up to hundreds of milliwatts, while monitoring transmitted power, modulation depth, and the mechanical resonance frequency; if the output rolls over, the resonance detunes, or the device damages before reaching that power class, the high-power-handling claim is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the commercial resonant phase-modulator baseline that the paper claims to beat by 15x in $V_\\pi$ and 100x in microwave power."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the watt-class bulk lithium-niobate modulator baseline used for the same efficiency comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Extends the silicon-nitride high-power evidence to near-visible wavelengths around 780 nm."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates tunable ring resonators in the same piezo-optomechanical platform, supporting the envisioned qubit control chip."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the moving-boundary perturbation theory used to compute the optomechanical coupling integrals."}],"review_version":1}