{"id":"40298888-938f-4eb4-b83f-850208da4e8e","arxiv_id":"2509.18834","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"An on-demand microwave-to-optical transducer based on a cold Rydberg ensemble stores and retrieves weak microwave pulses with area-normalized efficiency above 90 percent.","lead":"Scientists stored weak microwave pulses in a cloud of cold rubidium atoms and later turned them into optical photons, combining a quantum memory with a microwave-to-optical converter in one device. A general reader might care because such memory-capable conversion is a candidate building block for a quantum internet that links superconducting processors by optical fiber.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"ASE normalizes by a 66-µm sub-wavelength aperture for a 7.9-mm field; the omitted free-space mode-area overlap makes \"90% efficiency\" an aperture-bookkeeping result rather than a demonstrated end-to-end transduction efficiency.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the headline efficiency is expressed in an area-normalized metric that omits mode matching for realistic free-space microwave photons. I agree. The experiment is a credible proof of principle: the Maxwell-Bloch model reproduces the main trends, the storage-time decay and noise measurements are internally consistent, and the g(2) data support the thermal-noise interpretation. However, the strongest claim—'area-normalized storage efficiency greater than 90%'—is directly tied to a sub-wavelength normalization area. The paper acknowledges significant mode-matching losses but never quantifies them, and Supplementary SVI only equates two area-normalized definitions, not the coupling from a real source mode. This does not invalidate the demonstrated physics, but it means the device has not been shown to convert a single microwave photon from a realistic quantum node with 90% probability; the end-to-end efficiency could be orders of magnitude lower. Other potential concerns, such as partial circularity in fitting η0 and γ0 and then comparing with theory, are secondary and would not change the conditional verdict. No verdict change is needed beyond the reader's already-conditional assessment, but the manuscript should report either the absolute efficiency or the mode-overlap factor before the quantum-transducer framing can be accepted as stated.","tokens_in":19804,"tokens_out":7620,"duration_ms":66807,"concrete_test":"Perform a direct absolute-efficiency measurement: launch a pulse of known total photon number from the calibrated 20-dB horn at 30 cm (or from a characterized single-mode MW source), measure total retrieved optical counts N_L, and report η_abs = N_L/(P τ/ℏω_M) without any area normalization. In parallel, compute the mode-overlap integral between the incident free-space mode and the collective spin-wave mode (transverse extent 66 µm, length 20 mm). If η_abs ≈ η_ASE × S_M/A_mode, with A_mode the mode area at the ensemble, the missing factor is quantified. A second, cheaper check: recompute η_ASE using the full 4-mm × 4-mm atomic cloud cross-section or the horn footprint as S_M; if η drops by orders of magnitude, the headline value is normalization-dependent.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on the area-normalized storage efficiency η = N_L/N_M with N_M = I_M S_M/(ℏω_M) (Eq. 5). S_M is the transverse cross-section of the interaction volume, mean radius 66 µm, while the microwave wavelength is 7.9 mm. Thus S_M ≈ 1.4×10⁻⁸ m² is roughly 4×10³ times smaller than λ² ≈ 6×10⁻⁵ m². A freely propagating 37.5 GHz mode cannot be confined to this area over the 20-mm ensemble: the Rayleigh range of a 66-µm waist at 7.9 mm is πw₀²/λ ≈ 1.7 µm, far shorter than L. The quantity N_M is therefore the flux through a chosen bookkeeping aperture, not the number of photons in the mode that actually addresses the ensemble. For a plane-wave input, η_ASE scales inversely with the chosen S_M, so changing the normalization area changes the headline efficiency. The manuscript itself states that the small MW-reception solid angle 'introduces significant mode-matching losses for free-space signal photons' (Methods, ASE paragraph) but does not quantify the overlap between the incident MW mode and the stored spin-wave mode. Supplementary Sec. SVI only establishes equivalence between two area-normalized definitions (S_I = S_M); it never computes the overlap with a real source mode. Consequently, the 90% figure is not the probability that an incoming microwave photon in a waveguide or free-space single-mode beam is converted to an optical photon; it is an area-normalized cross-section. Because the quantum-network application requires mode-matched end-to-end conversion, this unquantified factor is the load-bearing gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a Rydberg-ensemble device that combines a quantum memory with microwave-to-optical transduction: a 37.5 GHz microwave pulse is stored as a collective Rydberg excitation via cascaded EIT and, after a programmable delay, retrieved as an optical photon. In a cold 87Rb ensemble the authors measure an area-normalized storage efficiency (ASE) greater than 90% in the single-photon regime, a bandwidth of 2.1 MHz, a 50%-efficiency storage time of about 0.56 microsecond, and a noise-equivalent temperature of 26 K under cavity-free conditions. The paper also presents an EIT-based Maxwell-Bloch model, a mean-field model of Rydberg dephasing, and a thermal-background-noise analysis. The central experimental result is the ASE itself, while the mode-matching interpretation of that ASE is the main point of concern.","tokens_in":20200,"tokens_out":12169,"duration_ms":104847,"significance":"If the ASE metric is accepted as the relevant figure of merit, this is an important proof-of-principle: an on-demand microwave-to-optical transducer with integrated memory, in a cavity-free Rydberg ensemble, with single-photon-level input. The experimental strengths are real: the ASE is obtained by direct photon counting, the g(2) data are compared with a no-free-parameter theory, the noise budget is decomposed into thermal, stray, and dark contributions, and the observed thermal noise count is reproduced by a thermal-radiation model. The limitations concern the mode-matching content of the headline efficiency: ASE is an area-normalized local conversion efficiency, not a mode-matched end-to-end transduction efficiency, and the manuscript does not quantify the mode overlap that would be needed for the proposed quantum-repeater application.","major_comments":[{"comment":"The central metric in Eq. (5) is not an end-to-end conversion efficiency. N_M = I_M S_M/(hbar omega_M) counts flux through the chosen aperture S_M ~ pi(66 um)^2, while the microwave wavelength is 7.9 mm. S_M is roughly 4x10^3 times smaller than lambda_M^2, and the Rayleigh range of a 66-um waist at 37.5 GHz is only ~1.7 um, far shorter than the 20-mm ensemble. For a plane-wave input, eta_ASE scales with the bookkeeping area S_M, so the 90% figure is an aperture-dependent local efficiency, not the probability that an incident photon in a realistic source mode is converted. The paper itself concedes that the small MW-reception solid angle 'introduces significant mode-matching losses for free-space signal photons' but does not quantify those losses. Supplementary Sec. SVI, Eq. (S45), only proves equivalence to another area-normalized intensity ratio under S_I = S_M; it does not compute the","section":"Methods — 'Area-normalized storage efficiency (ASE)', Eq. (5), and Supplementary Sec. SVI"},{"comment":"The agreement with theory is used as a validation, but eta0 and gamma0 are both free parameters fitted from the same ASE-versus-N data. With no reported uncertainties, the statement that the fitted gamma0/2pi = 12.8 kHz is 'close to' the mean-field value 10.8 kHz is not an independent test of the model. In addition, the main text uses gamma51/2pi = 10.8 kHz to obtain eta ~ 0.92, whereas Supplementary Sec. SI uses gamma51/2pi = 12.8 kHz to obtain eta_t ~ 0.9; the relation between these two theoretical estimates should be clarified. Please report parameter uncertainties, show the theoretical curve with an uncertainty band, and, if possible, test the predicted gamma51 proportional to sqrt(N) scaling directly. This is needed to support the 'minimal single-photon-level dephasing' claim, although the direct ASE measurement at N = 0.1 is not affected.","section":"Fig. 3c and Eq. (4)"},{"comment":"The thermal-noise agreement is presented as 'strong independent evidence' that ASE is a reliable metric, but Eq. (S35) takes eta_max equal to eta0 from the ASE analysis. The match to 0.109 noise photons per pulse is therefore a consistency check, not an independent validation. Either compute the thermal-noise contribution from independently measured parameters, with eta_max obtained from the Maxwell-Bloch model before fitting to the ASE data, or explicitly label the comparison as a consistency check.","section":"Supplementary Sec. SIV and Methods, ASE paragraph"}],"minor_comments":[{"comment":"The main-text citation 'Ref. 6' in the Methods and Supplementary sections refers to a free-space Rydberg converter, but the bibliography entry for [6] is given as 'Arquer et al., Semiconductor quantum dots'. Please correct the reference numbering.","section":"Reference list"},{"comment":"The noise-equivalent temperature T_NE = 26 K is stated without a definition or formula. Please provide the relation between the measured thermal noise count and T_NE.","section":"Methods and Results"},{"comment":"The Gaussian and exponential fits give different zero-time efficiencies (82% vs. 93%) and different 50%-efficiency storage times. Please state which fit is used for the headline 'storage time of 0.56 microseconds' and justify the choice.","section":"Fig. 2c"},{"comment":"The factors eta_s and eta_c are stated to approach unity but are not derived. A short derivation or quantitative values for the present parameters would strengthen the theoretical connection between Eq. (2) and the full Maxwell-Bloch model.","section":"Eq. (2) and preceding text"},{"comment":"The label 'parameter-free' for the mean-field dephasing prediction is overstated: the calculation uses a short-range cutoff at the blockade radius R_B, a measured pumping linewidth to set R_B, and a mean vdW coefficient C_35. Please qualify this language.","section":"Supplementary Sec. SIII"}],"recommendation":"major_revision","confidential_remarks":"The experimental result is promising, but the paper's headline efficiency is expressed in an area-normalized metric whose mode-matching content is not quantified. This is fixable by adding a mode-overlap analysis and reframing the claims, so I recommend major revision rather than rejection. The citation numbering error for Ref. 6 should be corrected. I did not see any indication of unsound data or misconduct."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper demonstrates a genuinely new capability: a cold Rydberg ensemble stores a weak microwave pulse and retrieves it as an optical photon on demand, combining memory and conversion in one device. The cascaded write-read scheme with temporal isolation is a real step beyond the earlier direct six-wave-mixing transducers and separate atomic memories. The data look credible, and the Maxwell-Bloch model reproduces the main trends, including the N-dependent dephasing and the thermal-noise counts. The noise characterization—HBT bunching, per-pulse noise budget, and the 26 K noise-equivalent temperature—is careful and reproducible.\n\nThat being said, the headline numbers are shakier than the abstract implies. The '>90% area-normalized storage efficiency' is normalized to a receiving area with a mean radius of about 66 µm, while the microwave wavelength is 7.9 mm. That is a sub-wavelength aperture, and the efficiency scales inversely with the chosen normalization area. As the stress-test correctly notes, the Rayleigh range of a 66 µm waist at 7.9 mm is ~1.7 µm, far shorter than the 20 mm ensemble, so the 90% is a cross-section for a bookkeeping aperture, not the probability that an incoming microwave photon in a well-defined mode gets converted. The paper itself admits 'significant mode-matching losses for free-space signal photons' in the Methods, but never quantifies the overlap. For the quantum-repeater application, this is the load-bearing gap.\n\nThere is also some circularity in the validation: they fit η0 and γ0 from the ASE-vs-N data and then take the closeness of γ0 to the mean-field prediction as confirmation. And in the thermal-noise model in SI Section IV, they use the fitted peak efficiency to predict the thermal count they then call agreement. The central ASE measurement is independent of the model, so this is not fatal, but it's worth flagging.\n\nMinor but annoying: the reference list is scrambled—ref 6 in the text for direct microwave-optical conversion points to a semiconductor quantum dots paper—and there are small inconsistencies in the storage-time values (0.56 vs 0.58 µs depending on fit). These should be cleaned up.\n\nOverall, this is a solid proof-of-principle with a genuine first. The efficiency claim needs to be re-framed as a cross-section, and the authors should be pushed to measure or bound the mode-matched end-to-end conversion efficiency. That does not diminish what they actually built. I'd send it to a good referee and ask for that clarification; the core result deserves to be in the literature.\n\nRecommendation: accept for peer review, with requests for quantified mode-matching and citation fixes.","headline":"Genuine first demonstration of on-demand microwave-to-optical transduction with an integrated Rydberg memory, but the headline '>90% efficiency' is an area-normalized cross-section, not a mode-matched end-to-end conversion efficiency.","tokens_in":20763,"tokens_out":2782,"would_cite":true,"duration_ms":20579,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["42.50.Gy","32.80.Ee","03.67.Hk"],"model":"deepseek-v4-flash","headline":"A single Rydberg ensemble can store a microwave pulse and later emit it as an optical photon on demand, with over 90% area-normalized storage efficiency and a noise-equivalent temperature of 26 K.","keywords":["microwave-optical transduction","quantum memory","Rydberg ensemble","electromagnetically induced transparency","area-normalized storage efficiency","on-demand photon storage","noise-equivalent temperature","quantum repeater"],"falsifier":"Measure the total converted optical photons per total input microwave photons for a defined free-space mode matched to the antenna and the full atomic ensemble (for example, a plane-wave mode covering the 4×4×20 mm cloud), rather than normalizing to the 66-µm-radius overlap area. If the mode-matched end-to-end efficiency is substantially below the reported ~90%, the headline area-normalized efficiency does not represent the conversion efficiency a real quantum link would experience.","tokens_in":19660,"feed_emoji":"📡","tokens_out":6820,"duration_ms":57953,"temperature":0.7,"pith_summary":"This paper proposes and demonstrates a single device that both stores a weak microwave pulse and converts it to an optical photon on demand, combining two functions needed for quantum repeaters. Using a cold Rydberg ensemble and cascaded electromagnetically induced transparency, the authors store microwave photons as collective Rydberg excitations and later read them out as 780-nm photons. They report over 90% area-normalized storage efficiency, 2.1 MHz bandwidth, 0.56 microseconds of storage at the 50% no-cloning threshold, and a noise-equivalent temperature of 26 K without an optical cavity. The high efficiency is credited to the very large optical depth that microwaves experience on Rydberg transitions, combined with storage control strategies adapted from optimal optical quantum memory. If correct, this integrates memory and transduction in one cryogenically compatible interface for distributing entanglement between solid-state quantum nodes.","feed_headline":"On-demand microwave-to-optical storage tops 90 percent","feed_subtitle":"A cold atomic cloud stores a microwave pulse and releases it as light on demand, bringing quantum repeaters a step closer.","key_machinery":"The carrying mechanism is cascaded electromagnetically induced transparency in a five-level Rydberg system: a write field maps a microwave pulse into a long-lived collective Rydberg spin wave, and a read field retrieves it as an optical photon. The key identity is the area-normalized storage efficiency η ≈ η0 exp(−2γ51 td), with η0 = 1/sqrt((1+αM/dM)(1+αL/dL)), where dM ~ 7.5×10^5 is the microwave optical depth and γ51 = sqrt(N̄)γ0 describes Rydberg dephasing from van der Waals interactions in a mean-field model. A receiving area SM of mean radius 66 µm sets the normalization in Eq. (5), and the paper argues this is equivalent to a flux-density efficiency when SM equals the effective beam ar","core_discovery":"The central claim is that an EIT-based Rydberg ensemble can perform on-demand microwave-to-optical transduction with integrated quantum memory, reaching an area-normalized storage efficiency above 90% at the single-photon level. In the experiment, a 37.5-GHz microwave pulse is directed onto a cigar-shaped 87Rb cloud, slowed by cascaded EIT, and mapped into a collective spin-wave excitation in a Rydberg state when the write field is switched off; after a controllable delay, a read field converts the excitation into an optical photon at 780 nm. The efficiency is governed by an expression of the form η ≃ η0 exp(−2γ51 td), with η0 near unity because the microwave transition has an optical depth","pith_inferences":["The reported >90% efficiency is area-normalized to a 66-µm-radius receiving region, far smaller than the 7.9-mm microwave wavelength; a fair comparison with cavity-based transducers would require a mode-matched end-to-end measurement that includes the antenna-to-ensemble coupling loss, which the paper notes but does not quantify.","The measured sqrt(N̄) scaling of dephasing implies the transducer is most efficient for true single photons; injecting multi-photon coherent pulses degrades efficiency, so the device naturally favors single-photon quantum applications over classical microwave detection.","The same storage-retrieval architecture could be recast as a temporal-mode converter: shaping the read field should allow the retrieved optical pulse to be produced with an arbitrary temporal envelope, adding flexibility for quantum-network synchronization beyond what the paper explicitly tests.","Because only paraxial thermal photons within a tiny solid angle are stored efficiently, the low noise-equivalent temperature partly reflects the same geometric rejection that limits signal collection; cryogenic operation would separate these two effects."],"forward_implications":["At the demonstrated parameters, a single node can store an incoming microwave photon and release it as an optical photon after a chosen delay, providing the synchronization that Bell-state measurements between distant nodes require.","Because the microwave transition's optical depth is orders of magnitude larger than typical optical transitions, the area-normalized efficiency can approach unity without an impedance-matched cavity, unlike direct transduction schemes.","At room temperature the dominant noise is thermal microwave photons about 0.109 per pulse, but at temperatures below 4 K it drops below 10^-3 per pulse, making the device suitable for cryogenic superconducting-qubit environments.","The storage time at the no-cloning threshold (η = 50%) is about 0.56 µs, limited by Rydberg spin-wave dephasing; transferring the excitation to a hyperfine clock state is proposed as a route to second-scale storage.","The transducer operates at single-photon input levels near N̄ = 0.1 with resolvable signal, whereas direct free-space Rydberg transduction at ambient temperature typically requires much larger photon numbers."],"fun_headline_variants":["Quantum memory and transducer combined: 90% efficient","On-demand light from stored microwaves: 90% efficiency","Rydberg cloud converts microwaves to optical on demand","Atomic ensemble remembers microwave, releases photon","Microwave-to-optical on-demand with 90% storage"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the area-normalized storage efficiency, defined against a 66-µm-radius receiving area for a 7.9-mm-wavelength microwave field, is a meaningful measure of transduction efficiency; the paper itself acknowledges that this small receiving solid angle introduces significant mode-matching losses for free-space signal photons, but it does not quantify those losses.","fun_headline_variants_meta":{"raw":{"variants":["Quantum memory and transducer combined: 90% efficient","On-demand light from stored microwaves: 90% efficiency","Rydberg cloud converts microwaves to optical on demand","Atomic ensemble remembers microwave, releases photon","Microwave-to-optical on-demand with 90% storage"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000245,"raw_usage":{"total_tokens":1346,"prompt_tokens":690,"completion_tokens":656,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":434,"completion_tokens_details":{"reasoning_tokens":578}},"tokens_in":434,"tokens_out":656,"duration_ms":6018,"temperature":1.0,"reasoning_tokens":578,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T15:40:55.809968+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the total converted optical photons per total input microwave photons for a defined free-space mode matched to the antenna and the full atomic ensemble (for example, a plane-wave mode covering the 4×4×20 mm cloud), rather than normalizing to the 66-µm-radius overlap area. If the mode-matched end-to-end efficiency is substantially below the reported ~90%, the headline area-normalized efficiency does not represent the conversion efficiency a real quantum link would experience.","supporting_citations":[],"review_version":1}