{"id":"ecd77f57-3778-4fa4-8e21-a00c96a0c3cc","arxiv_id":"2501.01301","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A silicon photonic chip with four integrated photon-pair sources executed a variational quantum eigensolver for the Hydrogen molecule and a variational quantum factoring algorithm for 35, at room temperature.","lead":"Researchers ran two quantum algorithms on a silicon photonic chip that generates entangled photons on-chip at room temperature. The work could make compact, integrated quantum processors more practical without extreme cooling.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (6) treats every twofold coincidence as a single-pair event, but accidental and multi-pair contamination at the 2 mW operating point is never quantified; a 1–10% bias is comparable to the claimed 0.003 Ha accuracy.","rationale":"The reader's weakest assumption points to the post-selected coincidence-basis model and the low-gain expansion. I agree that this is the right region of the argument, but I would sharpen it: the more dangerous term is not source distinguishability, which is partly supported by the measured visibilities in Fig. 3(a), but the accidental/multi-pair contamination that enters every CC used in Eq. (6). The paper estimates CAR as a function of pump power in Fig. A6, yet no CAR value at the operating point is quoted, and the data-processing pipeline does not state whether accidentals were subtracted before normalization. Since the reported VQE accuracy (0.003 Ha) is only about twice the standard chemical accuracy threshold, even a few percent bias in the normalized probabilities could change the conclusion. This is not an accusation of misconduct; the text simply leaves the key correction unquantified. The paper has genuine supporting evidence: agreement with theory across atomic distances, certified entanglement dimensions, and high measured visibilities. Those facts make it plausible that the demonstration is correct, but they do not settle the accidental-subtraction question because the same coincidence pipeline is used for all the presented data. The proposed raw-data check is decisive and feasible. If it shows no significant shift, the central claim survives; otherwise the reported accuracy and the 'successful execution' claim would need to be weakened. Since this is exactly the kind of condition already reflected in a CONDITIONAL verdict, I do not move the verdict.","tokens_in":43602,"tokens_out":11686,"duration_ms":128689,"concrete_test":"Take the raw time-tagging files for the VQE data in Fig. 4(c) and recompute the cost in Eq. (6) using CC_net = CC_raw − AC, where AC is the accidental level fit from the sidebands in Fig. A6; compare the resulting minimum energy and Table 1 values with those reported. If the shift exceeds 0.003 Ha or the per-particle chemical accuracy of 1.59e-3 Ha, the claimed accuracy is not established. As a second check, report the measured CAR at 2 mW per source; if CAR < 100, the O(ξ^2) bias exceeds 1% and must be propagated through the cost function.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the Si-PIC executes VQE and VQF—rests on Eq. (6), which converts raw coincidence counts CC into probabilities by normalizing with CCtot. This is valid only if each detected twofold coincidence is a single photon pair from one source and background/multi-pair events are negligible. The paper's nonlinear characterization (Fig. A6) estimates a CAR curve but never reports the measured CAR at the 2 mW-per-source operating point, nor states whether the accidental level obtained from the AC sidebands is subtracted from the data used in Secs. 3–4 and Figs. 4–6. At 2 mW on-chip power, with 15% detector efficiency, 20 ns gates, and 80 μs deadtime, the accidental fraction is not obviously negligible. In the low-gain expansion Eq. (A13), the first omitted term is O(ξ^2); if the mean pair number per source is ~1/CAR, the O(ξ^2) contribution to normalized probabilities is ~1/CAR. For CAR = 20, that is 5%, which in the H2 cost function would shift expectation values by order 0.005 Ha—larger than the reported 0.003 Ha agreement. The VQF cost is diagonal, so a uniform accidental floor partly cancels, but a source-power-dependent floor biases the gradient; the VQE includes off-diagonal terms, so it is directly affected. Because the claimed accuracy is close to chemical accuracy (1.59e-3 Ha per particle), this unquantified contamination is the weakest load-bearing point in the demonstration.","agreement_with_reader":"partial"},"referee_report":null,"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main take: this is the first demonstration of variational quantum algorithms on a silicon photonic chip with on-chip SFWM photon-pair sources, and the central claim holds up. The H2 VQE reaches 0.003 Ha agreement with UCC theory, and the 35 factorization converges to the correct (7,5) pair. The circuit design is thoughtful: converting time-energy entanglement to path entanglement, using 16 projective measurements with two detectors, and reporting 99.3% visibility and certified dimensions up to 4. The appendices give enough phase settings and calibration detail that a careful reader could in principle reproduce the measurement settings.\n\nThe soft spots are proportionate and addressable. First, the data-availability statement says 'No data were generated or analyzed'—that is simply wrong for an experimental paper and needs to be fixed. Second, no raw or processed data are provided, which matters for a claim this granular. Third, the VQF search space is reduced post hoc to four candidates by excluding trivial and symmetric factors; that is honest proof-of-principle, but it is a much weaker demonstration than a full factorization search.\n\nThe stress-test concern about Eq. (6) is real. The cost function is computed from normalized coincidence counts, and the paper shows a CAR measurement in App. D with a fit to extract accidental coincidences, but it never states whether that accidental level was subtracted from the data used in Secs. 3–4 and Figs. 4–6. At 2 mW per source, a 5% accidental floor would shift the H2 energy by roughly 0.005 Ha, which is comparable to the claimed 0.003 Ha accuracy. This is not a fatal flaw—the low-gain expansion and post-selection are standard—but it is a load-bearing detail that the authors need to clarify. Reporting the measured CAR at the operating point and stating explicitly how accidents were handled would resolve it.\n\nThe citation pattern looks fine. The self-citations (Refs. 104, 113, 145) are background on sources and indistinguishability, not load-bearing for the new result, and the independent benchmarks (PSI4 for H2, direct evaluation for 35) are appropriate.\n\nBottom line: this deserves a serious referee. I would send it to peer review with a request for data availability, raw data or clear processed data tables, and an explicit accidental-coincidence subtraction statement. It is a solid proof-of-principle result that will be useful to the integrated-photonics VQA community.","headline":"A genuine first demonstration of VQAs on a silicon PIC with integrated SFWM sources, with a real but fixable gap around accidental-coincidence subtraction and a contradictory data-availability statement.","tokens_in":44464,"tokens_out":1420,"would_cite":true,"duration_ms":18405,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68","81V55"],"pacs":[],"model":"deepseek-v4-flash","headline":"A silicon photonic chip with four integrated entangled-pair sources executes two variational quantum algorithms at room temperature, reaching chemical accuracy for the Hydrogen molecule and recovering the factors of 35.","keywords":["variational quantum eigensolver","variational quantum factoring","silicon photonics","integrated photon-pair sources","spontaneous four-wave mixing","path-entangled ququarts","room-temperature quantum processor","four-qubit photonic circuit"],"falsifier":"Repeat the Hydrogen VQE at increasing on-chip pump powers per source while keeping all phase settings fixed; if the low-gain coincidence model is valid the estimated energy should remain flat within shot noise, so a systematic drift beyond statistics, or a heralded-interference visibility meaningfully below the reported 99.3%, would falsify the central claim.","tokens_in":43424,"feed_emoji":"⚛️","tokens_out":8395,"duration_ms":78910,"temperature":0.7,"pith_summary":"This paper claims that a single silicon photonic integrated circuit can run the full hybrid quantum-classical loop of a variational quantum algorithm at room temperature using only on-chip entangled photon sources. The authors execute two proof-of-principle tasks on the same four-qubit device: a variational quantum eigensolver for the Hydrogen molecule and a variational quantum factorization of the semiprime 35. The chip prepares a two-ququart path-entangled state from four spontaneous-four-wave-mixing spiral-waveguide sources, and normalized coincidence counts directly supply the probabilities in the cost function. The measured Hydrogen ground-state energy agrees with theory to about 0.003 Ha, and the factorization converges to the correct factors (7,5). If the claim holds, it shows that integrated photon-pair sources are sufficient for small-scale variational quantum processors, removing the need for external sources or cryogenic hardware in this class of demonstrations.","feed_headline":"Photonic chip runs variational quantum algorithms at room temperature","feed_subtitle":"A four-qubit silicon photonic circuit with on-chip pair sources solves H2 and factors 35.","key_machinery":"The load-bearing object is the two-ququart path-entangled state $|\\psi^{(4)}_{\\mathrm{III}}\\rangle = \\sum_{m=1}^{4} \\alpha_m |m\\rangle_i |m\\rangle_s$, where each four-level ququart is encoded in the spatial mode of one photon and therefore carries two qubits. The amplitudes $\\alpha_m$ are set by the pump-splitting MZI network and serve as the variational parameters. The final MZI networks implement projective measurements onto $|2\\rangle_i|2\\rangle_s$, and the cost function is estimated by normalized coincidence counts, $C(\\boldsymbol\\alpha) = \\sum_k w_k \\sum_{m_1,m_2} \\pi[k,I,m_1,m_2]\\, \\mathrm{CC}[\\boldsymbol\\alpha,k,I,m_1,m_2]/\\mathrm{CC}_{\\mathrm{tot}}[\\boldsymbol\\alpha,k,I]$. Using the energy-time entanglement of the spontaneous-four-wave-mixing pairs, converted to spatial correlation by routing and asymmetric MZIs, this scheme prepares correlated trial states without probabilistic CNOT gates.","core_discovery":"The central claim is that a reconfigurable silicon photonic circuit with four integrated photon-pair sources can prepare and measure the trial states needed for two distinct variational algorithms, and that the measured outcomes match theory. For the Hydrogen molecule, the UCC trial state $\\cos\\theta|1010\\rangle - \\sin\\theta|0101\\rangle$ is prepared by pumping two sources with a controlled amplitude ratio, and the expectation value of the electronic Hamiltonian is reconstructed from projected coincidence counts in two commuting groups of observables. The experiment finds the equilibrium ground-state energy $-1.1340 \\pm 0.0124$ Ha at $\\theta = 0.11 \\pm 0.01$, compatible with the theoretical $-1.1373$ Ha, and a Bayesian optimizer reaches 0.003 Ha accuracy in 6 to 13 iterations. For factorization, the trial state is a superposition of four candidate bit strings for $N=35$, and gradient descent drives the amplitudes to $|1110\\rangle$, the encoding of (7,5). The authors state this is the first demonstration of variational quantum algorithms on a photonic quantum simulator with integrated photon-pair sources.","pith_inferences":["Editorial: The coincidence-basis model deliberately discards multi-pair events; at higher pump powers those events would either bias the estimated energy or could, if controlled, become a resource for larger entangled states.","Editorial: The same projector-and-coincidence strategy should transfer to other variational heuristics whose cost functions decompose into commuting groups of Pauli operators, as long as the trial state can be cast in the two-ququart Schmidt form.","Editorial: The paper quantifies source indistinguishability through visibility (99.3% for one pair) but does not propagate that visibility through to the energy uncertainty; doing so would give a direct error budget for the cost-function estimate.","Editorial: Because the two ququarts are entangled photon pairs, the scheme naturally produces the correlations that gate-based photonic processors obtain with probabilistic CNOT gates; connecting this preparation block to fusion-based or modular architectures is a plausible scaling path."],"forward_implications":["The same photonic processor solves both a chemistry problem and a factoring problem by changing only the classical coefficients $\\{w_k\\}$ and the measurement projectors, not the hardware.","Room-temperature operation means variational algorithms of this class do not inherently require cryogenic or vacuum environments.","With $d$ sources the detector count stays at two while the measurement settings grow as $d^2$; since up to 32 integrated spiral sources have already been reported, larger molecules and semiprimes are within reach of current technology.","The gradient-free Bayesian optimizer avoids the shot-noise-limited gradient problem observed for gradient descent on this chip, reaching 0.003 Ha accuracy in few iterations.","For the factorization instance, gradient descent on the three variational phases successfully drives the state to the correct factors, showing that a simple classical update is sufficient when the cost function has one commuting group."],"supporting_citations":[{"why":"Demonstrates the first photonic variational quantum eigensolver with an external photon-pair source, the baseline this work replaces with on-chip sources.","marker":"[22]"},{"why":"Provide the variational quantum factoring method and its cost-function construction, used for the semiprime 35 instance.","marker":"[31,32]"},{"why":"Shows variational factorization on superconducting hardware and supplies the factoring Hamiltonian form the paper adopts.","marker":"[33]"},{"why":"Very recent photonic VQF demonstration using an external pair source, serving as the direct comparator for integrated sources.","marker":"[34]"},{"why":"Establish multidimensional entangled states and ququart manipulation in large-scale silicon photonics, the platform the chip builds on.","marker":"[36,42]"},{"why":"Supplies the Hydrogen molecule Hamiltonian coefficients for each internuclear distance used in the VQE cost function.","marker":"[114]"},{"why":"Gives the theoretical Hydrogen ground-state energy used to judge the experimental agreement.","marker":"[121]"},{"why":"Reports up to 32 integrated spiral-waveguide pair sources, used to argue that larger instances are within reach.","marker":"[133]"}],"fun_headline_variants":["Integrated photonic chip runs variational quantum algorithms","Four-qubit silicon photonic chip solves H2 and factors 35","Room-temp photonic chip with on-chip pair sources runs quantum algorithms","First variational quantum algorithms on a photonic chip with integrated sources"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the four on-chip sources are sufficiently indistinguishable and the pump power low enough that normalized coincidence counts equal the ideal single-pair probabilities; if multi-pair generation or source distinguishability biases those counts, every cost function in the paper is biased.","fun_headline_variants_meta":{"raw":{"variants":["Integrated photonic chip runs variational quantum algorithms","Four-qubit silicon photonic chip solves H2 and factors 35","Room-temp photonic chip with on-chip pair sources runs quantum algorithms","First variational quantum algorithms on a photonic chip with integrated sources"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000511,"raw_usage":{"total_tokens":2457,"prompt_tokens":886,"completion_tokens":1571,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":502,"completion_tokens_details":{"reasoning_tokens":1500}},"tokens_in":502,"tokens_out":1571,"duration_ms":11743,"temperature":1.0,"reasoning_tokens":1500,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:29:45.932002+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the Hydrogen VQE at increasing on-chip pump powers per source while keeping all phase settings fixed; if the low-gain coincidence model is valid the estimated energy should remain flat within shot noise, so a systematic drift beyond statistics, or a heralded-interference visibility meaningfully below the reported 99.3%, would falsify the central claim.","supporting_citations":[{"cited_title":"PSI41.4: Open-sourcesoftwareforhigh-throughputquantum chemistry,","cited_arxiv_id":null,"evidence_quote":"Supplies the Hydrogen molecule Hamiltonian coefficients for each internuclear distance used in the VQE cost function."},{"cited_title":"Very-large-scale integrated quantum graph photonics,","cited_arxiv_id":null,"evidence_quote":"Reports up to 32 integrated spiral-waveguide pair sources, used to argue that larger instances are within reach."}],"review_version":1}