Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Quantum mechanics can find a needle in a haystack every time

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Deterministic Grover search hits 99.77% success on a programmable photonic chip.

desk verdict A solid, same-hardware demonstration that deterministic Grover beats original Grover for N=4–10 and looks more robust to imperfections, but the headline success probability depends on efficiency corrections that aren't fully documented for the cloud processor. read the letter →

arxiv 2506.06435 v1 pith:X6I7B3AY submitted 2025-06-06 quant-ph

Fatemeh Mohit , Joshua Guanzon , Jaden McKinlay , Till J. Weinhold , Casey R. Myers , Marcelo P. Almeida , Markus Rambach , Andrew G. White This is my paper · ORCID
classification quant-ph
keywords quantumsearchGrover'salgorithmdeterministicphotonicintegratedcircuitamplitudeamplificationsingle-photonsourceprogrammablephotonicsdevicecalibration
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper reports an experimental realization of the deterministic form of Grover's algorithm on programmable photonic integrated circuits, searching databases of 4 to 10 elements with every choice of a single marked element. The deterministic algorithm, which uses two alternating diffusion axes instead of the original fixed axis, reaches an average success probability of $99.77 \pm 0.05\%$ on a fully optimized 12-mode processor. Across all three implementations (unoptimised, sequentially optimised, and clear-box optimised), the deterministic algorithm beats the original Grover's algorithm in success probability and is more tolerant of circuit imperfections. The authors argue this robustness is a reason to prefer the deterministic formulation even where raw speedup is not the main concern.

What carries the argument

The object that carries the argument is the pair of alternate diffusion axes $A$ and $B$ used in the deterministic Grover algorithm: rather than reflecting about the same axis every iteration, the algorithm alternates between two axes chosen so that the state lands exactly on the target superposition after an integer number of steps. On the experimental side, the workhorse is a programmable silicon-nitride photonic mesh of Mach-Zehnder interferometers, together with two compensation strategies: a sequential heater-by-heater optimisation that minimises the total variation distance (TVD) between expected and measured output distributions, and a clear-box model relating phase values to heater voltages through a cross-talk matrix, trained by stochastic gradient descent.

What would settle it

Re-measure the same searches using independently calibrated absolute-efficiency detectors without normalising to one reference channel; if the average success probability drops materially below $99.77\%$, the correction procedure is responsible.

Watch

Extended reading notes

Core claim

The central discovery is that the deterministic Grover's algorithm, proposed in 2022, not only achieves the theoretical promise of unit success probability but also shows markedly less degradation than the original algorithm when implemented on imperfect hardware. In the original algorithm, the diffusion operator reflects about a fixed axis, so the state after an integer number of iterations fails to align with the target; the deterministic version computes two alternative rotation axes from the constraint that the final state align with the target, and applies them alternately. The experiment programs both algorithms onto the same processor and compares them directly, finding that the deterministic algorithm outperforms the original for every database size $N=4$ through $10$ under all optimisation levels. With clear-box optimisation, the average success probability over all marked elements and database sizes is $99.77 \pm 0.05\%$, exceeding previous demonstrations of Grover's algorithm.

Load-bearing premise

The headline success probabilities assume that the per-channel detection-efficiency corrections (which vary from 1.00 to 5.53 across the eight output channels) are accurate; biased corrections would inflate the reported averages.

Editorial extensions

If this is right

  • The deterministic algorithm gives quantum search a clear experimental advantage: users can expect near-unity success probability across database sizes where the original algorithm's success probability dips.
  • Because the deterministic algorithm degrades less under imperfect beamsplitters and cross-talk, it is a better choice for near-term noisy hardware.
  • The $99.77\%$ average success probability sets a new benchmark for realised Grover's algorithm demonstrations, including earlier bulk-optics and integrated-photonics realisations of the deterministic version.
  • The clear-box optimisation technique transfers to other programmable photonic processors, making high-fidelity unitary implementation achievable without exhaustive calibration.
  • The result supports extending the deterministic algorithm to larger searches and to other qubit encodings, where its robustness may yield similar benefits.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The robustness trend likely extrapolates to larger $N$: because deterministic Grover's final state is constrained rather than merely close, small coherent errors should translate less directly into success-probability loss.
  • The efficiency-correction dependence of the headline number suggests that an independent absolute calibration of the detectors would strengthen the claim; the relative corrections vary by more than a factor of five across channels.
  • The same alternating-two-axis construction may be generalisable to amplitude amplification inside other quantum algorithms, not just standalone search, which could improve their noise tolerance.
  • One testable extension is to implement the deterministic algorithm with two- or multi-photon entangled inputs, where the search photon is entangled with other computational photons, and check whether the robustness advantage persists.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper reports an experimental implementation of the deterministic Grover algorithm of Roy et al. on programmable photonic integrated circuits, benchmarking it against the original Grover algorithm on the same hardware. The authors use an in-house 20-mode SiN processor with a sequential heater optimisation and a 12-mode Quandela cloud processor with a clear-box voltage-to-phase model. For database sizes N = 4 to 10 with a single marked element, they report that the deterministic algorithm yields higher average success probabilities than the original, with an overall average of 99.77 ± 0.05% on the cloud processor. They conclude that the deterministic algorithm is markedly more robust to circuit imperfections.

Significance. Demonstrating the Roy et al. deterministic Grover algorithm on a programmable integrated photonic platform, with every marked element tested for N = 4–10, is a useful step beyond earlier bulk-optics and fixed-integrated demonstrations. A particular strength is that both algorithms are run on the same processor under the same optimisation procedure, making the comparison direct. The near-unity success probability would be a meaningful benchmark if the calibration is reliable. However, the central numbers rest on per-channel detector-efficiency corrections that are not fully documented for the cloud processor, and the robustness claim is not defined or statistically quantified. These points need to be resolved before the headline claims can be accepted.

major comments (3)
  1. [Supplemental Material, 'Calibrating channel efficiencies' and Table II; main text Fig. 4(c)] The headline average success probability of 99.77 ± 0.05% is measured on the 12-mode Quandela cloud processor, but the calibration procedure and correction factors in Table II are reported only for the in-house 20-mode processor. Because the per-channel correction factors 1/η_tot range from 1.00 to 5.53 even for the in-house system, and because η_d is 'calculated' rather than independently verified, a systematic error in one channel's correction factor could shift the corrected success probabilities by more than the quoted ±0.05% (which appears to be a dispersion measure over database sizes and marked elements, not a systematic uncertainty). The validation that corrected single-photon data reproduce CW-laser results is not an independent check of η_d, since the CW measurements were used to set η_c and were taken with a different detector chain. Please provide the cloud processor's per-channel η_c, η_d and 1/η_tot values, the raw uncorrected counts, and a systematic uncertainty budget for all reported success probabilities.
  2. [Table III and Fig. 4; abstract] The claim that the deterministic algorithm is 'markedly more robust against technological imperfections' is not supported by a defined robustness metric. The improvements from optimisation in Table III are 4%, 10%, 9% and 9% for the original algorithm versus 1%, 1%, 4% and 6% for the deterministic algorithm for N = 5–8; only the N = 6 difference (9 percentage points) is clearly larger than the quoted ±0.03 uncertainties, and this comparison is confounded by a ceiling effect because the deterministic algorithm starts at a higher absolute success probability. Under a measured-to-theoretical success ratio, the optimised data in Table III actually favour the original algorithm at N = 6–8. Please define a robustness metric, apply it consistently, and report confidence intervals for the deterministic-versus-original differences.
  3. [Table III; 'Two trends are evident' (main text)] The statement that 'the deterministic algorithm always outperforms the original algorithm for all three implementations' is not supported by Table III: for unoptimised N = 4 both algorithms give 0.94 ± 0.03, and for sequentially optimised N = 8 both give 0.88 ± 0.03. The claim should be softened to 'never worse' or supported by a paired statistical test across all marked elements and N.
minor comments (5)
  1. [Throughout] There are numerous typographical errors, including 'sequentual optimisation', 'beampslitter', 'deteced', 'non-interger', 'sequential-otpimised' and 'Clement's decomposition' (should be Clements); please proofread carefully.
  2. [Introduction (original algorithm description)] The number of iterations is given as '(π−θ)/2π', which is dimensionally inconsistent and cannot be correct (for N = 4 it would be less than one iteration); the intended expression is presumably (π/2 − θ)/(2θ).
  3. [Fig. 4(a) caption] The caption says 'classical light ... detected via avalanche photodiodes', but the main text describes APDs only for single-photon-level light and photodiodes for the bright CW laser; please clarify which source and detector configuration produced each panel.
  4. [Eq. (2)] The symbol ⊙2 is not defined; please state that it denotes element-wise squaring of the voltage vector.
  5. [Reference [25] and data availability] The Supplemental Material link is given as 'http://tbd'; the actual URL should be provided. In addition, no raw count data or data availability statement are included, and these would be valuable given the sensitivity of the results to per-channel corrections.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the experimental claim is an implementation and measurement of an external deterministic-Grover construction, with standard device calibration applied equally to both algorithms.

full rationale

The paper's central derivation chain is: cite Roy et al. [18] for the deterministic Grover construction, program the corresponding unitaries onto two photonic processors, optimize the devices to approximate those unitaries, and then measure success probabilities for all marked elements for N=4-10. None of these steps defines a central quantity in terms of the claimed result. The deterministic algorithm's ideal 100% success probability is imported from external theory, not derived from the experiment. The TVD-based sequential heater optimisation and the clear-box voltage model are device-calibration procedures: they tune phase settings so the measured intensity distribution approaches the ideal unitary's distribution, and the same calibration is applied to both the original and deterministic algorithms, so the comparison is fair. The per-channel efficiency corrections are measured calibration constants, not fitted predictions; the validation that corrected single-photon data reproduce CW-laser data is a consistency check of the linear-efficiency model, not a circular derivation of the success probabilities. The unoptimised data (Fig. 4a) already show the deterministic algorithm degrading less than the original, so the robustness claim does not depend solely on optimised data. The wide spread of correction factors (1.00-5.53 in Supplementary Table II) raises an accuracy concern, but that concern is about systematic error in absolute success probabilities, not about a derivation that reduces to its inputs. No load-bearing self-citation is present: refs. [18] and [21] are external, and refs. [24] and [35] include some present authors but are cited only as a photon-source reference and an outlook item, not as authority for the main result. One completeness flag, not a circularity: the Supplemental Material pointer is the placeholder 'http://tbd', so the calibration tables and optimisation details for the cloud processor are not independently inspectable in the current version.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the correctness of the cited deterministic Grover's algorithm, on the standard Clements decomposition, and on the accuracy of several calibration procedures (efficiency corrections, cross-talk compensation, and the clear-box voltage-phase model). No new physical entities are introduced. The fitted calibration parameters are disclosed only in aggregate as TVD and success probabilities, not as raw datasets.

free parameters (2)
  • clear-box model matrix C2 and offset c0 = unknown
    These parameters in Eq. (2) are trained on randomly generated voltage-phase pairs via an Adam optimizer to model cross-talk and insertion loss. They are fitted to device response data and are necessary to achieve the reported near-unity success probabilities. They are calibration parameters, not physics constants, and their values are not disclosed.
  • sequential optimization increment and accuracy threshold = not specified
    The sequential heater optimization varies each phase by a small increment and repeats within a pre-specified accuracy to avoid the noise floor. These values are hand-chosen and not disclosed, affecting reproducibility.
assumptions (5)
  • domain assumption Roy et al. deterministic Grover's algorithm always finds the marked element when N/M >= 4
    The paper does not re-derive this algorithm; it is cited as reference [18] and used as the basis for the experiment. If this algorithm were incorrect, the experiment's target unitary would not correspond to deterministic search.
  • standard math The Clements decomposition can express the combined unitary UG in terms of the chip's phase shifters
    The paper uses the Clements decomposition [27] to program the unitary. This is a standard result for universal linear optics.
  • domain assumption The detector efficiency and coupling correction factors accurately compensate channel-dependent losses
    The success probabilities are computed from efficiency-corrected counts. The correction factors are measured using CW light and weak coherent states (Supplementary Table II). If these corrections are inaccurate, the quoted success probabilities could be biased.
  • domain assumption The clear-box voltage-phase model C2 V^2 + c0 from [21] accurately predicts the required voltages
    The 12-mode cloud processor is calibrated using this model trained on random voltages. The accuracy of this model limits how closely the implemented unitary matches the ideal.
  • domain assumption The quantum-dot single-photon source produces single photons with negligible multi-photon contamination
    Final data are taken with the single-photon source; weak coherent and CW results are reported as consistent, but any multi-photon events could alter the inferred success probabilities.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantum mechanics can find a needle in a haystack every time." pith.science (2026). https://pith.science/paper/X6I7B3AY

@misc{pith2026250606435,
  author       = {Pith},
  title        = {Pith review of: Quantum mechanics can find a needle in a haystack every time},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X6I7B3AY}},
  note         = {Machine review of arXiv:2506.06435}
}
abstract

Grover's algorithm is one of the pioneering demonstrations of the advantages of quantum computing over its classical counterpart, providing - at most - a quadratic speed-up over the classical solution for unstructured database search. The original formulation of Grover's algorithm is non-deterministic, finding the answer with a probability that varies with the size of the search space and the number of marked elements. A recent reformulation introduced a deterministic form of Grover's algorithm that - in principle - finds the answer with certainty. Here we realise the deterministic Grover's algorithm on a programmable photonic integrated circuit, finding that it not only outperforms the original Grover's algorithm as predicted, but is also markedly more robust against technological imperfections. We explore databases of 4 to 10 elements, with every choice of a single marked element, achieving an average success probability of $99.77 \pm 0.05\%$.

Figures

Figures reproduced from arXiv: 2506.06435 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Bloch sphere representations of: [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Schematic for in-house experimental platform. Three sources, all at [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: FIG. 5. Frequency of total variation distance (TVD) for all [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 4
Figure 4. Figure 4: FIG. 4. Assessment of original vs deterministic Grover’s al [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 7
Figure 7. Figure 7: Starting with a unitary transformation, the initial [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Logarithm heatmap of normalised transmission [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Sequential optimisation process flow. The blue box [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Exact and Fixed-Point Grover Search with Qudits

    quant-ph 2026-07 conditional novelty 4.5 of 10

    A hardware-oriented framework implements standard, deterministic, and fixed-point Grover search on homogeneous and heterogeneous qudit registers via explicit oracles, diffusion operators, and phase matching.

Reference graph

Works this paper leans on

35 extracted references · 30 canonical work pages · cited by 1 Pith paper

  1. [1]

    M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information(Cambridge University Press, 2010)

    2010

  2. [2]

    Preskill, Quantum computing and the entanglement frontier (2012), arXiv:1203.5813 [quant-ph]

    J. Preskill, Quantum computing and the entanglement frontier (2012), arXiv:1203.5813 [quant-ph]

    arXiv 2012

  3. [3]

    L. K. Grover, Phys. Rev. Lett. 79, 325 (1997)

    1997

  4. [4]

    Zalka, Phys

    C. Zalka, Phys. Rev. A 60, 2746 (1999)

    work page 1999

  5. [5]

    Wang and P

    Y. Wang and P. S. Krstic, Phys. Rev. A 102, 042609 (2020)

    work page 2020

  6. [6]

    P. G. Kwiat, J. R. Mitchell, P. D. D. Schwindt, and A. G. White, Journal of Modern Optics 47, 257 (2000)

    work page 2000

  7. [7]

    Bhattacharya, H

    N. Bhattacharya, H. B. van Linden van den Heuvell, and R. J. C. Spreeuw, Phys. Rev. Lett. 88, 137901 (2002)

    work page 2002

  8. [8]

    R. Das, T. Mahesh, and A. Kumar, Chemical Physics Letters 369, 8 (2003)

    work page 2003

Show all 35 references
  1. [9]

    Brickman, P

    K.-A. Brickman, P. C. Haljan, P. J. Lee, M. Acton, L. Deslauriers, and C. Monroe, Phys. Rev. A 72, 050306 (2005)

    2005

  2. [10]

    Hosten, M

    O. Hosten, M. T. Rakher, J. T. Barreiro, N. A. Peters, and P. G. Kwiat, Nature 439, 949 (2006)

    2006

  3. [11]

    T. W. Hijmans, T. N. Huussen, and R. J. Spreeuw, J. Opt. Soc. Am. B 24, 214 (2007)

    2007

  4. [12]

    DiCarlo, J

    L. DiCarlo, J. M. Chow, J. M. Gambetta, L. S. Bishop, B. R. Johnson, D. I. Schuster, J. Majer, A. Blais, L. Frun- zio, S. M. Girvin, and R. J. Schoelkopf, Nature 460, 240 (2009)

    2009

  5. [13]

    A. F. Benjamin Perez-Garcia, Raul I. Hernandez-Aranda and T. Konrad, Journal of Modern Optics 65, 1942 (2018)

    2018

  6. [14]

    Heurtel, A

    N. Heurtel, A. Fyrillas, G. d. Gliniasty, R. Le Bihan, S. Malherbe, M. Pailhas, E. Bertasi, B. Bourdoncle, P.- E. Emeriau, R. Mezher, L. Music, N. Belabas, B. Valiron, P. Senellart, S. Mansfield, and J. Senellart, Quantum 7, 931 (2023)

    2023

  7. [15]

    G. L. Long, Physical Review A 64, 022307 (2001)

    2001

  8. [16]

    Brassard, P

    G. Brassard, P. Hoyer, M. Mosca, and A. Tapp, Quantum Amplitude Amplification and Estimation (2002)

    2002

  9. [17]

    Høyer, Physical Review A 62, 052304 (2000)

    P. Høyer, Physical Review A 62, 052304 (2000)

    2000

  10. [18]

    T. Roy, L. Jiang, and D. I. Schuster, Physical Review Research 4, L022013 (2022)

    2022

  11. [19]

    He, W.-T

    X. He, W.-T. Zhao, W.-C. Lv, C.-H. Peng, Z. Sun, Y.-N. Sun, Q.-P. Su, and C.-P. Yang, Optics Letters 48, 4428 (2023)

    2023

  12. [20]

    Li, G.-F

    Z.-H. Li, G.-F. Yu, Y.-X. Wang, Z.-Y. Xing, L.-W. Kong, and X.-Q. Zhou, Science China Physics, Mechanics & Astronomy 66, 290311 (2023)

    2023

  13. [21]

    Fyrillas, O

    A. Fyrillas, O. Faure, N. Maring, J. Senellart, and N. Be- labas, Optica 11, 427 (2024)

    2024

  14. [22]

    Taballione, R

    C. Taballione, R. van der Meer, H. J. Snijders, P. Hooi- jschuur, J. P. Epping, M. de Goede, B. Kassenberg, P. Venderbosch, C. Toebes, H. van den Vlekkert, P. W. H. Pinkse, and J. J. Renema, Materials for Quantum Tech- nology 1, 035002 (2021)

    2021

  15. [23]

    Taballione, M

    C. Taballione, M. C. Anguita, M. de Goede, P. Vender- bosch, B. Kassenberg, H. Snijders, N. Kannan, W. L. Vleeshouwers, D. Smith, J. P. Epping, R. van der Meer, P. W. H. Pinkse, H. van den Vlekkert, and J. J. Renema, Quantum 7, 1071 (2023)

    2023

  16. [24]

    Somaschi, V

    N. Somaschi, V. Giesz, L. De Santis, J. C. Loredo, M. P. Almeida, G. Hornecker, S. L. Portalupi, T. Grange, C. Ant´ on, J. Demory, C. G´ omez, I. Sagnes, N. D. Lanzillotti-Kimura, A. Lema ´ ıtre, A. Auffeves, A. G. White, L. Lanco, and P. Senellart, Nature Photonics 10, 340 (2016)

    2016

  17. [25]

    See Supplemental Material at http://tbd

  18. [26]

    M. Reck, A. Zeilinger, H. J. Bernstein, and P. Bertani, Phys. Rev. Lett. 73, 58 (1994)

    1994

  19. [27]

    W. R. Clements, P. C. Humphreys, B. J. Metcalf, W. S. Kolthammer, and I. A. Walsmley, Optica 3, 1460 (2016)

    2016

  20. [28]

    Str¨ omberg and V

    P. Str¨ omberg and V. Blomkvist Karlsson, 4- qubit Grover’s algorithm implemented for the ib- mqx5 architecture , Ph.D. thesis, KTH (2018), https://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva- 229797

    2018

  21. [29]

    Mandviwalla, K

    A. Mandviwalla, K. Ohshiro, and B. Ji, in 2018 IEEE International Conference on Big Data (Big Data)(2018) pp. 2531–2537

    2018

  22. [30]

    Hlembotskyi, R

    V. Hlembotskyi, R. Burczy´ nski, W. Jarnicki, A. Szady, and J. Tu lowiecki, Efficient unstructured search im- plementation on current ion-trap quantum processors (2020), arXiv:2010.03841 [quant-ph]

    2020 arXiv

  23. [31]

    Gwinner, M

    J. Gwinner, M. Bria´ nski, W. Burkot, L. Czerwi´ nski, and V. Hlembotskyi, Benchmarking 16-element quantum search algorithms on superconducting quantum proces- sors (2021), arXiv:2007.06539 [quant-ph]

    2021 arXiv

  24. [32]

    Zhang, P

    K. Zhang, P. Rao, K. Yu, H. Lim, and V. Korepin, Quan- tum Information Processing 20, 233 (2021)

    2021

  25. [33]

    A. J., A. Adedoyin, J. Ambrosiano, P. Anisimov, W. Casper, G. Chennupati, C. Coffrin, H. Djidjev, D. Gunter, S. Karra, N. Lemons, S. Lin, A. Malyzhenkov, D. Mascarenas, S. Mniszewski, B. Nadiga, D. O’malley, D. Oyen, S. Pakin, L. Prasad, R. Roberts, P. Romero, N. Santhi, N. Si...

    2022

  26. [34]

    Zhang, K

    K. Zhang, K. Yu, and V. Korepin, Europhysics Letters 140, 18002 (2022)

    2022

  27. [35]

    Thorvaldson, D

    I. Thorvaldson, D. Poulos, C. M. Moehle, S. H. Misha, H. Edlbauer, J. Reiner, H. Geng, B. Voisin, M. T. Jones, M. B. Donnelly, L. F. Pe˜ na, C. D. Hill, C. R. Myers, J. G. Keizer, Y. Chung, S. K. Gorman, L. Kranz, and M. Y. Simmons, Nature Nanotechnology 20, 472 (2025). 7 SUPP...

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.