REVIEW 4 major objections 4 minor
Quantum random number generation using spatial quantum noise of light
T0 review · 4 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash
Pith's one-line read A camera's frame-subtracted spatial shot noise from a laser yields raw random bits at 5.92 Gbps, passing NIST and Diehard without algorithmic extraction.
desk verdict Camera-based QRNG idea is new and the experiment is real, but the 5.92 Gbps burst rate treats 30 correlated difference images as independent entropy sources. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The mechanism is the frame-subtraction of kinetic-mode EMCCD images of two equal-power coherent pulses: subtracting consecutive frames removes the classical beam profile and leaves δI(m,n), the spatial quantum fluctuation. Its quantum character is established by the noise ratio NR = Var(δI)/mean photon number ≈ 1, invariant under spatial binning, and by the absence of correlation peaks in cross- and auto-correlation maps. The bit extraction takes the three lowest-order bits of |δI|, relying on the variance of the shot-noise distribution being much larger than the digitization step, making those bits equiprobable.
What would settle it
Compute the mutual information between two difference images that share a frame (for instance, F1−F2 and F1−F3). If the shared-frame mutual information is large relative to each image's min-entropy, the 30-pair summation overcounts independent entropy; likewise, rerunning the NIST and Diehard suites on the five consecutive differences only (F1−F2, ..., F5−F6) would reveal whether the extra pairs were masking correlations.
Extended reading notes
Core claim
The central claim is that the spatial shot noise of coherent states, once static backgrounds are removed by subtracting consecutive EMCCD frames, is a spatially independent, Poissonian fluctuation field. Verified via spatial cross- and auto-correlation, a noise ratio that stays near unity under spatial binning, and a 29 dB shot-noise-to-background SNR, this field provides the entropy. Taking the three least-significant bits of the absolute fluctuation values yields a raw bitstream with min-entropy 0.9966 per bit that passes NIST SP 800-22 and Diehard without post-processing; the 30 frame-pair combinations from a single six-frame acquisition give 2.025e6 bits per capture, corresponding to 5.9
Load-bearing premise
The load-bearing premise is that the 30 frame-difference images generated from one six-frame acquisition are statistically independent entropy sources, even though every image shares frames with others, so any shared-frame correlation would lower the true entropy below the claimed bit count.
Editorial extensions
If this is right
- If the claim holds, spatial shot noise provides a random source that needs no algorithmic extractor, eliminating a traditional entropy bottleneck.
- The instantaneous rate scales with pixel count and frame combinations: the paper projects 11.45 Gbps by illuminating the full 170x512 sensor region.
- The sustained throughput is limited by the EMCCD readout electronics, not the quantum source, so faster low-noise readout would directly raise the continuous rate.
- The method's spatial parallelism and positive statistical test results suggest it could be combined with source-independent QRNG security proofs.
- The noise-ratio-versus-binning analysis provides a reusable diagnostic for verifying that detected fluctuations are shot-noise-limited and free of detector-induced correlations.
Reading between the lines
- The paper's 5.92 Gbps figure hinges on treating all 30 pairwise frame differences as independent; a direct measurement of mutual information between differences sharing a frame (e.g., F1−F2 vs F1−F3) would test whether the effective independent entropy per acquisition is lower than 2.025e6 bits.
- The three-LSB extraction depth was chosen from min-entropy measurements; a systematic scan of illumination power, exposure time, and binning would delimit the region where extractor-free operation survives, which the paper does not fully map.
- The combinatorial frame-subtraction idea is not specific to EMCCDs; any low-noise array detector with kinetic or frame-straddling acquisition could apply the same 30-image trick, with the noise-ratio-invariance test as a quality gate.
- The paper leaves the source-independent certification as future work; pairing its spatial-parallel scheme with a security proof would be the natural next step.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports a quantum random number generator based on spatial intensity fluctuations of coherent states detected with an EMCCD camera in kinetic mode. Two balanced coherent pulses are imaged; frame subtraction removes the mean profile; the residual fluctuations are attributed to quantum shot noise. The authors define a noise ratio (NR), verify its invariance with binning, and report NR≈1 as evidence of Poissonian statistics. Random bits are obtained by taking the absolute value of each fluctuation image and extracting the three least significant bits per pixel. From a six-frame acquisition they form 30 pairwise frame-difference images, and from a 150×150 ROI per image they claim 2.025×10^6 raw bits, corresponding to 5.92 Gbps instantaneous rate without extractor. The bitstream passes NIST SP 800-22 and Diehard suites; min-entropy is reported as 0.9966 per bit. The sustained rate is stated as 7.5 Mbps due to serial readout.
Significance. If the claims hold, this is a notable contribution: it demonstrates a massively parallel, extractor-free QRNG based on spatial shot noise, with a high instantaneous bit rate from a relatively simple setup. The use of EMCCD kinetic-mode imaging to obtain many spatial entropy sources is an interesting direction, and the shot-noise analysis and statistical testing are in the right spirit. The paper also gives a projected 11.45 Gbps rate with full sensor illumination. The significance of the result, however, depends critically on the independence of the 30 difference images and on the soundness of the entropy accounting, because the headline 5.92 Gbps rate is obtained by summing bits from those images as if they were independent sources.
major comments (4)
- [Eq. (4) and Fig. 2(c)] The noise ratio as defined in Eq. (4) does not equal 1 for independent Poisson fluctuations. If N_{L,n}, N_{L,n+1}, N_{R,n}, N_{R,n+1} are independent Poisson variables each with mean μ, the numerator has variance 4μ while the denominator ⟨N_L⟩+⟨N_R⟩ = 2μ, giving NR = 2. The paper instead reports NR≈1 as the Poissonian expectation (Fig. 2(c)). Either a factor of 1/2 is missing, the numerator is intended to be the variance of a single frame difference, or an unstated normalization is used. This needs clarification and correction; as written the central shot-noise validation is off by a factor of two.
- [Random bit extraction (combinatorial frame subtraction)] The 5.92 Gbps rate assumes that the 30 pairwise frame-difference images produced from a single six-frame acquisition are independent entropy sources. They are not independent: each raw frame participates in five different difference images, so for common frame i, Cov(D_{ij}, D_{ik}) = Var(F_i) > 0. The joint entropy of the 30 difference images is bounded by the entropy of the six underlying frames, not 30 times the per-image entropy. The text sums 2.025×10^6 bits from 30 images and divides by 342 μs, implicitly treating them as independent. Neither the per-bit min-entropy nor the NIST/Diehard passes tests for cross-image independence. Provide an entropy accounting that shows how six frames can support 2.025×10^6 extractor-free bits, or reduce the rate claim accordingly.
- [Min-entropy and bit extraction depth] The choice of extracting three LSBs is made from the min-entropy curve computed on the same dataset (Fig. 3(a)), which introduces post-hoc selection bias; the reported 0.9966 per-bit min-entropy is not a prediction but a retroactive fit. More importantly, the claim that 'the absolute-value operation ... does not degrade entropy quality' is asserted without proof. For a symmetric fluctuation distribution, absolute value folds the distribution and can change the statistics of the lower-order bits. Since the extractor-free claim rests on the per-bit min-entropy of the processed (absolute-value, LSB) stream, this step must be justified analytically or with a clear empirical demonstration that a conservative min-entropy bound holds for the exact preprocessing used.
- [Statistical validation (Sec. on NIST/Diehard)] The paper states that NIST tests were passed based on global p-values ≥ 0.0001 and reports only global p-values in Fig. 4(a). It does not report the proportion of passing sequences, which is a required NIST criterion, nor does it give the number of p-values per test. For a 400-sequence analysis, the minimum pass proportion is 0.9750; the manuscript should provide the proportion for each test. Without this, the claim of passing NIST SP 800-22 is not fully supported.
minor comments (4)
- [Abstract and Introduction] Typographical and grammatical issues: 'a extractor-free QRNG' should be 'an extractor-free QRNG'; 'quantum optical intensity fluctuations' is vague; 'the requirement for a highly radioactive source' could be more precise. The phrase 'large-scale stochastic simulations' is fine but the sentence about banking systems is somewhat informal.
- [Fig. 2 caption and Sec. on noise analysis] In Fig. 2(c), error bars are described as 'standard deviation of the mean for the NR over 100 acquired images' but the text above says '100 acquired images' while later the analysis is scaled to 2,400 acquisitions. Please clarify the number of images used for each figure and whether the NR data are from the left, right, or both beams.
- [Sustained rate calculation] The sustained rate is given as 'approximately 7.5 Mbps' based on a 270 ms readout period. Since 2.025×10^6 bits per acquisition divided by 0.27 s gives 7.5e6 bit/s, this is arithmetically consistent, but the text also says 'the digitization period is approximately 270 ms per acquisition cycle' while later referring to '342 μs acquisition time.' Please define which time interval corresponds to the instantaneous rate and which to the sustained rate.
- [References] Reference [11] is to Appl. Phys. Lett. 127, 104002 (2025) but the volume/issue may be incorrect; please verify. Also reference [38] duplicates the title of [37]; please check the intended citation.
Circularity Check
No significant circularity: the shot-noise and randomness claims are supported by in-paper falsifiable measurements and external statistical benchmarks; self-citations are corroborative, while the 30-image independence issue is a correctness risk rather than a circular reduction.
full rationale
The derivation chain is self-contained for the claims actually made. Coherent-state statistics (Eq. 2) are textbook; the frame-subtraction residual is characterized in-paper through the noise-ratio test (NR≈1 invariant under binning, Fig. 2c), which is a falsifiable check that the fluctuations obey Poissonian scaling and does not use any fitted constant. Spatial auto-/cross-correlation (Fig. 2a,b) and the 29 dB shot-noise SNR give additional in-paper evidence. The min-entropy and NIST/Diehard validation are applied to the generated bitstream against external benchmarks, not against the model inputs. The choice of 3 LSBs is informed by the same dataset's min-entropy curve, a mild in-sample selection, but it is conservative (six extraction depths showed near-ideal min-entropy) and the headline rate is an arithmetic consequence of the chosen configuration, not a fitted quantity renamed as a prediction. Refs [31,32] are self-cited characterization/method references (co-author Kumar with Marino), but the paper re-derives the key Poissonian and correlation checks, so the self-citation is not load-bearing. The strongest concern—that the 30 pairwise difference images derived from six shared raw frames are correlated, so counting 2.025×10^6 bits per acquisition may overstate extractable entropy—is a statistical independence/correctness risk, not circularity: the reported rate is forced by the counting convention, not by a fitted input that is then re-derived. Because the central quantum-noise and randomness claims reduce neither to their own definitions nor to a self-citation chain, the paper exhibits no significant circularity.
Assumptions & free parameters
free parameters (2)
- Bit extraction depth =
3 LSBs per pixel
- ROI sizes =
150 × 150 (left), 100 × 100 (right)
assumptions (6)
- standard math Coherent state photon statistics are Poissonian: Var(n) = ⟨n⟩.
- domain assumption The EMCCD operates in a shot-noise-limited regime with SNR ≈ 29 dB, making detector read noise and clock-induced charge negligible.
- domain assumption Subtracting consecutive frames cancels all static and common-mode classical noise, leaving only quantum shot noise.
- domain assumption Pixels are statistically independent entropy sources.
- ad hoc to paper The 30 combinatorial frame-difference images are independent sources of bits.
- ad hoc to paper The absolute-value operation preserves the per-bit min-entropy of the differential signal.
Cite this review
Pith. "Pith review of Quantum random number generation using spatial quantum noise of light." pith.science (2026). https://pith.science/paper/ISLC4KZO
@misc{pith2026260726486,
author = {Pith},
title = {Pith review of: Quantum random number generation using spatial quantum noise of light},
year = {2026},
howpublished = {\url{https://pith.science/paper/ISLC4KZO}},
note = {Machine review of arXiv:2607.26486}
}
read the original abstract
Generating high-speed, verifiable random numbers is a fundamental requirement for cryptography, large-scale stochastic simulations, and secure quantum communication. Here, we present a robust quantum random number generator that utilizes the spatial distribution of quantum fluctuations captured by an electron-multiplying charge-coupled device operated in high-speed kinetic mode. Through rigorous detector calibration and shot-noise analysis, we characterize the spatial quantum noise obtained from the spatial intensity fluctuations of the coherent states of light. Such quantum noise serves as a high-entropy source, enabling an instantaneous random bit generation rate of 5.92 Gbps without algorithmic randomness extraction in the present configuration. The sustained output rate is, however, limited to 7.5 Mbps by the bandwidth of the serial electronic readout. The generated sequences successfully pass the NIST SP 800-22 and Marsaglia Diehard statistical test suites, confirming the high quality and unpredictability of the entropy source.
Figures
Figures from the paper (1 more)
Reviewed August 1, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.