{"id":"5fa33b87-0dd5-40b3-a181-c30d4d5e32d9","arxiv_id":"2509.14950","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"First demonstration of two-dimensional ghost imaging using electron-photon pairs in a TEM, achieving 2 micrometer resolution on complex patterns.","lead":"This paper demonstrates ghost imaging, a correlation-based imaging technique, inside a transmission electron microscope using pairs of electrons and photons. It reconstructs a cat-shaped pattern and a test grating with micrometer-scale resolution, the first two-dimensional ghost image from electron-photon pairs.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Electron-photon position correlation is assumed from prior work, not directly measured; a pinhole ghost-image test is needed to validate the 2 μm resolution claim.","rationale":"The paper convincingly demonstrates ghost imaging with electron-photon pairs, and the cat image is strong evidence that a spatial correlation exists. The weakest point is the quantitative resolution claim, which rests on the tightness of the electron-photon position correlation. This correlation is not directly measured; it is assumed from prior work and indirectly inferred from a grating fit with several adjustable parameters. The pinhole test directly measures the joint PSF and would settle whether the 2 μm resolution is real or an artifact of the fitting model. Since the reader already identified the same assumption as the weakest point and recommended a conditional acceptance, my analysis does not change the verdict: the paper is acceptable if the authors provide such a direct calibration or otherwise strengthen the evidence for the correlation width.","tokens_in":10933,"tokens_out":9300,"duration_ms":102453,"concrete_test":"Replace the cat mask with a pinhole of ~1 μm diameter placed in the photon image plane. Record the ghost image of this pinhole using the same electron-photon coincidence protocol. Measure the FWHM of the resulting spot in the electron coincidence image. If the spot size is ~2 μm, the resolution claim is supported; if it is significantly larger or elongated, the grating-based fit likely underestimates the PSF width, and the correlation is not as tight as assumed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the CL photon be emitted from the exact electron impact point on the 50 nm Si membrane, and that this correlation survive energy filtering and the ±25 ns coincidence window. This assumption is stated in the Experimental Setup and Fig. 1 caption but is not directly verified in this manuscript; it is inherited from refs. [18] and [42]. The end-to-end spatial resolution is inferred from a grating-target fit (Fig. 3) using an analytical model with multiple free parameters (sample position, Gaussian σ, distortion parameters). If the true electron-photon position correlation were broader than modeled, the fit could still converge by adjusting these parameters, yielding a misleadingly small resolution. The ghost image of the cat demonstrates qualitative correlation but does not quantify its width. A direct measurement of the joint point-spread function would strengthen the quantitative claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a coincidence (ghost) imaging experiment using electron–cathodoluminescence photon pairs in a transmission electron microscope. A 200 keV electron beam passes through a 50 nm Si membrane; the emitted CL photons are collected by a custom parabolic mirror, transmitted through a mask in an external image plane, and detected by a single-photon counting module. The transmitted electrons are energy-filtered and detected by a time-stamped Timepix3 camera. Coincidence filtering between electron and photon timestamps reconstructs the mask pattern in the electron image. The authors demonstrate a clear ghost image of a cat-shaped mask and quantify the resolution with a grating target fitted by a Gaussian PSF model, obtaining 2.03 ± 0.06 μm FWHM. They claim this is the first two-dimensional ghost image formed from the spatial correlations between an electron and its corresponding photon.","tokens_in":11145,"tokens_out":4924,"duration_ms":56640,"significance":"The central qualitative claim—that coincidence selection of electron–photon pairs reconstructs a complex 2D pattern—is well supported by the cat image and the grating line structure. The experiment integrates a substantial custom apparatus (parabolic mirror, optical viewport, energy filter, Timepix3 detector) and collects more than 10^5 coincidence events. The resolution claim, however, rests on a fit with several open parameters and on an inherited assumption about the electron–photon position correlation. The resolution figure is an output, not an input, and should be substantiated by a direct PSF measurement or a robustness analysis. If confirmed, the work is a meaningful step toward quantum-inspired imaging in electron microscopy.","major_comments":[{"comment":"The reported FWHM of 2.03 ± 0.06 μm is obtained by convolving an ideal grating image with a Gaussian PSF and fitting the model with several free parameters, including 'the precise location of the sample relative to the mirror and the standard deviation of the Gaussian function' as well as distortion parameters. The paper does not present a model-independent measurement of the joint electron-photon point-spread function, e.g., a pinhole or sharp-edge ghost-image scan. A broader intrinsic correlation could be absorbed by other fit parameters, so the quoted statistical uncertainty likely underestimates the systematic error. Please add a direct PSF measurement or a sensitivity analysis demonstrating that the fitted σ is robust to parameter degeneracy.","section":"RESULTS / Fig. 3"},{"comment":"The key assumption that the CL photon is emitted from the precise point at which the electron enters or exits the membrane is stated but not tested in this manuscript; it is inherited from refs. [18,42]. Because the resolution claim is presented at the TEM sample plane and linked to the joint electron-photon state, this assumption is load-bearing. An in-situ verification (e.g., imaging a small emitter or pinhole target in coincidence) would remove the risk of circularity in relying on a fitted model to validate the correlation width.","section":"Experimental Setup / Fig. 1"}],"minor_comments":[{"comment":"The g(2)(τ) temporal cross-correlation function is mentioned but never defined. Please provide a definition or equation, and state the coincidence window explicitly in relation to the histogram in Fig. 1(C).","section":"Results"},{"comment":"The caption states that the image was rotated and smoothed with a Gaussian filter for presentation after fitting. Please specify the rotation angle and the smoothing kernel width, and clarify that the fit was performed on the unprocessed data.","section":"Fig. 3(A)"},{"comment":"The demagnification factor is quoted as ~19× for the cat mask and 16× for the grating target. The discrepancy is not explained; please clarify whether this arises from different mask-to-mirror distances or an effective optical magnification.","section":"Results"},{"comment":"The phrase 'quantum ghost images' is used without defining the nonclassical character of the electron-photon pairs. If the term is intended loosely, consider rephrasing to 'coincidence ghost imaging' or add a sentence justifying the quantum terminology.","section":"Conclusion"},{"comment":"Ref. [42] is cited as an arXiv preprint for the back-projection model and post-processing. If this work has been published or updated, please update the reference.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The experimental demonstration is impressive and likely of interest to the journal, but the quantitative resolution claim needs strengthening. I would not necessarily require a full pinhole PSF measurement if the authors can show fitting robustness and clearly separate the system PSF from the joint-state correlation. Also, the use of 'quantum' should be disciplined."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Clean, solid proof-of-principle. The paper demonstrates the first two-dimensional ghost image of a complex mask formed from electron-cathodoluminescence photon pairs in a TEM, and the cat reconstruction is convincing. The grating calibration with a Gaussian PSF gives a plausible resolution of about 2 um. This is a real step beyond the same group's earlier coincidence work [42], which used the correlations to probe uncertainty relations rather than reconstruct a pattern. What the paper does well: the experimental setup is substantial and described in enough detail to be reproduced - custom parabolic mirror, free-space CL extraction, energy filtering, time-stamped Timepix3 detection. The post-processing back-projection model is reasonable, and the supplementary on mirror fabrication is thorough. The self-citations are appropriate. Soft spots, in order of severity: 1. The quantitative resolution rests on an assumption and a fit. The tight electron-photon position correlation is inherited from earlier work, not directly measured here. The cat image demonstrates correlation at the few-micrometer level, so the qualitative ghost-imaging claim is safe. But the 2.03 +/- 0.06 um value comes from a Gaussian PSF convolved with a model that has open parameters (sample position, distortion terms). A broader underlying correlation could be partially absorbed by the fit. A direct PSF measurement - a pinhole or knife-edge target, or a measured joint correlation width - would make the number much more robust. I'd call the resolution an estimate, not a rigorous result. 2. The conclusion calls this 'the first realization of quantum ghost images.' That overreaches. Nothing in the data requires entanglement or shows a quantum advantage; it's ghost imaging with correlated pairs. The title and abstract are carefully worded, but the conclusion should be toned down. 3. Minor: no data or code availability statement. Not a blocker, but annoying for a methods paper. Who should read this: people in electron-photon correlations, quantum optics in TEM, and low-dose imaging. The 2 um resolution is far from atomic-scale TEM, but that's not the point - it's the first demonstration of the technique. Recommendation: send to peer review. Ask for a direct measurement or calibration that supports the fitted resolution, and soften the 'quantum ghost image' language. With those changes, this is a solid contribution.","headline":"First 2D ghost image from electron-photon pairs in a TEM, with a credible few-micron resolution; needs a direct PSF check and a toned-down 'quantum' label before it's solid.","tokens_in":11659,"tokens_out":3306,"would_cite":true,"duration_ms":33401,"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":"Ghost imaging inside a transmission electron microscope reconstructs complex patterns from electron–photon coincidences at about 2-micrometer resolution.","keywords":["ghost imaging","coincidence imaging","cathodoluminescence","transition radiation","transmission electron microscopy","electron-photon pairs","quantum imaging","time-correlated single-photon counting"],"falsifier":"Remove the mask and record the photon image on a camera simultaneously with the position-resolved electron detection. If the conditional spread of photon position given electron position is larger than the 0.87 µm standard deviation implied by the fitted point-spread function, then the claimed 2 µm resolution cannot be attributed to the electron–photon correlation and must come from something else, such as the optics.","tokens_in":10855,"feed_emoji":"🐱","tokens_out":5083,"duration_ms":51335,"temperature":0.7,"pith_summary":"The paper sets out to show that ghost imaging—reconstructing an image from correlations between two particles, only one of which has touched the object—can be carried out inside a transmission electron microscope using an electron and the single photon it emits. The authors generate electron–cathodoluminescence photon pairs from a thin silicon membrane, place a transmission mask in the photon path, and use time-stamped coincidence filtering to keep only electrons whose partner photon passed through the mask. They reconstruct the mask shape with a spatial resolution of 2.03 ± 0.06 micrometers, including fine features of a cat-shaped pattern. If correct, this is the first two-dimensional ghost image formed by an electron and its corresponding photon, and it extends quantum-optics imaging techniques into electron microscopy.","feed_headline":"Electron–photon pairs image a cat at 2-micron resolution","feed_subtitle":"First 2D ghost image from correlated electrons and photons, opening a route to quantum-enhanced electron microscopy.","key_machinery":"The load-bearing object is the electron–cathodoluminescence photon pair generated by transition radiation in a thin silicon membrane. Because the photon is emitted from the exact electron entrance/exit point, the photon's position in the optical image plane is a proxy for the electron's position in the sample plane. The experiment isolates these pairs by energy filtering the electrons to 2-3 eV loss and by requiring photon and electron detection times to match within ±25 ns, removing uncorrelated background. A parabolic mirror with numerical aperture about 0.58, a free-space optical relay, and a transmission mask in the photon image plane turn 'did the photon pass?' into a binary spatial tes","core_discovery":"The central claim is that transition-radiation cathodoluminescence photons carry a tight position correlation with the electron that emitted them—the photon is emitted from the precise point where the electron enters or exits the membrane—and that this correlation is strong enough to form a useful two-dimensional image. A 200 keV electron beam passes through a 50 nm silicon membrane; the emitted photons are collected by a parabolic mirror and imaged onto a mask, so a photon reaching the bucket detector tells which spatial region the parent electron passed through. The electrons are energy-filtered to select those that lost 2-3 eV (i.e., emitted a visible photon) and detected on a position-re","pith_inferences":["Editorial inference: The stated resolution is set at the TEM sample plane, but the mask is in the photon image plane; if the electron–photon position correlation is tighter than the measured 0.87 µm standard deviation, then the 2 µm resolution is dominated by aberrations in the parabolic mirror and relay optics, so better optics should improve the image directly.","Editorial inference: The paper does not measure the joint position correlation itself; one could test the scheme's premise by imaging the photon beam onto a camera while simultaneously recording electron positions with the mask removed, and checking that the conditional photon-position spread matches the fit.","Editorial inference: The ±25 ns coincidence window and 50 ns timing resolution mean the demonstrated pairs are not shown to be quantum-correlated in a way that violates classical bounds; the same imaging protocol would likely work with classically correlated pairs, so the 'quantum' enhancement aspect remains a separate question the paper does not settle.","Editorial inference: A natural next experiment would replace the static cat mask with a programmable spatial light modulator, enabling rapid switching and computational post-selection schemes without re-aligning the microscope."],"forward_implications":["The technique produces the first two-dimensional quantum ghost image of a non-trivial mask inside a TEM, with enough resolution to resolve the cat's eyes, ears, and tail (features of order a few micrometers on the sample).","Because only the photon touches the mask and only the electron is spatially resolved, the image is formed without placing any absorber in the electron path, a property relevant for radiation-sensitive samples.","Energy filtering and coincidence selection suppress background from electrons that did not emit a photon, improving signal-to-noise and allowing higher beam currents and shorter acquisition times.","Adaptive masks (e.g., digital micromirror devices) and spatially resolved photon detection should allow post-selection shaping of electron wave functions, with implications for programmed and low-dose imaging.","The electron optics contribution to blur is negligible compared with the photon optics, so resolution gains are expected from improving the photon collection path."],"fun_headline_variants":["Ghost imaging with electron–photon pairs hits 2-micron resolution","Electron–photon ghost imaging achieves 2-micron resolution","Ghost images from electron–photon pairs at 2-micron scale","First 2-micron ghost image with electron–photon pairs"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The entire reconstruction relies on the assumption, taken from earlier work rather than measured in this experiment, that the cathodoluminescence photon is emitted from the precise point where the electron enters or exits the membrane, so that the photon's position faithfully reports the electron's position; if that correlation were not tight, the ghost image would blur or disappear.","fun_headline_variants_meta":{"raw":{"variants":["Ghost imaging with electron–photon pairs hits 2-micron resolution","Electron–photon ghost imaging achieves 2-micron resolution","Ghost images from electron–photon pairs at 2-micron scale","First 2-micron ghost image with electron–photon pairs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000768,"raw_usage":{"total_tokens":3211,"prompt_tokens":684,"completion_tokens":2527,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":428,"completion_tokens_details":{"reasoning_tokens":2446}},"tokens_in":428,"tokens_out":2527,"duration_ms":19110,"temperature":1.0,"reasoning_tokens":2446,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T16:12:26.928166+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Remove the mask and record the photon image on a camera simultaneously with the position-resolved electron detection. If the conditional spread of photon position given electron position is larger than the 0.87 µm standard deviation implied by the fitted point-spread function, then the claimed 2 µm resolution cannot be attributed to the electron–photon correlation and must come from something else, such as the optics.","supporting_citations":[],"review_version":1}