Pith. sign in

REVIEW 2 major objections 6 minor 22 references

Tomographic Phase Imaging with Randomized Probe Imaging

T0 review · 2 major / 6 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read Tomographic phase imaging of gold nanoparticles from one diffraction pattern per projection, at 62 nm resolution.

desk verdict First RPI-tomography demonstration works as a feasibility proof, but the 62 nm half-pitch resolution claim is likely inflated by shared missing-wedge artifacts in the FSC analyses. read the letter →

arxiv 2607.29471 v1 pith:XLLVMVBI submitted 2026-07-31 physics.optics physics.data-an

Jordan T. O'Neal , Byeongdu Lee , Soenke Seifert , Hanna Ruth , Michael Wojcik , Haidan Wen , Yanqi Luo , Yi Jiang
show 1 more author
Junjing Deng
This is my paper
classification physics.opticsphysics.data-an PACS 42.30.Rx07.85.Qe42.30.Wb
keywords randomizedprobeimagingtomographicphasesingle-shotcoherentdiffractionzoneplatenanoscaleX-raytomographyptychographycomparisongoldnanoparticlestime-resolved
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 demonstrates that a technique called randomized probe imaging (RPI) can supply the two-dimensional phase projections needed for X-ray nanotomography, using just one far-field coherent diffraction pattern per viewing angle instead of the many overlapping scans required by ptychography. The authors reconstruct a three-dimensional volume of a cluster of cubic gold nanoparticles from 220 single-shot diffraction patterns, achieving a half-pitch resolution of about 62 nm. They compare this volume with a higher-resolution ptychography-based reconstruction of the same sample and find the larger structural features match well. The core motivation is speed: with one pattern per projection, a tomographic scan that took roughly four hours with ptychography could, in principle, be collected in about 73 seconds with a step scan, or about 7 seconds with a fly scan. If this holds, RPI-tomography becomes a practical path to time-resolved nanoscale phase imaging for in-situ studies of crystal formation, fuel-cell operation, and similar dynamic processes.

What carries the argument

Randomized probe imaging (RPI), with a randomized zone plate (RZP) as the central optical element. The RZP focuses X-rays into a random speckle pattern with features of roughly 50 nm, so that each diffraction pattern is a scrambled encoding of the sample over a wide angular spread of illumination directions. The reconstruction algorithm uses a one-time ptychographic calibration of the complex probe (measured on a Siemens star) and then, for each single diffraction pattern, applies an iterative phase-retrieval loop with a bandwidth-limiting constraint that enlarges the effective reconstructed pixel size and stabilizes the inversion. The bandwidth-limiting constraint is what turns a single far

What would settle it

Run an actual fast RPI-tomography scan on the same type of gold nanoparticle sample, with the probe calibrated once before the scan and the full 73-second or 7-second acquisition, then compare the resulting volume to a ptychography-tomography volume of the same sample: if the 3D resolution degrades by more than the expected noise level, or if individual cubes appear in one volume but not the other in a way inconsistent with the reported 62 nm half-pitch, the fixed-probe/stability assumption is violated. A second check: simulate RPI projections from a known 3D phantom with random probe drift be

Watch

Extended reading notes

Core claim

The central claim is that tomographic phase imaging at sub-100 nm resolution does not require a scanned probe: a single far-field coherent diffraction pattern per projection, illuminated by a randomized zone plate, contains enough information to reconstruct the complex-valued projection if the probe is already known from a one-time calibration. The paper validates this by selecting one diffraction pattern from each angle of a four-hour ptychography-tomography dataset, reconstructing each projection with the RPI algorithm, feeding the phase images into a standard SART/FBP tomographic pipeline, and obtaining a 3D volume whose half-pitch resolution is 62 nm by the half-bit Fourier shell correla

Load-bearing premise

The entire pipeline depends on the assumption that the RZP probe, calibrated once before the tomography scan, remains a perfectly fixed and faithful illumination model for every projection, and that each single-frame RPI reconstruction is a clean coherent far-field intensity of a static sample — in this data, that assumption was never tested under genuinely fast scanning conditions, because the 'RPI' frames were actually extracted from a slow four-hour ptychography dataset.

Editorial extensions

If this is right

  • A tomographic phase scan can be collected in roughly 73 seconds with an angular step scan at 3 Hz, and potentially in 7.3 seconds with a fly scan at the detector's 30 Hz limit, opening a path toward time-resolved nanoscale phase tomography of dynamic samples.
  • RPI-tomography works with an ordinary far-field setup, requiring no coded-aperture motion or synchronized sample-aperture mechanics, so it can be retrofitted to existing coherent X-ray beamlines with a randomized zone plate.
  • Because each projection is reconstructed independently, standard tomographic alignment and reconstruction software apply unchanged, and any future algorithmic improvement to single-frame RPI phase retrieval directly improves the tomographic volume.
  • The consistent resolution between 2D projections and the 3D volume (62 nm half-pitch by FSC, 76 nm by FRC) suggests that tomographic integration itself does not degrade RPI reconstructions, and may in fact compensate for single-frame noise.
  • The method's resolution is bounded by the RZP outer zone width (here 50 nm) and by the bandwidth-limiting constraint, so fabricating RZPs with finer outer zones would directly push the achievable tomographic resolution toward the 30 nm Crowther limit of the current geometry.

Reading between the lines

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

  • The paper's two-step workflow (per-projection RPI, then tomographic reconstruction) discards information: the same diffraction patterns could in principle be fed into a joint 3D phase-retrieval algorithm that shares the known probe across all projections, likely improving resolution and robustness beyond the demonstrated 62 nm; the authors themselves hint at this possibility.
  • Because the probe is calibrated once and held fixed, the method's true failure point in a fast scan is mechanical and thermal stability of the sample and RZP over the full 73-second or 7-second acquisition; the paper's alignment data (1-2 μm total drift over four hours) is encouraging, but the single-shot reconstruction sensitivity to sub-pixel drift was not directly tested with genuinely fast-acq
  • The demonstrated resolution is in the 50-100 nm band, which is the same regime as the recently proposed coded-aperture single-distance nanotomography method; a direct experimental comparison of speed, dose, field of view, and reconstruction robustness between the two approaches would clarify which method scales better for in-situ studies.
  • The observation that tomographic resolution matches single-projection resolution despite a 70-degree missing wedge suggests that the missing-wedge penalty may be partially offset by the angular diversity of the single-shot reconstructions; this is a claim worth testing with simulated data before relying on RPI-tomography for anisotropic samples.
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

2 major / 6 minor

Summary. The paper demonstrates tomographic phase imaging of a cluster of cubic gold nanoparticles using randomized probe imaging (RPI). A random zone plate creates a speckled illumination; for each tomographic projection, a single far-field diffraction pattern is used to retrieve a complex-valued real-space image via the RPI algorithm with a ptychographically calibrated probe. The resulting 2D projections are fed into a standard SART/FBP tomographic workflow. The data are extracted from a four-hour ptychography-tomography scan, using one diffraction pattern per projection. The authors report a 3D half-pitch resolution of 62 nm by Fourier shell correlation, compare the RPI volume with a ptychography-tomography volume of the same sample, and estimate that a dedicated RPI scan could be performed in 73 s, a factor of 200 faster than the ptychography scan.

Significance. If the resolution and speed claims are robust, this is a valuable contribution to fast nanoscale X-ray phase tomography. The experimental demonstration is nontrivial: it uses a real heterogeneous sample, a random zone plate probe calibrated by ptychography on a Siemens star, and cross-validates the RPI volume against an independent ptychography-tomography reconstruction. The use of established software (Pty-Chi, CDTools, pyxalign) and the comparison of RPI with ptychography on the same dataset are strengths. The main weakness is that the central resolution claim rests on FSC analyses that may be inflated by common missing-wedge artifacts, and the speed improvement is extrapolated from a ptychography dataset rather than demonstrated in a dedicated fast RPI acquisition. The paper is the first application of RPI to tomography, but the method itself is prior work [12], so the novelty lies in the demonstration and in identifying the practical potential and limitations.

major comments (2)
  1. [Results, Fig. 1(f); Methods, tomography scan paragraph] The 62 nm half-pitch 3D resolution is not convincingly established because both FSC curves share confounds. The RPI-vs-ptychography FSC uses a reference volume reconstructed from the same 110 degree angular range with the same 70 degree missing wedge; common missing-wedge streaks and shared low-frequency artifacts can inflate the correlation. The split-RPI FSC compares two reconstructions made with the same probe model, the same reconstruction schedule, and the same missing wedge; common artifacts are as reproducible as signal. The half-bit criterion does not correct for this common-mode bias. The paper's observation that the 3D FSC resolution is no worse than the 2D FRC resolution despite the missing wedge is a red flag and is more simply explained by FSC inflation than by tomography overcoming the Nyquist limit. Please report directional FSC, wedge-masked FSC, or a phantom simulation w
  2. [Results, data acquisition speed paragraph] The factor-200 speed improvement and the 73 s / 7.3 s scan times are extrapolated from a ptychography-tomography dataset acquired at 2.5 Hz with a 750 nm step scan. No dedicated RPI fast scan was performed, and the manuscript does not demonstrate that the fixed calibrated probe remains valid under a fast angular step or fly scan, where vibration, drift, and partial coherence may differ from the slow ptychography scan. The speed claim is important to the paper's central motivation, so it should either be demonstrated in a proof-of-principle fast acquisition or clearly labeled as a projected capability. If the 3 Hz step-scan rate is based on measured overhead, state that explicitly and specify whether it was achieved in this experiment or in a separate test.
minor comments (6)
  1. [Abstract and Results, first paragraph] The abstract says sub-100 nm resolution, but the Results state that the 62 nm half-pitch resolution is slightly too large to resolve individual AuCs with side length approximately 100 nm. Please clarify that resolution refers to a half-pitch FSC metric and does not imply that the 100 nm cubes are individually resolved.
  2. [Fig. 1(f) caption] The caption says Fourier shell correlations but does not identify which curve corresponds to RPI-vs-ptychography and which to the split-RPI comparison. Please label the curves directly in the figure or in the caption.
  3. [Methods, RPI data selection] The sentence The RPI scan was assembled by selecting one diffraction pattern from each ptychography projection should specify how the single pattern was chosen from the raster grid, for example center position, brightest position, or random. This matters for the single diffraction pattern per projection claim and for reproducibility.
  4. [Methods, beam scan description] Minor formatting: '10x8 um2' should be written as '10 um x 8 um' or '10 x 8 um squared'.
  5. [References] Reference [22] contains the typo 'reconstuction' in the title; it should be 'reconstruction'.
  6. [Data availability] The data are not publicly available. Given that the paper makes quantitative FSC claims, a public release of the diffraction patterns and reconstructions would strengthen reproducibility. If this is not possible, state the reason.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: RPI probe is independently calibrated and the comparison with ptychography is a cross-validation.

full rationale

The derivation chain is: (1) calibrate the RZP probe by ptychography on a Siemens star; (2) use that fixed probe in the RPI algorithm to reconstruct each projection from a single diffraction pattern; (3) feed the projections into SART/FBP; (4) compare with a ptychography-tomography volume. No step defines its output in terms of the claimed result. The probe is not fitted to the tomographic data, and the RPI reconstructions do not use the ptychography volume as a constraint. The RPI-vs-ptychography FSC and the split-RPI FSC are standard cross-validation/resolution metrics; the shared missing wedge may bias the resolution estimate, but that is an accuracy concern, not a circularity. Citations to prior RPI work [12,13] are by different authors and are external support. No self-citation chain is load-bearing. The paper does not rename a known result or smuggle an ansatz via citation: it applies an established method to tomography. Therefore no circular step can be exhibited, and the appropriate score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper contributes an experimental demonstration; it does not introduce new theoretical entities. Its central claim rests on the RPI forward model and convergence assumptions from [12], a fixed probe assumption, and standard tomography; no invented entities or data-fitted physics parameters appear.

free parameters (3)
  • RPI bandwidth-limiting pixel enlargement / image crop = 200x200 px from 512x512 diffraction crop
    The real-space pixel size and field of view of the RPI reconstruction are set by this choice, directly setting the resolution scale; chosen by method rather than fitted to data.
  • Iteration schedule and object modes = 20 L-BFGS iterations with 2 incoherent modes, then 60 with 1 mode; mild regularization
    Convergence choices that affect image quality and the resolution claim.
  • Spot fill factor = 80% of field of view
    Spot size was chosen to fill 80% of the FOV, a tradeoff between imaging area and maximum achievable RPI resolution.
assumptions (5)
  • domain assumption Single far-field diffraction pattern is a coherent intensity |FT(probe x object)|^2 under the thin-sample, single-scattering approximation.
    RPI forward model from [12]; invoked in Methods Sec. 2.
  • domain assumption The ptychographically calibrated RZP probe remains fixed and valid for all projections.
    Methods: 'using the calibrated probe from the initial scan dropping the OPR modes'; no in-situ probe update in RPI.
  • domain assumption RPI phase retrieval with bandwidth-limiting constraint recovers the true complex object from a single pattern.
    Borrowed from Levitan et al. [12]; not re-derived here.
  • standard math Standard tomographic reconstruction (FBP seed + SART) and alignment recover the 3D volume from the 2D projections.
    Standard tomography relation; missing wedge factor from Radermacher [20].
  • standard math The half-bit FSC/FRC criterion gives a meaningful resolution estimate.
    Standard resolution metric; assumed in Results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tomographic Phase Imaging with Randomized Probe Imaging." pith.science (2026). https://pith.science/paper/XLLVMVBI

@misc{pith2026260729471,
  author       = {Pith},
  title        = {Pith review of: Tomographic Phase Imaging with Randomized Probe Imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XLLVMVBI}},
  note         = {Machine review of arXiv:2607.29471}
}
read the original abstract

We demonstrate tomographic phase imaging of cubic gold nanoparticles with sub-100 nm resolution requiring just a single far-field coherent diffraction pattern per projection. By using randomized probe imaging (RPI), a real-space amplitude and phase image can be reconstructed from a single diffraction pattern. These phase images are then fed into a standard tomographic workflow to retrieve the 3D volume. Nanoscale X-ray phase tomography has often relied on ptychography, which requires slow 2D scanning for each projection. By using RPI, the data collection is greatly sped up at the cost of some resolution. We compare an RPI-tomography volume to a ptychography-tomography volume of the same sample, finding high consistency between the two methods. To the best of our knowledge, this is the first application of RPI to tomography.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

22 extracted references

  1. [12]

    A. L. Levitan, K. Keskinbora, U. T. Sanli, M. Weigand, and R. Comin, Single-frame far-field diffractive imaging with randomized illumination, Opt. Express, OE 28, 37103 (2020)

    2020

  2. [1]

    P. I. C. Claro, E. P. B. S. Borges, G. R. Schleder, N. L. Archilha, A. Pinto, M. Carvalho, C. E. Driemeier, A. Fazzio, and R. F. Gouveia, From micro- to nano- and time-resolved x-ray computed tomography: Bio-based applications, synchrotron capabilities, and data-driven processing, Appl. Phys. Rev. 10, 021302 (2023)

    2023

  3. [2]

    Dewanckele, M

    J. Dewanckele, M. A. Boone, F. Coppens, D. Van Loo, and A. P. Merkle, Innovations in laboratory‐based dynamic micro‐CT to accelerate in situ research, J. Microsc. 277, 197 (2020)

    2020

  4. [3]

    Jiang et al., In situ ptychographic x-ray nanotomography of temperature-controlled crystallization processes, Nat Commun (2026)

    Z. Jiang et al., In situ ptychographic x-ray nanotomography of temperature-controlled crystallization processes, Nat Commun (2026)

    2026

  5. [4]

    Y. P. Wijaya, F. P. Orfino, and E. Kjeang, X-ray Computed Tomography Visualization of Liquid Water in Proton Exchange Membrane Fuel Cells: A State-of-the-Art Review, ACS Appl. Energy Mater. 8, 12997 (2025)

    2025

  6. [5]

    Martoïa, T

    F. Martoïa, T. Cochereau, P. J. J. Dumont, L. Orgéas, M. Terrien, and M. N. Belgacem, Cellulose nanofibril foams: Links between ice-templating conditions, microstructures and mechanical properties, Materials & Design 104, 376 (2016)

    2016

  7. [6]

    Flenner, M

    S. Flenner, M. Storm, A. Kubec, E. Longo, F. Döring, D. M. Pelt, C. David, M. Müller, and I. Greving, Pushing the temporal resolution in absorption and Zernike phase contrast nanotomography: enabling fast in situ experiments, J Synchrotron Rad 27, 1339 (2020)

    2020

  8. [7]

    Longo, L

    E. Longo, L. Sancey, S. Flenner, A. Kubec, A. Bonnin, C. David, M. Müller, and I. Greving, X-ray Zernike phase contrast tomography: 3D ROI visualization of mm-sized mice organ tissues down to sub-cellular components, Biomed. Opt. Express, BOE 11, 5506 (2020)

    2020

Show all 22 references
  1. [8]

    Maire, C

    E. Maire, C. Le Bourlot, J. Adrien, A. Mortensen, and R. Mokso, 20 Hz X-ray tomography during an in situ tensile test, Int J Fract 200, 3 (2016)

    2016

  2. [9]

    Aidukas, N

    T. Aidukas, N. W. Phillips, A. Diaz, E. Poghosyan, E. Müller, A. F. J. Levi, G. Aeppli, M. Guizar- Sicairos, and M. Holler, High-performance 4-nm-resolution X-ray tomography using burst ptychography, Nature 632, 81 (2024)

    2024

  3. [10]

    Nikitin, M

    V. Nikitin, M. Carlsson, D. Gürsoy, R. Mokso, and P. Cloetens, X-ray nano-holotomography reconstruction with simultaneous probe retrieval, Opt. Express, OE 32, 41905 (2024)

    2024

  4. [11]

    Nikitin, M

    V. Nikitin, M. Carlsson, R. Mokso, P. Cloetens, and D. Gürsoy, Single-distance nano-holotomography with coded apertures, Opt. Lett., OL 50, 574 (2025)

    2025

  5. [13]

    A. L. Levitan, Studying Electronic Textures with Coherent Lensless Imaging, Massachusetts Institute of Technology, 2023

    2023

  6. [14]

    Gleber, M

    S.-C. Gleber, M. Wojcik, J. Liu, C. Roehrig, M. Cummings, J. Vila-Comamala, K. Li, B. Lai, D. Shu, and S. Vogt, Fresnel zone plate stacking in the intermediate field for high efficiency focusing in the hard X-ray regime, Opt. Express, OE 22, 28142 (2014)

    2014

  7. [15]

    M. N. O’Brien, M. R. Jones, K. A. Brown, and C. A. Mirkin, Universal Noble Metal Nanoparticle Seeds Realized Through Iterative Reductive Growth and Oxidative Dissolution Reactions, J. Am. Chem. Soc. 136, 7603 (2014)

    2014

  8. [16]

    Auyeung, R

    E. Auyeung, R. J. Macfarlane, C. H. J. Choi, J. I. Cutler, and C. A. Mirkin, Transitioning DNA‐ Engineered Nanoparticle Superlattices from Solution to the Solid State, Advanced Materials 24, 5181 (2012)

    2012

  9. [17]

    Du et al., Pty-Chi: A PyTorch-Based Modern Ptychographic Data Analysis Package, arXiv:2510.20929v1 [physics.optics]

    M. Du et al., Pty-Chi: A PyTorch-Based Modern Ptychographic Data Analysis Package, arXiv:2510.20929v1 [physics.optics]

  10. [18]

    Odstrcil, P

    M. Odstrcil, P. Baksh, S. A. Boden, R. Card, J. E. Chad, J. G. Frey, and W. S. Brocklesby, Ptychographic coherent diffractive imaging with orthogonal probe relaxation, Opt. Express, OE 24, 8360 (2016)

    2016

  11. [19]

    Levitan, M

    A. Levitan, M. Cain, and A. Kutakh, CDTools, https://github.com/cdtools-developers/cdtools

  12. [20]

    Radermacher, Three‐Dimensional reconstruction of single particles from random and nonrandom tilt series, J

    M. Radermacher, Three‐Dimensional reconstruction of single particles from random and nonrandom tilt series, J. Electron Microsc. Tech. 9, 359 (1988)

    1988

  13. [21]

    Ruth, Pyxalign, https://github.com/AdvancedPhotonSource/pyxalign

    H. Ruth, Pyxalign, https://github.com/AdvancedPhotonSource/pyxalign

  14. [22]

    A. H. Andersen and A. C. Kak, Simultaneous algebraic reconstuction technique (SART): a superior implementation of the art algorithm, Ultrason. Imaging 6, 81 (1984). Supplementary Material This supplementary material contains additional views of the tomographic reconstructions ...

    1984

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.