{"id":"1fc3be9c-a6e0-4ce3-bf5e-96a37edc189e","arxiv_id":"2504.19762","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"St4DeM bundles 4D-STEM, EELS/EDS spectrum imaging, and tomography into a Digital Micrograph suite, with a proof-of-principle 7D-STEM reconstruction pipeline.","lead":"St4DeM is a free software suite that lets electron microscopes combine 4D-STEM with EELS/EDS and tomography, and it demonstrates a full 7D acquisition-and-reconstruction workflow. A generalist reader would check this to see whether multi-modal STEM data collection is now practical in one package.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 7D reconstruction claim is unvalidated: Section III.B back-projects local CBED patterns along rays with no known-structure test, and the energy axis is never truly reconstructed.","rationale":"The reader's CONDITIONAL verdict is appropriate. Re-reading Section III.B and the stated limitations, the load-bearing issue is that the only evidence for the analysis half of the headline claim is a reconstruction pipeline whose central physical assumption is untested: a focused-probe CBED pattern is back-projected along a full ray through the sample. This is not equivalent to a tomographic line integral for crystalline or even strongly scattering specimens, and the authors admit it introduces problematic dynamical diffraction behavior. The paper also reports that the TiNi sample did not diffract sufficiently to form meaningful dark-field virtual images, which is exactly the signal that would validate the 6D diffraction volume. No known-structure test or resolution metric is supplied. Independently, the claimed 7D analysis is incomplete because the energy axis was collapsed to zero-loss and plasmon sums before reconstruction. None of this undermines the acquisition suite, alignment methods, iDPC implementation, or the availability of code and data; those are real and reproducible contributions that justify publication. But the advertised 7D analysis claim is not yet supported, so a conditional acceptance with a request for validation on a simulated or strongly diffracting known structure is the right call. The reader already identified the central assumption as unvalidated, and my additional energy-axis point does not change the verdict.","tokens_in":9372,"tokens_out":7932,"duration_ms":78906,"concrete_test":"Build a simulated 4D-STEM tilt series of a known nanocrystal (e.g., a 40-60 nm gold or Fe3O4 cube) with a multislice package (abTEM or Prismatic) over the same -64 to +74 degree range with 2-degree steps, matching the experimental probe and camera settings; run the St4DeM Method-1 and Method-2 reconstruction scripts (or the public Jupyter/SimpleITK pipeline) on this ideal data, and compare the resulting 3D diffraction volume and detected peak list with the known kinematical peak positions and intensities. If the reconstructed reciprocal lattice is not recovered for a strong diffractor, the Section III.B back-projection is not a valid 6D reconstruction; if it is recovered, the principal objection is answered. Also check whether the full energy axis can be reconstructed from the stored EELS tilt series rather than only the summed zero-loss and plasmon slices.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section III.B's 6D/7D reconstruction is the only evidence for the 'analysis' half of the headline claim, and it is not established. The pipeline assigns the 2D CBED pattern recorded at each probe position and tilt to every voxel along an inverse-Radon ray (rotating the pattern by the goniometer angle) and accumulates these into a 3D diffraction volume per real-space voxel. For a focused STEM probe this is not a valid line-projection of the scattered intensity: the CBED pattern is a dynamical, coherent signal that depends on local orientation, thickness, and defocus, so back-projecting the same pattern across the entire ray distributes one localized measurement over many voxels. The authors acknowledge the approximation ('problematic dynamical diffraction behavior', nearest-neighbor interpolation, no precession), and also state that the TiNi specimen 'was not diffracting enough to form meaningful dark field virtual images.' No comparison to a known structure, no simulated phantom, and no resolution or peak-position metric is provided, so the reconstructed 6D diffraction volume in Figure 7 cannot be distinguished from software plumbing. Additionally, the '7D' claim is incomplete: EELS data were reduced to zero-loss and plasmon sums before reconstruction, so a full energy dimension was never reconstructed. The acquisition and iDPC contributions are well supported by publicly available code and data, but the 7D-analysis claim is stronger than the evidence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents St4DeM, a Digital Micrograph-based software suite for multi-modal 4D-STEM acquisition. The suite combines 4D-STEM acquisition with EELS and EDS spectrum imaging, electron tomography, ptychography, real-time visualization, alignment routines, and incoherent differential phase contrast (iDPC) analysis. Acquisition is demonstrated on two microscope platforms (Thermo Fisher Themis and JEOL JEM-2100F), with frame rates up to 100 fps via synchronized server/client control. The authors further describe a tomographic reconstruction workflow that back-projects the 2D diffraction patterns from a 4D-STEM tilt series into a 3D real-space volume, yielding 6D diffraction volumes or sparse peak lists, and they present this as a proof-of-principle 7D-STEM pipeline when combined with EELS spectrum-image tilt series. The software, acquired data, and analysis code are stated to be available on GitHub and Zenodo.","tokens_in":9603,"tokens_out":3243,"duration_ms":35221,"significance":"If the acquisition and analysis claims hold, St4DeM would be a useful contribution to the electron microscopy software ecosystem, because it integrates several multimodal acquisition modes in a widely used commercial environment. The paper benefits from explicit, publicly available software and data, and it demonstrates reproducible multi-instrument acquisition, including automated tilt-series acquisition and iDPC processing. The reconstruction component, however, is only a proof of principle and is not quantitatively validated against a known structure or a simulated phantom; the authors themselves note that the TiNi specimen was not diffracting enough for meaningful dark-field images and that only nearest-neighbor interpolation was used. The significance is therefore conditional: the acquisition and visualization aspects are credible, while the 7D-analysis claim is stronger than the evidence presented.","major_comments":[{"comment":"The 6D/7D reconstruction claim is not validated. The method back-projects each 2D diffraction pattern along a straight ray through the 3D real-space volume, but a focused STEM probe's CBED pattern is a dynamical, coherent signal that depends on local thickness, orientation, and defocus; it is not a line integral of the object's diffracted intensity. The authors acknowledge limitations (nearest-neighbor interpolation, no precession, weak diffraction from the TiNi sample), but no test against a known structure, simulated phantom, or quantitative metric (peak positions, resolution, missing-wedge effects) is provided. Without such a test, the reconstructed diffraction volume in Figure 7 cannot be distinguished from software plumbing. I recommend adding a validation experiment or simulation that compares the back-projected diffraction volume against a known structure or phantom, and reporting a quantitative agreement metric.","section":"III.B, Figure 7"},{"comment":"The '7-dimensional data' claim is broader than the demonstrated result. The EELS energy dimension was reduced to sums of the zero-loss and plasmon peaks before reconstruction, so a full energy-resolved volume was never reconstructed. The paper should either present a truly energy-resolved reconstruction or explicitly qualify the demonstration as 6D diffraction reconstruction plus 2D energy-filtered projections; the current wording in the Abstract and Section III.B overstates the dimensionality that was actually achieved.","section":"III.B, Conclusions"},{"comment":"The registration and alignment chain for the 4D-STEM tilt series is described but not quantitatively assessed. The method relies on registering the 4D-STEM tilt series to a re-projected SIRT reconstruction of a HAADF tilt series, yet no metric is given for the registration accuracy or for the propagated effect of SIRT reconstruction artifacts and the missing wedge on the final diffraction-volume alignment. Since the entire back-projection pipeline depends on this alignment, the absence of any error metric is a load-bearing gap.","section":"III.B, Methods 1 and 2"}],"minor_comments":[{"comment":"The callout to 'Figure 2(b)' in the description of the Main tab of the user interface should be 'Figure 3(b)', since Figure 2 already refers to STEM images.","section":"III, figure callout"},{"comment":"The caption of Figure 1 contains a stray parenthesis: 'diameter less than )100 nm)' should be 'diameter less than 100 nm'.","section":"II, FIB sample preparation"},{"comment":"There are several typographical errors, including 'defocuse's' (should be 'defocus values' or 'defoci'), 'tomogaphy' (should be 'tomography'), and 'compromising as a tilt series' (likely 'composing' or 'comprising').","section":"Various"},{"comment":"The term 'd-scanning effects' is used without definition; please clarify what is meant by this term, as it is central to the automatic plane-subtraction step.","section":"III.A"}],"recommendation":"major_revision","confidential_remarks":"The paper is a software/methods contribution, and the acquisition portion is credible and useful. The main risk is that the 7D-analysis claim will be taken at face value despite the lack of validation; I would encourage the editor to require either a phantom/known-structure test or a clearly softened claim in the title, abstract, and conclusions. The public code and data are a genuine strength and should be preserved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper is best read as a software contribution, not as a tomography result. St4DeM gives the electron microscopy community a free, integrated way to drive 4D-STEM, EELS/EDS SI, and tilt series from Digital Micrograph, with real data shown from two instruments, 100 fps synchronized acquisition, iDPC analysis, and an open GitHub/Zenodo release. That is a useful and reproducible contribution, and the authors are honest about what they did and did not show.\n\nWhat is new: the integration itself, plus a 6D/7D reconstruction pipeline that back-projects 2D diffraction patterns along rays into a 3D real-space volume. The pipeline is novel but unvalidated. The stress test is right that Section III.B never compares against a known structure or phantom, and the linear ray projection of a coherent CBED pattern ignores dynamical diffraction and the missing wedge. The authors more or less concede this when they mention the \"problematic dynamical diffraction behavior\" and their choice of nearest-neighbor interpolation. To make matters worse, the TiNi specimen \"was not diffracting enough to form meaningful dark field virtual images,\" so the central figure for the 6D/7D method shows plumbing, not science. And the energy dimension in the \"7D\" reconstruction was never actually reconstructed: they used the sum of zero-loss and plasmon peaks, not a full energy axis.\n\nNone of that kills the acquisition contribution. The iDPC example on noisy carbon nanotubes works, the auto-focus and drift correction are useful, and the code is out there to try. The paper's real weakness is that the title and abstract sell the 7D reconstruction as a headline feature, while the body is appropriately cautious. A referee should push for either a proper validation (phantom or known nanocrystal) or a more modest title, and for dropping the 7D framing unless the energy dimension is actually reconstructed.\n\nWho gets value from this: anyone running 4D-STEM on a Gatan DM system who wants a ready-made acquisition and alignment toolbox. Also useful as a template for combining modalities. I would cite it for the software in a methods section, but not for the tomography claim.\n\nRecommendation: yes, send to peer review. It is a real software contribution with public code and data, and the flaws are in overclaiming rather than in fabrication or sloppy data handling. A serious referee can fix the framing and ask for a validation experiment. If the journal wants a clean story, ask the authors to resubmit as an acquisition-suite paper with the 7D part clearly marked as a preliminary pipeline.","headline":"A genuinely useful acquisition suite, but the 7D reconstruction is a proof-of-principle that the paper itself shows to be weak — referee it for the software, not the 7D claim.","tokens_in":10201,"tokens_out":2285,"would_cite":true,"duration_ms":22955,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"St4DeM lets one microscope session record 7D data by combining 4D-STEM, spectroscopy, and tomography.","keywords":["4D-STEM","nanodiffraction","7D-STEM","EELS spectrum imaging","EDS spectrum imaging","electron tomography","Digital Micrograph","iDPC"],"falsifier":"Record the same 4D-STEM tilt series from a crystal whose structure and orientation are already known, apply Method 1 or Method 2, and compare the reconstructed per-voxel diffraction volumes or peak lists with multislice or Bloch-wave simulations of the same tilt geometry; if peak positions or relative intensities differ beyond interpolation error, the straight-ray back-projection is not faithful.","tokens_in":9170,"feed_emoji":"🔬","tokens_out":9225,"duration_ms":82159,"temperature":0.7,"pith_summary":"St4DeM is a software suite, written in the Digital Micrograph scripting environment, that brings several scanning transmission electron microscopy modes into one interface: 4D-STEM diffraction-pattern acquisition, EELS and EDS spectrum imaging, ptychography, and tilt-series electron tomography. The paper's central claim is that these modes can be synchronized so a single experiment records diffraction and chemical spectra at every probe position across many tilt angles, producing a genuinely 7-dimensional dataset. A proof-of-principle 7D-STEM reconstruction is shown, with two real-space lateral axes, one tomographic depth axis, three diffraction axes, and one energy-loss axis. If the software performs as described, a standard microscope computer can acquire and analyze multi-modal nanoscale data without stitching together separate programs.","feed_headline":"St4DeM makes 7D-STEM acquisition practical inside Digital Micrograph","feed_subtitle":"One suite combines 4D-STEM, EELS/EDS mapping, and tilt tomography into 7D data.","key_machinery":"The load-bearing mechanism is the inverse Radon index map: for each tilt angle, the map stores, at every voxel of the 3D real-space reconstruction, the x-axis index of the 4D-STEM tilt image whose ray passes through that voxel. The software then rotates each 2D diffraction pattern by the goniometer tilt angle and accumulates it into every voxel whose index matches, producing a 3D diffraction volume per voxel (Method 1) or a sparse list of 3D peak coordinates and intensities (Method 2). The real-space volume itself is obtained beforehand from a conventional HAADF tilt series with SIRT, re-projected to form a virtual tilt series, and used to register the 4D-STEM tilt series. No weighting or Fourier filtering is applied during the diffraction back-projection; only nearest-neighbor interpolation is used, deliberately, because the authors note that interpolating between adjacent diffraction patterns would mimic dynamical diffraction artifacts.","core_discovery":"The core discovery is a reconstruction pipeline that converts a 4D-STEM tilt series into per-voxel 3D diffraction volumes. After a conventional HAADF tilt series is reconstructed with SIRT to give a real-space volume, that volume is down-sampled and re-projected to create a virtual tilt series, which is used to register the 4D-STEM tilt series. Then, guided by an inverse Radon index map, every 2D diffraction pattern is rotated by the goniometer tilt angle and accumulated along the straight ray through the real-space volume, so each voxel ends up with its own 3D diffraction pattern (Method 1), or with a sparse list of 3D peak coordinates and intensities (Method 2). Adding an EELS energy dimension makes the full dataset 7-dimensional. The authors present this as a proof of principle for the acquisition and analysis software, not as a refined tomographic study of the test samples.","pith_inferences":["If the linear back-projection is replaced by an iterative scheme that enforces consistency between the accumulated diffraction volumes, the 7D reconstruction could tolerate sparser tilt sampling and reduce the missing-wedge artifacts the authors leave uncorrected.","The same inverse-Radon-index machinery could be applied directly to momentum-resolved EELS tilt series, adding a fourth signal dimension without first reducing each tilt image to a scalar.","Online peak prediction combined with double-tilt zone-axis tracking, which the authors propose, could make the acquisition dose-efficient enough for beam-sensitive materials if implemented as a real-time feedback loop.","Benchmarking Method 1 against Method 3, where virtual images are formed from the 4D-STEM cubes and reconstructed with scalar tomography, would quantify how much fidelity is lost or gained in the straight-ray diffraction projection."],"forward_implications":["A single acquisition session can output a 7D dataset, combining two real-space axes, one tilt/depth axis, three diffraction axes, and one energy axis, from standard microscope hardware without a dedicated direct detector.","The sparse peak-list reconstruction (Method 2) makes 6D and 7D analysis practical on a laptop: days instead of roughly two months and about 1 GB instead of nearly 4 TB of data.","Virtual 3D bright-field, dark-field, and chemical maps can be computed after the fact from the reconstructed volume, so one experiment serves multiple contrast mechanisms.","The alignment strategy, re-projecting a SIRT real-space volume to register the 4D-STEM tilt series, can be reused for other multi-modal tilt series.","Because the suite runs inside the Digital Micrograph environment, it lowers the barrier for laboratories already using that platform to adopt 4D-STEM tomography and spectrum imaging."],"supporting_citations":[{"why":"Defines the 4D-STEM acquisition family whose data structure the suite records.","marker":"[5]"},{"why":"Supplies the Digital Micrograph scripting environment in which the suite is implemented.","marker":"[17]"},{"why":"Provides the iDPC implementation that the suite's phase-contrast analysis follows.","marker":"[26]"},{"why":"Supplies the weighted back-projection formalism that the 6D/7D reconstruction adapts.","marker":"[27]"},{"why":"Provides the 3D electron diffraction analogy that justifies accumulating diffraction patterns into a 3D volume.","marker":"[28]"},{"why":"Used for HAADF tilt-series processing, SIRT reconstruction, and creating the virtual tilt series for registration.","marker":"[29]"},{"why":"Used for registering the 4D-STEM tilt series to the re-projected real-space volume.","marker":"[30]"},{"why":"Used for processing the multi-dimensional datasets in the reconstruction workflow.","marker":"[32]"},{"why":"Used for diffraction peak finding to build the sparse peak-list reconstruction.","marker":"[33]"}],"fun_headline_variants":["St4DeM turns 4D-STEM tilt series into per-voxel 3D diffraction volumes","St4DeM integrates 4D-STEM, EELS, and tilt into 7D data","Per-voxel diffraction volumes from 4D-STEM tilt series","St4DeM: 7D acquisition and analysis in Digital Micrograph","Software suite St4DeM delivers 7D-STEM data with per-voxel diffraction"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 7D reconstruction assumes each measured 2D diffraction pattern can be back-projected along a straight ray through a coarsely sampled volume with no correction for dynamical diffraction, no treatment of the missing wedge, and only nearest-neighbor interpolation, and this assumption is never checked against a known structure.","fun_headline_variants_meta":{"raw":{"variants":["St4DeM turns 4D-STEM tilt series into per-voxel 3D diffraction volumes","St4DeM integrates 4D-STEM, EELS, and tilt into 7D data","Per-voxel diffraction volumes from 4D-STEM tilt series","St4DeM: 7D acquisition and analysis in Digital Micrograph","Software suite St4DeM delivers 7D-STEM data with per-voxel diffraction"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000726,"raw_usage":{"total_tokens":3211,"prompt_tokens":858,"completion_tokens":2353,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":474,"completion_tokens_details":{"reasoning_tokens":2236}},"tokens_in":474,"tokens_out":2353,"duration_ms":15571,"temperature":1.0,"reasoning_tokens":2236,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:43:34.550123+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Record the same 4D-STEM tilt series from a crystal whose structure and orientation are already known, apply Method 1 or Method 2, and compare the reconstructed per-voxel diffraction volumes or peak lists with multislice or Bloch-wave simulations of the same tilt geometry; if peak positions or relative intensities differ beyond interpolation error, the straight-ray back-projection is not faithful.","supporting_citations":[{"cited_title":"Cautaerts , author P","cited_arxiv_id":null,"evidence_quote":"Used for diffraction peak finding to build the sparse peak-list reconstruction."},{"cited_title":"Ophus ,\\ @noop journal journal Microscopy and Microanalysis \\ volume 25 ( year 2019 ) NoStop","cited_arxiv_id":null,"evidence_quote":"Defines the 4D-STEM acquisition family whose data structure the suite records."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Digital Micrograph scripting environment in which the suite is implemented."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the iDPC implementation that the suite's phase-contrast analysis follows."},{"cited_title":"Radermacher ,\\ @noop title Weighted Back-projection Methods \\ ( publisher Springer, New York ,\\ year 2007 ) NoStop","cited_arxiv_id":null,"evidence_quote":"Supplies the weighted back-projection formalism that the 6D/7D reconstruction adapts."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the 3D electron diffraction analogy that justifies accumulating diffraction patterns into a 3D volume."},{"cited_title":"Schwartz , author C","cited_arxiv_id":null,"evidence_quote":"Used for HAADF tilt-series processing, SIRT reconstruction, and creating the virtual tilt series for registration."},{"cited_title":"Yaniv , author B","cited_arxiv_id":null,"evidence_quote":"Used for registering the 4D-STEM tilt series to the re-projected real-space volume."},{"cited_title":"de la Peña , author V","cited_arxiv_id":null,"evidence_quote":"Used for processing the multi-dimensional datasets in the reconstruction workflow."}],"review_version":1}