_GaAs −
_AlGaAs| + ε), Eq. (3), is built from mean COMs in the two field-free regions, and the same field-free regions are then used to evaluate whether the weighted potential is flat. For small ε, tilts with large field-free-region COM mismatch are strongly downweighted, so the weighted average at the AlGaAs field-free region is biased toward the GaAs reference regardless of the true field. The stated choice ε≈1 µrad is selected after seeing the data to make these regions agree. Thus the observed flattening of the AlGaAs side is partly a construction of the weighting scheme, not an independent prediction. The further claim that the processed profile almost exactly matches the off-axis profile is qualitative: no error bars, no uncertainty propagation, and only one line scan per orientation are shown. In addition, the zone-axis case, where no weighting improves on simple averaging, shows that the method does not generically improve 'strong' diffraction conditions, so the conclusion overstates the regime of demonstrated improvement. The acquisition concept and calibration work are sound, but the flagship quantitative improvement needs validation with fixed analysis parameters and independent evaluation regions.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript introduces a sequential-tilting variant of momentum-resolved STEM in which a full diffraction pattern is acquired for each beam tilt and scan position, yielding 6D datasets. The authors describe a two-step calibration procedure for beam tilt, detilt, and tilt-induced beam shift, and demonstrate residual diffraction shifts of about 4 µrad and beam shifts of about 1 nm. They use the per-tilt diffraction information to construct virtual LACBED patterns and per-tilt residual-COM maps, and they propose weighted tilt-averaging schemes, including a residual-COM weight in Eq. (3), to suppress dynamical diffraction artifacts. The method is tested on an AlGaAs/GaAs heterojunction at three specimen orientations: off-axis, systematic-row, and zone-axis. The main claim is that sequential tilting MRSTEM improves field mapping under medium and strong diffraction conditions, with the systematic-row orientation as the key demonstration.","tokens_in":10505,"tokens_out":6931,"duration_ms":65977,"significance":"If the improvement is real, the acquisition concept is a valuable, hardware-free extension of precession MRSTEM, and the per-tilt diagnostics (vLACBED continuity, residual-COM maps, tilt-dependent total intensity) are genuinely useful for data-quality assessment. The calibration results are a clear strength: the residual tilt-induced diffraction shift of about 4 µrad and the tilt-induced beam shift of about 1 nm, smaller than the probe size, are quantitative and convincing. The claim of improved field mapping under dynamical diffraction, however, rests on a weighting scheme whose construction shares the same field-free regions used for evaluation and on a regularization parameter chosen after seeing the data. The zone-axis result further limits the demonstrated regime. The paper is therefore a solid methods contribution whose central quantitative improvement still needs independent validation.","major_comments":[{"comment":"The residual-COM weight a(K) = 1/(|
_GaAs -
_AlGaAs| + epsilon) is constructed from the same field-free regions on the two sides of the junction that are later used to judge the flatness of the systematic-row potential. For small epsilon, this weighting selectively suppresses tilts for which the AlGaAs field-free COM differs from the GaAs reference, so the weighted average in the AlGaAs field-free region is biased toward the GaAs value. Consequently, the observed flattening of the AlGaAs side in Fig. 5h is partly a consequence of the weighting scheme rather than an independent confirmation that the diffraction artifact has been removed. Please validate the approach with weights fixed on one scan or region and evaluated on an independent scan or region, or with a hold-out field-free region.","section":"Sec. 3.3, Eq. (3), Fig. 5h"},{"comment":"The paper does not specify how epsilon ≈ 1 µrad is selected, and the phrase \"with the proper choice of epsilon\" indicates a post hoc choice made after inspecting the data to make the field-free regions agree. Because the unweighted and intensity-weighted systematic-row profiles in Fig. 5b and 5e show a clear linear artifact on the AlGaAs side, the claimed improvement is contingent on this tuned parameter. Please provide an a priori or cross-validated selection rule for epsilon, and show a quantitative flatness metric (for example, the residual slope in the field-free AlGaAs region) as a function of epsilon so the sensitivity of the conclusion can be assessed.","section":"Sec. 3.3, Fig. 5h"},{"comment":"The conclusion that \"for medium and strong diffraction conditions, we improved the sequential tilting MRSTEM results\" overstates the demonstrated regime. For the zone-axis orientation, the authors themselves show that no weighting improves on simple averaging and that the residual-COM weighting with large epsilon reduces to equal weights. Thus the demonstrated improvement is limited to one systematic-row orientation with a tuned weight. Please narrow the claim accordingly or add additional medium- and strong-diffraction examples, otherwise the central claim is broader than the evidence.","section":"Sec. 4 (Conclusion), Sec. 3.3, Fig. 5i"}],"minor_comments":[{"comment":"There are several typos: \"approache\" at the end of Sec. 1, \"high-resoltuion\" in Sec. 2.2, and \"convetional\" in Sec. 3.2.","section":"Sec. 1, Sec. 2.2, Sec. 3.2"},{"comment":"The regularization constant epsilon is quoted in µrad, but the denominator is a COM difference, which is in momentum or detector units; please specify the units of
and the conversion used so that Eq. (3) is dimensionally consistent.","section":"Eq. (3)"},{"comment":"The field-free region on the AlGaAs side used to define
_AlGaAs in Eq. (3) is not specified in the text; please state its coordinate range or mark it in Fig. 4, just as the GaAs reference region is described as -30 to -20 nm.","section":"Eq. (3), Fig. 4"},{"comment":"No error bars or repeat measurements are shown for the potential profiles, despite the 30 nm position averaging perpendicular to the scan direction; a statement of the measurement uncertainty or noise floor would make the quantitative comparison to the expected potential more convincing.","section":"Fig. 5"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a honest methods paper that deserves a referee, but the flagship quantitative claim is sturdier on paper than in the data.\n\nThe acquisition concept is real and useful: sequential tilting with a full diffraction pattern per tilt and scan position is a natural generalization of tilt-scan DPC, and the per-tilt information enables vLACBED-based calibration and data-quality assessment. The calibration results are genuinely good: residual diffraction shift around 4 micro rad (about a quarter pixel) and tilt-induced beam shift under 1 nm, below the probe size. The vLACBED diagnostic for checking orientation and calibration continuity is a nice addition. The paper also cites the prior tilt-scan DPC work fairly, and the experimental setup is described in enough detail to reproduce in principle.\n\nWhere it softens is exactly where your stress-test lands. Equation (3) constructs weights from the COM difference between the field-free GaAs and AlGaAs regions, and the same regions are then used to judge whether the weighted potential is flat. With epsilon tuned after inspecting the data to about 1 micro rad, the flattening is partly a construction of the weighting, not an independent prediction. The claim that the processed potential \"almost exactly matches\" the off-axis profile is qualitative: one line scan per orientation, no error bars, no uncertainty propagation. And the conclusion overreaches slightly: for the zone-axis orientation, the data show no weighting improves on simple averaging, so the demonstrated improvement is for the systematic-row case, not for \"medium and strong\" diffraction conditions generally.\n\nThat said, these issues do not sink the acquisition concept. The calibration methodology and the vLACBED quality metric are contributions worth publishing on their own, and the residual-COM weighting is a reasonable heuristic even if the current validation is not fully independent. The thickness inference and the constant-field-along-beam assumption are conventional in the field and not the main concern.\n\nBottom line: I would send this to peer review, not desk reject. The paper deserves a serious referee, but I would ask for validation with fixed analysis parameters and independent evaluation regions, uncertainty quantification, and a toned-down conclusion. If I were working on 4D-STEM junction mapping, I would cite it for the calibration and the vLACBED idea.","headline":"Solid acquisition/calibration work with a clear caveat: the systematic-row improvement rests on a partially circular weighting scheme and post hoc choice of epsilon, so the strongest claim needs validation.","tokens_in":11075,"tokens_out":3322,"would_cite":true,"duration_ms":32050,"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":"Recording a full diffraction pattern for each beam tilt lets users detect and downweight beam directions corrupted by dynamical diffraction, improving nanoscale electric-field maps across junctions.","keywords":["4D-STEM","momentum-resolved STEM","differential phase contrast","precession electron diffraction","dynamical diffraction","electric field mapping","AlGaAs/GaAs heterojunction","center-of-mass analysis"],"falsifier":"Take the same AlGaAs/GaAs lamella in the systematic-row orientation and split the field-free AlGaAs side into two disjoint regions, using one to build the residual-center-of-mass weights and the other only to evaluate the resulting potential profile; if the held-out region still shows a linear slope, the flattening is produced by the tuning of $\\epsilon$ rather than by a genuine suppression of dynamical diffraction.","tokens_in":10073,"feed_emoji":"🔬","tokens_out":7060,"duration_ms":65278,"temperature":0.7,"pith_summary":"This paper tries to show that momentum-resolved scanning transmission electron microscopy (MRSTEM) can map nanoscale electric fields more reliably if, instead of continuously precessing the beam and averaging diffraction patterns, the beam is tilted through a sequence of discrete directions and a full diffraction pattern is saved for every tilt at every scan position. The extra information, which beam directions suffer from dynamical diffraction, can then be used to reweight or discard corrupted measurements. On AlGaAs/GaAs heterojunctions the authors show that for a systematic-row crystal orientation, weighting beam tilts by the residual center-of-mass shift of field-free regions removes a spurious sloping potential and reproduces the profile obtained under weak diffraction conditions. The significance is that the improvement is inaccessible to prior MRSTEM techniques that only record an averaged diffraction pattern, and the approach requires no hardware modification.","feed_headline":"Per-tilt diffraction data fix STEM field maps under strong diffraction","feed_subtitle":"Recording a full diffraction pattern for every tilt lets users discard directions corrupted by dynamical diffraction.","key_machinery":"The central object is the six-dimensional sequential-tilting data set, a full diffraction pattern at each scan position and each incident-beam tilt. From it the paper constructs a virtual large-angle convergent beam electron diffraction (vLACBED) pattern by shifting each tilt's diffraction pattern by its tilt angle, which visualizes HOLZ lines and Bragg deficiency lines and checks calibration, and a position- and tilt-resolved residual center of mass $\\delta C(r,K)$ that isolates center-of-mass changes due to fields from those due to changing diffraction conditions. The decisive processing step is a tilt-dependent weighting $a(r,K)$; intensity weighting reproduces the implicit weighting of precession MRSTEM, while residual-center-of-mass weighting $a(K)=1/(|\\langle p\\rangle_{\\mathrm{GaAs}}-\\langle p\\rangle_{\\mathrm{AlGaAs}}|+\\epsilon)$ suppresses tilts whose field-free center of mass differs strongly across the junction.","core_discovery":"The paper claims that acquiring a complete six-dimensional data set, a full diffraction pattern at every scan position for every beam tilt, makes MRSTEM field mapping more reliable under dynamical diffraction than precession-averaged MRSTEM, because the per-tilt information reveals which beam directions are corrupted and allows them to be downweighted or excluded. Using AlGaAs/GaAs heterojunctions, it demonstrates that with no weighting or with intensity weighting, the systematic-row orientation shows a spurious linear potential slope on the AlGaAs side, whereas weighting by the residual center of mass of field-free regions, with a regularization parameter near one microradian, removes that slope and reproduces the off-axis potential profile. For the zone-axis orientation, no weighting scheme improves on simple averaging because all tilts are strongly affected, and intensity weighting fails because the correlation between direct-disc intensity and dynamical artifact is absent. The contribution is therefore not a new detector or hardware but a data-acquisition and post-processing strategy that exploits the per-tilt diffraction information.","pith_inferences":["An extension of the paper's logic suggests that on samples without clear field-free regions on both sides of a junction, the same residual-center-of-mass weighting could be built from a robust local baseline or an iterative estimation procedure, though the paper does not develop this.","Because the data set contains the full diffraction pattern per tilt, one could go beyond scalar center-of-mass weights and use pattern-level dissimilarity or learned classifiers to identify dynamical diffraction corruption without assuming field-free regions, an avenue the paper leaves open.","The data-dependent choice of the regularization parameter implies that a validation protocol with held-out field-free regions would strengthen the method before it becomes a turnkey quantitative tool for arbitrary samples.","The same six-dimensional acquisition could, in principle, separate electric and magnetic field contributions by exploiting symmetry of the center-of-mass signal under tilt reversal, which the paper does not address."],"forward_implications":["For weak diffraction conditions, such as the off-axis orientation, sequential tilting MRSTEM gives reliable potential profiles under all tested weighting schemes, validating the technique as a baseline.","For medium diffraction conditions, such as the systematic-row orientation, per-tilt residual-center-of-mass weighting can recover a potential profile that matches the off-axis result, something not accessible to precession MRSTEM with only averaged intensity information.","Intensity weighting helps in the systematic-row geometry because direct-disc intensity decreases for tilts with strong diffraction, but in the zone-axis geometry it does not help because that correlation is absent.","The acquisition concept requires no hardware modification and can be transferred to other 4D-STEM applications such as strain, domain, or magnetic field mapping.","The calibration procedure reduces residual tilt-induced diffraction shift to about four microradians and tilt-induced beam shift to about one nanometer, smaller than the probe size, enabling precise arbitrary tilt patterns."],"supporting_citations":[{"why":"Supplies the precession-DPC method whose averaged diffraction pattern lacks the per-tilt information the paper exploits.","marker":"[29]"},{"why":"Establishes 4D-STEM with precession for measuring heterointerface potential drops, the comparison standard for the improved processing.","marker":"[26]"},{"why":"Demonstrates quantitative MRSTEM field mapping on the nanometer scale using center-of-mass analysis, the technique whose reliability is being improved.","marker":"[24]"},{"why":"Provides another precession-based electric field measurement baseline that sequential tilting is designed to outperform.","marker":"[23]"},{"why":"Gives the relation between the center of mass of the diffraction pattern and electric fields in the sample.","marker":"[16]"},{"why":"Grounds the conversion from center of mass to projected potential gradient used for the final potential profiles.","marker":"[17]"},{"why":"Supplies the precession path segmentation approach adapted here to calibrate tilt-induced beam shift.","marker":"[37]"},{"why":"Provides the inelastic mean free path used to estimate lamella thickness for converting center of mass to electric field.","marker":"[38]"}],"fun_headline_variants":["Sequential tilts beat precession for STEM field mapping","Tilt-by-tilt diffraction rescues STEM field maps","Per-tilt data reveals and corrects STEM artifacts","Per-tilt diffraction data fix STEM field maps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claimed improvement depends on assuming that the shift of the diffraction-pattern center between two field-free regions faithfully measures how badly each beam tilt is corrupted by dynamical diffraction, and on choosing the weighting parameter $\\epsilon\\approx1\\,\\mu\\mathrm{rad}$ from the same data used to judge the result.","fun_headline_variants_meta":{"raw":{"variants":["Sequential tilts beat precession for STEM field mapping","Tilt-by-tilt diffraction rescues STEM field maps","Per-tilt data reveals and corrects STEM artifacts","Per-tilt diffraction data fix STEM field maps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000518,"raw_usage":{"total_tokens":2535,"prompt_tokens":997,"completion_tokens":1538,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":613,"completion_tokens_details":{"reasoning_tokens":1474}},"tokens_in":613,"tokens_out":1538,"duration_ms":11188,"temperature":1.0,"reasoning_tokens":1474,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:43:03.817179+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the same AlGaAs/GaAs lamella in the systematic-row orientation and split the field-free AlGaAs side into two disjoint regions, using one to build the residual-center-of-mass weights and the other only to evaluate the resulting potential profile; if the held-out region still shows a linear slope, the flattening is produced by the tuning of $\\epsilon$ rather than by a genuine suppression of dynamical diffraction.","supporting_citations":[{"cited_title":"Mawson, A","cited_arxiv_id":null,"evidence_quote":"Supplies the precession-DPC method whose averaged diffraction pattern lacks the per-tilt information the paper exploits."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes 4D-STEM with precession for measuring heterointerface potential drops, the comparison standard for the improved processing."},{"cited_title":"Beyer, M","cited_arxiv_id":null,"evidence_quote":"Demonstrates quantitative MRSTEM field mapping on the nanometer scale using center-of-mass analysis, the technique whose reliability is being improved."},{"cited_title":"Bruas, V","cited_arxiv_id":null,"evidence_quote":"Provides another precession-based electric field measurement baseline that sequential tilting is designed to outperform."},{"cited_title":"M¨ uller, F","cited_arxiv_id":null,"evidence_quote":"Gives the relation between the center of mass of the diffraction pattern and electric fields in the sample."},{"cited_title":"Winkler, J","cited_arxiv_id":null,"evidence_quote":"Grounds the conversion from center of mass to projected potential gradient used for the final potential profiles."},{"cited_title":"Nordahl, L","cited_arxiv_id":null,"evidence_quote":"Supplies the precession path segmentation approach adapted here to calibrate tilt-induced beam shift."},{"cited_title":"Iakoubovskii, K","cited_arxiv_id":null,"evidence_quote":"Provides the inelastic mean free path used to estimate lamella thickness for converting center of mass to electric field."}],"review_version":1}