{"id":"ec942f2b-dde0-4efd-a04d-055f7dead46f","arxiv_id":"1908.08436","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Sparse Field Focusing measures a small set of transmission matrix rows in the Fourier plane and superposes them with adjustable phases to scan a focus through scattering media beyond the optical memory effect range.","lead":"This paper demonstrates a new technique, Sparse Field Focusing, that scans a focused light spot through a strongly scattering material by measuring only a small subset of the transmission matrix. It offers a faster way to steer a corrected focus in opaque media where the optical memory effect is too small to be useful.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The enhancement and gain formulas (Eqs. 2-3) rest on an unverified orthogonality of the M target-optimized wavefronts (Eq. 6); if target spacing is below the speckle size, the claimed gain of 6.0 is not established.","rationale":"The reader's weakest-assumption analysis correctly identifies the orthogonality of the M target-optimized wavefronts as the load-bearing step. The paper's derivation of Eq. (2) explicitly depends on this orthogonality in the normalization prefactor D of Eq. (6), yet the experimental section does not provide the target spacings or any direct measurement of the independence of the 500 targets. The claim that each target is smaller than the speckle size addresses target resolution, not target separation, so it does not by itself establish that the optimized wavefronts are orthogonal. If the targets are spaced closer than the speckle correlation length, the enhancement curve in Fig. 3(d) would overestimate the number of usable scan points, and the central gain of 6.0 would need to be revised. At the same time, the empirical demonstration that a focus can be formed and scanned over 30 µm is credible and does not collapse under this concern; the issue is specifically with the quantitative efficiency claim and the theoretical model. The reader's verdict of CONDITIONAL remains appropriate, so no change to the verdict is recommended.","tokens_in":6617,"tokens_out":22124,"duration_ms":247483,"concrete_test":"Report the 500 target coordinates (for example, as a histogram of pairwise separations) and recompute the normalized overlap matrix of the corresponding TM rows. For each target m, form the optimized input vector u_m from the measured partial TM or from the recorded SLM patterns, and compute C_ml=|⟨u_m|u_l⟩|^2/(||u_m||^2||u_l||^2) for all m≠l. If the average off-diagonal overlap is not much less than 1, the orthogonality assumption fails. As an independent experimental check, run the SFF superposition with M=1,2,4,8,...,500 targets at the same total power and verify that the central enhancement grows linearly in M; sublinear growth would confirm the assumption fails and would require re-evaluating the stated gain.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation (6) of Appendix A sets the normalization D=(N M Σ_b⟨|t_ba|^2⟩)^{-1/2} by assuming that the M target-optimized input fields are mutually orthogonal. This is the step that converts the coherent superposition into the enhancement formula η=|γ|^2 M(N−1)F/N_s+1 (Eq. 2) and hence into the gain A=n/M (Eq. 3). The manuscript states only that each of the 500 targets is smaller than the 30 µm speckle; it reports neither the target separations nor the Gram matrix of the optimized wavefronts. If many target separations fall below the speckle correlation length, the input-space overlaps ⟨u_m|u_l⟩ are not δ_{ml}, so the normalization prefactor D is too large by an amount depending on Σ_{m≠l}⟨u_m|u_l⟩. The predicted enhancement for a given M then becomes an upper bound rather than the true value. The measured agreement with Eq. (2) cannot fully resolve this issue because F(x,y) is taken from the measured diffuse envelope and |γ|^2=0.54 is effectively determined from experiment, making the comparison in Fig. 3(d) partly a consistency check. Since the headline quantitative result (6.0±0.8 scan points per measured row) is obtained by thresholding the measured enhancement, a failure of orthogonality would lower the usable n and the gain, directly weakening the central efficiency claim of the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces Sparse Field Focusing (SFF), a wavefront-shaping method for focusing and scanning light through a strongly scattering medium without relying on the optical memory effect. Instead of measuring the full transmission matrix, the authors optimize a set of M target spots in the back focal plane, measure the relative phases among the resulting optimized wavefronts, and superpose them to form a focus in the image plane. By applying phase ramps to the individual wavefronts, the focus can be scanned over a range of about 30 micrometers through a zinc-oxide sample whose isoplanatic patch is stated to be below 0.38 micrometers. Experimental results show a focus 47 times brighter than the background speckle, and a gain of 6.0 +/- 0.8 scan points per measured matrix row. An analytical model in Appendix A predicts the enhancement as a function of scan position, Eq. (2), and is compared with the measurements in Fig. 3(d).","tokens_in":6958,"tokens_out":8612,"duration_ms":91934,"significance":"If the method holds up, it is a practically useful alternative to full transmission-matrix measurement in regimes where the optical memory effect is too small to exploit, and it may be adaptable to guide-star-based imaging. The paper's strengths are a clear experimental demonstration, a quantitative model, and a well-defined efficiency metric (gain). The central experimental fact, that a focus can be scanned well beyond the isoplanatic patch by superposing a subset of optimized wavefronts, is convincingly shown. However, the theoretical model rests on an orthogonality assumption for the optimized wavefronts that is not verified in the manuscript, and the agreement with Eq. (2) is partly a consistency check because the diffuse intensity envelope and the fidelity are extracted from the same sample. These issues affect the strength of the quantitative efficiency claim but not the basic feasibility of the method.","major_comments":[{"comment":"The derivation of the enhancement formula assumes that the M target-optimized input fields are mutually orthogonal, which sets the normalization D = (N M sum_b <|t_ba|^2>)^(-1/2). The manuscript states only that each target is smaller than the 30-micrometer speckle in the back focal plane; it does not report the separations between the 500 targets or the Gram matrix of the optimized wavefronts. If many target separations are smaller than the speckle correlation length, the overlaps <u_m|u_l> are not delta_ml, the normalization is incorrect, and Eq. (2) becomes an upper bound rather than a prediction. This is load-bearing for the model-experiment agreement in Fig. 3(d) and for the interpretation of the gain A as a reduction in measurement effort relative to full TM measurement. Please report the target positions and separations and either compute the Gram matrix of the optimized fields from the measured rows or provide another direct test of the orthogonality assumption.","section":"Appendix A, Eq. (6)"},{"comment":"The claim that 'we obtained the full transmission matrix by performing 6 times less measurement than measuring the full transmission matrix' is not precisely defined. The gain A = n/M is a measured ratio, but its interpretation as a factor of reduction in measurement effort depends on what is meant by measuring the full transmission matrix: the number of input modes, output modes, and phase-stepping steps all matter. It also depends on the M measured rows being independent, which connects to the orthogonality issue above. Please specify the reference full-TM measurement cost and clarify how the number of individual scan points, n = 2984 +/- 390, is obtained from the two-dimensional scan.","section":"Section 'In our experiment' (p. 4)"},{"comment":"The experimental protocol for the two-dimensional scan is underdocumented. The text describes a vertical scan and Fig. 3(d) shows enhancement along one axis, but the two-dimensional scan that yields 2984 +/- 390 individual scanning points is not described. It is unclear how the scan grid is chosen, how the enhancement threshold of 10 is applied, and how the 30-micrometer scan range is defined given that the enhancement decreases with displacement. Please provide the 2D enhancement map or a precise description of the scanning and thresholding procedure, including how the uncertainty in n is estimated.","section":"Section 'Next, we tried to test...' and Fig. 3(d)"}],"minor_comments":[{"comment":"The plot would benefit from error bars and a statement of how many measurements each point represents; without them, the agreement between experiment and the model cannot be fully assessed.","section":"Fig. 3(d)"},{"comment":"The definition of the 30-micrometer scan range should be stated explicitly; it is not clear whether this is the full range over which an enhancement above a threshold is maintained, or the range over which the focus can be visually tracked.","section":"Section 'Next, we tried to test...'"},{"comment":"The value |gamma|^2 = 0.54 is said to be the average fidelity, but the text does not report how the SNR was measured for the 500 targets or how the average was taken; please add this detail.","section":"Section 'We present an analytical model...' and Eq. (7)"},{"comment":"The model treats the optimized incident field as a complex field with a normalization factor D, but the experiment uses a phase-only SLM; the text should clarify how the phase-only constraint is represented in the complex-field model, particularly when the M wavefronts are superposed by displaying only the phase of the sum.","section":"Appendix A, Eq. (6)"},{"comment":"The statement that the isoplanatic patch is less than 0.38 micrometers is not derived or measured in the body; please indicate whether this is an estimate from the sample parameters or a measured quantity, and if measured, how.","section":"Summary and body"},{"comment":"The inset showing the 500 target locations would be more informative with a scale bar or axis labels so that the separations between targets can be evaluated by the reader.","section":"Fig. 2 inset"},{"comment":"The reference to 'Visualization 1' should be accompanied by a description in the text, since the video file is not part of the manuscript and the reader cannot otherwise assess the scanning behavior.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The experimental demonstration is solid and likely worth publishing after revision. The main technical risk is the unverified orthogonality assumption in Eq. (6); I would like the editor to require the authors to provide either the target separation statistics or a direct computation of the overlaps from the measured rows. The paper would also be strengthened by a clearer definition of the full-TM measurement cost used in the gain comparison. The topic fits the journal's scope and the result is of interest to the wavefront-shaping community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the two-sentence version. The idea is to measure a small set of transmission-matrix rows corresponding to Fourier-plane targets, then superpose those optimized wavefronts with programmable phase offsets to scan the focus independent of the optical memory effect. That is a real and useful contribution, and the experiments show a clean 30-micrometer scan through a strongly scattering zinc-oxide layer, far beyond the ~0.38-micrometer isoplanatic patch.\n\nWhat earns credit: the method genuinely breaks the memory-effect limit without requiring the full transmission matrix; the measured focus is 47x above speckle and near diffraction-limited; the scan behavior follows the diffuse envelope as expected; and the write-up is honest about many practical details. The Appendix A model is a reasonable Gaussian-TM treatment.\n\nThe soft spots are real but not fatal. The derivation's normalization D in Eq. (6) assumes the M target-optimized wavefronts are mutually orthogonal. The paper reports each target is smaller than a speckle but never reports the target separations, so we cannot check independence. If targets overlap in the Fourier plane, the input-space overlap is nonzero, D is too large, and the enhancement formula becomes an upper bound. Since the headline gain of 6.0 +/- 0.8 is obtained by thresholding that enhancement, the central efficiency claim is not fully established. At minimum the authors need to show the Gram matrix or target separations, or soften the claim to a lower bound. Relatedly, the comparison to 'full transmission matrix' is not direct: they never measured a full TM, only asserted similar resolution, and F(x,y) is taken from the measured envelope, making the fit in Fig. 3(d) partly a consistency check. No data or code is provided, which is a minor issue for an experimental techniques paper but worth noting.\n\nWho this is for: wavefront shaping and deep-tissue imaging people. It deserves a serious referee: the concept is new and the experiment is believable, even if the quantitative gain needs to be pinned down. I would send it out, with a request for the orthogonality check and a more measured abstract.","headline":"New trick for scanning a focus through scattering media, demonstrated convincingly, but the efficiency gain number is weaker than advertised without a check on target orthogonality.","tokens_in":7450,"tokens_out":2672,"would_cite":true,"duration_ms":28004,"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":"A wavefront-shaping approach called Sparse Field Focusing forms a focus behind a strongly scattering medium and scans it over tens of micrometers by measuring only a subset of the transmission matrix, bypassing both the optical memory…","keywords":["wavefront shaping","transmission matrix","sparse field focusing","optical memory effect","scattering media","focus scanning","isoplanatic patch","enhancement"],"falsifier":"Measure the transmission-matrix rows for pairs of targets separated by less than one speckle diameter (30 µm in this setup) and compute their overlap or the singular-value spectrum of the 500-target subset; if the rows are not approximately orthogonal, the enhancement will deviate from Eq. (2) and the gain of 6.0 ± 0.8 will not be reproducible.","tokens_in":6445,"feed_emoji":"🔬","tokens_out":6728,"duration_ms":56275,"temperature":0.7,"pith_summary":"This paper introduces Sparse Field Focusing (SFF), a wavefront-shaping method that forms and scans an optical focus through a strongly scattering medium without relying on the optical memory effect and without measuring the full transmission matrix. Instead of optimizing a new wavefront for each focus position, SFF optimizes a set of M wavefronts, each of which emerges as a plane wave, then superposes them and scans the focus by applying relative phase shifts. The authors demonstrate a focus scanned over roughly 30 micrometers through a zinc-oxide layer whose isoplanatic patch is below 0.38 micrometers, with about 6.0 ± 0.8 resolvable scan points per measured transmission-matrix row. The measured enhancement matches an analytical model in which the focus intensity follows the diffuse background intensity distribution. This matters because full transmission-matrix measurements are slow and memory-effect range is often tiny in biological tissue.","feed_headline":"Sparse wavefront method scans a focus 30 µm through scattering media","feed_subtitle":"Focus can be moved over 30 µm even when the memory-effect range is below 0.38 µm.","key_machinery":"The central object is the sparse set of optimized wavefronts, i.e., M rows of the transmission matrix measured instead of the full matrix. Each optimized wavefront is shaped so that, after scattering, it produces a plane wave with constant phase across the image plane; the superposition of M such plane waves creates a focus by constructive interference. Scanning works by applying a linear phase ramp across the target positions in the back focal plane, adding $\\phi_m = k_{mx}\\Delta x + k_{my}\\Delta y$ to each wavefront; because the phase shift is global across the spatial light modulator for each wave, the wave's propagation through the medium is unchanged. The analytical model predicts enhancement $\\eta(x,y) = |\\gamma|^2 M(N-1)/N_s \\, F(x,y) + 1$, where $|\\gamma|^2$ is the wavefront-shaping fidelity, $N$ the number of controlled segments, $N_s$ the number of speckles in the background, and $F$ the normalized ensemble-averaged intensity distribution; the model shows enhancement follows the diffuse intensity envelope, defining the scan range. The prefactor $D$ in the derivation normalizes the superposed incident intensity by assuming the M target modes and N segments are orthogonal.","core_discovery":"The central claim is that scanning a corrected focus through a scattering medium does not require the memory effect or a full transmission matrix; a sparse subset of rows suffices. The authors construct a focus by optimizing M incident wavefronts so each produces an apodized plane wave at the image plane, measuring their relative phases interferometrically, and coherently superposing them. Shifting the relative phases by $\\phi_m = k_{mx}\\Delta x + k_{my}\\Delta y$ moves the constructive-interference peak to $(\\Delta x, \\Delta y)$, because each shaped wave propagates through the medium in the same way and only its overall phase changes. In the experiment, 500 optimized waves produced a focus 47 times brighter than the unshaped speckle, and the focus could be scanned over 30 µm, while tilting or shifting the incident wavefront (the memory-effect approach) did not move it. The paper claims this yields a gain of 6.0 ± 0.8 scan points per measured matrix row, meaning roughly six times fewer measurements than a full transmission matrix for the same number of focus positions.","pith_inferences":["If the target rows are independent, the same superposition principle should extend to three-dimensional scanning by applying phase ramps along additional axes, so the number of measurements would scale with the number of distinct output speckles rather than the number of scan positions.","A direct test of the orthogonality assumption would be to place feedback targets closer than one speckle diameter and measure the gain; the model predicts the gain should drop, which would give design rules for target spacing.","Because the method controls only relative phases, it might be combined with polarization or wavelength multiplexing to scan multiple foci simultaneously, an extension the paper does not discuss.","The scan range is ultimately bounded by the diffuse intensity envelope, so engineering the illumination or the sample's angular spread could extend the range well beyond the demonstrated 30 µm."],"forward_implications":["Imaging through scattering media can be performed with sparse transmission-matrix measurements, reducing acquisition time and memory requirements relative to full-matrix methods.","The scan range is set by the diffuse intensity envelope rather than by the isoplanatic patch, so the method applies to samples such as deep biological tissue where the memory effect is too small to use.","There is a quantifiable trade-off: increasing the number M of optimized wavefronts raises the focus enhancement but lowers the gain A = n/M, so the method can be tuned for either brightness or speed.","The focus resolution approaches the diffraction limit (measured FWHM 0.520 µm versus 0.407 µm theoretical), so sparse sampling does not sacrifice spatial resolution for scan range.","Adapting SFF to embedded guide stars could enable focusing inside scattering media at positions where a guide star is not available, as the paper envisions."],"supporting_citations":[{"why":"Establishes that an optimized wavefront can focus light through scattering media, the foundation of SFF's per-wavefront shaping.","marker":"[1]"},{"why":"Defines the optical memory-effect range and its limitations in deep tissue, the constraint SFF bypasses.","marker":"[4]"},{"why":"Introduces full transmission-matrix measurement for arbitrary focusing, the baseline that SFF compares against.","marker":"[9]"},{"why":"Supplies the stepwise sequential algorithm used to optimize each of the 500 target wavefronts.","marker":"[15]"},{"why":"Provides the fidelity parameter |γ|² for phase-only modulation, which enters the enhancement model.","marker":"[17]"},{"why":"Gives the transport mean free path of similar zinc-oxide samples, establishing the sample is optically thick.","marker":"[14]"}],"fun_headline_variants":["Sparse matrix method scans focus without memory effect","Beating memory effect limits: sparse wavefront scans focus","Sparse wavefront method: focus scan without memory effect or full TM","No memory effect needed: sparse wavefront scanning through scatter","Sparse matrix rows instead of full: scan focus 30 µm through scatter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation assumes that the M targets used for feedback are spaced far enough apart that their transmission-matrix rows are independent and orthogonal; if two targets lie within the same speckle, the superposition formula and the gain estimate no longer hold.","fun_headline_variants_meta":{"raw":{"variants":["Sparse matrix method scans focus without memory effect","Beating memory effect limits: sparse wavefront scans focus","Sparse wavefront method: focus scan without memory effect or full TM","No memory effect needed: sparse wavefront scanning through scatter","Sparse matrix rows instead of full: scan focus 30 µm through scatter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000579,"raw_usage":{"total_tokens":2727,"prompt_tokens":939,"completion_tokens":1788,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":555,"completion_tokens_details":{"reasoning_tokens":1702}},"tokens_in":555,"tokens_out":1788,"duration_ms":11634,"temperature":1.0,"reasoning_tokens":1702,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:42:37.146764+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the transmission-matrix rows for pairs of targets separated by less than one speckle diameter (30 µm in this setup) and compute their overlap or the singular-value spectrum of the 500-target subset; if the rows are not approximately orthogonal, the enhancement will deviate from Eq. (2) and the gain of 6.0 ± 0.8 will not be reproducible.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that an optimized wavefront can focus light through scattering media, the foundation of SFF's per-wavefront shaping."},{"cited_title":"Judkewitz, R","cited_arxiv_id":null,"evidence_quote":"Defines the optical memory-effect range and its limitations in deep tissue, the constraint SFF bypasses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces full transmission-matrix measurement for arbitrary focusing, the baseline that SFF compares against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the stepwise sequential algorithm used to optimize each of the 500 target wavefronts."},{"cited_title":"Yılmaz, W","cited_arxiv_id":null,"evidence_quote":"Provides the fidelity parameter |γ|² for phase-only modulation, which enters the enhancement model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the transport mean free path of similar zinc-oxide samples, establishing the sample is optically thick."}],"review_version":1}