{"id":"b129efd3-830c-488f-9ef3-4600ec5e7e92","arxiv_id":"1908.04180","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"MRFM imaging scans achieve 0.9 nm resolution in one dimension using a hyperbolic-secant spin inversion protocol for a sharply defined slice.","lead":"This paper demonstrates magnetic resonance force microscopy (MRFM) scans with a one-dimensional resolution of 0.9 nanometers and a localization precision of 0.6 nanometers, roughly twice as good as prior MRFM records. The improvement comes from a new hyperbolic-secant spin inversion protocol that creates a sharply defined imaging slice, bringing sub-nanometer MRI for biological structures closer to reach.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. 8's slope-to-noise metric yields single-edge localization precision, not resolution; with a fitted edge width of ~10 nm, two features 0.9 nm apart would be unresolved, so the headline resolution claim is not established by the scans.","rationale":"The reader's verdict correctly identifies the weakest assumption. I checked the manuscript's derivation and usage of Eq. 8. The full text defines sigma_x,slope from the maximum slope of the fitted signal onset and the noise variance, then equates it to 'uncertainty-limited spatial resolution.' In estimation theory, this is the asymptotic standard deviation of a maximum-likelihood edge-position estimator; it answers 'how well can we find a single edge?', not 'how close can two features be and still be seen as two?'. The fitted edge width w ~ 10 nm, which is explicitly stated to be dominated by sample/slice convolution, directly contradicts the idea that the point-spread function is sub-nanometer in these scans; a wide PSF blurs two nearby edges into one. Therefore the central claim's strength depends on treating edge-localization precision as resolution. This is common in MRFM literature sometimes, but the abstract and conclusions make a strong milestone claim ('sub-nanometer MRI', 'macromolecular structures') that requires two-point separation. The paper includes enough data (11 scans, slope, noise, fits) to perform the proposed two-edge analysis, so the concern is testable without new experiments. Since the reader already issued CONDITIONAL, I do not adjust the verdict; the condition should explicitly require either two-point resolution evidence or corrected wording.","tokens_in":8930,"tokens_out":4567,"duration_ms":43110,"concrete_test":"Re-analyze the 11 lateral scans with a two-edge model based on Eq. 7: F^2(x) = A*s(x-x0) + B*s(x-x0-d), with s a hyperbolic tangent step, and compare the fit likelihood for d = 0 versus d = 0.9 nm, or compute a profile-likelihood confidence interval for d. If the data cannot exclude d = 0, or if d is only constrained to within a range much larger than 0.9 nm, then Eq. 8 is estimating edge-localization precision, not two-point resolution. An even simpler analytic check: using the reported w ~ 10 nm edge width, the separation of 0.9 nm is 0.09w, far below the Rayleigh-type resolution limit for a 10 nm point-spread function; the claim would require demonstrating that the effective PSF is far sharper than the fitted edge width, e.g., by deconvolving the known sample geometry.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—0.9 nm one-dimensional resolution—rests on Eq. 8: sigma_x,slope = F_noise^2 / |dF_spin^2/dx|_max. That expression is the standard-error estimate for locating a single sharp edge under Gaussian noise; it is a localization precision, not an imaging resolution. The scans are fit to a single hyperbolic-tangent step (Eq. 7), and the authors report w ~ 10 nm, which they attribute to sample/slice convolution. If the effective point-spread function has ~10 nm width, two edges separated by 0.9 nm cannot be distinguished; their signals merge into one step whose position can be estimated to sub-nanometer precision only because of high SNR and dense sampling. The separately reported localization precision sigma_x0 = 0.6 nm (fit uncertainty of x0) is the same type of single-edge estimator. Thus neither number demonstrates the ability to resolve adjacent spin structures at 0.9 nm, which is the milestone claimed for macromolecular imaging. The minimum slice width of 0.7 nm is a physical property of the excitation slice, not a demonstrated two-point image resolution. The manuscript should either reframe the claim as sub-nanometer edge-localization precision or add a two-point resolvability analysis.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports MRFM lateral scans of a hydrogen-containing adsorbate layer on a silicon nanorod, claiming a one-dimensional spatial resolution of 0.9 ± 0.2 nm and a localization precision of 0.6 ± 0.1 nm. The authors introduce hyperbolic secant (HS^n) adiabatic pulses for sharp slice definition, characterize the slice profile experimentally and by simulation, and demonstrate long-term instrument stability. The resolution metric is obtained from the maximum slope-to-noise ratio of the signal onset (Eq. 8). The paper also derives a slice width of 0.7 nm from the tip model and predicts a possible 0.3 nm resolution with improved SNR. The authors claim the first sub-nanometer MRI and a milestone toward 3D macromolecular imaging.","tokens_in":9301,"tokens_out":6725,"duration_ms":68717,"significance":"The experimental work is of high quality: the HSn pulse protocol is a clear improvement over conventional trapezoidal modulation, the drift of 1.8 nm over 31 h is excellent, and the gradient calibration appears careful. The central quantitative claim, however, rests on the interpretation of the slope-to-noise ratio as 'resolution.' If the claim is reframed as sub-nanometer edge-localization precision, the result is still remarkable and of interest to the MRFM community. The paper also provides a reasonable model-based extrapolation to 0.3 nm, which is clearly labeled as an expectation. No computational or derivational circularity was apparent.","major_comments":[{"comment":"The quantity σ_x,slope defined in Eq. (8) is the uncertainty in estimating the position of a single step edge from its slope and the noise level; it is a localization precision, not a two-point resolution. The paper's own fits yield an edge width w ≈ 10 nm (Eq. (7)), and the scans in Fig. 4 are single-step onsets. Two features separated by 0.9 nm would produce a merged step whose position could still be estimated to sub-nanometer precision, so the scans do not demonstrate a resolution of 0.9 nm. The abstract and conclusions should either use the term 'edge localization precision' or add a two-point experimental demonstration (e.g., two adsorbate edges at a known separation) to support the resolution claim.","section":"§4 (Results and Discussion), Eq. (8)"},{"comment":"The statement 'minimum measured slice width of 0.7 nm' in the conclusion is not directly measured; the value is computed from the simulated HSn slice function and the tip-model gradient G_x ≈ 2.3 × 10^6 T/m. Since the slice width is a key figure for the sub-nanometer claim, the text should explicitly label it as a model-derived estimate, or better, provide a direct measurement by scanning the slice across a sharp feature.","section":"Conclusion"}],"minor_comments":[{"comment":"The word 'suffcient' in the abstract is a typo for 'sufficient'.","section":"Abstract"},{"comment":"The citation 'gradient generation,12,12–16' should read '12–16'; the duplicate '12' appears to be a numbering error.","section":"Introduction"},{"comment":"The inset of Fig. 4b shows a confidence interval for one scan; it would be helpful to state how the error bars of σ_x,slope (0.2 nm) were computed across the 11 scans (e.g., standard deviation or standard error).","section":"§4 (Results and Discussion)"}],"recommendation":"major_revision","confidential_remarks":"This is a strong experimental paper from a leading group. The main barrier to acceptance is the correct labeling of the resolution metric. The distinction between edge-localization precision and two-point resolution is not merely semantic for a paper that aims to establish a milestone toward macromolecular imaging; readers in structural biology will interpret 'resolution' in the imaging sense. A careful revision that either adds a two-point experiment or explicitly limits the claim to precision would resolve the concern. I believe the work is suitable for the journal after such a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First sub-nanometer MRFM scans are real, and the HSn pulse work is a genuine step forward. But the 0.9 nm figure is an edge-localization precision, not an image resolution; the paper overstates what it demonstrates.\n\nThe authors show line scans with a sharp signal onset and use Eq. 8 to convert slope-to-noise into a spatial uncertainty of 0.9 nm. That is a legitimate way to characterize how well you can find an edge, and the measured stability (1.8 nm drift over 31 h) is impressive. The new HSn pulse protocol clearly produces much sharper slice edges than the old trapezoidal modulation, and the minumum slice width of 0.7 nm is a real physical property of the excitation. This is the best MRFM result to date and a solid advance.\n\nThe problem is the word \"resolution.\" With a fitted edge width w~10 nm (Eq. 7), two structures 0.9 nm apart would not be separated in the scan; you would see one step at the average position. The 0.9 nm number tells you the statistical uncertainty of that single step's location, not how close two features can be. The authors use \"localization precision\" for sigma_x0 but also call sigma_x,slope \"uncertainty-limited spatial resolution,\" and the abstract says \"resolution.\" That is not a fatal error, but it is a mismatch between headline and what the measurement actually proves. The extrapolation to 0.3 nm relies on the modeled gradient and slice shape, with calibration details in an SI we can't check; that's fine as an expectation, but it shouldn't be presented as part of the demonstrated resolution.\n\nThe citation pattern looks fine, with appropriate comparisons to Rose et al. and Degen et al. No data or code are included, which is common for this type of experimental paper but makes reproducibility harder; not a reason to reject on its own.\n\nThis paper is for the MRFM and nanoscale MRI community. It deserves serious peer review; referees should push for a clearer distinction between resolution and localization, and for either a two-point demonstration or a revised claim. I would engage with it and would want to see the SI details for the slice shape and tip model before fully trusting the numbers, but the experimental work is genuine.","headline":"Impressive sub-nanometer MRFM scans with a new HSn pulse protocol, but the 0.9 nm headline is edge-localization precision, not a demonstrated two-point spatial resolution.","tokens_in":9813,"tokens_out":1859,"would_cite":true,"duration_ms":18112,"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":"The paper reports magnetic resonance force microscopy scans with a one-dimensional resolution of 0.9 nm, the first sub-nanometer MRI measurement.","keywords":["magnetic resonance force microscopy","nanoscale magnetic resonance imaging","spatial resolution","nuclear magnetic resonance","scanning probe microscopy","hyperbolic secant pulses","adiabatic spin inversion","proton imaging"],"falsifier":"Scan a sample with two parallel 1H-rich edges separated by a known distance around 1 nm: if the measured signal shows one unresolved combined step instead of two distinguishable onsets, then the slope-based 0.9 nm figure would not correspond to two-point imaging resolution. An additional check would be to vary the averaging time and see whether σx,slope improves as expected from noise averaging and whether the fitted onset positions remain stable.","tokens_in":8731,"feed_emoji":"🧲","tokens_out":3317,"duration_ms":36156,"temperature":0.7,"pith_summary":"The paper tries to show that magnetic resonance force microscopy (MRFM) can image with sub-nanometer spatial resolution, reporting a one-dimensional resolution of 0.9 ± 0.2 nm and a localization precision of 0.6 ± 0.1 nm. A sympathetic reader would care because this is the first demonstrated sub-nanometer MRI measurement, moving the technique toward three-dimensional imaging of protein and macromolecular structures. The advance comes from a sharper spin-excitation protocol, a magnetic gradient above $10^{6}$ T/m, a well-defined nanorod sample geometry, and instrument drift below 2 nm over 24 hours. The authors argue that at maximum modeled gradient the arrangement supports resolutions down to 0.3 nm given sufficient signal-to-noise ratio.","feed_headline":"MRFM scans hit 0.9 nm one-dimensional resolution","feed_subtitle":"A sharper spin-inversion slice pushes magnetic resonance force microscopy into sub-nanometer imaging territory.","key_machinery":"The central mechanism is the hyperbolic secant (HSn) pulse protocol, an amplitude- and frequency-modulated radio-frequency excitation that adiabatically inverts nuclear spins in a narrow, well-defined detuning window. The slice function is ξ(Δf) = a₁(Δf) F(Δf) ≈ (4/π) F(Δf), where F is the spin-inversion fidelity over hundreds of cantilever cycles. This sharp slice, combined with a tangential nanorod geometry and a measured lateral gradient of 0.56 × $10^{6}$ T/m, produces a signal step whose maximum slope divided by noise (Eq. 8) gives the sub-nanometer resolution value.","core_discovery":"The paper's central claim is that MRFM, using optimized hyperbolic secant (HSn) adiabatic pulses to create a sharply defined resonant slice, achieves a one-dimensional spatial resolution of 0.9 ± 0.2 nm and a localization precision of 0.6 ± 0.1 nm, as measured from the slope-to-noise ratio of the signal onset (Eq. 8). This is presented as the first sub-nanometer MRI measurement. The authors find that the imaging slice is sharply bounded—about 30 kHz between full spin inversion and no signal—and that all technical sources of blur are reduced below 1 nm, so resolution is limited only by sensor noise. A model of the tip field indicates that the same arrangement could reach roughly 0.3 nm slice width at maximum gradient.","pith_inferences":["Beyond the paper's claim, the 0.9 nm figure is an edge-localization uncertainty from a single sharp signal step, not a demonstrated two-point resolution; a two-point phantom measurement would be needed to equate it with the smallest resolvable separation between two features.","Since the detected signal arises from statistical rather than thermal spin polarization, the effective resolution likely improves with averaging time and signal-to-noise; this could be tested systematically by measuring σx,slope at different integration times.","The sharp edge produced by HSn pulses may be transferable to other MRFM and MRI configurations, potentially improving slice definition wherever adiabatic inversion is used.","If the tangential-slice geometry is pushed to maximum gradient, a 0.3 nm resolution would bring MRFM close to imaging secondary protein structure directly, though three-dimensional scanning remains an unsolved step."],"forward_implications":["If the central claim holds, MRFM can localize nuclear spin density features with sub-nanometer precision in one dimension, a prerequisite for meaningful structural imaging of macromolecules.","At the maximum modeled gradient of about 6 × 10^6 T/m, the same instrumentation is expected to support slice widths around 0.3 nm, provided the signal-to-noise ratio can be improved.","With a more sensitive force transducer, the authors expect three-dimensional images with 1 nm voxel size, corresponding to roughly 100 hydrogen atoms, to become possible.","The sharp HSn slice protocol removes the previously limiting slice-edge blur, and the demonstrated drift below 2 nm over 31 hours makes long, undistorted scans feasible.","Because MRFM can handle objects larger than 20 nm, sub-nanometer resolution would extend MRI-style imaging to structures that are difficult for other nanoscale MRI techniques."],"supporting_citations":[{"why":"Establishes the MRFM force-detection concept that the whole measurement builds on.","marker":"[1]"},{"why":"Demonstrates prior MRFM imaging on tobacco mosaic virus with a best-effort resolution of 4 nm, the benchmark this work improves on.","marker":"[3]"},{"why":"Provides earlier MRFM sensitivity and voxel-size estimates used to contextualize the sub-nanometer result.","marker":"[17]"},{"why":"Reports a nominal 2 nm resolution using Fourier encoding, another baseline this work compares against.","marker":"[18]"},{"why":"Describes the MRFM apparatus and field-map calibration approach used for the gradient and slice modeling.","marker":"[21]"},{"why":"Supplies the numerical magnetic field model for the FeCo nanomagnet used to calculate gradients and slice widths.","marker":"[22]"},{"why":"Introduces the hyperbolic secant pulse family from which the improved HSn excitation protocol is derived.","marker":"[26]"},{"why":"Provides the general HSn pulse formalism and its properties that the paper adapts for MRFM.","marker":"[27]"}],"fun_headline_variants":["MRFM achieves 0.9 nm resolution in one dimension","Sub-nanometer MRI: MRFM hits 0.9 nm resolution","MRFM pushes MRI to sub-nanometer scale with 0.9 nm resolution","0.9 nm resolution MRFM scan demonstrates sub-nanometer MRI"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 0.9 nm resolution is computed from the steepest slope of the signal step divided by the measurement noise, which assumes that this edge-localization uncertainty equals the smallest separation between two resolvable image features; the paper does not demonstrate two-point resolution.","fun_headline_variants_meta":{"raw":{"variants":["MRFM achieves 0.9 nm resolution in one dimension","Sub-nanometer MRI: MRFM hits 0.9 nm resolution","MRFM pushes MRI to sub-nanometer scale with 0.9 nm resolution","0.9 nm resolution MRFM scan demonstrates sub-nanometer MRI"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000317,"raw_usage":{"total_tokens":1750,"prompt_tokens":856,"completion_tokens":894,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":472,"completion_tokens_details":{"reasoning_tokens":813}},"tokens_in":472,"tokens_out":894,"duration_ms":8379,"temperature":1.0,"reasoning_tokens":813,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:48:24.991777+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Scan a sample with two parallel 1H-rich edges separated by a known distance around 1 nm: if the measured signal shows one unresolved combined step instead of two distinguishable onsets, then the slope-based 0.9 nm figure would not correspond to two-point imaging resolution. An additional check would be to vary the averaging time and see whether σx,slope improves as expected from noise averaging and whether the fitted onset positions remain stable.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the MRFM force-detection concept that the whole measurement builds on."},{"cited_title":"L.; Poggio, M.; Mamin, H","cited_arxiv_id":null,"evidence_quote":"Demonstrates prior MRFM imaging on tobacco mosaic virus with a best-effort resolution of 4 nm, the benchmark this work improves on."},{"cited_title":"J.; Oosterkamp, T","cited_arxiv_id":null,"evidence_quote":"Provides earlier MRFM sensitivity and voxel-size estimates used to contextualize the sub-nanometer result."},{"cited_title":"Q.; Jeon, N.; Lauhon, L","cited_arxiv_id":null,"evidence_quote":"Reports a nominal 2 nm resolution using Fourier encoding, another baseline this work compares against."},{"cited_title":"A.; Eichler, A.; Tao, Y.; Takahashi, H.; Navaretti, P.; Degen, C","cited_arxiv_id":null,"evidence_quote":"Describes the MRFM apparatus and field-map calibration approach used for the gradient and slice modeling."},{"cited_title":"L.; Mamin, H","cited_arxiv_id":null,"evidence_quote":"Supplies the numerical magnetic field model for the FeCo nanomagnet used to calculate gradients and slice widths."},{"cited_title":"S.; Joseph, R","cited_arxiv_id":null,"evidence_quote":"Introduces the hyperbolic secant pulse family from which the improved HSn excitation protocol is derived."},{"cited_title":"Journal of Magnetic Resonance","cited_arxiv_id":null,"evidence_quote":"Provides the general HSn pulse formalism and its properties that the paper adapts for MRFM."}],"review_version":1}