{"id":"5cf174be-5207-47c4-97a3-8d2d15398ec0","arxiv_id":"2506.18163","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A kinetic Monte Carlo-style stochastic dynamics model for FEBID, parameterized by short molecular dynamics simulations, reproduces experimental growth rates and heights but only after selecting the parameter set that best matches those experiments.","lead":"This paper builds a coarse-grained stochastic dynamics model of focused electron beam induced deposition (FEBID) and uses it to simulate the growth of a tungsten-rich nanostructure from W(CO)6 on a silica substrate. The simulation reaches millisecond and submicron scales that atomistic molecular dynamics cannot, and the resulting structures are compared with experimental FEBID deposits.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The validation claim depends on a parameter set chosen after matching the experimental targets; the reported 15 nm / 15% agreement is therefore not an independent predictive test.","rationale":"The paper presents a genuine methodological advance: an SD framework that scales to millisecond and sub-micrometer FEBID conditions, with rates connected to MD simulations. The growth morphology and the emergence of a crown-shaped metal distribution are qualitatively plausible and merit attention. However, the central validation claim—that the simulations 'align well with known experimental measurements'—is not supported as stated because the parameter set used for the final comparison was selected after inspecting how each set matched the chosen experimental targets. This is the load-bearing weak point. The reader's verdict identified the same concern, and my independent reading of Section II.D.3 and Section III confirms it: the paper explicitly introduces scaling factors to explore parameter influence and then selects the set that best matches experiment, calling it the most realistic. That is a circular validation procedure. The issue is not that parameter exploration is forbidden, but that the paper presents the result as a validation rather than as a calibrated fit. The proposed test—re-running with Set 1 and comparing to experiment—would directly reveal whether the agreement is an artifact of tuning. The internal inconsistency between the stated experimental range (25-40%) and the simulated value (~15%) further weakens the claim, though it is secondary to the selection issue. A CONDITIONAL verdict is appropriate: the methodology is promising, but the validation should be redone with a pre-specified parameter set and with uncertainty estimates on the MD-derived rates. The authors should also clarify the W-content mismatch rather than describing 15% as consistent with a 25-40% range.","tokens_in":24070,"tokens_out":2190,"duration_ms":26404,"concrete_test":"Re-run the identical 100-cycle FEBID simulation using parameter Set 1 from Table V (the unscaled, MD-derived detachment probabilities) and report the resulting deposit height and W content. If Set 1 yields values within or near the cited experimental ranges (height ~15 nm and W content 25-40%, or at least qualitatively similar), the agreement is robust to parameter choice; if Set 1 gives significantly different height or W content (e.g., outside a factor of two) and only Set 3 matches, the claimed validation depends on the ad hoc 1.460 scaling factor. A complementary test is to fix Set 3 before simulating a different experimental condition (e.g., a different beam current or precursor pressure) and compare those predictions to available experimental data.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that the SD model reproduces experimental FEBID growth characteristics and is thereby validated. This claim rests on parameter Set 3 from Table V, whose pairwise detachment probabilities p_dis(α_i,α_6) were obtained by scaling the MD-derived Set 1 values by an ad hoc factor of 1.460 (Section II.D.3), explicitly 'to explore the influence of these parameters.' After running all three sets, the paper selects Set 3 as 'the most physically realistic and experimentally consistent' (Section III) because it yields ~15 nm height and ~15% W content. This is a post-hoc selection of parameters based on the target observables, not an a priori prediction. The intermediate-fragment desorption rates in Table I are also linearly extrapolated between W(CO)6 and W without MD support, so the chosen Set 3 inherits this unsupported interpolation. Consequently, the claimed alignment with experiment is at least partly a consequence of parameter tuning; the method's predictive power is not established by the agreement shown. Additionally, the text states that experimental W contents are typically 25-40% while Set 3 gives ~15%, described as 'consistent'—an internal inconsistency that further weakens the validation statement.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a stochastic dynamics (SD) model of focused electron beam-induced deposition (FEBID) for W(CO)6 on a hydroxylated SiO2-H substrate, implemented in the MBN Explorer/MBN Studio software. Input parameters—precursor–substrate and precursor–precursor detachment energies, vibrational frequencies, and diffusion probabilities—are derived from molecular dynamics (MD) simulations using Morse-potential fits and Arrhenius rates, while the electron-induced fragmentation profile is taken from a prior Monte Carlo track-structure study. The model simulates 100 irradiation/replenishment cycles (43 ms total) on a cubic grid, producing a dome-shaped deposit with a peak height of about 15 nm and a tungsten content of about 15%. The authors claim these results 'align well with known experimental measurements' and that the SD approach reaches temporal and spatial scales typical of FEBID experiments (milliseconds, submicrometers).","tokens_in":24339,"tokens_out":5129,"duration_ms":55394,"significance":"If the central claim is established, the SD approach would be a valuable multiscale tool, bridging the gap between atomistic irradiation-driven MD (limited to tens of nanometers and nanoseconds) and continuum reaction–diffusion models. The paper's strengths are the explicit step-by-step parameterization from MD, the detailed tabulation of all input parameters, and the demonstration that a lattice SD simulation can follow FEBID growth over tens of milliseconds while retaining particle-type-level composition information. The claim of predictive, experimentally consistent atomistic insight is, however, currently not fully supported: the agreement with experiment is obtained after selecting one of three parameter sets based on how closely it reproduces the target observables, and several input parameters are extrapolated or self-referenced rather than independently validated. With revision to remove the post hoc selection and to clarify the provenance of key inputs, the methodology could be a useful contribution to FEBID modeling.","major_comments":[{"comment":"The validation claim is weakened by post hoc selection of parameter Set 3. The paper states that Set 1, Set 2, and Set 3 were generated by scaling pairwise detachment probabilities with ad hoc factors of 2.065 and 1.460 'to explore the influence of these parameters' (Section II.D.3), and then Set 3 is declared 'the most physically realistic and experimentally consistent' (Section III) because it yields approximately 15 nm height and 15% W content. This is a calibration procedure, not an independent predictive test. To support the stated claim that the SD framework provides predictive atomistic insight, the authors should either derive Set 3 (or the scaling factors) from a physical argument before comparison with experiment, or validate the chosen parameters against observables not used in the selection (e.g., the radial profile shape, time evolution, or composition distribution) and show that the agreement is not merely a consequence of fitting.","section":"Section III, Figs. 8–9; Section II.D.3, Table V"},{"comment":"There is an internal inconsistency in the reported metal content. The text states that the simulated W content stabilizes at about 15% and that 'This value is consistent with experimental measurements of tungsten concentrations in FEBID structures grown from W(CO)6, where typical metal contents fall within 25–40%.' Since 15% lies outside the cited 25–40% range, the statement is contradictory. The authors should either provide a specific experimental reference supporting W contents near 15% for their particular beam parameters and precursor flux, or explicitly acknowledge that the simulated composition falls below the typical experimental range and discuss possible reasons.","section":"Section III, Fig. 9B and accompanying text"},{"comment":"The detachment energies and vibrational frequencies for intermediate fragments W(CO)i (i=1..5) are linearly extrapolated between the simulated W(CO)6 and W values, and the resulting rate constants are then used in Eqs. (7)–(8) to define the pairwise detachment and diffusion probabilities in Tables II, III, and V. This is a load-bearing approximation that is not tested against MD simulations or quantum chemistry for the intermediate species. The paper claims that 'the parameters for the new SD-based FEBID model were determined using molecular dynamics (MD) simulations' (Abstract), but for the great majority of fragment species the parameters are assumed, not derived. The authors should justify the linear extrapolation or provide additional MD data for at least some intermediate fragments (e.g., W(CO)3 or W(CO)) to demonstrate that the trend is physically reasonable.","section":"Section II.D.1, Table I; Section II.D.3, Table IV"},{"comment":"The provenance of the surface diffusion coefficients is not adequately established. The text states that D_W6 = 29.10 µm²/s and D_W0 = 0.21 µm²/s were 'adapted from the literature, see [21,22,24,25,34,35]', but references [21], [22], [24], [25], [34], and [35] are the authors' own software papers, books, and MBN Explorer user guide and tutorials. No independent experimental or first-principles source for these values is provided. Since precursor diffusion is a fast process that sets the time step (Eq. (9)) and directly enters the diffusion probabilities in Table III, the authors should clarify the original source of these coefficients and, ideally, benchmark them against independent experimental measurements or density-functional-theory-based estimates.","section":"Section II.D.2"}],"minor_comments":[{"comment":"In the paragraph reporting the vertical growth rate, the phrase 'consistent with with experimentally reported growth rates' contains a duplicated 'with'.","section":"Section III"},{"comment":"The caption reads 'have been used to in the FEBID simulations'; the word 'to' should be removed.","section":"Table V caption"},{"comment":"The sentence 'The fragmentation rates shown in Fig. 5 are obtained by dividing the values used in simulations by 3.13×10^-7' is ambiguous. It should specify whether the plotted rates are normalized, and how the values from Eq. (19) are converted into the per-step fragmentation probabilities used in the SD simulations.","section":"Section II.F, Eq. (19)"},{"comment":"The sentence 'The probabilities of the diffusion of particles describing the precursor, the tungsten, and precursor fragments over the substrate are summarised in Table III' is awkwardly phrased; 'the tungsten' should be 'the tungsten atom' (or 'a W atom').","section":"Section II.D.2"},{"comment":"The supporting video is referred to as 'Supplementary video 1' in Section III and as 'Supplementary video S1' in Section V; the naming should be consistent.","section":"Section III and Section V"}],"recommendation":"major_revision","confidential_remarks":"The manuscript would benefit from a clearer separation between 'calibration' and 'validation.' As written, the central claim of experimentally consistent predictive modeling is weakened by the post hoc selection of Set 3 and by the internal inconsistency in the W-content comparison (15% vs. 25–40%). I also note the heavy reliance on the authors' own software and prior publications for key input parameters (diffusion coefficients, fragmentation rates); this is not itself disqualifying, but the provenance of those inputs should be independently verifiable. If the authors reframe the work as a demonstration of the SD methodology with calibrated parameters, rather than as a validated predictive tool, the paper could be acceptable after revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper does something genuinely new: it bridges atomistic MD and continuum FEBID models by building a stochastic dynamics (SD) engine whose rate parameters are derived from short MD runs, and it reaches millisecond/submicron scales. That is a real step forward for the FEBID simulation community. The MD-to-SD workflow is described in enough detail to reproduce conceptually, and the morphological features—bell-shaped deposit, crown-like metal distribution—are physically plausible.\n\nBut the validation claim doesn't hold up as stated. The 'agreement' with experiment relies on parameter Set 3 of Table V, which was chosen after seeing which set best matched the experimental height and W content. That is post hoc tuning, not prediction. The paper acknowledges Sets 2 and 3 were generated 'to explore influence', then selects Set 3 because it gives the 'right' numbers. The linear extrapolation of intermediate fragment desorption energies and frequencies (Table I) is another soft spot; it is a reasonable first approximation but not MD-derived.\n\nThe internal inconsistency is also hard to ignore: the text says typical experimental W contents are 25-40%, then states Set 3's ~15% 'aligns well.' Those are not the same. Either the experimental range quoted is too broad, or the simulation misses the metal content by a factor of ~2. The authors need to address this directly.\n\nThe self-referential parameter sources (diffusion coefficients from their own MBN Explorer manuals, fragmentation rates from their own earlier MC study) don't by themselves invalidate anything—reusing your own well-tested constants is normal. But combined with the post hoc parameter selection, it means the paper hasn't yet demonstrated predictive power.\n\nWho should read this: anyone modeling FEBID or similar beam-induced growth processes, and anyone building SD/MD multiscale workflows. It is a solid methodological template. For peer review: yes, send it out. The framework deserves scrutiny and the validation flaws are fixable—rerun the target case with a parameter set fixed a priori, report uncertainties on the MD-derived rates, and provide input files so others can reproduce the SD runs. As it stands, the abstract's 'validated' is too strong; the paper is a promising approach with a plausible but not yet convincing benchmark.\n\nI'd bring it to a reading group focused on multiscale simulation methods, and I'd cite it if I worked in FEBID. But I wouldn't take the 15 nm/15% numbers as a benchmark until the post hoc selection is addressed.","headline":"Useful multiscale framework for FEBID, but the validation rests on a post-selected parameter set and an internally inconsistent metal-content claim; worth refereeing, not for unconditional acceptance.","tokens_in":24947,"tokens_out":2469,"would_cite":false,"duration_ms":26761,"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":"The paper claims that a stochastic dynamics model with molecular-dynamics-derived rates can simulate focused electron beam-induced deposition of W(CO)$_6$ on silica at experimental time and length scales, producing a 100-cycle deposit of…","keywords":["focused electron beam induced deposition","FEBID","stochastic dynamics","kinetic Monte Carlo","multiscale modeling","molecular dynamics parameterization","W(CO)6 precursor","nanostructure growth"],"falsifier":"Measure the desorption rate or surface diffusion coefficient of an intermediate fragment such as W(CO)$_3$ on SiO$_2$-H, for example by temperature-programmed desorption or single-molecule tracking, and compare with the linearly extrapolated values used here; a large disagreement would collapse the parameter chain. Alternatively, run the same SD protocol without parameter rescaling for a different precursor, such as Fe(CO)$_5$, and compare predicted deposit height and iron content against published experiments; a systematic mismatch would indicate that the present agreement with W(CO)$_6$ data is a fitting outcome rather than a prediction.","tokens_in":23775,"feed_emoji":"⚛️","tokens_out":9963,"duration_ms":109773,"temperature":0.7,"pith_summary":"This paper claims that stochastic dynamics, with rates taken from atomistic molecular dynamics, can simulate focused electron beam-induced deposition (FEBID) of W(CO)$_6$ on a hydroxylated silica substrate over the millisecond and sub-micrometer scales typical of real experiments. The model treats intact precursors, partially fragmented precursors, tungsten atoms, CO ligands, and the substrate as particles moving on a cubic grid, with adsorption, diffusion, desorption, and electron-induced fragmentation occurring at prescribed rates. After 100 irradiation/replenishment cycles the simulated deposit is roughly 15 nm tall and contains about 15\\% tungsten, which the authors state aligns with experimental W(CO)$_6$-based FEBID deposits. If correct, this would provide a predictive atomistic-level tool for deposit composition, morphology, and growth rate, and a path toward simulating 3D nanoprinting.","feed_headline":"New stochastic model reproduces FEBID growth to 15 nm","feed_subtitle":"A 100-cycle simulation of W(CO)6 deposition lands at about 15% tungsten content, matching lab-built nanostructures.","key_machinery":"The machinery is a lattice stochastic-dynamics (kinetic Monte Carlo type) model: a cubic grid with cell size 0.63 nm and one particle per cell, where particle types include the intact precursor W(CO)$_6$, fragments W(CO)$_i$ ($i=1,\\dots,5$), tungsten atoms, CO ligands, and immobile substrate cells. The load-bearing identity is the factorization of a particle's translocation probability into pairwise bond-break and bond-maintenance probabilities, $p_{\\mathrm{dis}}(\\alpha,\\beta)$ and $p_{\\mathrm{dif}}(\\alpha,\\beta)$, with exponents set by the number of broken and maintained bonds on the lattice. These pairwise probabilities are inverted from desorption and diffusion rates obtained by fitting Morse potentials to MD trajectories, and the fastest process, precursor surface diffusion, fixes the simulation time step $\\Delta t = 0.097$ ns; every other event occurs per step with probability $\\Delta t \\Gamma$. Electron-induced fragmentation uses a fitted radial profile of primary, secondary, and backscattered electron flux multiplied by a fragmentation cross section, with the flux distribution taken from a prior track-structure Monte Carlo study.","core_discovery":"The central claim, stated as the authors would state it, is that the stochastic dynamics approach overcomes the size and time limitations of atomistic irradiation-driven molecular dynamics: it reaches millisecond and sub-micrometer scales while still resolving the elemental composition and morphology of the growing deposit. For a 30 keV electron beam on W(CO)$_6$ over SiO$_2$-H, the model predicts a dome-shaped deposit that grows to about 15 nm in height after 100 cycles (43 ms of simulated time, corresponding to a vertical growth rate near 0.35 nm/ms) and stabilizes at about 15\\% tungsten content. These values are reported as consistent with experimental W(CO)$_6$-based FEBID measurements. The simulations also predict a crown-shaped radial metal profile, with the tungsten fraction peaking away from the beam center, which the authors attribute to hindered CO-ligand escape in the dense central region of the deposit.","pith_inferences":["Because the final parameter Set 3 was adopted after comparing simulated height and tungsten content with experimental values, a prospective test would strengthen the method: fix all rates from MD and experiment without rescaling, then predict a new FEBID system, such as Fe(CO)$_5$ or Me$_2$Au(tfac), and compare with measurements.","The single-particle-per-cell lattice treats each CO ligand as one grid cell; a finer or multi-occupancy grid could alter the crown-shaped metal profile, whose magnitude depends on whether CO can escape the dense central region.","The conversion of particle counts into deposit height uses van der Waals volumes, so the predicted 15 nm height and 15\\% tungsten content could be checked for sensitivity by repeating the same simulations with a modestly different cell size or volume assignment.","If the growth-rate scaling holds, the same parameter chain could be used to design deposition recipes, such as beam current, pressure, or cycle timing, to position or suppress the metal-enriched ring in real FEBID fabrication."],"forward_implications":["The method can follow FEBID over tens of milliseconds of simulated time and sub-micrometer lateral distances, capturing hundreds of irradiation/replenishment cycles while retaining sub-nanometer compositional and morphological detail.","The computed vertical growth rate of about 0.35 nm/ms falls within the experimentally reported 200--500 nm/s range for W(CO)$_6$ FEBID, supporting the use of the same SD parameter chain for growth-rate prediction.","The crown-shaped radial tungsten profile emerges from transport-limited CO escape in the dense central deposit, offering a mechanistic explanation for metal-enriched rings observed around experimental FEBID structures.","The workflow of deriving SD rate constants from MD and then simulating full deposition cycles is transferable to other precursors and substrates, and can be extended toward modeling 3D nanoprinting and related growth techniques."],"supporting_citations":[{"why":"Defines the stochastic-dynamics framework, the rate-probability relations used throughout, and the lattice implementation this work parameterizes for FEBID.","marker":"[20]"},{"why":"Supplies the electron track-structure Monte Carlo fragmentation-rate distribution for W(CO)$_6$ under a 30 keV beam that the SD simulations adopt as the irradiation input.","marker":"[16]"},{"why":"Establishes the irradiation-driven molecular dynamics approach whose CHARMM force field and W(CO)$_6$ parameters the MD calculations here use, and whose time and length limits the SD method extends.","marker":"[11]"},{"why":"Provides the experimental W(CO)$_6$ FEBID reference, including beam conditions and pressure, used to set simulated fluxes and to validate deposit height and growth rate.","marker":"[29]"},{"why":"Supplies experimental data on the properties and tungsten content of W-based FEBID structures used as the validation target for the simulated metal fraction.","marker":"[33]"},{"why":"Reviews FEBID growth and gives the experimentally observed range of metal content and deposit shapes against which the simulated 15\\% tungsten and dome/crown morphology are compared.","marker":"[6]"}],"fun_headline_variants":["Stochastic FEBID model grows 15 nm in 43 ms","FEBID simulation matches experiments at 15% tungsten","Multiscale model predicts 15 nm FEBID dome","Stochastic dynamics reaches ms timescales for FEBID"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the desorption rates for intermediate fragments W(CO)$_i$ ($i=1,\\dots,5$) can be linearly interpolated between the simulated intact-precursor and pure-tungsten values, and that the ad hoc scaling used to produce parameter Set 3 represents real physics rather than being tuned to match the chosen experimental targets.","fun_headline_variants_meta":{"raw":{"variants":["Stochastic FEBID model grows 15 nm in 43 ms","FEBID simulation matches experiments at 15% tungsten","Multiscale model predicts 15 nm FEBID dome","Stochastic dynamics reaches ms timescales for FEBID"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001014,"raw_usage":{"total_tokens":4284,"prompt_tokens":947,"completion_tokens":3337,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":3265}},"tokens_in":563,"tokens_out":3337,"duration_ms":24106,"temperature":1.0,"reasoning_tokens":3265,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T23:22:57.565016+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the desorption rate or surface diffusion coefficient of an intermediate fragment such as W(CO)$_3$ on SiO$_2$-H, for example by temperature-programmed desorption or single-molecule tracking, and compare with the linearly extrapolated values used here; a large disagreement would collapse the parameter chain. Alternatively, run the same SD protocol without parameter rescaling for a different precursor, such as Fe(CO)$_5$, and compare predicted deposit height and iron content against published experiments; a systematic mismatch would indicate that the present agreement with W(CO)$_6$ data is a fitting outcome rather than a prediction.","supporting_citations":[{"cited_title":"V., Sushko, G., and Solo v’yov, A","cited_arxiv_id":null,"evidence_quote":"Defines the stochastic-dynamics framework, the rate-probability relations used throughout, and the lattice implementation this work parameterizes for FEBID."},{"cited_title":"D., and Plank, H","cited_arxiv_id":null,"evidence_quote":"Supplies the electron track-structure Monte Carlo fragmentation-rate distribution for W(CO)$_6$ under a 30 keV beam that the SD simulations adopt as the irradiation input."},{"cited_title":"Living up to its potent ial—Direct-write nanofabrication with focused electron beams","cited_arxiv_id":null,"evidence_quote":"Establishes the irradiation-driven molecular dynamics approach whose CHARMM force field and W(CO)$_6$ parameters the MD calculations here use, and whose time and length limits the SD method extends."},{"cited_title":"A., Solov’yov, A","cited_arxiv_id":null,"evidence_quote":"Provides the experimental W(CO)$_6$ FEBID reference, including beam conditions and pressure, used to set simulated fluxes and to validate deposit height and growth rate."},{"cited_title":"D., Fitzek, H., Rauch, S., Ratten- berger, J., Rack, P","cited_arxiv_id":null,"evidence_quote":"Supplies experimental data on the properties and tungsten content of W-based FEBID structures used as the validation target for the simulated metal fraction."},{"cited_title":"B., Fowlkes, J","cited_arxiv_id":null,"evidence_quote":"Reviews FEBID growth and gives the experimentally observed range of metal content and deposit shapes against which the simulated 15\\% tungsten and dome/crown morphology are compared."}],"review_version":1}