{"id":"138cc78f-68f9-4dd1-bd63-76de7a9f32f7","arxiv_id":"2608.08647","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A retrofitted magnetron sputtering chamber used Gaussian-process-guided experiments to semi-autonomously locate a 2.3 MV/cm coercive field in Al1-x-yScxByN thin films.","lead":"This paper describes how to retrofit an older sputtering deposition machine with automated controls and data logging, then demonstrates the upgraded system finding a lower-coercive-field recipe for ferroelectric AlScBN thin films. It matters because it offers a lower-cost path for academic labs to participate in data-driven, autonomous materials experiments.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The coercive-field demonstration rests on unreported electrical measurements; without a stated P-E protocol, uncertainty, or repeat growths, the 2.3 MV/cm champion value and the GPR minimization claim are unverifiable.","rationale":"The read identifies the same critical weakness: the coercive field measurements are the quantitative foundation of the demonstration, and they are not described with enough detail to assess reproducibility. The automation and data-infrastructure parts of the paper have independent support from the hardware inventory, tables, pseudocode, and stated data availability, so they are not the main risk. The demonstration, however, makes a quantitative claim about a material property whose measurement protocol, uncertainty, and baseline are absent; this is exactly the kind of missing support that warrants a conditional verdict. The manuscript itself includes a passage suggesting that the optimum may lie at or beyond the Sc power safety limit, which further weakens the stopping rationale. A concrete repeat-measurement test would settle whether the 2.3 MV/cm value is robust or an artifact of measurement noise.","tokens_in":7745,"tokens_out":3746,"duration_ms":44450,"concrete_test":"Regrow the champion recipe (2.3 MV/cm) and at least one higher-EC recipe three times each, and measure EC with a stated protocol (PUND or frequency-dependent P-E, fixed top-electrode area, maximum field, and leakage correction). Report mean and standard deviation per recipe. If the champion EC is not reproduced within, say, 10–20%, or if run-to-run spread exceeds the 5% improvement threshold in Eq. 3, the GPR loop was fitting noise and the minimization claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central demonstration claim—that Bayesian optimization found a film with minimum coercive field 2.3 MV/cm—depends entirely on the coercive field values fed into the GPR loop. The paper never states how EC was measured: no P-E loop frequency, maximum applied field, top-electrode geometry, leakage correction, or thickness determination method. Since EC is a derived quantity rather than a direct instrument readout, run-to-run and film-to-film differences in measurement protocol can easily be comparable to the 5% improvement threshold in Eq. 3 (κ = 0.95). With no repeat growths and no error bars, the acquisition function may be chasing measurement noise rather than process-property trends. The stopping argument is also internally weakened: Section 3 itself states that 'increasing Sc power could further decrease the coercive field,' so the assertion that the minimum had been located at Iteration 7 is not supported by the GP posterior shown. These issues do not undermine the automation template, which is well documented, but they are load-bearing for the headline demonstration.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper describes a retrofit of an existing four-cathode confocal magnetron sputtering chamber to enable automated, semi-autonomous thin-film deposition. The retrofit centralizes control in a LabVIEW-based DAQ system, replaces manual pneumatic valves and lead screws with solenoid valves and stepper motors, implements time- and event-based recipe execution, and ingests deposition logs and metadata into the LiST data platform. As a demonstration, the authors run a campaign on wurtzite Al1-x-yScxByN in which Gaussian Process Regression (GPR) with a rational quadratic kernel suggests subsequent processing conditions (N2 flow, Sc power, B power) to minimize coercive field EC. They report a champion film with EC = 2.3 MV/cm and composition Al0.668Sc0.33B0.002N, and they claim that the Bayesian loop located the field minimum by iteration 7. The main contribution is a transferable template for upgrading legacy deposition chambers for data-driven experimentation.","tokens_in":7856,"tokens_out":4653,"duration_ms":49572,"significance":"If taken at face value, the paper provides a practical, lower-cost route to bring existing academic deposition tools into the loop of automated, Bayesian-optimization-driven materials discovery. The hardware narrative is coherent and unusually specific, with component inventories (Tables 1-2), control pseudocode (Algorithm 1), GUI and wiring documentation, and publicly available recipes, deposition logs, and XRD data. The demonstration also illustrates a legitimate closed loop: GPR predictions are checked against new measurements at each iteration, so the workflow itself is sound. The significance of the materials-specific result, however, is currently limited by the absence of any metrological detail on the coercive-field measurements that drive the optimization and validate the 2.3 MV/cm claim.","major_comments":[{"comment":"The coercive-field data that drive the GPR loop are not metrologically characterized. The manuscript never states how EC was measured: no P-E loop frequency, maximum applied field, top-electrode geometry or material, leakage/displacement-current correction, or thickness determination method is given. Because EC is a derived quantity, uncontrolled run-to-run or film-to-film differences in measurement protocol can shift values by amounts comparable to the 5% improvement threshold encoded in Eq. (3) through kappa = 0.95. Without repeat growths, error bars, or a baseline comparison against previously reported AlScN or AlScBN values, the optimizer may be fitting measurement noise rather than process-property trends. Please add the full electrical measurement protocol, thickness determination, repeated growths at least for the champion recipe, and a statement of measurement uncertainty, or explicitly reframe the demonstration as a workflow proof rather than a validated material optimization.","section":"Section 3, Eqs. (1)-(3), Figures 4-5"},{"comment":"The claim that the field minimum had been located at Iteration 7 is not supported by the paper's own posterior. The text states that increasing Sc power could further decrease EC and that the maximum Sc power was selected only for system safety, which means the reported optimum sits at or near the boundary of the searched range. With an optimum at the bound, the GP posterior does not justify a convergence claim. Please either extend the parameter domain or relax the safety bound in a supplementary experiment, or temper the conclusion to say that the minimum was located within the imposed safety-bounded processing window.","section":"Section 3, stopping criterion and Figure 4"},{"comment":"The number of films in the demonstration is inconsistent. The text says four films were produced with randomly selected parameters and ten subsequent films were produced with GPR-guided recipes (14 total), while the Figure 4b caption refers to \"the ten films produced at each iteration\" and the text later says \"the growths were stopped at 10 films.\" Please clarify how many iterations are shown in each panel, which iterations correspond to random seeding versus GPR guidance, and make the caption and axis labels consistent with the text.","section":"Section 3, Figure 4 caption and text"},{"comment":"The acquisition function treats the GP predictive variance as if it represented uncertainty in the process-property relationship, but the model receives only point values of EC with no measurement noise. The variance maps in Figure 5b therefore conflate sparse sampling of the parameter space with experimental reproducibility. Please state this limitation explicitly, or include an additive noise term in the GP and propagate the measurement uncertainty through the acquisition loop.","section":"Section 3, Eq. (3) and GP variance"}],"minor_comments":[{"comment":"The values of the GPR kernel hyperparameter alpha = 0.05 and the acquisition threshold kappa = 0.95 are stated without justification or sensitivity analysis; a short paragraph on how these were chosen and how sensitive the results are to them would strengthen the reproducibility of the method.","section":"Section 3, parameter choices"},{"comment":"The text says \"The film heights and coercive fields were then input into the GPR model,\" but Eq. (1) defines the input p as processing parameters only; please clarify whether film height (thickness) is an input feature, an output, or used only to derive EC.","section":"Section 3, first paragraph after Eq. (3)"},{"comment":"In Algorithm 1, the Event Mode loop is written as \"while |[sysVals] - [sysTargets]| < [tolerances] do,\" which appears inverted: the loop should continue while the system is outside the tolerance band, not inside it. Please correct the pseudocode.","section":"Algorithm 1, Event Mode loop condition"},{"comment":"The captions for Figure 5 do not define the color scale or units for EC and sigma in the three-dimensional plots, nor state how the two-dimensional slices are selected; adding colormap legends and slice descriptions would aid interpretation.","section":"Figure 5"},{"comment":"The composition of the champion film is reported as Al0.668Sc0.33B0.002N from XPS, but no detection limit, fitting procedure, or uncertainty is given; a one-sentence note on precision would be useful given the very small boron fraction.","section":"Section 3, XPS composition"}],"recommendation":"major_revision","confidential_remarks":"The paper is best viewed as an instrumentation and workflow contribution rather than a validated materials-discovery result. The automation template is solid and the public data availability is a genuine strength. The main risk is the unverified coercive-field demonstration; if the authors can supply measurement details, error bars, and one or two repeat growths, the central claim becomes supportable. If they cannot, they should soften the material-optimization claim to a workflow demonstration, which would still be publishable. I see no reason to question the integrity of the work; the missing electrical characterization details look like routine omissions rather than evidence of a deeper problem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The automation piece is the real contribution here. The authors give a concrete, reproducible recipe for taking a manual multi-cathode sputter chamber and making it remote-controllable, recipe-driven, and data-logged: hardware inventory, relay and stepper upgrades, LabVIEW control logic in pseudocode, and a documented path into the LiST data platform. That is exactly the kind of practical template many academic labs need, and it is presented with enough specificity that a competent grad student could follow it. The GPR-guided campaign on AlScBN is a reasonable illustration of the workflow, and the paper is honest that the Bayesian optimization machinery is standard and cited to prior work. I believe the system works as described and that the demonstration is not a Potemkin experiment; the data are even linked publicly.\n\nThe soft spot is precisely where the stress-test note lands. The coercive field is the quantity being minimized, but the paper never says how it was measured. No P-E loop frequency, no maximum applied field, no electrode geometry, no leakage treatment, no thickness method. Since coercive field is derived from a measurement, not read directly, run-to-run and film-to-film protocol differences can easily be as large as the 5% improvement threshold in the acquisition function. With no repeat growths and no error bars, the GPR loop may be chasing measurement noise. And the stopping argument is internally weakened: the text itself says that higher Sc power might lower the coercive field further, but the top of the Sc range was capped for safety. So calling Iteration 7 the minimum is an assertion, not a demonstrated optimum. That matters because the minimization claim is the paper's evidence that the autonomous loop works.\n\nIs this fatal? Not for the main purpose of the paper. The retrofitting template does not depend on whether 2.3 MV/cm is the global optimum. But the demonstration should not be oversold. If the authors add a measurement protocol section, uncertainty quantification from repeated or duplicate growths, and a clear statement that the campaign stopped at a practical operating constraint rather than a proven minimum, the paper would stand. Without those, the coercive-field result should be read as illustrative, not as a verified process-property optimum.\n\nWho is this for? Any lab running legacy deposition hardware that wants to move toward data-driven experimentation. It deserves peer review, but the referee should push on the electrical characterization details. I would take it to a reading group and probably cite the automation template; I would not cite the 2.3 MV/cm number in my own work.","headline":"A genuinely useful retrofitting template for autonomous sputter deposition, with a demonstration whose coercive-field numbers are not yet backed by a stated measurement protocol.","tokens_in":8484,"tokens_out":907,"would_cite":true,"duration_ms":12466,"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 retrofitted sputter chamber finds a 2.3 MV/cm ferroelectric film","keywords":["magnetron sputtering","thin film deposition","automation","Bayesian optimization","Gaussian process regression","ferroelectric thin films","AlScBN","coercive field"],"falsifier":"Repeatedly grow the same nominal recipe (or remeasure the same film) and show that coercive field values vary by more than the 5% improvement threshold used by the acquisition function; alternatively, rerun the same Bayesian loop from a different random seed and observe that the minimum coercive field and the composition achieving it shift significantly. If either happens, the optimization loop is fitting measurement noise rather than a stable process-property relationship.","tokens_in":7482,"feed_emoji":"⚙️","tokens_out":3771,"duration_ms":42558,"temperature":0.7,"pith_summary":"The paper shows how an existing, manually operated magnetron sputtering chamber can be upgraded with modest hardware additions and a central control program to become a semi-autonomous deposition platform. The authors replace manual valves and knobs with addressable actuators, connect all instruments to a data acquisition chassis, and run deposition through human-readable recipe files that also log sensor data. They then close the loop with a Gaussian process regression that proposes the next growth conditions, balancing the lowest predicted coercive field against the most uncertain region of parameter space. In a demonstration with wurtzite Al1−x−yScxByN thin films, the system produced a film with a coercive field of 2.3 MV/cm after four seeding runs and ten guided runs. The template is presented as a general route for making legacy deposition infrastructure compatible with data-driven, autonomous experimentation.","feed_headline":"Retrofitted sputter chamber finds 2.3 MV/cm ferroelectric film","feed_subtitle":"A bolt-on automation template turns a manual deposition chamber into a self-guiding materials lab.","key_machinery":"The central object is a recipe-executing control loop built around a data acquisition chassis and a customized control program that reads spreadsheet-formatted recipe files, sets instrument setpoints, logs sensor values on a defined cadence, and advances through either time-based or event-based steps. On top of this control layer sits Gaussian Process Regression with a rational quadratic kernel, used to model the unknown function from process parameters to coercive field and to propose the next experiment: the acquisition function returns the arg minimum of the predicted mean if that mean improves at least 5% over the current best, and otherwise returns the point of maximum predictive variance. Together these pieces couple automated sputter source power, gas flow, shutters, substrate stage motion, and data logging into a closed loop of synthesis, characterization, and suggestion.","core_discovery":"The paper claims that a legacy, academic-style sputter deposition chamber can be retrofitted into a semi-autonomous experimental platform by following a three-part strategy: automate the hardware (pneumatic shutters, substrate stage, and shutter motion), centralize control and data logging in a recipe-driven program, and automatically ingest all deposition logs into a sample-tracking repository. As proof of concept, the upgraded system is used to minimize the coercive field of wurtzite Al1−x−yScxByN thin films as a function of N2 gas flow, Sc target power, and B target power. A Gaussian process with a rational quadratic kernel suggests the next set of parameters—picking the predicted minimum if it improves by at least 5% over the current best, or the point with the highest predictive variance otherwise. After fourteen total growths, the best film has a coercive field of 2.3 MV/cm and composition Al0.668Sc0.33B0.002N, and the paper concludes that the retrofitting template can generalize to other custom research tools.","pith_inferences":["Because the Bayesian loop was seeded with only four random points, the reported 2.3 MV/cm is a statement about the algorithm's trajectory under one initialization, not a guarantee that this is the global minimum in the bounded parameter space.","A stronger demonstration of the closed loop would compare the converged minimum against a random-sampling baseline under the same measurement protocol, which would separate the benefit of the acquisition function from the benefit of simply growing many films.","The general retrofitting template should extend naturally to other synthesis systems (e.g., pulsed laser deposition, chemical vapor deposition) that share the same pattern of manual valves, stepper-motor stages, and serial-controlled instruments.","The paper does not provide error bars or repeat growths on coercive field, so a follow-up with replicated runs under the same recipe would clarify whether the optimization is tracking true process-property variation or run-to-run measurement noise."],"forward_implications":["Academic and small-scale labs can convert existing manual deposition chambers into autonomous experimentation platforms without purchasing a new instrument, provided the components have serial or analog interfaces that can be remotely addressed.","Bayesian-guided parameter search can locate a low coercive field in a three-dimensional deposition parameter space with roughly fourteen growths instead of a dense grid search.","Centralized recipe files and synchronized sensor logs make each growth reproducible and machine-readable, so the resulting datasets can support later analysis, digital twins, or further machine learning.","The same acquisition logic can be ported to optimize other process-property relationships, such as minimizing leakage current or maximizing piezoelectric response, by swapping the measured figure of merit.","Forcing all instrument data through a single controller and into a sample-tracking repository creates a permanent, queryable record linking each film's recipe, in situ logs, and ex situ characterization."],"supporting_citations":[{"why":"Supplies the fixed growth parameters (substrate temperature, Al target power, substrate bias, deposition time) that define the baseline recipe used for all films.","marker":"[19]"},{"why":"Provides the Gaussian Process Regression methodology used to model the process-property relationship and propose the next deposition parameters.","marker":"[20]"},{"why":"Gives the rational quadratic covariance kernel, which sets the distance-weighted similarity structure of the Gaussian process.","marker":"[21]"},{"why":"Describes the Lifetime Sample Tracking (LiST) platform used for automated ingestion and storage of deposition logs and metadata.","marker":"[18]"},{"why":"Establishes AlScN as a wurtzite ferroelectric, the material family from which the AlScBN films in this study derive.","marker":"[15]"},{"why":"Contextualizes growth of wurtzite ferroelectrics and frames the material system being optimized.","marker":"[16]"}],"fun_headline_variants":["Automation retrofit for sputter chambers finds 2.3 MV/cm ferroelectric","Legacy sputter tool becomes semi-autonomous for ferroelectric films","Bolt-on automation for magnetron sputtering yields low-coercivity AlScBN","Self-driving sputter chamber cuts coercive field to 2.3 MV/cm","Closing loop on old deposition tool minimizes ferroelectric coercive field"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The coercive field values used to train the Gaussian process are accurate, reproducible, and directly comparable across films, even though the paper reports no error bars, no repeat growths, and no benchmark against prior AlScN or AlScBN values.","fun_headline_variants_meta":{"raw":{"variants":["Automation retrofit for sputter chambers finds 2.3 MV/cm ferroelectric","Legacy sputter tool becomes semi-autonomous for ferroelectric films","Bolt-on automation for magnetron sputtering yields low-coercivity AlScBN","Self-driving sputter chamber cuts coercive field to 2.3 MV/cm","Closing loop on old deposition tool minimizes ferroelectric coercive field"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000203,"raw_usage":{"total_tokens":1365,"prompt_tokens":906,"completion_tokens":459,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":522,"completion_tokens_details":{"reasoning_tokens":356}},"tokens_in":522,"tokens_out":459,"duration_ms":5316,"temperature":1.0,"reasoning_tokens":356,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:28:29.183838+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeatedly grow the same nominal recipe (or remeasure the same film) and show that coercive field values vary by more than the 5% improvement threshold used by the acquisition function; alternatively, rerun the same Bayesian loop from a different random seed and observe that the minimum coercive field and the composition achieving it shift significantly. If either happens, the optimization loop is fitting measurement noise rather than a stable process-property relationship.","supporting_citations":[{"cited_title":"Mercer, C","cited_arxiv_id":null,"evidence_quote":"Supplies the fixed growth parameters (substrate temperature, Al target power, substrate bias, deposition time) that define the baseline recipe used for all films."},{"cited_title":"Williams, C","cited_arxiv_id":null,"evidence_quote":"Provides the Gaussian Process Regression methodology used to model the process-property relationship and propose the next deposition parameters."},{"cited_title":"Duvenaud, Automatic model construction with Gaussian processes, Ph.D","cited_arxiv_id":null,"evidence_quote":"Gives the rational quadratic covariance kernel, which sets the distance-weighted similarity structure of the Gaussian process."},{"cited_title":"Lifetime Sample Tracking (LiST): A Data Platform for Materials Science","cited_arxiv_id":"2606.17469","evidence_quote":"Describes the Lifetime Sample Tracking (LiST) platform used for automated ingestion and storage of deposition logs and metadata."},{"cited_title":"Fichtner, N","cited_arxiv_id":null,"evidence_quote":"Establishes AlScN as a wurtzite ferroelectric, the material family from which the AlScBN films in this study derive."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contextualizes growth of wurtzite ferroelectrics and frames the material system being optimized."}],"review_version":1}