npj-computational-materials - SKILL.md Agent Skill

name: npj-computational-materials description: Use when targeting npj Computational Materials (npj Comput. Mater.) or deciding whether a computational or data-driven materials manuscript fits this open-access Springer Nature venue. Encodes the journal's fit, framing, method-and-evidence bar, house style, official-submission re-check, and desk-reject heuristics.

npj Computational Materials (npj-computational-materials)

Journal positioning

npj Computational Materials is an open-access Springer Nature journal (part of the Nature Partner Journals series) dedicated to computational and data-driven materials science. Its defining character is rigorous computation that delivers materials insight or discovery: first-principles and electronic-structure methods (DFT and beyond), molecular dynamics, multiscale modeling, machine learning and materials informatics, high-throughput screening, and the data infrastructure that supports them. The journal rewards work where the computational approach yields a generalizable conclusion, a predictive capability, or a discovery of broad materials interest — not routine single-system calculations or method applications without a clear advance. It values methodological rigor, reproducibility, and FAIR data practices, and increasingly experimental validation or testable predictions strengthen a submission. Readership spans computational materials scientists, condensed-matter physicists, materials chemists, and the growing materials-ML and informatics community. This skill is a fit / venue-selection / re-framing tool. It does not replace the journal's current official submission guidelines. Before submitting, re-check the live author instructions on the npj Computational Materials site.

When to trigger

The author names npj Computational Materials as the target for a computational or data-driven materials study of broad interest.
A manuscript uses DFT, MD, machine learning, or high-throughput screening to reveal a generalizable materials principle, predict new materials, or develop a method, and the author is choosing between this venue and Nature Materials or Advanced Materials.
A paper's primary contribution is computational discovery, a predictive model, or materials informatics rather than an experimental result.
The author needs the journal's rigor, reproducibility, and FAIR-data bar plus desk-reject criteria before submission.

Scope & topic fit

First-principles and electronic-structure studies (DFT, GW, DMFT, beyond-DFT) that establish a materials principle, mechanism, or design rule of broad relevance.
Molecular dynamics and multiscale modeling resolving structure-property or kinetic questions in materials.
Machine-learning interatomic potentials, surrogate models, and ML-accelerated simulation with demonstrated accuracy and transferability.
Materials informatics and data-driven discovery: descriptor design, property prediction, generative/inverse design, and active learning that yield validated materials insight.
High-throughput screening and computational materials databases that produce actionable candidates or design principles.
Methodological and workflow advances (new functionals, algorithms, automation, benchmarking) that demonstrably improve computational materials science.

Method & evidence bar

The generalizable conclusion, prediction, or methodological advance must be stated in one or two sentences; a single-system calculation without broader insight is misfit.
Computational rigor is mandatory: convergence tests, justified functionals/parameters, error estimates, and validation against known references or experiment where available.
Machine-learning work must report proper train/validation/test splits, out-of-distribution behavior, uncertainty quantification, and baselines; performance claims compared to relevant prior models.
High-throughput and screening studies must state the search space, filters, and confidence in candidates, with experimental validation or testable predictions strengthening the case.
Reproducibility is central: input files, structures, code/workflows, and datasets should be deposited following FAIR principles in recognized repositories.
Claims of accuracy, transferability, or discovery must be benchmarked against the best current methods and data, not the authors' baseline alone.

Structure & house style

npj Computational Materials uses a Nature-style format; Articles are the primary type, with concise abstracts and integrated narrative — re-check current types and limits on the live site.
The introduction frames the materials question and the gap the computation resolves; the readership is expert, so background is minimal and the advance is stated early.
Figures must be efficient and quantitative: each carries a key computational result, with validation and benchmark comparisons made explicit.
Methods describe the computational setup completely — codes, functionals/force fields, parameters, convergence criteria, ML architectures, and data provenance — for full reproducibility.
Supplementary Information carries extended computational details, additional results, and validation data; data and code availability statements are expected.
Claims of predictive power or discovery must be supported by validation against experiment or by clearly testable predictions.

Official-submission checklist

Before giving submission-ready advice, read ../../resources/source-basis.md and ../../resources/official-source-map.md; start from the official source anchors for this journal family, then cite the current journal-specific page you checked.
Search the live site for "npj Computational Materials author guidelines" and follow the current Springer Nature version.
Re-check current article types, length/figure expectations, and abstract format; confirm Methods and Supplementary Information conventions.
Re-check the data- and code-availability requirements and FAIR/repository expectations central to this journal.
Re-check open-access/APC, licensing, competing-interests, funding, and AI-use disclosure requirements; confirm preprint policy (arXiv posting compatibility).
If the live official instructions conflict with this skill, the official instructions win.

Pre-submission self-check

One sentence — the generalizable conclusion, prediction, or method advance, and why it matters across materials.
Computational rigor (convergence, functional/parameter choices, error estimates, reference validation) is documented.
ML work reports proper splits, OOD behavior, uncertainty, and baselines; screening states search space and candidate confidence.
Inputs, structures, code/workflows, and datasets are deposited following FAIR principles.
Predictions are validated against experiment or stated as clearly testable.
The paper is positioned against recent npj Computational Materials / Nature Materials computational work on this question.

Common desk-reject triggers

A routine single-system DFT or MD calculation with no generalizable principle, prediction, or method advance.
A machine-learning model with no proper validation, baselines, uncertainty, or out-of-distribution assessment.
A high-throughput screen with no stated confidence in candidates and no validation or testable predictions.
Computational results with missing convergence tests, unjustified parameters, or no reference/experimental validation.
Absent or inadequate data/code availability inconsistent with the journal's FAIR-data expectations.

Re-routing decision

A fundamental materials discovery whose primary impact is the new physics/chemistry of the material itself, computation supporting: nature-materials.
A combined computation-plus-experiment study where a synthesized, characterized functional material is the headline result: advanced-materials.
An energy-materials computational study where energy-device metrics are central: an energy-materials venue.
A broad cross-domain methods or data paper without materials-specific framing: a general computational-science venue.

Output format

[Fit] High / Medium / Low (one-line reason)
[Target] npj Computational Materials
[Topic tags] <2–3 closest topics>
[Method/evidence] <does the computation yield a generalizable conclusion or prediction with rigor, validation, and FAIR-data reproducibility?>
[Top risk] <the single most likely reason for rejection>
[Official items to re-check] <article types/limits / data-code FAIR deposition / open-access & licensing / disclosure / preprint policy>
[Re-route suggestion] <if not a fit, a better-matched venue>