科研 Skill
材料信息学性质预测
materials-informatics-ml
Covers ML for materials property prediction and generative design: descriptor-based models, crystal GNNs (CGCNN/MEGNet-class). Use for materials informatics distinct from interatomic potentials.
SKILL.md
Materials Informatics and ML
Overview
Materials informatics applies machine learning to predict material properties from composition/
structure (screening candidates faster than DFT or experiment) or to generate novel candidate
structures. This is distinct from computational-materials/references/ml-interatomic-potentials.md,
which uses ML to approximate a potential energy surface for simulation — here, ML predicts a target
property (band gap, formation energy, hardness) directly, or generates new structures, as an end
goal in itself.
When to use this skill
Activate when the request involves:
- Materials property prediction, ML for materials, CGCNN, MEGNet, graph neural network for crystals, generative model for materials, materials descriptors/featurization
- Terms: matminer, crystal graph, composition-based feature vector, train/test split for materials data
- "Predict this property from composition/structure," "featurize these materials for ML," "generate candidate structures with property X"
Core usage
Descriptor-based property prediction (matminer featurization + scikit-learn)
from matminer.featurizers.composition import ElementProperty
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
import pandas as pd
featurizer = ElementProperty.from_preset("magpie")
df = featurizer.featurize_dataframe(df, col_id="composition") # df has a pymatgen Composition column
feature_cols = [c for c in df.columns if c.startswith("MagpieData")]
X_train, X_test, y_train, y_test = train_test_split(df[feature_cols], df["target_property"], test_size=0.2, random_state=42)
model = RandomForestRegressor(n_estimators=200, random_state=42)
model.fit(X_train, y_train)
Graph neural network property prediction (conceptual — CGCNN-class model)
# Crystal graph convolutional networks represent a structure as a graph (atoms
# as nodes, bonds/neighbors as edges) and learn representations directly from
# structure rather than hand-engineered descriptors — typically outperforming
# descriptor-based models when enough training structures are available.
# Training requires a graph-construction step (e.g. via pymatgen's structure
# graph tools) feeding into a PyTorch Geometric-style GNN — see the specific
# package (CGCNN, MEGNet, ALIGNN) for its current data-pipeline conventions,
# which differ meaningfully between packages.
Model evaluation with materials-appropriate cross-validation
from sklearn.model_selection import GroupKFold
def composition_aware_cv(X, y, chemical_system_groups, n_splits=5):
"""Random K-fold on materials data can leak information when chemically
similar/related compounds end up split across train and test — grouping by
chemical system (or another structural-similarity grouping) gives a more
honest estimate of generalization to genuinely new chemistry."""
cv = GroupKFold(n_splits=n_splits)
return list(cv.split(X, y, groups=chemical_system_groups))
Validation & Pitfalls
Canonical references: Ward et al. (2016), "A general-purpose machine learning framework for predicting properties of inorganic materials," npj Computational Materials (the Magpie featurization approach); Xie & Grossman (2018), "Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties," Physical Review Letters (CGCNN).
- Random train/test splitting on materials data typically overstates real-world generalization, the same core problem as in other domains' ML but with a materials-specific cause: many datasets contain multiple compositions/structures from the same chemical system or even the same base structure with minor substitutions. A model can "predict" a held-out entry partly by having seen a near-identical relative in training. Use grouped/chemical-system-aware splitting (as above) for an honest generalization estimate, especially when the claim is about predicting genuinely novel chemistry.
- A model trained on DFT-computed labels inherits DFT's own systematic errors (e.g. underestimated band gaps for GGA-trained models) — it will confidently reproduce those errors, not correct them. State what labels a model was trained on, and don't present ML-predicted properties as more accurate than the ground truth they were trained against.
- Descriptor-based models are only as good as the descriptors' ability to capture what determines the target property — a composition-only descriptor (like the Magpie example above) cannot distinguish between polymorphs with identical composition but different structure/properties; use structure-aware descriptors or graph-based models when polymorph-level distinction matters for the question.
- Extrapolation outside a model's training distribution (new element combinations, property ranges far from training data) is unreliable and often silent — a model doesn't know it's extrapolating unless specifically designed to flag it (e.g. via an ensemble-based uncertainty estimate or a distance-to-training-data metric); check whether a candidate material is actually within the training distribution before trusting a screening prediction for it.
- Generative models can produce chemically implausible or synthetically inaccessible structures
— a generated candidate passing a property-prediction filter still needs independent validation
(stability via
materials-structure-analysis's phase-diagram tools, and realistic synthesizability assessment) before being treated as a real candidate, not just a model output that looks reasonable.
示例提示词
我想训练一个图神经网络从晶体结构预测带隙并探索新候选材料。
所属插件
安装该插件会一并获得同领域的 42 个 Skill / MCP。
同插件的其他科研 Skill
- 贝叶斯优化材料发现Covers Bayesian optimization for materials and chemistry design: surrogate models, acquisition functions, batch/constrained BO. Use when planning sample-efficient experiments or simulations.
- CALPHAD热力学计算工作流Covers CALPHAD computational thermodynamics: database selection, equilibrium/metastable calculations, binary/ternary phase diagrams. Use when predicting phase stability or solidification paths.
- 材料第一性原理与分子动力学模拟Routes first-principles and classical materials simulation (DFT, MD, ML potentials) to the right specialized workflow. Use as an entry point for any simulation-based materials task.
- 蠕变力学建模与验证Selects, calibrates, and validates creep deformation and rupture models under sustained thermomechanical load, covering rate prediction. Use when assessing long-term high-temperature components.
- 晶体塑性有限元建模Covers crystal plasticity and CPFE modeling of metals: slip systems for FCC/BCC/HCP, Schmid factors. Use when simulating texture- and orientation-dependent plastic deformation.
- 衍射结构表征Covers diffraction-based structure characterization (XRD, neutron, SAED): space group determination, Rietveld refinement, phase identification. Use for any diffraction-pattern analysis technique.
- EBSD织构分析Analyzes EBSD orientation maps for grain size, misorientation, phase, and texture, with pole figures and twin-boundary identification. Use for electron backscatter diffraction workflows.
- 电化学性能测试分析Covers electrochemical characterization: cyclic voltammetry, battery cycling analysis, EIS. Use for electrochemistry-specific data analysis distinct from general materials characterization.
- 金属疲劳寿命评估Builds, calibrates, and validates fatigue assessments for metals and components: stress-life, strain-life, multiaxial critical-plane. Use when predicting fatigue life or crack growth.
- 多物理场有限元建模Covers finite element modeling for coupled materials, mechanics, and transport problems: weak-form construction, boundary conditions, mesh choice. Use for multiphysics FEM simulations.
- 连续尺度有限元力学建模Covers continuum-scale mechanical, thermal, and multiphysics simulation with Abaqus/COMSOL conventions, mesh generation. Use for macroscale FEM consuming measured or computed material properties.
- 断裂力学评估与建模Selects, constructs, and validates fracture-mechanics assessments: LEFM checks, elastic-plastic fracture mechanics. Use when assessing cracked materials or components.
- 高斯过程材料回归建模Covers Gaussian process regression for materials and chemistry datasets: descriptor selection, kernel design. Use when a probabilistic surrogate model with uncertainty is needed.
- 自驱动实验室闭环材料发现Covers combinatorial synthesis, robotic experimentation. Use for autonomous or robotics-driven experimental campaigns, a less standardized domain.
- 材料数据库可复现访问Covers reproducible API access to materials databases (Materials Project, AFLOW, OQMD, NOMAD, COD, ICSD-style sources): authentication. Use when programmatically querying these databases.
- 材料数据库跨库查询Covers querying public materials databases beyond Materials Project: AFLOW, OQMD, NOMAD, JARVIS. Use when pulling or comparing data from these live, API-integrated resources.
- 材料数据标准与互操作Covers data interoperability standards for materials science: CIF crystal files, the OPTIMADE query API. Use when structuring or exchanging materials data between tools.
- 材料结构特征化Covers converting structures, compositions, and trajectories into ML feature vectors: Magpie-style elemental statistics. Use when building inputs for materials ML models.
- 材料科学图表绘制规范Covers publication-quality figure conventions for materials science plots: phase diagrams, pole figures, stress-strain curves. Use when the deliverable is a figure, not just an analysis result.
- 材料生命周期评估Covers life-cycle assessment for materials and manufacturing routes: goal/scope definition, functional units, inventory construction. Use for environmental-impact studies of materials.
- 材料科学文献检索Covers formulating literature search queries for materials science questions and synthesizing/citing results responsibly. Requires a web search tool; it cannot substitute for one.
- 工程材料选型设计Selects engineering materials via material indices and Ashby charts, ranking candidates by performance per cost or mass. Use when picking a material class for a design constraint.
- 材料统计方法与实验设计Covers statistics shared across materials characterization and simulation: design of experiments. Use for cross-cutting statistical treatment of materials data.
- 材料结构与成分核心分析Covers core structure and composition analysis: reading/writing crystal structures, composition handling, symmetry analysis, phase diagrams. Use as a general-purpose structure-analysis toolkit.
常见问题
材料信息学性质预测能做什么?
Covers ML for materials property prediction and generative design: descriptor-based models, crystal GNNs (CGCNN/MEGNet-class). Use for materials informatics distinct from interatomic potentials。
如何安装材料信息学性质预测?
在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「材料信息学性质预测」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。
材料信息学性质预测需要哪些权限?
该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。
材料信息学性质预测经过官方审核吗?
是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。