materials-and-engineering

name	materials-and-engineering
description	Use for materials and device work: structure-property-processing reasoning, characterisation (XRD, SEM/TEM/EDS, AFM, DSC/TGA), mechanical testing and Weibull statistics for brittle failure, FEA mesh convergence and material models, SPICE and CFD sanity checks, tolerances and GD&T, ASTM/ISO/IEC method numbers, datasheets and derating, measurement uncertainty and calibration traceability, and open materials data.
version	1.0.0
author	Remedy
tags	["research","materials","engineering","characterisation","simulation","standards"]
requires	[]
tools	["analysis_env","analysis_run","analysis_ledger","data_profile","data_diff","stats_assumptions","stats_effect_size","power_analysis","manuscript_build","lit_search","lit_fetch","cite_add","cite_check","skill_activate","file_read","file_write"]
triggers	["\\b(stress[- ]strain|Young.?s modulus|yield strength|fatigue (?:life|limit|test)|fracture toughness|hardness test)\\b","\\b(\\bSEM\\b imaging|\\bTEM\\b imaging|\\bEDS\\b|\\bEBSD\\b|nanoindentation|tensile test|creep test)\\b","\\b(finite element (?:analysis|model|mesh)|\\bFEA\\b|ASTM [A-Z]?\\d{2,4}|ISO \\d{3,5})\\b","\\b(microstructure|grain (?:size|boundar\\w+)|sintering|anneal(?:ing)? (?:step|temperature)|phase diagram)\\b"]

Materials and engineering

Activate research-method first: skill_activate(skill="research-method"). It owns question framing, evidence standards, citation honesty and how to say "we do not know". Do not restate it. This pack is what materials and device work does differently.

The spine: processing - structure - property - performance

Every claim sits somewhere on that chain, and a claim only travels one link at a time. Processing history (composition, thermal path, deformation, atmosphere, cooling rate) produces a structure; structure produces properties; properties, plus geometry and loading, produce performance. A result reported without its processing history cannot be reproduced, and a property measured on one geometry does not transfer to another without an argument. Write the processing history into the methods section first, then the characterisation, then the property.

Measure with a named method

Give the standard's number, its issuing body and its year in the methods: ASTM International, ISO, IEC, ANSI, DIN, JIS, NIST. The number tells a reader the specimen geometry, conditioning, strain rate, fixture, sample count and reporting format in one token — which is exactly why it belongs in the paper and why "a tensile test" does not. Standards are revised: quote the edition you followed and check the issuing body's catalogue for the current one rather than trusting a remembered designation. If you deviate from the standard, say which clause and why.

Characterisation: match the technique to the claim

Each technique answers a bounded question, and a claim beyond that bound needs a second technique. XRD gives phases and lattice parameters, not morphology. SEM gives surface morphology; EDS gives semi-quantitative composition with poor light-element sensitivity and a micron-scale interaction volume, not trace analysis. TEM gives local structure at the cost of a thinned, possibly damaged, and certainly unrepresentative sample. AFM gives topography convolved with the tip. DSC/TGA give transitions and mass loss at a stated ramp rate and atmosphere. See references/characterisation-methods.md. Report instrument, settings and how many regions or specimens; one micrograph is an anecdote.

Statistics of failure

Strength of a brittle material is a distribution, not a value — the weakest-link statistics mean the mean alone is misleading and the size of the specimen matters. Fit Weibull, report the modulus, the characteristic strength, the confidence intervals and the number of specimens, and state the estimator. Use power_analysis to size the sample before testing and stats_assumptions on the data before choosing a test. references/mechanical-testing-and-weibull.md.

Simulation is a claim until it is validated

Verification asks whether the equations are solved correctly; validation asks whether they are the right equations for this system. Both are required, and they are separate statements. For FEA: mesh convergence with the observed convergence rate, element type and order justified, boundary conditions and contact stated, and the material model named with its validity range — linear-elastic results past yield are not results. For circuits: SPICE model provenance, corner and Monte-Carlo runs, convergence settings. For CFD: y+, turbulence model, grid-convergence index. Validate against a measurement or an analytic case and show the comparison. , .