Research Thermoelectric Semiconductors
Build falsifiable, evidence-traceable research from transport physics to a working device. Treat peak material zT as one link, never as proof of useful device performance.
Route the task
- Use this Skill for broad field maps, proposals, experiment plans, mechanism audits, metrology, contacts, modules, cooling, and material-to-device translation.
- Activate
develop-flexible-bismuth-telluride for bendable/conformal Bi2Te3-family films, composites, or wearable devices.
- Activate
engineer-gete-thermoelectrics for GeTe chemistry, phase transition, defects, contacts, legs, or modules.
- Read foundations.md for equations, units, coupling, and device translation.
- Read research-methodology.md before designing experiments or diagnosing results.
- Read evidence-and-reporting.md before comparing papers, values, or device claims.
Frame the question
Write six items first: function; hot/cold boundary conditions; material/film/leg/module scale; controllable variables; primary observables; and a result that supports, weakens, or falsifies the hypothesis. Reject “raise zT” until temperature window, carrier type, process, geometry, and device target are defined.
Build this chain without skipping links:
chemistry/process → phase + defects + texture + interfaces → carriers + phonons → S, sigma, kappa versus T → contacts/legs → module → system value
For every mechanism, report intervention, predicted structural change, independent discriminator, predicted transport signature, confounders, control, and falsification condition.
Use stage gates
- Reproduce a credible baseline and quantify batch variance.
- Screen one mechanistic axis or a justified factorial design.
- Connect verified chemistry/structure to independent transport observables.
- Repeat the winner in independent batches through the service window.
- Build junction, leg, or device coupons and measure actual operating temperatures and load curves.
- Stress under relevant temperature, atmosphere, current, thermal cycling, and mechanics.
Combine S, sigma, and kappa into zT only for the same material state, direction, temperature, and equivalent specimens. Preserve raw dimensions, density, calibration, fitting assumptions, specimen identity, and uncertainty.
Deliver decisions
Return the smallest useful artifact: field map; hypothesis/DoE matrix; claim–evidence audit; symptom/cause/discriminator diagnosis; or device translation plan. End with what is measured, what is inferred, what is unknown, the next decisive experiment, and the stopping rule.
Never infer efficiency from peak zT, lattice thermal conductivity from an unjustified Lorenz number, scalability from a process label, or device value from coupon power factor alone.
1---2name: research-thermoelectric-semiconductors3description: Route and execute rigorous thermoelectric semiconductor and device research, including field mapping, transport-physics reasoning, literature audits, hypotheses, experiment design, metrology, uncertainty, materials-to-device translation, and failure analysis; use for broad thermoelectric questions spanning materials, contacts, modules, cooling, or power generation, and route flexible Bi2Te3 or GeTe work to their dedicated Skills.4---5
6# Research Thermoelectric Semiconductors
7
8Build falsifiable, evidence-traceable research from transport physics to a working device. Treat peak material `zT` as one link, never as proof of useful device performance.
9
10## Route the task
11
12- Use this Skill for broad field maps, proposals, experiment plans, mechanism audits, metrology, contacts, modules, cooling, and material-to-device translation.
13- Activate `develop-flexible-bismuth-telluride` for bendable/conformal Bi2Te3-family films, composites, or wearable devices.
14- Activate `engineer-gete-thermoelectrics` for GeTe chemistry, phase transition, defects, contacts, legs, or modules.
15- Read [foundations.md](references/foundations.md) for equations, units, coupling, and device translation.
16- Read [research-methodology.md](references/research-methodology.md) before designing experiments or diagnosing results.
17- Read [evidence-and-reporting.md](references/evidence-and-reporting.md) before comparing papers, values, or device claims.
18
19## Frame the question
20
21Write six items first: function; hot/cold boundary conditions; material/film/leg/module scale; controllable variables; primary observables; and a result that supports, weakens, or falsifies the hypothesis. Reject “raise zT” until temperature window, carrier type, process, geometry, and device target are defined.
22
23Build this chain without skipping links:
24
25`chemistry/process → phase + defects + texture + interfaces → carriers + phonons → S, sigma, kappa versus T → contacts/legs → module → system value`
26
27For every mechanism, report intervention, predicted structural change, independent discriminator, predicted transport signature, confounders, control, and falsification condition.
28
29## Use stage gates
30
311. Reproduce a credible baseline and quantify batch variance.
322. Screen one mechanistic axis or a justified factorial design.
333. Connect verified chemistry/structure to independent transport observables.
344. Repeat the winner in independent batches through the service window.
355. Build junction, leg, or device coupons and measure actual operating temperatures and load curves.
366. Stress under relevant temperature, atmosphere, current, thermal cycling, and mechanics.
37
38Combine `S`, `sigma`, and `kappa` into `zT` only for the same material state, direction, temperature, and equivalent specimens. Preserve raw dimensions, density, calibration, fitting assumptions, specimen identity, and uncertainty.
39
40## Deliver decisions
41
42Return the smallest useful artifact: field map; hypothesis/DoE matrix; claim–evidence audit; symptom/cause/discriminator diagnosis; or device translation plan. End with what is measured, what is inferred, what is unknown, the next decisive experiment, and the stopping rule.
43
44Never infer efficiency from peak `zT`, lattice thermal conductivity from an unjustified Lorenz number, scalability from a process label, or device value from coupon power factor alone.
gabrielmoreira