From material selection to discovery

Elixir, the reasoning layer for materials.

Elixir reads a material problem in plain language, breaks it into the properties that actually matter, and narrows a knowledge base of +600,000 materials down to a short, ranked shortlist. Every candidate comes with its evidence, what would rule it out, and what to test next.

The material universe

You don’t choose from every material that exists. You choose from what you can see.

A material that was never measured was never rejected. It simply never appeared as an option. Elixir predicts the properties nobody has measured, then screens what survives against the constraints a real decision has to clear.

Live demo

See how Elixir reasons, end to end.

Elixirlive demo
Selection
Mode
Selection
Field
Aerospace alloys
The question

Find replacements for titanium alloys that stay strong above 500 °C, hold yield strength over 600 MPa, and resist creep and oxidation.

Illustrative demo · representative values

Building proxy chains
yield strength proxy ← bulk modulus K and shear modulus G (E = 9KG / (3K+G))

How Elixir reasons

Every answer follows the same path. Elixir reads the question, breaks it into the properties that actually matter, then narrows a knowledge base of +600,000 materials down to a short, ranked list. Each candidate carries its evidence, what would rule it out, and what to test next. Transparency, never guarantees.

1 · Decompose the plain-language problem into measurable property proxies.
2 · Scan +600,000 materials linked to processes and properties.
3 · Rank a shortlist of named candidates, each with evidence and failure modes.
4 · Plan the next tests, checks, and supplier questions worth running first.

Reasoning, not retrieval

Elixir reasons across materials, properties, conditions, and failure modes. It does not just search a database.

Evidence on every claim

Each property and recommendation carries its evidence level, its uncertainty, and what would rule it out.

Your data, isolated

Bring your own constraints, specs, and results. They stay in a namespace that is yours alone.

Open data: thermoelectrics

A worked example: 500 thermoelectric materials distilled from more than 40,000 candidates, with Seebeck coefficient, thermal conductivity, and zT.

Open data: critical material alternatives

Substitutes for export-controlled and supply-constrained elements, lined up with their cost and performance trade-offs.

How it reasons

Every ranking carries the graph behind it.

A property value on its own does not tell you whether a material survives your application. Elixir connects each candidate to the conditions it has to withstand and the failure modes tied to those conditions, so the reasoning behind a rank is there to check, not just the rank itself.

CandidateNb-silicidePropertyStrength @ 800 °CPropertyOxidation resistanceConditionSustained heatFailure modeHigh-temp oxidationMitigationProtective coating

The Digital Lab

Bring a candidate. Leave with a validated one.

Upload a structure and Digital Lab runs it through the same benchmark and backfill pipeline behind every property on this page. You get a confidence-scored result back, not a black box.

1
Upload structure
2
Benchmark + backfill
3
Compare to precedent
4
Confidence-scored output
Nb-silicide variant, uploaded structure
Bulk modulus228 GPaTested
Band gap0.41 eVPredicted
Formation energy-0.62 eV/atomModeled