Turn decades of materials data into sourced shortlists your engineers can test.
Lattice Graph brings 85 materials data sources (DFT, experimental structures, synthesis recipes, patents) into one graph and filters compositions against your constraints. Database values link to their sources.
Why now: the open DFT databases are now large enough to screen with, and they disagree. For each pair of Materials Project, OQMD and JARVIS, 17–18% of all the compositions the pair shares sit on the convex hull (within 1 meV/atom) in one database and off it in the other (our study, 62,803 compositions).
Ehull = 0.000 eV/atom
Look up a real compound before you sign up.
Built for teams who ship atoms, not just apps.
What is in the graph today, from our public API — the substrate behind every shortlist, from battery cathodes to next-generation ceramics.
Three steps from fragmented data to a sourced, ranked shortlist.
Lattice reconciles formulas across sources, scores how well they agree, and shows the source of each database value.
85 sources, one schema
DFT databases, experimental crystal structures, synthesis literature, and patents — normalized so LiFePO4 and FeLiO4P find each other.
Linked into a graph
Lattice links compositions, properties, synthesis routes, and patent families into a knowledge graph.
Rank with evidence
Filter by stability and structure, check how well the databases agree, and follow a candidate to its recipes where the literature has one.
The decisions it is built for.
Lattice is designed for materials screening decisions like these.
Shortlist candidates against multi-objective constraints.
Rank battery materials on capacity, voltage and supply risk at once, and see which ones sit on the Pareto front.
Find chemical neighbours.
Start from a formula you know — Lattice returns related compositions that share its elements, with how many sources and recipes back each one.
Know the risk before procurement asks.
Element-level concentration, top-producer and geopolitical scoring.
Screen a whole list at once.
Send up to 500 formulas and get synthesis-route counts, supply risk, and patent density back for each.
Start from a known route.
Literature synthesis recipes, searchable by target composition, each linked to the paper it came from.
Screen by the properties you need.
Filter by stability, space group, and crystal system — then check how well the databases that hold a formula agree on it.
Take the results with you.
Export results as CSV or JSON, or pull the same data into notebooks and your own systems through the REST API and Python SDK.
Familiar tools, purpose-built for materials.
Built for engineers, not data scientists. Use it in the browser, or pipe it into your existing notebooks via the SDK. The screen below is an illustrative mock-up with sample values.
Stability / band gap trade-off
One line from notebook to knowledge graph.
A Python SDK with typed sub-clients for each API domain, automatic retries, and lazy pagination. Use it in a Jupyter cell the same way you'd query a database.
from latticegraph import LatticeGraphClient client = LatticeGraphClient( api_key="mat_…", base_url="https://api.latticegraph.com", ) hits = client.materials.search( formula="LiFePO4", max_e_hull=0.05, limit=12, ) print(hits["total"], hits["has_more"]) # /v1/health takes no key — it is where the # counts at the top of this page are read from print(client.health.status()["version"]) # → 1.3.1
From mine deposit to candidate substitute.
Search 304,632 USGS mineral-site records in 166 countries, then go one step further: 24 computational candidates for critical-mineral recovery, rare-earth separation and battery recycling.
Candidates the platform has screened, for licence.
The same screens also propose candidate uses for known and computed compounds; many of the compounds are already known. None has a Lattice Graph lab measurement yet: each is a computational candidate, rated on our market estimate and the depth of its computational evidence, with a preliminary IP screen (not a freedom-to-operate opinion). Available for licensing or co-development.
Mapped to public business lines.
Our own analysis from public information: each company's business lines, mapped to the portfolios that could serve them, with the best-fit assets ranked.
Syensqo
Specialty materials — batteries, hydrogen, fluoropolymers, critical-mineral separations
View fit →Applied Materials
Semiconductor process equipment & materials engineering
View fit →Eaton
Intelligent power management — distribution, power electronics, data-center power
View fit →Lam Research
Deposition & etch equipment for advanced packaging and memory
View fit →Intel
Logic, advanced packaging & glass-core substrates
View fit →Samsung Electro-Mechanics
Advanced substrates, MLCC dielectrics & glass interposers
View fit →Absolics (SKC)
First-mover glass-substrate commercialization
View fit →Chemours
Performance chemicals & two-phase cooling fluids (Opteon)
View fit →Nick Harris, founder. CTO of VivaMed BioPharma; co-founder of MedSim.ai, FastRead.io and Formulai.
Judge the work as well as the bio: everything below is public today.
- A live API whose row counts you can read without a key at api.latticegraph.com/v1/health.
- Three open data studies, each with the rows behind it as a CSV, in the Data Atlas.
- Write-ups of what that data shows on the blog, including an optical-only high-k screen graded on two public datasets, and kill reports on the approaches that did not survive our own screens.
Currently in early access. Contact: nick@latticegraph.com
Ship the material, not the search for it.
We're onboarding early teams now. Sign up for early access and help shape the platform.