Computational screening for R&D laboratories
From a candidate to a decision.
Your laboratory has more candidates than it can ever test, and every real experiment costs months. We screen the space cheaply, rank it on the metric that actually decides, and tell you where the models stop being trustworthy — without a cluster, a specialist, or a six-figure licence.
Live today: acidic oxygen-evolution catalysts for PEM water electrolysis — activity, stability and iridium intensity, closed to cost per kilogram of hydrogen. The engine is built for more →
Calculator is free and needs no account. Early access includes free credits. New to this? Start with how it works — no chemistry background needed.
The chain in the hydrogen pack
Most tools stop at a predicted property
A number in eV is not a decision. The question a lab or a supplier actually has is whether a candidate is worth a synthesis slot — and that depends on activity, on whether it dissolves, on how much iridium it needs, and on what all of that does to cost.
verna.science runs the chain end to end and shows you every step, so you can check the reasoning rather than trust a black box.
The full explanation: what this replaces in a research workflow →
General-purpose interatomic potentials are trained mostly on bulk crystals and are measurably least accurate on oxide surfaces with oxygen adsorbates — the OER regime. We fine-tune for that regime.
Outside the validated composition domain we refuse to answer rather than return a plausible wrong number.
Every result carries its model version, method parameters and input/output hashes, and can be re-run to the same answer.
One engine, focused packs
The science is specific. The problem isn't.
A tool that is vague about the science is useless, and one that only ever works for a single chemistry is a dead end. So the engine is general and every pack is deliberately narrow.
A space too big to test
Thousands or millions of candidates. A lab gets through a handful a year.
Experiments costing months
A synthesis batch or a growing season — either way you cannot brute-force it.
Models needing infrastructure
The methods exist and are often open. Clusters and specialists are what most labs lack.
The deciding metric, too late
Cost or yield gets assessed after the work, so the search optimises the easy variable.
PEM electrolysis catalysts
Acidic oxygen-evolution catalysts for green hydrogen — activity, stability and iridium intensity, closed to cost per kilogram.
Explore this pack →
Ammonia cracking
Catalysts for releasing hydrogen from ammonia — the carrier problem, where hydrogen is shipped as ammonia and cracked on arrival.
See what it would do →
Methane pyrolysis
Splitting methane into hydrogen and solid carbon, avoiding CO2 at the point of production.
See what it would do →
CO2 conversion and e-fuels
Turning captured CO2 into fuels and chemicals — selectivity is the hard constraint, not just activity.
See what it would do →
Inside the hydrogen pack
Four things, one run
OER activity
Relaxation, slab generation, OH*/O*/OOH* adsorption and a thermodynamic overpotential descriptor.
Acidic stability
A dissolution-risk estimate under acidic OER conditions — the thing that actually kills PEM catalysts in service.
Cost closure
Properties propagated through a PEM cell and system model to kWh/kg, iridium g/kW and LCOH.
Provenance
Model version, method, hashes and a one-command re-run on every result. Citable.
Free tools
Useful before you sign up for anything
No account, no credits, no email. They run entirely in your browser.
LCOH calculator →
Turn catalyst performance and an operating point into kWh/kg H2, iridium intensity and levelised cost, with a full breakdown.
Iridium intensity →
Convert catalyst loading and power density to grams of iridium per kilowatt — and check it against GW-scale supply.
Overpotential converter →
Extrapolate a measured overpotential to a different current density with a Tafel slope, so published claims become comparable.
Product preview
Activity, stability and cost on one screen
Every run shows the whole chain — what was predicted, how, and what it means for cost per kilogram. Design preview: the platform is in early access and has not shipped. The calculator, however, is live today.
Built for the laboratories priced out of computational science
Academic groups
Screen a composition space before committing a student to months of synthesis. Free tier for academic use; results are citable.
Energy-materials startups
Triage your candidate list against cost, not just activity — and show an investment committee the reasoning.
Suppliers & manufacturers
Check a claim, yours or someone else's, against the operating point you actually sell into.
Questions
What does verna.science actually do?
You give it a catalyst composition or structure and your operating conditions. It predicts OER activity and acidic stability with a machine-learned interatomic potential, then propagates those properties through a PEM electrolyser model to energy consumption (kWh/kg H2), iridium intensity (g/kW) and levelised cost of hydrogen. Every number carries its method version, input hashes and uncertainty.
How is this different from a general AI materials platform?
Two ways. First, specialisation: general-purpose interatomic potentials are trained mostly on bulk crystals and are measurably least accurate on oxide surfaces with oxygen adsorbates — which is exactly the OER regime. We fine-tune for that regime instead of covering all materials. Second, closure: most platforms stop at a predicted property. We continue to cost per kilogram, because that is where the decision actually gets made.
Do I need to know DFT or set up a cluster?
No. That is the point. Workflows ship with defaults chosen for acidic OER, run on our hardware, and return in minutes. Expert controls exist but are collapsed by default.
How accurate is it?
Screening accuracy, not benchmark accuracy. It is built to rank and triage candidates so you spend lab time well, not to replace a converged DFT study or an experiment. Where a composition falls outside the validated domain we refuse to answer rather than return a plausible wrong number.
What does it cost?
The LCOH calculator is free and needs no account. Computational jobs are priced in credits per job, with free credits on signup and a free tier for academic groups. See pricing.
Which chemistries are supported?
The first domain pack is acidic OER for PEM water electrolysis — the iridium-reduction problem. Ammonia cracking, methane pyrolysis and CO2 conversion are planned as further packs on the same engine.
Test it against your own chemistry
We are onboarding a first group of labs and companies. Early access includes free credits, direct input into which workflows ship first, and no obligation.