A discovery campaign ends with a delivery, not a verdict. MS Lattice hands your team a ranked set of candidates, each with its scores and its proof, and from that point the decisions are yours: which to trust, which to make first and what to ask for next. This guide walks through what you receive and how to read it without mistaking a rank for a promise or an estimate for a measurement.
What lands on your desk
Every campaign closes with the same kind of delivery, whatever the domain or the mode. A cathode campaign reports voltage and capacity, a hard-coating campaign reports hardness and oxidation resistance, but the shape holds.
- A ranked shortlist of candidates, each scored against every target you set.
- Compositions and crystal structures, with a standard CIF file for every candidate, ready for your tools.
- Estimated properties for each candidate, every estimate with its uncertainty attached.
- The record of every check: what each candidate passed, what it failed and where it stopped, with the reason.
- Stability against every rival phase of its chemistry, and a check against the known crystals, so you see what is new.
- The whole campaign exported as CSV or JSON, a candidate browser and a 3D view of every structure.
Behind that sit two things you can call on later: reproducibility from the record, and first-principles verification on demand for the candidates you choose. Each domain page also states what a campaign there hands back. For hard coatings, it is ranked nitride, carbide, diboride and MAX-phase coatings, with estimated hardness, mechanical stability, oxidation resistance, expansion match and cost. Read your domain’s page before the campaign starts, so the delivery holds no surprises.


How to read a ranking
A rank is an ordering against the targets you gave the campaign. The candidate at the top is the one whose estimated properties best meet your requirements and push your preferences furthest. It is not a forecast that the top candidate will win in your lab, and a ranking built on a loose brief will rank loosely.
Start with requirements. In MS Lattice, requirements are held inside the search: a proposal that breaks them is turned away before a structure is built. So every candidate on the shortlist has already cleared your must-haves on its estimates. What separates them is how far each pushes your preferences.
Then look across the targets, not only at the overall position. Because every candidate is scored against every target you set, you can see where each one earns its place. A candidate ranked fourth may be the best on the property your product cares about most. A candidate ranked first may sit close to a limit on one target. The candidate browser lets you filter the campaign your way, and that is often where the real shortlist forms.
Finally, read the ranking for spread. MS Lattice aims for diversity on purpose: different strong answers, not one family of near-copies. If one family fails for a reason the estimates could not see, another may not.
The record shows what a top rank has meant where the answer was known. From a chemical formula alone, MS Lattice ranked the correct crystal structure first for 13 of 15 compounds. Asked in one sentence for a coating for dry, high-speed machining, it put (Ti,Al)N and (Cr,Al)N, the coatings the industry runs on today, in its top five. The ranking points the right way. No single rank is a promise.
How to read a screening estimate
The property figures in a delivery are screening estimates for the bulk crystal. They exist to tell strong candidates from weak ones before anyone spends a day at the bench. They are not measurements.
Every estimate arrives with its uncertainty attached. Use it. Two candidates whose estimates overlap within their uncertainty are, for practical purposes, tied on that property, and other factors should decide between them. A candidate whose advantage is larger than its uncertainty is a stronger signal. As one reference point on the record, MS Lattice’s hardness estimates show a 10.6% average error across 20 reference crystals.
The larger gap is not the estimate’s error but what the estimate does not cover. A bulk crystal is not a coating, a sintered part, a doped wafer or a working cell. Processing, microstructure, interfaces and service conditions shape the measured property, sometimes more than the composition does. Each domain page says where its figures stop.
Read the limits as a test plan, not a disclaimer. If your domain page says tool life depends on the deposition process, then the first cutting test is the experiment that matters, and the estimates have done their job by deciding which compositions deserve it.
The record of every check
A score is a claim. The record is the proof behind it. Every candidate MS Lattice evaluates is judged in stages: a valid structure, stable, on target, recorded. A miss stops on the spot, and the reason stays on its record. You can see what each candidate passed, what it failed and where it stopped.
A reviewer can check why a candidate is on the list without rerunning anything. The near misses help too: candidates that stopped at one check tell you which target is doing the most work, which matters when you decide whether to loosen or tighten the brief.
Reproducibility is on the record. From three production-size campaigns, all 73 audited candidates were rebuilt from the record MS Lattice keeps of each, every property within its declared tolerance. For a team that has to defend a material choice to a program review or a customer, that means the reasoning behind a candidate can be rebuilt from the record.
- 73 of 73candidates rebuilt from their record
- 4 of 4first-principles stability verdicts right


First-principles verification, on demand
Screening estimates sort a large field quickly. When a verdict matters, MS Lattice checks it from first principles. You send the candidates you choose for the deepest check: stability against every rival phase of their chemistry, elastic constants for how they carry load and phonons for how their lattice vibrates.
The record shows what this check has done. It called Mg₂Si and MoSi₂ stable and deliberately wrong arrangements of each unstable, four verdicts out of four. It reproduced silicon’s lattice constant, its three elastic constants and its highest optical phonon frequency, five reference values out of five, each within the tolerance set before the run.
Use it at the boundary between screening and the lab. A candidate that is about to take real synthesis effort is worth the deeper check first. What holds is confirmed. What does not is caught before a single day in the lab.
Choosing the first few to make
The shortlist is built so your lab can make it. The question is which few go first, and rank is the starting point, not the whole answer.
- Confirm the hidden requirementsCheck each candidate against limits that never made it into the brief: precursors you cannot source, processes your lab cannot run. Anything that fails here comes off, whatever its rank.
- Read the marginsPrefer candidates whose advantage on your key property is larger than its uncertainty, sitting inside your limits rather than at their edge.
- Pick for spreadChoose from different families near the top, so one failure the estimates could not see does not take out the whole round.
- Check what is newA known crystal may be quicker to make; a new one may be the real prize. A good first round often holds both.
- Verify before you commitSend the candidates that will take the most lab effort through first-principles verification for stability, and for elastic constants or phonons where your application needs them.
- Keep a baselineMake your current material, or a well-known reference, alongside the new candidates, so your measurements have a comparison your team trusts.
What to send back for the next round
A campaign is rarely the last one. MS Lattice is built to go again from what you kept: keep the winner, tighten the target, run again.
- The measured values for what you made, beside the estimates, including the misses.
- Why a candidate failed, if it did: it would not form, it degraded, it would not process, or it measured short of the target.
- Targets and limits revised from what you learned, such as a tighter cost ceiling or a new element to avoid.
- Your best made material, as the starting structure for Prototype-Guided Optimization, which searches around it and ranks every variant against the original.
What your own lab measures can go into your next campaign, where it sharpens the search where it matters to you. With Norck’s validation scopes, the loop is part of the service: real-world results feed straight back into the next round of design.
Questions teams ask
Is the top-ranked candidate the one we should make?
Not necessarily. The rank orders candidates against your targets on screening estimates. Read the scores, the uncertainty and the record, then pick for margin and spread.
Are the property figures measured values?
No. They are screening estimates for the bulk crystal, each with its uncertainty. Testing in the lab and in the application decides.
Can we check a result ourselves?
Yes. Every candidate carries the record of every check and comes as a standard CIF file, and 73 of 73 audited candidates were rebuilt from their record.
What if nothing on the shortlist works in the lab?
Send back what you measured and why each candidate failed, revise the targets and limits, and run again. A clear account of the misses is the most useful brief the next campaign can get.


MatterSpace’s first domain engine
MatterSpace built MS Lattice for the scientists and engineers who discover and design new materials, starting from the properties their product needs.









