Conference announcement
Lupine Science is heading to TMS2027
How measuring simulation error can help us choose the next experiment.
When14–18 March 2027
WhereOrlando, Florida
ContributionPoster presentation
Before a researcher commits to the next experiment, a simulation can help them decide where to look. But how do we know when its answer is reliable enough to guide that choice?
That question is taking Lupine Science to the TMS 2027 Annual Meeting & Exhibition, held March 14–18 at the Orlando World Center Marriott in Orlando, Florida. Alex Welcing will present a poster in Accelerating Innovation in Materials and Manufacturing. The abstract received acceptance on August 31, 2026. Conference details.
A convincing simulation still needs a reference
Materials researchers use computer models to explore how atoms interact: how a crystal responds to pressure, how strongly its atoms bind, or what happens around a defect. An interatomic potential is the mathematical model that makes these calculations practical.
Every such model is an approximation. A model that reproduces one property well can miss another. If a simulation guides the choice of an alloy or a battery material, that gap can matter to the next experiment. This is why resources such as the NIST Interatomic Potentials Repository provide both models and tools for judging where they apply.
Measuring error means comparing a prediction with a clearly specified reference. That reference might be an experiment or a calculation made at another level of approximation. Its physical conditions matter. So does the property being measured.
Learn from the pattern of mistakes
Our poster asks whether errors across existing potentials contain patterns that can help us anticipate where another prediction may need closer examination.
The research starts with archived material-property predictions and their reference values. It examines how errors vary together, then tests prediction on held-out materials while excluding related publication groups from the fitting data. The immediate focus is equilibrium properties of cubic metals, especially their elastic response: how they resist small deformations.
The practical question is simple: can an error we already know help us decide what to measure next? Answering it requires checking the reference, comparing with simple baselines, and keeping unsuccessful tests in view. A pattern is useful only within the conditions where it has been tested.
What would we need to measure before we could trust the next prediction?
What this milestone means
The accepted poster is titled “From Transferability to Prediction: The Error Geometry of Interatomic Potentials.” It gives us a place to discuss error measurement with the materials and manufacturing community, share the boundaries of the evidence, and sharpen the next validation experiment.
Our longer-term goal is to help researchers spend their experimental and computational effort on better-informed choices. Establishing whether this approach saves work in a prospective materials campaign is a next test. The announcement marks acceptance of a poster abstract; manuscript publication is a separate process.
The animation above expresses that journey as a survey: scattered traces, reference landmarks, and a path that reaches toward an unfinished landscape. It is an original conceptual illustration. Its shapes do not encode measurements or imply that uncertainty has disappeared.
Meet us in Orlando
If you build atomistic models, measure material properties, or use simulations to choose experiments, we would welcome a conversation at TMS2027. We are especially interested in the measurements you rely on when a simulation gives a plausible answer and the next decision matters.
Presenter: Alex Welcing, Lupine Science. Poster session time and location: to be announced by TMS. We will add the session details here when they are assigned.
Explore the conference · Read the Lupine research collection · Explore LUPI