MARVEL ends, leaving behind a lasting community and a vibrant research field

By Nicola Nosengo - EPFL

The day started with Nicola Marzari reminding everyone that the vision put forward by MARVEL more than 12 years ago may have seemed far-fetched at the time but was in fact spot-on: materials simulation is now the bread and butter of major tech companies, not to mention the materials and chemical industry. Marzari also noted that computational materials science is in many ways “big science”: it is done by a large and interconnected community, it tackles big problems and its papers are regularly ranked among the most cited papers in the entire scientific literature. But it needs considerably less funds than other “big science” fields and perhaps yields more broadly useful results.

Marzari’s opening was followed by the morning session, The quantum of materials, that explored how quantum mechanics continues to transform our understanding of matter.

Chaired by Joost VandeVondele (ETH Zurich / CSCS), the session featured four former MARVEL members who highlighted the convergence of quantum simulations and artificial intelligence as powerful tools for accelerating materials discovery.

Matthias Troyer (Microsoft) opened the scientific program by discussing how quantum computing and AI can work together to unlock a new generation of "materials marvels," enabling researchers to tackle problems that remain beyond the reach of conventional computational approaches. Ana Akrap (University of Zagreb) followed with insights into magnetically controlled optical effects in europium-based conductors, illustrating the rich physical phenomena emerging in quantum materials. Quansheng Wu (Chinese Academy of Sciences) presented advances in first-principles magnetotransport calculations using WannierTools, demonstrating how sophisticated computational methods are providing increasingly accurate descriptions of real materials. Closing the session, Antimo Marrazzo (SISSA, Trieste) showcased the broad potential of quantum materials, from fundamental topological properties to applications in technology and industry.

The focus then shifted from understanding materials to designing them. Chaired by Anirudh Raju Natarajan (EPFL), The design of materials examined how artificial intelligence is reshaping the discovery process. Anatole von Lilienfeld (University of Toronto) traced the evolution of materials design from traditional intuition-driven approaches to modern AI techniques, arguing that the future lies in combining data-driven methods with scientific insight. Julia Wiktor (Chalmers University of Technology) addressed the challenge of translating promising computational candidates into realistic energy materials suitable for practical applications, while Rubén Laplaza Solanas (University of Sevilla) demonstrated how machine-learning interatomic potentials can reveal the stability and behavior of complex subnanoclusters.

A highlight of the day came just before lunch with the formal signature of a collaboration agreement between CECAM and the Paul Scherrer Institute (PSI), represented by Andrea Cavalli, Barbara Montanari and Christian Rüegg. The agreement reinforces the shared commitment of leading European research organizations to strengthen scientific cooperation and support the computational science community.

The afternoon opened with a round-table discussion on The future of materials, moderated by science journalist Nicola Nosengo (EPFL). Bringing together perspectives from academia and industry, the panel featured Raffaela Cabriolu (Norwegian University of Science and Technology), Ganna (Anya) Gryn'ova (University of Birmingham), Giulia Mangione (Stellantis) and Senja Barthel (Vrije Universiteit Amsterdam). The discussion explored emerging research directions, the growing importance of interdisciplinary collaboration and the skills required to prepare the next generation of materials scientists.

Machine learning remained at the heart of the following session, chaired by Michele Ceriotti (EPFL). Jigyasa Nigam (MIT) described the rapid evolution of atomistic machine learning and its transformative impact on computational materials science. Philippe Schwaller (EPFL) presented new approaches for the sample-efficient design of molecules and materials, bringing computational predictions closer to experimental synthesis. Sandip De (BASF) concluded the session by illustrating how machine learning is accelerating catalyst innovation through the integration of high-throughput computation and experimental validation, highlighting the increasing role of AI in industrial research.

The conference concluded with a session dedicated to the growth of the materials simulation community itself. Chaired by Sara Bonella (EPFL), it emphasized that scientific excellence depends not only on technological advances but also on strong networks of people. Barbara Montanari (STFC Scientific Computing) reflected on how AI is transforming research communities while reinforcing the importance of collaboration and shared knowledge. Garu Gebreyesus (University of Ghana) offered a personal perspective on building electronic-structure research across sub-Saharan Africa through the various editions of the ASESMA School. Finally, Leopold Talirz (SCHOTT) presented Project GlasAgent, demonstrating how collaborative, agent-based approaches can support innovation in glass development.

In the end it was hard not to agree with what Garu Gebreyesus wrote in their slides: “sustained investment in people builds lasting scientific communities”, and the investment in MARVEL definitely built one.