Highlights

In a study just published in ACS Nano, MARVEL researchers  systematically analyzed hundreds of atomically thin materials to understand how magnetism emerges and stabilizes at the nanoscale. Predicting the magnetic behavior of 2D materials is challenging because they often exhibit complex energy landscapes with multiple local minima, meaning a single material can settle into many competing magnetic states. To tackle this, the team developed an automated computational approach designed to explore this complexity more thoroughly than conventional methods. Among the most exciting results is the discovery of 12 new half-metals—materials that conduct electrons of one spin type while blocking the other. Such behavior is highly sought after in spintronics, an emerging technology that aims to use electron spin, rather than charge, to process and store information more efficiently. These materials could form the basis of faster, lower-power electronic devices.

The AiiDAlab software was originally developed to simplify computer simulations in materials science. In a new paper in Digital Discovery, a  team of MARVEL researchers at Empa and PSI have now demonstrated that it is also highly useful in a number of other applications, ranging from the simulation of the Earth’s atmosphere to battery development. At Empa, for example, atmospheric transport simulations controlled by AiiDAlab are being used to quantify Swiss and European greenhouse gas emissions from atmospheric measurements. Empa also uses AiiDAlab for the automated characterization of batteries, where it can coordinate not only simulations but also the execution and analysis of experiments.

In a new article in Energy and Environmental Science (highlighted as a “hot” article by the journal’s editors), scientists in Nicola Marzari’s lab at EPFL present a computational workflow that they used to screen thousands of materials from vastly different families and simulate their electrolytic properties, in the end identifying a handful of promising candidates for solid lithium-ion batteries. The workflow starts from  over 30,000 compounds containing lithium from three of the main databases of experimental structures, filters out the unsuitable ones, and combines DFT, a "pinball model" approximation of Molecular Dynamics and first-principle MD to identify the best potential fast ion conductors. The final list includes 34 materials that deserve further studies.  Solid-state batteries could be a game-changer for electric vehicles and other applications. 

In 2015, MARVEL scientists collaborated with the Swiss jewellery company Varinor to develop and test a computational workflow to predict the color of metals and metal alloys from first principles. The workflow had four main computational steps: electronic structure calculation with DFT; calculation of the dielectric function with the independent particle approximation (IPA); absorption coefficient and reflectivity; photorealistic rendering. The method could well reproduce the optical properties of most elemental metals, with a few exceptions. When it came to alloys, some were more difficult to simulate than others and new solutions were developed to  simulate some of gold-based alloys that are most interesting for jewellery. The work resulted in multiple publications, and the code and the workflow remain available via AiiDA and GitHub and are open-source. This is the fourth  article in a series about MARVEL's success stories from its 12 years of research. You can read the previous ones here, here, and here.

An accurate description of how heat flows through diamond at the nanoscale is challenging for both theorists and experimentalists. A new article in Physical Review Letters addresses both challenges presents a joint effort involving MARVEL members from the THEOS lab at EPFL. Experimentalists fabricated long suspended diamond cantilevers with a triangular section less than a micrometer in width, and used some luminescent defects in the diamond as nano-thermometers to measure changes in temperature. Theorists then applied viscous heat equations to explain that transport arises in the cantilever from the interplay between different types of phonon interactions – “hydrodynamic” events that conserve the momentum of the crystal - and extrinsic effects determined by sample size and geometry.

A team of NCCR MARVEL scientists, led by researchers at EPFL and PSI, has introduced the Materials Cloud Three-Dimensional Structure Database (MC3D), a systematically curated database of quantum-mechanical calculations for inorganic materials derived from experimental crystal structures. The database contains more than 32 000 structures whose relaxed geometry and electronic structure were computed using carefully standardized DFT workflows, using three different functionals and/or computational protocols. Beyond providing a consistent reference dataset for computational materials science, MC3D also supports emerging data-driven approaches. For example, it served as a starting point for the MAD dataset used by Michele Ceriotti’s group at EPFL to train the PET-MAD machine-learning interatomic potential.

Researchers at EPFL’s Laboratory of Computational Science and Modeling (COSMO) have reached a significant milestone in material science, reaching the top position on Matbench Discovery, the leading benchmarking platform for machine-learning interatomic potentials.  The achievement was made possible by PET-OAM-XL, a new model, which builds on the foundations laid by the recently published PET-MAD model. 

Powerful atomistic simulation tools have transformed materials research, but their complexity still limits who can use them and how easily results can be reproduced: a gap that a new web-based app now helps close. The AiiDAlab Quantum ESPRESSO app, described in a recent publication in npj Computational Materials, can run not only isolated calculations but also complete, end-to-end computational workflows involving multiple passages over several different materials, lowering the barrier for both experimentalists and computational experts. This is achieved through the tight integration of the widely-used Quantum ESPRESSO simulation software package with the AiiDA engine, a workflow-management system to help automate complex simulations in materials science and a core pillar of NCCR MARVEL, which has provided substantial support to its development alongside contributions from the broader community.

In a new article in Physical Review Letters, MARVEL scientists from the THEOS lab at EPFL have made a leap forward in modelling and explaining phonon hydrodynamics, a heat conduction regime where heat flows like water in a pipeline. Their brand new mathematical description makes the phenomenon easier to test experimentally and clarifies the physics behind it. When applied to the in-plane section of graphite at a temperature of 70 K, it also points to a bizarre phenomenon that can emerge with hydrodynamic transport and by which heat can flow in reverse, from a colder region towards a hotter one. Being able to insert such a system into consumer electronics products would have huge applications, for example hydrodynamic heat management could help prevent batteries or other devices from overheating.

A research team led by Mathieu Luisier from ETH Zurich and MARVEL has introduced a software package called QuaTrEx (Quantum Transport Simulations at the Exascale and Beyond), that combines in a new way DFT, the GW approximation and the non-equilibrium Green's function (NEGF). By running it on two supercomputers - one in Switzerland and one in the USA – the Alps supercomputer at the Swiss National Supercomputing Center in Lugano, and on the Frontiers machine at the Oak Ridge Leadership Computing Facility in Tennesse, USA – they were able to simulate the behavior of a nanoribbon, the fundamental component of next-generation transistors, made out of over 42,000 atoms. For this work, the group was awarded an Honorable Mention (or second place) at the 2025 ACM Gordon Bell Prize.

In a new study in npj Computational Materials, scientists in Michael Herbst’s laboratory at EPFL have combined Density Functional Perturbation Theory (DFPT) with algorithmic differentiation (AD), a mathematical technique to compute derivatives of virtually any calculation codified in a computer program. The result is a new computational framework that makes DFPT easy by automatically deriving all necessary derivatives of DFT outputs for any input parameter, thus simplifying calculations of properties like elasticity of materials. The method also proved efficient for inverse materials design, where one starts from a desired set of properties and seeks atomic structures satisfying these properties most closely.

A study in Nature Communications by Michele Ceriotti’s group at EPFL has introduced a new dataset and model that greatly improve the efficiency of machine-learning interatomic potentials (MLIPs) and their applicability for different chemical elements and material classes.  The first key innovation of the study is a brand-new dataset called Massive Atomistic Diversity (MAD), covering both organic and inorganic materials and ranging from 3D bulk structures to nanoclusters and molecules. The other  element is the network architecture itself, that unlike other models has few a priori assumptions and learns chemical symmetries during training.