Scientific computing is an often-overlooked source of carbon emissions in research [1]. Inefficiencies such as redundant parameter sweeps, duplicated datasets, poorly targeted simulations, and short-lived, non-reusable workflows waste CPU hours and electricity, increasing carbon footprints without yielding proportional scientific benefits [2]. The 4R strategy—Refuse, Reuse, Reduce, and Repair—guides circular product development, yet is seldom applied to scientific data production. Applying the 4Rs to computational research means refusing unnecessary runs and oversized models, reducing simulation campaigns through careful design and uncertainty quantification, reusing valuable data and validated models, and repairing or maintaining community codes and workflows to keep them interoperable and reproducible. These practices cut energy use and costs while also accelerating scientific discovery.

Simulating fracture processes of amorphous materials, which is one of the most challenging and expensive tasks in scientific computing, offers a clear opportunity to implement these changes. Although fracture has been studied for centuries and failure of materials typically has major economic consequences [3], its material-specific mechanisms are only recently being systematically uncovered [4, 5]. Furthermore, it has become evident and a scientific consensus that fracture is a genuinely multiscale phenomenon: whether and how a structure fails depends on interconnected processes, ranging from bond rupture and local chemistry to microstructural heterogeneity, viscoelastic or plastic dissipation, and large-scale driving forces. This complexity means the field spans physics, chemistry, materials science, mechanics, and data science. However, interactions between these disciplines remain limited, resulting in siloed models, repeated “from-scratch” simulations, and a slow transfer of knowledge into engineering practice with all consequences regarding carbon emissions in scientific computing    

Particularly, we identify the following challenges to be addressed:

1. Communication across disciplines. How can we create a shared problem definition and set clear, minimal reporting standards across physics, chemistry, materials science, mechanics, and data science to ensure that results are both comparable and reproducible? How can we identify shared and conflicting objectives early on to better align project scope and prevent redundant work or unnecessary computational effort?

2. Collaborative research. How can we jointly develop interoperable, open workflows so that a single data-rich simulation produces reusable outputs for multiple research communities? Can we construct multiscale models that predict macroscale fracture while preserving molecular mechanisms and quantifying uncertainties? Which strategies—such as adaptive design, multi-fidelity approaches, or surrogate models—can minimize computational resources and emissions without compromising model accuracy?

3. Standardized data sharing. Given the limited availability of curated databases, which unified data formats and metadata standards will make amorphous-fracture datasets truly FAIR—especially Findable—across atomistic, mesoscale, and continuum levels? How can we curate, version, benchmark, and offer machine-actionable access to these datasets to ensure provenance, comparability, and real reuse?

The main aim of the workshop is to tackle these challenges by bringing together researchers with diverse backgrounds, expertise, and experience. The Lorentz Center provides an ideal setting for these discussions. We will identify the community’s next key steps and share knowledge through plenary discussions, brainstorming sessions, technical talks, and hands-on tutorials.    

This workshop will be considered a success if...
...we reach agreement on the next steps to address the identified challenges, strengthen connections between different communities by sharing knowledge as well as collecting and distributing tools and techniques that support collaboration and multidisciplinary projects.

To address Challenge (1), senior participants will deliver plenary talks on “perspectives and challenges,” followed by roundtable brainstorming sessions. For Challenge (2), technical talks will spark in-depth discussions, and hands-on tutorials will offer participants practical experience with various modeling techniques, including MD, FE, and ML-based approaches. These tutorials will help participants understand the requirements of each technique and inspire new collaborative research ideas. To address Challenge (3), participants will discuss the standardized publication and indexing of data sets, focusing on the specific needs of each discipline and sharing best practices. These discussions will occur during the roundtable sessions. More details are outlined in the program.

Workshop outcomes

By the end of the workshop, we aim to significantly improve communication among attendees, fostering future collaborations and joint projects. Within two weeks after the workshop, we will prepare a report summarizing the brainstorming outcomes, outlining recommended next steps for the community, and providing a list of hyperlinks to key tools and papers. This report will be shared with all participants and uploaded to the research website of the key organizer(s) as the main outcome of the CECAM-Lorentz workshop. To further increase its visibility, the report will be promoted on social media, mailing lists, and similar platforms. Key organizers will also highlight it during their talks at international scientific conferences in the months following the workshop. Materials from the computational tutorials and experimental use cases will be made available online in a public GitHub or Zenodo repository, complete with documentation. These resources will be shared both with participants and publicly via a link in the final report. We will encourage the community to contribute original content to this repository. Senior participants and the broader community will be invited to use these materials for their courses or as training resources for new PhD students, helping to ensure that the workshop outcomes serve as a foundation for ongoing and expanding efforts.

Finally, we will make a video of the event, including short interviews to the participants, highlights of the workshop, and information on how to find the tools discussed, the final report, and how to get in touch with the expert attendees. We will share this video in the form of a vlog entry on social media platforms within one month of the end of the workshop, and advertise it via targeted mailing lists and via the organizers’ talks at international conferences in the months following the workshop.