Digital Infrastructures for 21^st Century Science—Opportunities, Benefits and Needs

The scientific community has long established the need for major national and international long-term efforts to support ambitious and unique capabilities - from synchrotrons and colliders to telescopes, from computing and sequencing to fusion. Intriguingly, it's computational science and especially computational condensed-matter physics, chemistry, and materials science that have led across fields the indicators for publications and impact - one might even surmise relevance. Crucially, these computational capabilities are most often available under open-source and open-access models, meaning that they can be replicated effortlessly and at a flick of a switch worldwide, with scaling costs that are profoundly different from those of physical infrastructures, and with a most democratic model of dissemination. And they are supported by IC technologies where throughput capacity still doubles every 18 months; where unexpected accelerators - from machine learning to big data to large-language models - have appeared; and where novel paradigms - from memcomputing to quantum computing - might emerge. I'll present my own vision of an ecosystem and a digital infrastructure of open-source simulation codes and open-access data, of automated workflows, of externalizable capabilities driven by universal APIs that can be integrated by human or non-human orchestrators, and that can be dedicated to the most pressing societal needs addressed by materials innovations. Most notably, I'll underscore that what is most needed for these digital infrastructures are software scientists and engineers and long-term career opportunities, at a cost that is negligible with respect to the traditional investments of big science, and with multipliers that no physical infrastructure can match. Disclaimer: No artificial intelligence was employed in the preparation of this abstract.