Modeling of star formation on massively parallel computers

Very young stars form in collapsing interstellar clouds, initially pancaking into massive gas and dusty disks from which matter flows down onto the central star. Inside these disks, planets also begin to form and grow, potentially leading to the birth of life. It is therefore important to better understand the dynamics and composition of such disks, known as protostellar disks. Thanks to the advent of advanced telescopes, protostellar disks can now be observed with unprecedented detail. The progress in observations requires the development of numerical codes capable of interpreting the flow of new observational data. This is however not an easy task. Various physical processes at work described by conceptually different mathematical tools makes numerical simulations so complex that innovative solutions and deep expertise in applied math is required to achieve the goals. At the same time, a matching expertise in astrophysics is needed to guide the efforts of applied math experts towards the development of codes that can be efficiently and successfully applied to topical astrophysical problems. The goal of this multidisciplinary project is to join the expertise in applied math provided by the Russian scientists and in astrophysics of star formation provided by the Austrian team and develop a high-performance three-dimensional numerical (magneto)hydrodynamics code, which will be used for studying the formation and long-term evolution of protostars and protostellar disks. The code will be fine-tuned to utilize the power of recent Intel Xeon Scalable processors (e.g., Skylake-SP/Cascade Lake SP), which use vectorization of parallel operations (AVX512) to accelerate the numerical computations, and NVIDIA graphics processing units (GPUs), which allow for significant acceleration of computations. The current project can enrich the available repository of numerical codes with a tool that will make use of these innovative hardware products. The strength and novelty of our project is in synergy between applied math and astrophysics, and in various feedback loops that this synergy implies. The code will be applied to solving specific astrophysical problems which in turn will impose certain requirements on numerical methods, instigating the search for non-trivial solutions. We are confident that developing a numerical code sharpened for the use with the Xeon Scalable processors and NVIDIA GPUs can generate invaluable experience and added value in the form of new solutions and numerical techniques that can further be employed by the mathematical and astrophysical communities in both countries and worldwide.