Federal Project Grant Award Summary

The University of Maryland, College Park received a $266,000 Project Grant from the National Science Foundation's Division of Information and Intelligent Systems under the Computer and Information Science and Engineering (CISE) program (CFDA 47.070) for a collaborative research initiative titled "Formally Verified and Performance-Optimized Tensor Contraction Sequences in Quantum Many-Body Computations." Awarded on May 1, 2026, with a completion date of April 30, 2029, this grant supports the development of a scalable formal verification framework designed to ensure the correctness and reliability of optimized tensor contraction sequences used in quantum chemistry, quantum physics, and materials science research. The project addresses critical gaps in scientific computing by establishing automated methods to verify tensor computations that employ performance optimization techniques such as mixed-precision arithmetic, operation reordering, and kernel fusion—optimization strategies that, while effective at reducing runtime and memory costs, can introduce subtle numerical errors that compromise scientific reproducibility.

The deliverables include the creation of domain-specific satisfiability modulo theories (SMT) encodings using quantifier-free floating point logic for mixed-precision verification, automated formula generation and counterexample-guided abstraction refinement methods, incremental SMT solving with solver-state reuse and equality-saturation techniques, and integration of verification capabilities into multi-objective design space exploration. The project will produce open-source tooling for the high-performance computing and formal methods communities, research publications, and interdisciplinary training opportunities that bridge formal verification, numerical analysis, and high-performance computing—thereby advancing more reliable and reproducible scientific computing across quantum and computational sciences domains.