Georgia Tech researchers have engineered a groundbreaking algorithm poised to revolutionize the way scientists study interactions between electrons, accelerating discoveries in fields ranging from physics and chemistry to materials science.
The novel algorithm, which stands out for its speed and accuracy, breaks through current computational limitations by demonstrating exceptional scalability across a wide range of chemical system sizes.
Computer scientists and engineers stand to gain from the algorithm's ability to balance processor loads effectively, enabling the tackling of larger, more complex problems without the high costs associated with previous methodologies.
The key to the algorithm's success lies in its innovative approach to solving block linear systems. This method marks the first known application of a block linear system solver for calculating electronic correlation energy, a critical measure in many-body systems like atoms and molecules.
"The combination of solving large problems with high accuracy can enable density functional theory simulation to tackle new problems in science and engineering," Edmond Chow, professor and associate chair of Georgia Tech's School of Computational Science and Engineering (CSE), said in a news release.
By leveraging Density Functional Theory (DFT), the algorithm enhances simulations of electronic structures, offering profound improvements within the random phase approximation (RPA) framework, a method known for its computational intensity as system size increases.
Georgia Tech's innovation overcomes these inefficiencies, reducing computation times and scaling effectively even for smaller chemical systems.
The research team integrated their algorithm into SPARC, a real-space electronic structure software tailored for precise, efficient and scalable DFT solutions. Phanish Suryanarayana, a professor from the School of Civil and Environmental Engineering, leads the SPARC initiative.
Upon testing small chemical systems such as silicon crystals with just eight atoms, the algorithm delivered rapid calculations and scaled efficiently to larger systems.
"This algorithm will enable SPARC to perform electronic structure calculations for realistic systems with a level of accuracy that is the gold standard in chemical and materials science research," added Suryanarayana.
What sets the algorithm apart is its cubic scaling capability in solving block linear systems, as opposed to the quartic scaling of traditional RPA computations. This advancement significantly reduces computational costs when system sizes double.
Addressing the challenge of dynamic block size selection, Georgia Tech's solution allows each processor to independently choose block sizes, enhancing scalability, processor load balance and parallel efficiency.
"The new algorithm has many forms of parallelism, making it suitable for immense numbers of processors. The algorithm works in a real-space, finite-difference DFT code, which can scale efficiently on the largest supercomputers," Chow added.
The Georgia Tech team will present their pioneering work at SC24, the 2024 International Conference for High Performance Computing, Networking, Storage and Analysis, taking place Nov. 17-22 in Atlanta.