GPU speedup for an exploration of non-Markovianity and qubit systems
PI: Sarah Elizabeth Shandera (ECoS, Physics)
The purpose of this project is to update a code for quantum circuits, developed in Professor Sarah Shandera’s group, to run on GPUs. The goal is to increase the number of qubits that can be efficiently simulated from 12 to about 20 or more. The code should also enable simulations of long sequences of the quantum gates (>1500 circuit layers), which is essential for statistical analysis of non-Markovianity.
The improved code will enable Shandera, the PI, and Professor Dezhe Jin to investigate a new possible class of non-scrambling dynamics. When qubits undergo evolution by a random circuit, information about the initial state generically becomes delocalized, spread out among correlations between the qubits. A problem of current interest is to understand the classes of dynamics for which this delocalization does not occur. Perhaps surprisingly, a circuit that mimics a typical, time-independent Hamiltonian also generically scrambles information. On the other hand, one well-studied means of preserving local information is to intersperse measurements of some qubits with the action of quantum gates. When the probability of measurement is high enough, local information can be preserved. This project will analyze a different technique, studying the extent to which non-Markovian gate sequences can prevent scrambling. Our previous work with the existing code provided preliminary results (arXiv:2505.00116) that suggest this approach.
Our team consists of Prof. Shandera (expertise in quantum systems) and Professor Dezhe Jin (expertise in classical non-Markovian sequences as partially observable Markov processes in the context of birdsong). Prof. Jin also has experience with GPU programming. Both Shandera and Jin, are faculty in Physics, will mentor the ICDS Junior researcher. Professor Shandera’s graduate students will also participate in the project.
We are seeking to recruit a post-comps junior researcher with demonstrated expertise in GPU programming and basic familiarity with qubits and density matrices. The effort requested is 25% RA, for two semesters. Numerical and statistical studies of quantum systems require efficient codes. The full state of system of N qubits is described by a 2N x 2N matrix, and to parameterize finite system-size effects requires performing the same suite of simulations at several N, up to the largest practical N. The current (Python) code implements symmetric circuits that result in sparse matrices, but we need an additional speedup to get beyond 12 qubits and 500 circuit layers.
This project enables a new collaboration between Shandera (particle theory/cosmology) and Jin (biological physics) around their common interest in non-Markovian processes, supporting ICDS’s mission to catalyze interdisciplinary research. We anticipate the project leading to at least one publication, positioning the team to apply for a ‘New Initiative’ grant from the Kaufman Foundation next spring.
This effort is not associated with any ICDS-affiliated Center. PI Shandera helped organize and host a panel on quantum computing at the 2023 ICDS fall symposium (Discovery to Impact: Scientific Storytelling with Data). She will engage with the activities of the new ICDS quantum hub.