Integrating Quantum Computers into a Radio Telescope

This project investigated in what way quantum computers can be integrated into radio telescopes. We started by identifying computational hotspots in the processing pipeline that could benefit from acceleration on quantum computers, but quickly found out that this was not a viable approach. There are only a fairly limited number of (classes of) quantum algorithms that are known to be faster than conventional algorithms, so it is better to see which one of those can be applied to this use-case.

We invited a number experts in the field to brainstorm on the possible applications, which yielded two possible approaches: machine learning for pulsar search and HHL to solve sets of linear equations for calibration. Both were implemented during this project.

We investigated the use of quantum machine learning to classify radio pulsars in a well known standard dataset. This was succesfully demonstrated using a single qbit quantum classifier, which unfortunately confirmed the findings presented by Kordzanganeh et. al.. This work was extended to cover fast radio bursts (FRBs), but we did not have sufficient time to test that approach rigorously.

We also implemented a quantum linear solver to mitigate computational hotspots in radio astronomic calibration. We selected a fairly straight-forward calibration approach, redundancy calibration, for its relative simplicity. This was tested using simulated data representing the HERA telescope. Two approaches were tested, using a Variational Quantum Linear Solver targeting gate-based quantum computers and a similar approach rewritten as a Quadratic Unconstrained Boolean Optimization problem to fit Quantum Annealing machines. Both were evaluated on simulated machines, and the IBM-Q Gualdalupe gate-based quantum computer. Results were encouraging but clearly showed that with the current state of the hardware, we are not able to achieve quantum advantage. However, we did show that algorithms that should result in a quantum advantage can be effectively used in a radio telescope. This work was documented in a scientific paper that was submitted for peer-review. The generalized VQLS implementation was offered as a prototype to the Qiskit commity.

We also supervised two student projects. These were focused on efficient representation of radio astronomy data in a quantum computer. Efficient representations that use orders of magnitude fewer qbits than a naive approach were suggested for both visibility data from the correlator, as well as for image data. These were presented in a Bachelor’s thesis and a Master’s semester project report. This work was extended with discussion on the efficient use of quantum fourier transforms and a very simple calibration pipeline and documented in a scientific paper that was submitted for peer-review.

The radio astronomy use-case was also added as a challenge in the Dutch Quantum Application Lab.