IQM's VIME method cuts quantum circuit complexity for molecules by up to 26 times

[Karen50]

Researchers at IQM Quantum Computers have developed a method called Variational Imaginary-time Majorana Evolution, or VIME, for simulating complex molecules on quantum computers more efficiently, according to Quantum Zeitgeist. The key idea is to do a lot of the work classically first. A pre-training step using a technique called Majorana Propagation prepares the circuit before it runs on quantum hardware, combined with a compressed circuit template known as a tiled Unitary Product State ansatz

The results are eye-catching. For a ruthenium complex called TLD-1411, energy calculations were about 400 times more accurate than with the earlier ADAPT-VQE approach. For acene molecules needing up to 60 qubits, VIME reached similar precision with "up to twenty-six times fewer CNOT gates." Two qubit gate depth fell 167 fold compared with earlier methods

Those reductions matter because today's quantum hardware is noisy. Every gate adds a chance of error, so fewer and shallower circuits mean results that are more likely to survive the noise. The article describes the classical pre-training as assessing expectation values "at a computational cost linked directly to circuit depth rather than system size." In other words, the classical side handles the heavy lifting where it can

This is part of a wider trend. Several recent results, including work from Oak Ridge on fluid dynamics, show classical preprocessing dramatically cutting what is asked of the quantum computer. Some see this as the most realistic route to useful quantum chemistry in the near term. Others point out that if the classical side does so much, it raises the question of what the quantum computer adds

Is hybrid classical and quantum computing the most realistic path to useful results? And at what point does clever classical preprocessing make the quantum part unnecessary?

Researchers Cut Quantum Circuit Complexity By 26x For Moleculesquantumzeitgeist.com

[Reacher Quarry]

A 167 fold reduction in depth is huge. On noisy hardware, that could be the difference between a result that means something and pure noise. Big if it holds up on real devices

[Sentinel68]

Interesting that IQM is putting effort into algorithms as well as hardware. A hardware company that also builds efficient software has a better chance of showing useful results on its own machines

Most hardware companies now have algorithm teams. The software can be as valuable as the chips

[Nicola72]

Majorana Propagation is a new one for me. Would be good if someone could explain it simply. The name alone makes it sound like Microsoft's topological qubits, which I assume it is not

[Dean78]

Ruthenium complexes are relevant to catalysis and some light activated drug research, so it is a sensible test molecule. Better than the usual hydrogen chain examples. Real molecules make for more convincing demonstrations