Deeptech Pasqal draws with BMW to use quantum calculation for metal forming

We are not stopping Pasqal anymore. Two months after the launch of its quantum-based quantum processor on Azure Quantum, Microsoft’s quantum cloud offering, and the day after a similar announcement with French company OVHcloud (as Industri & Technologies exclusively mentioned in March), the French start-up underway on a new collaboration with BMW.

The goal: to utilize quantum computers to improve the formation of the metals that make up auto parts, ie. “test the resistance to deformation, and in particular to bending”, begins Benno Broer, co-founder of Qu & Co, a Dutch quantum software start-up which, following the acquisition of Pasqal in January 2022, became the commercial director of the French start-up.

“The ultimate goal of the manufacturer is to perform as many simulation tests as possible to reduce physical tests – that is, to break fewer parts and cars”, he continues. Gold, “current classical computer methods are unable to cope with the complexity of simulating a complete vehicle with the desired accuracy”, completes the press release. “Such simulations will also help the BMW Group create lighter parts that will make cars more fuel efficient,” it is also indicated.

This collaboration is the result of Qu & Co’s victory in one of the four use cases submitted to quantum computer players during the BMW Quantum Computing Challenge, the results of which were communicated in December 2021. “In our category, between 5 and 10 candidates presented a solution to solve metal bending, notes Benno Broer. Interestingly, we found that many of them actually used an algorithm developed by Qu & Co and the University of Exeter, UK, and owned by Pasqal. Fortunately, BMW preferred to work with the original. »

A quantum version of physics-informed machine learning (PIML)

This algorithm is inspired by physics-informed machine learning (PIML), a type of neural network “which occurred in the years 2015-2016” and “is used to solve complex differential equations”. A mathematical problem “found in countless applications, such as meteorology, but also in all multi-physical simulation problems, in metal forming, fluid dynamics, heat transfer, fields electric …”, leads Benno Broer.

“The quantum version of this algorithm provides better accuracy of the results, which is very difficult to achieve for these complex problems, insists the commercial director. We have already tested it on several industrial cases, but not yet on metal forming. »

Qu & Co and Pasqal are not the only ones testing a quantum algorithm for this type of problem. “Various actors have proposed methods to solve a set of differential equations. All require fault-tolerant hardware – something that will not exist for at least five to ten years. Except one, the one developed by the University of Oxford, but after testing it, we think it does not resist scaling. Become ours. ”

Avoid two-qubit ports

Now that the Qu & Co and Pasqal teams have merged, it’s time to move from theory to practice. “For two months we have been working on configuring Pascal’s quantum processor to be ‘aligned’ with this use case and the use of our algorithm”, develops Benno Broer. The former CEO of Qu & CO is particularly proud to have found “a means of interleaving quantum bits (qubits) without a two-qubit gate, which is very difficult to implement with the coherence time allowed by the prior art”.

Pascal’s hardware calculates according to a so-called analog mode: the system of cold atoms – qubits – is placed in an initial configuration corresponding to the given problem, then it is allowed to converge towards its minimum energy state, giving the solution to problem. “We showed that we only needed one-qubit gates to successfully wrap our qubits, a necessary step to improve the outcome of our experiment.”

Encouraging first test

According to initial tests conducted with BMW’s quantum team, Pasqal claims results “30 to 50 higher than what we achieve with other quantum hardware, for a method that is much more resistant to noise”, excited the engineer. Performances that make Georges-Olivier Reymond, CEO of Pasqal, say it “Pasqal currently offers the only method on the market to solve these types of differential equations with quantum technology, which is essential for running efficient and accurate simulations,” says the press release.

This project “macro scale door” a collaboration between Pasqal and BMW dating back to 2019, “but which until now were only about nuclear-scale experiments”, says Benno Broer. The two partners have agreed on an initial six-month exploration phase, after which they will assess which direction the project should take. “But I do not see this cooperation stopping for the time being,” assures the commercial director.

With regard to the possible result, “if it is satisfactory, we should offer our quantum solver as one of the possible tools in classic platforms, such as our chemical simulation platform Quantum Back-End Chemistry (QUBEC), which has been integrated into Schrödinger’s solution, a leader in chemical simulation software publishing”, details Benno Broer, before concluding: “And we are sure that we can achieve the quantum advantage of this type of case within a short time, hopefully two to three years. »

Chose for you

Fujitsu presents a 36-qubit quantum simulator based on the technologies of its Fugaku supercomputer

Leave a Comment