In recent years, quantum technologies have emerged as a game changer for a wide range of industries, impacting potentially all aspects of our lives. Quantum sensors in particular provide a new ability to measure the invisible. In semiconductors, this means informing the development of new materials and more efficient chips and controlling their quality at the core. In healthcare, this means detecting diseases at the earliest stage.
From data to valuable information
While measuring with exquisite precision is key to bringing new data, the analysis of the output is equally important to bring useful information to the user. Qnami’s CIO Alexander Stark explains: “the challenge is not just to measure the signal from a tiny source, but to use the data to infer the very properties of that source.”
For Qnami the question is important. In 2020, Qnami launched ProteusQ, a microscope leveraging diamond quantum technologies, allowing to measure magnetic properties at the nanoscale. R&D engineers and scientists use this instrument for the development of new materials and electronic chips. To inform their process developments, quantitative information about their devices is required. “This information can represent a significant gain of time for their process development”, says Stark.
The AI solution
Yet, this task is difficult as we are talking about ill-posed problems involving complex mathematical modeling and a large number of variables. “Solving the problem ‘manually’ requires deep expertise and a lot of patience,” says Patrick Maletinsky, professor at the University of Basel and Chief Scientific Officer for Qnami. “We have spent a lot of time in my group refining our models, but their application remains limited to a certain class of materials and when the input data is of high quality.”
Artificial intelligence, on the other hand, is well suited for addressing such tasks. Stark explains “Machine learning offers a near-ideal route to solving such task, where the problem is difficult to formulate analytically. It is also to-date the most successful technique for extracting key features from the large noisy data sets and generalizing this knowledge to new large-scale inputs”.
A collaboration between Qnami, the University of Basel, and the ETH Zurich
In 2021, Qnami therefore decided to team up with the Quantum Sensing Lab of Prof. Maletinsky (Basel University) and the Condensed Matter Theory group of Prof. Huber’s group (ETH Zurich) to explore the idea. The team reports key progress in a paper published in Physics Review Applied.
Adrien Dubois, first author of the paper, shares his excitement: “the problem turned out to be even more challenging than anticipated. The solution we have found is very elegant and already yields excellent results”.
Indeed, the team had to face the absence of the large set of data typically used to train AI algorithms and, instead, developed physically informed neural networks to achieve excellent results. Maletinsky shares this enthusiasm “The novel machine learning approach represents a key advancement that allows for quantitative reconstructions of magnetization with a reliability and accuracy that already exceeds prior state of the art”.
Innovation for Qnami products
Qnami is now working to integrate these new tools into their products. Stark comments “our mission is to bring valuable information into the hands of our users. It is important not only to focus on improving data acquisition but also on how to use the data to provide meaningful information, which our customers can use to solve their problems.”
And the story does not end here. “We are very excited to apply our learnings to new problems” explains Stark, who is now looking to expand his software team, as the work and the collaboration go on into the exploration of new applications.