Certifying the unpredictable: a key step in quantum computing

For the past several years, Argonne National Laboratory computational mathematician Jeffrey Larson has been developing algorithms and software to solve challenging scientific problems. Recently, he was part of a team comprising JPMorgan Chase, Quantinuum, the University of Texas and the U.S. Department of Energy’s (DOE) Argonne and Oak Ridge national laboratories investigating randomness in quantum computing.

Random number generation is a natural fit for quantum computers because the laws of quantum mechanics are based on probability. Unlike traditional computers, quantum computers can generate random numbers that are difficult — or even impossible — for outsiders to predict.

To test a randomness-generation protocol, the research team used the 56-qubit Quantinuum System Model H2 trapped-ion quantum computer. They sent specially designed quantum circuits to the quantum computer, which ran each circuit and generated a random number in about two and a half seconds (see Fig. 1).

However, ensuring that random numbers are truly unpredictable — and not secretly calculated in advance by someone else —requires more than just a fast quantum chip. It requires certification, or proof that it’s truly random. Certification takes a lot of computing power. In this case, the researchers were able to use DOE supercomputers to ensure that the numbers generated by the quantum computer matched the initial circuits.

Larson’s work involved helping implement and fine-tune the optimization algorithm used to efficiently perform a process called tensor contraction. Tensor contraction simplifies complex data structures known as tensors, or multidimensional arrays, in order to reduce the number of their dimensions. This is important for simulating large quantum systems. In this study it allowed the researchers to determine the best-possible classical computing strategy to either verify or attempt to fake the quantum results.

By integrating cutting-edge quantum hardware from Quantinuum with the combined power of Oak Ridge’s Frontier and Argonne’s Polaris supercomputers, the researchers were able to confirm that the quantum computer produced more than 70,000 truly unpredictable bits.

“Certified random numbers are critical for secure communication, online privacy and fairness in systems such as lotteries or elections,” Larson said. ​“This research delivers a secure protocol that allows even classical users to verify, in real time, that a remote quantum device produced true randomness. It is an important step toward demonstrating the practical applicability of present-day quantum computers.”

For the full report of this groundbreaking work, see the paper by M. Liu et al., ​“Certified randomness using a trapped-ion quantum processor,” Nature 640, 343-348 (2025).