Quantum computers promise to solve problems far beyond the reach of today’s machines, but there is a stubborn obstacle standing in the way: errors.
Unlike ordinary computers, quantum devices are extremely sensitive to their environment, and even tiny disturbances can throw calculations off track.
Now, a team of Australian and international scientists has uncovered a deeper reason why these errors are so hard to control—quantum computers can “remember” them over time.
In a new study led by Dr. Christina Giarmatzi at Macquarie University, researchers have, for the first time, mapped how errors inside a quantum computer evolve and connect across different moments in time.
The findings, published in Quantum, show that errors don’t simply appear and disappear at random. Instead, they can linger, change, and even influence future errors.
“We can think of it as quantum computers retaining a memory of past errors,” Giarmatzi explains. “Many existing methods assume that errors have no memory at all, but that assumption just doesn’t hold in real machines.”
This kind of behavior, known as non-Markovian noise, is a major challenge for building reliable, large-scale quantum computers. If errors are linked across time, they become much harder to predict and correct.
To uncover this hidden structure, the researchers developed a new experimental approach that allowed them to reconstruct the full history of a quantum process. In simple terms, they found a way to see not just when errors occur, but how one error can affect what happens later.
The experiments were carried out on advanced superconducting quantum processors, some hosted at the University of Queensland and others accessed through cloud-based machines run by IBM. These processors use tiny electrical circuits cooled to near absolute zero, where quantum effects can be harnessed—but also easily disrupted.
A long-standing problem in studying quantum systems is that measuring them changes their state. Once you look at a qubit mid-calculation, you can’t simply reset it and continue as if nothing happened.
The team overcame this by using a clever statistical trick. They assumed that, on average, measurement outcomes were evenly split between possible results, and then used software to work backward and reconstruct what the system must have been doing.
As co-author Dr. Fabio Costa from Nordita put it, the hardware was already capable—the missing piece was knowing how to prepare the system again after a measurement.
What the researchers found was striking. Even the best current quantum machines show subtle time-linked noise, including noise that spreads between nearby qubits and has genuinely quantum properties. These effects quietly build up and interfere with calculations over time.
Understanding this hidden memory is a crucial step toward better quantum error correction. With more accurate models of how errors behave, scientists can design smarter ways to detect and cancel them out.
The work marks an important milestone. By revealing how quantum computers actually behave in the real world, it brings researchers closer to the long-term goal of building quantum machines that are not just powerful, but dependable.
