Scientists find quantum computers forget most of their work

Quantum circuits operate in a similar way. They consist of many small steps, called ("operations"), that work together to process information in a highly coordinated manner.

Now picture those dominoes slightly unsteady. In quantum systems, that instability is known as "noise." It may seem minor at first, but even small disturbances can build up over time and interfere with the entire sequence.

How Noise Limits Quantum Computing Power

This raises an important question. If every step in a quantum circuit is affected by noise, does increasing complexity still provide an advantage? Quantum circuits are essential to technologies such as quantum computers, which aim to solve problems beyond the reach of classical machines.

A new theoretical study has explored this issue in depth. The researchers found that noise places a strict practical limit on how deep a quantum circuit can be, meaning how many steps can be performed in sequence. They also showed that noise can make parts of these circuits easier to simulate using classical computers.

The study was led by Armando Angrisani and Yihui Quek at EPFL, Antonio Anna Mele at the Free University of Berlin, and Daniel Stilck França at the University of Copenhagen. The results were published in Nature Physics.

Why Only the Final Steps Matter

To understand the effect of noise, the team examined large groups of quantum circuits built from simple two-qubit operations. Their model included realistic conditions, where each qubit experiences noise after every step.

Using mathematical analysis, they tracked how the influence of each layer moves through the circuit. The results revealed that in most noisy quantum circuits, only the last few steps significantly affect the outcome.

Even when circuits are designed to be very deep, the impact of earlier operations gradually disappears. In the domino comparison, it is as if only the final pieces determine the end result.

This has real implications. When a quantum computer is used to calculate properties such as the energy or the state of a qubit, the result is largely determined by the final layers. Earlier operations effectively "fade from memory" as noise accumulates.

Why Noisy Quantum Circuits Can Still Be Trained

The findings also help explain why noisy quantum circuits can still be adjusted or "trained" for certain tasks. Changing the circuit's settings can influence the result, but mainly because the final layers continue to play an active role.

As a result, a deep circuit affected by noise behaves much like a shallower one. Adding more steps does not necessarily increase performance, since most earlier steps no longer contribute in a meaningful way.

What This Means for Future Quantum Technology

This work provides a clearer picture of what current quantum machines can realistically achieve. Simply increasing circuit depth is unlikely to deliver better results for many common tasks, especially those based on local measurements.

Future progress will likely depend on reducing noise or designing circuits that can function effectively despite it. The study also highlights a potential misconception. Noisy circuits may appear trainable, but this is partly because noise has already reduced their effective complexity. Treating noise as a simple blur can lead to unrealistic expectations about the true capabilities of quantum computing.

Contributors

• Free University of Berlin
• EPFL
• University of Sorbonne
• University of Chicago
• Fraunhofer Heinrich Hertz Institute
• ENS Lyon
• MIT