Scheme checks answers to quantum computation

CQT's Joe Fitzsimons and his collaborators show way to check that a quantum computer is functioning as it should
01 October 2013

Research by CQT's Joe Fitzsimons and his collaborators shows a way to check that a quantum computer is working as it should. This is an artist's representation of the working parts of the computer - clusters of quantum entangled photons that carry information. Image: Equinox Graphics.


Here's a conundrum: you build a quantum computer to solve a problem that you have no other way to answer. How do you know if the computer gets the answer right?

This is the challenge that CQT's Joe Fitzsimons and his collaborators tackle in research published 29 September in Nature Physics. They ran a calculation on a small quantum computer in a way that tested the quantum computer at the same time.

"You're checking that the device functions as you expect it to function. It sounds like it should be easy but it's not," says Joe, a Research Assistant Professor at CQT and Assistant Professor at the Singapore University of Technology and Design. He has received a $3-million fellowship from Singapore's National Research Foundation to work on such problems.

The promise of quantum computers is that they can quickly solve problems that today's supercomputers could not complete in the age of the Universe. That's all well and good when the problem is something like factoring large numbers – a famous example of quantum computing because it threatens the encryption used in many of today's secure transactions. The answer can be checked simply by multiplying the numbers back together again. When the problem is something like a simulation of material properties, the correct answer isn't known. How, then, can you be sure you trust your computer?

Be careful who you trust

Joe and Elham Kashefi, a researcher at the University of Edinburgh, UK, proposed a way to hide test calculations as 'traps' in a program. If the computer is malfunctioning, the answer to the trap calculations will show that something is going wrong. The scheme, described in a preprint last year, builds on their earlier idea with Anne Broadbent from the Institute of Quantum Computing in Waterloo, Canada, for 'blind' quantum computing.

In blind quantum computing both the program and the data inputs are disguised. Using this trick in their new scheme means that the embedded traps are indistinguishable from the rest of the input. This is important because the computer may be under someone else's control. You might just use the verification step to check your trusted computer for random errors or imperfections in construction. But you might also want to use it in a situation where you are suspicious of the quantum computer – if, say, you'd bought the device from a dodgy dealer or were just renting time on someone else's machine. Then you not only need to test the quantum computer but also be sure that it had not been designed or programmed to deceive your test.

Joe and Elham collaborated with experimental physicists Philip Walther and Stefanie Barz of the University of Vienna, Austria, to design a verification scheme that could be tested in the Vienna labs. The team had already worked together to perform blind quantum computing. The new experiment demonstrated the verification procedure on a quantum computer that uses photons as information carriers. The proof-of-principle experiment used four quantum bits (or qubits) and alternated runs of the calculation with runs of the trap. With a different setup and larger number of bits, Joe and Elham's scheme would allow the program and test steps to run in parallel on the quantum computer.

For more details, see "Experimental verification of quantum computations", Nature Physics doi:10.1038/nphys2763 (2013); arXiv:1309.0005. The trap idea was first described in "Unconditionally verifiable blind computation", arXiv:1203.517.

Coverage in the media:

Learn more