What good is a fast computer if you can’t trust it? Thanks to half a century of research on getting computers to do their job correctly even in the presence of mechanical errors, our modern machines tend to be pretty reliable.
Unfortunately, the laws of quantum mechanics render all that research useless for quantum computers, the sheer complexity of which leaves them prone to errors. Now, we finally have the first demonstration of a quantum program that can detect data corruption.
Two research groups – one from the University of Maryland and Georgia Tech and the other from IBM – have demonstrated the same quantum error-detecting program, albeit implemented with different hardware.
“Quantum computers can never be practical without error correction,” says Daniel Lidar at the University of Southern California. As we build bigger quantum computers, “errors add up to the point that they wash out the quantum effects… which obviates the need for the quantum computer,” says Lidar.
The telltale qubit
In classical computers, error detection and correction are done with duplicated data – any mistakes can be remedied by reconstructing the erroneous bits from uncorrupted parts of the machine.
But in quantum computers, it’s impossible to duplicate quantum states without measuring them, and measurement causes loss of information. So, without any means to back up intermediate results, quantum computers simply cannot use classical techniques of error detection and correction.
The solution the teams are proposing consists of five qubits, each of which can be in two states: one or zero. For every two qubits’ worth of information, there are four possible combinations: zero-zero, zero-one, one-zero, and one-one.
The program uses four qubits to record these states, while the fifth ancillary qubit catches errors in the first four qubits. For example, when four qubits represent a two-qubit state that should be zero-zero, the information is in a superposition where the four qubits are either showing four ones or four zeros, or an equal number of each digit. If there’s an error in one qubit, the fifth qubit will note the uneven distribution of ones or zeros and change its state.
Only one error
This verification system reduces the error rate to 0.1 per cent, compared with about 10 to 15 per cent potential error for quantum programs of about this size, says Norbert Linke of the University of Maryland. The IBM group’s implementation shows similarly reduced error rates as well.
However, there are limitations to the approach. For example, if one error changes the ancillary qubit from zero to one, and a second changes it back to zero, then the program will not detect these two consecutive errors. Fortunately, experiments suggests such a scenario is rare.
Moreover, the program merely demonstrates the existence of an error. Locating the error precisely requires more qubits. Linke says his group plans to scale up the experiment and implement an error-correction feature, which requires more qubits. Andrew Cross of IBM Watson Research Center says his group plans to first perfect the five-qubit program before moving on to error correction.