Chip corrects quantum errors

To safeguard one quantum bit (qubit) against errors, start by building five qubits on a chip. That’s what researcher Dr. Leo DiCarlo and his team from the Kavli Institute of Nanoscience (Faculty Applied Sciences) and the QuTech Institute have done. They published a photo of their 2 by 7 mm chip in Nature, together with the test results.

The logic behind the manifold is this: three qubits (called top, middle and bottom) are used to encode one qubit’s worth of data in special, so-called ‘entangled’ states of the three. Direct measurement of the state of each qubit would collapse the encoded information. So instead of direct measurements, the researchers perform a parity check between middle and top qubits, and another between middle and bottom qubit. The parity check produces a ‘0’ when both qubits are in the same state or a ‘1’ when they are in a different state. Each parity check requires one extra qubit, bringing to five the total number of qubits on the chip.

“We can detect flip errors on any one qubit and still preserve the encoded information”, DiCarlo explains. Take a simple example. If the original state is ‘1’, the three qubits are encoded into ‘111’. Now suppose the top parity check detects an error, then we know that either the middle or the top qubit flipped. Now, if the bottom parity check does not detect an error, we know the top qubit must have flipped. Because the double parity check not only tells us that an error has occurred, but also which qubit it is in, we can preserve the original information.”

Such bit flips are one of three types of disturbances that happen to qubits, the other two types carry even stranger names. To protect one qubit’s worth of quantum data against all possible types of error, one can use nine instead of three qubits to store the data, and eight instead of two qubits to perform the parity check. Building and operating 17 qubits on a chip is the next goal for the team.

Although clear in concept, the current quantum error corrector is far from perfect, as Di Carlo is quick to admit. What’s even worse: the parity error checking is itself a source of errors. There is 25 % chance of a false positive: a parity check detecting an error where there was none.

Error checking of quantum bits is a subtle game. If you do it too slow, errors will accumulate, and you will no longer be able to reconstruct the original quantum state. If you check too often, the error checking will introduce additional errors.

So will future quantum computers be fickle instruments the output of which can never be trusted? Di Carlo thinks not. To begin with: if an output is erroneous, the machine will detect and signal the errors. Di Carlo also puts his trust in technological progress to take the quantum computer beyond the threshold of fault tolerance. Improvements in error correction devices and increasing qubit stability are crucial for that progress to happen.

⇒ D. Riste, S. Poletto, L. DiCarlo et. al., Detecting bit-flip errors in a logical qubit using stabilizer measurements, Nature Communications, 29 April 2015, doi: 10.1038/ncomms7983