 |
| A view of the inner components of a quantum computer Ben Fourmi |
Special quantum bits called “cat qubits” could make it possible for quantum computers to make fewer errors and more efficiently break common encryption algorithms.
Conventional computers process information by manipulating bits, which take on values of 0 or 1, based on whether a transistor is “on” or “off”. For quantum computers, however, a standardised counterpart for the transistor has not yet been settled upon. Quantum bits, or qubits, can be made from charged atoms, particles of light that have quantum properties and more.
One important characteristic of any qubit is how likely it is to make errors during calculations. Currently, quantum computers aren’t broadly practical because they either make too many errors or are not big enough, i.e., don’t contain enough qubits to self-correct those errors. Jérémie Guillaud at Alice&Bob, a quantum computing company in Paris, found that making quantum computers with so-called “cat qubits” may enable them to correct their own errors while retaining an unexpectedly small size.
These qubits are named after Erwin Schrödinger’ famous thought experiment where a cat is in a combined state of being dead and alive. Cat qubits are also a combination of two states, but they describe two different ways in which light trapped inside of a small hole in a superconducting circuit can move back and forth. Guillaud and his colleagues mathematically analysed a quantum computer made of many such circuits.
They evaluated the number of qubits this computer would need to break the encryption used to secure Bitcoin transactions. They assumed some qubits would be dedicated to the calculation, and some to correcting errors in that calculation. They found that 126,133 cat qubits and 9 hours of computation would be enough.
Though this may seem like a large number of qubits, the value is about 160 times smaller than the previous lowest estimate of qubits necessary to perform such a task, says Guillaud. “Four years ago, another team calculated that you would need 20 million superconducting qubits. Before that the number was believed to be in the billions,” he says.
Guillaud says that the necessary number of qubits is smaller for cat qubits because they are engineered to make significantly fewer errors equivalent to a bit “flipping” from a 0 to a 1. Their two states are chosen so that it would take a lot of energy to force them to flip. Researchers have estimated that such errors happen about 1 million times more rarely for cat qubits, says Guillaud. In a cat qubit-based computer then, very few or no qubits would have to be set aside to correct bit flip errors, decreasing the total number of qubits needed for error correction.
It is really important to evaluate how cat qubits’ relative resilience to bit flips changes outside of the conditions in the new analysis, says Martin Ekerå at KTH Royal Institute of Technology in Sweden. The assumption that cat qubits will experience no bit-flip errors even when there are many of them and they are used for hours at a time is pretty extreme, Craig Gidney at Google wrote on Twitter.
“The challenge now really is on the hardware side. We do need to do better than what has been done in labs so far,” says Guillaud. He and his colleagues are working on improving their cat qubits in experiments, like figuring out how to best shield them from cosmic rays that can cause bit flips.
Reference
0 Comments