By Andre Saraiva, UNSW

Ambitious quantum algorithms require many qubits. Many more than quantum computer manufacturers are offering currently. The most qubit-hungry applications are those that require quantum error correction, such as the notorious Shor algorithm that threatens our cryptosystems.

Exactly how many qubits are needed is a complicated question. The more qubits we cram in a chip, the harder it is to control them in an accurate fashion. And the worse the accuracy, the more qubits are necessary. The most advanced error correction strategies will only become sustainable against errors at the mark of thousands of qubits (or millions for most useful applications).

Amazon is betting that they can shrink this number to less than 100 qubits. In order to do this, they are rethinking qubits from scratch, putting all of their efforts in a unique type of qubit, based on the so-called cat states. This name is a tip-of-the-hat to Schrodinger’s all-enduring feline, that illustrates the weirdness of quantum mechanics when you think about the superposition of multiple particles, such as those that compose a life-sized cat. Using a much smaller number of particles (dozens, instead of moles), Amazon’s favourite qubits can be made.

Theoretically, their plan is sound – qubits made from these cat states can be monitored for errors without directly measuring the qubit state, which would cause it to collapse and defeat the purpose of quantum computing. In a recent paper, Amazon brought out all its guns and calculated the exact number of qubits they need to tap into quantum error correction. This number is an encouraging 97 qubits. With canonical qubits, this number would be at least 1457. Reading the small print under this number, however, is a technically challenging task.

Fully exposed in a 40-page long preprint manuscript, the team led by Dr. Fernando Brandão describes details of their model for how errors creep into quantum calculations, how those errors would be handled in an error correcting algorithm and how computer simulations were used to guarantee that fault tolerance is achieved. This is done for a general class of quantum processing architectures called GKP (after Gottesman-Kitaev-Preskill). Following all the details requires a deeper understanding of how quantum computers work and theoretical physics than even most scientists and engineers would lack.

To reduce to practice this theoretical idea, Amazon will need to push the fidelity of cat state control to the level currently achieved in other qubits. The 15-fold improvement that the theorists have calculated assumes cat state qubits of quality comparable to regular qubits, which is not the current status of the technology. A few months ago, they have laid out a plan for improving cat states using a particular combination of technologies, but the technical milestones required to make their plan come true are numerous.

So, what’s next?

It is refreshing to see a quantum computing program being led by error correction theory. Looking ahead is the way to go for this type of complex project, especially when the single qubit characteristics can hugely affect what the final full-scale architecture will look like. However, Amazon’s chips keep piling on a technology that is behind others in terms of experimental demonstrations. The real attention grabber now will be Amazon’s first in-house qubit and confirm that they are able to control it with small error rates, preserving the properties of the cat state that make it easier to correct.

Dr. Saraiva has worked for over a decade providing theoretical solutions to problems in silicon spin quantum computation, as well as other quantum technologies. He works on MOS Silicon Quantum Dot research and commercially-oriented projects at the University of New South Wales (UNSW).

July 9, 2021