By Andre Saraiva, UNSW

All roads lead to silicon. The unparalleled power of integration developed by the CMOS industry attracts all qubit-makers. Superconductors, photons and ion traps have now been retrofitted to accommodate fabrication in foundries. But spins are naturally born silicon qubits, and as such there is a lot of expectation for this technology. Now a larger-than-99% two-qubit fidelity, the key milestone for proving the viability of spin in silicon as qubits, was achieved in three corners of the world at the same time – the Netherlands, Australia and Japan.

A question immediately comes to mind: “is 99% a lot?”. While our human brains tend to think so, it still means that one error occurs every hundred operations. Put together thousands of qubits and millions of operations, and soon you find yourself in a catastrophic situation. For instance, it is estimated only one error in a quadrillion (1/1015) is tolerable in order to implement Shor’s algorithm to decrypt RSA-2048 (and win a $200,000 cash prize). That is why some critics of quantum computing say it will be impossible to create such a device.

No amount of careful engineering can push the error rate of individual qubits this low. Instead, scientists and engineers focus on creating logical qubits – a redundant arrangement of tens or hundreds of thousands of physical qubits that collectively have a much-reduced error rate. The higher the fidelity, the less overhead is necessary. That is why companies such as Google, IBM and IonQ, who have long ago hit the mark of 99%, keep pushing for ever higher fidelities.

And why would 99% in spins be newsworthy? Two reasons: spins in silicon are tiny (tens of nanometres), so this is the only technology that can put billions of qubits on the same chip. All other technologies need to either connect separate chips or content themselves with NISQ applications only (no Shor, Grover, universality, etc). So, however behind spins might be in control fidelity, they make up for it in prospects for scalability and error correction.

The second reason is that 99% is the minimum fidelity needed to enter the game of error correction. It doesn’t matter how much overhead you are willing to pay, error correction fails if the individual operations are too faulty. Hardly any error correction code can cope with error rates above 1%, which sets a “99% club” of qubits. This club is quite select – only superconductors and ion traps have reached this fidelity in a fabricated system (some systems like NMR molecules and NV centres in diamond can achieve this fidelity, but they relied on fortuitous arrangements of atoms).

So, what’s next?

As discussed, this fidelity is good but not great. Some control techniques can help fend off noise and create more robust pulses. Pushing another “9” in fidelity can make a world of difference when it comes to realistically achieving error correction.

Another demonstration that remains to be seen is the extent to which this fidelity can be achieved in a device with several qubits. Crosstalk (the unintended interaction between qubits that are very close to each other) will kick in and the challenges multiply. Ions and superconductors have been achieving 99% average fidelities for a while now, so silicon still hasn’t entirely caught up with the leading technologies.

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).

January 28, 2022