Scientists at Atom Computing have developed a method to reuse error-tracking qubits in quantum computers made from cold atoms, recycling them up to 41 times without errors. This approach aims to make quantum computations more reliable and efficient by reducing the need for vast numbers of qubits. The technique addresses a key challenge in scaling up quantum technology.
Quantum computers using qubits from extremely cold atoms are expanding rapidly, but error rates currently hinder their practical use. To overcome this, researchers led by Matt Norcia at Atom Computing, a US firm, have created a system that reuses or replaces ancilla qubits, which monitor errors during computations.
In their setup, qubits are ytterbium atoms cooled near absolute zero using lasers and electromagnetic pulses. These are manipulated with 'optical tweezers' into three zones: one with 128 tweezers for data computation, another with 80 for error measurement and swapping faulty qubits, and a storage zone for 75 fresh qubits. This allowed the team to recycle ancilla qubits 41 times consecutively.
“Any computation of use is likely to be a very long computation, so you’d have to do many rounds of measurements. Ideally, you want to be able to reuse the qubits throughout multiple rounds so that you don’t have to continue providing more qubits into the system,” Norcia explained.
Challenges included preventing stray laser light from disturbing qubits, requiring precise laser control and tuning data qubits to avoid interference. Yuval Boger at QuEra emphasized the importance: “Ancilla reuse is fundamentally important for quantum computing progress.” Without it, even modest calculations would demand millions or billions of qubits, which current hardware cannot support.
Similar techniques appear elsewhere: a Harvard and MIT team sustained a 3000 rubidium atom quantum computer for hours, and Quantinuum's Helios ion-based machine also reuses qubits. Norcia noted that the neutral atom community recognizes the need for resetting and reloading atoms during computations.
The work is detailed in Physical Review X (DOI: 10.1103/v7ny-fg31).