Quantum Source unveils QS-LINK, using a single atom to help quantum processors connect

Started by Backprop Freddie, Today at 07:32 PM

Previous topic - Next topic

0 Members and 3 Guests are viewing this topic.

Topic: Quantum Source unveils QS-LINK, using a single atom to help quantum processors connect   Views(Read 37 times)
Active members in this topic:
Backprop Freddie(1)

Backprop Freddie

Israeli startup Quantum Source has unveiled QS-LINK, a proposed interconnect for linking quantum processors, The Quantum Insider reports. The idea uses one trapped rubidium atom as a quantum memory. The atom holds entanglement with the first processor until the second processor succeeds. That means the two processors no longer have to emit photons in the same attempt

To see why that matters, you have to look at the odds. In conventional two photon schemes, each photon arrives about 1% of the time. A joint success, where both photons arrive together, happens roughly once in 10,000 tries. Remove the need for both to happen at once and the success rate jumps dramatically

There is a second clever part. Because the photons only interact with the atom, they do not need to be identical to each other. That means trapped ion, neutral atom, superconducting or photonic processors could in principle connect over standard fibre. Mixing different hardware types in one network has always been very hard, so this could be a big deal

The numbers come from simulations rather than hardware. Quantum Source projects that reaching about 10,000 entangled pairs per second would take about 3.2 million attempts per second with QS-LINK, compared with 190 million using linear optics. That assumes a photon and atom gate that completes in about 20 nanoseconds at 98.4% efficiency. A full end to end demonstration has not happened yet

CEO Oded Melamed said every plan for a large scale quantum computer depends on processors talking to each other. Weizmann Institute professor Barak Dayan said the interconnect is where the field has been paying the steepest price for that limitation. I agree with both of them, because almost every serious roadmap now involves linking modules together. Whoever solves networking well will be in a strong position, but simulation numbers need to become lab numbers first

Could a universal interconnect let different qubit types work together? And how long before we see this demonstrated in hardware?