Buried in a sliver of silicon sits a single atom of antimony, and inside it, a nucleus spinning in one of eight possible states. Those eight states are doing a job. They are holding quantum information, the fragile raw material of a future computer, in a form physicists rather poetically call a Schrödinger cat state. And here is the trouble: to keep the machine running, you have to keep checking whether that cat is still alive. Look too hard, and you kill it.
That, more or less, is the central headache of quantum error correction. A useful quantum computer will need to measure its own innards constantly, hunting for mistakes mid-calculation, without smudging the very data it is trying to protect.
A team at the University of New South Wales has now found a neat way around the problem, and they have explained it with a metaphor that practically writes itself. “Imagine you’re trying to find your cat hiding in one of eight identical cardboard boxes, in a dark and noisy room,” says Andrea Morello, a Scientia Professor at UNSW. You cannot open the boxes, because opening a door may kill the cat. So how do you find it? You could rig up eight sprinklers, one per box, spray each in turn, and listen for an indignant meow.
The snag is the noise. In a racket, you might hear a meow from an empty box, or miss a real one from the right box.
The usual fix is brute repetition: spray everything over and over, then bet on whichever box yelped most often. But each spray is a gamble. “Repeatedly sprinkling the boxes risks changing the very thing you are trying to observe,” says Morello. Spray too much and the startled cat bolts to a different box. In the lab, the “cat” is the antimony nucleus, and the “sprinkler” is an electron shoved onto the atom and then pulled off again depending on the nuclear state. That tugging is what can knock the nucleus into a new state, an effect with the rather brutal name of ionization shock.
Morello’s team, with lead author Arjen Vaartjes, asked a different question. What if you stop the moment you hear the first meow?
The trick is to treat that first squeak as a working guess, then switch tactics entirely. Instead of carrying on spraying everything, you sprinkle only the boxes where the cat supposedly isn’t. If all of them stay quiet, your guess gets stronger, and crucially you have learned this without ever poking the box you think the cat is in. Quantum mechanically, those silent boxes are what the researchers call the dark-state subspace, and probing them is a “negative-result” measurement: you extract information from the absence of a signal, leaving the system’s underlying physics untouched. “The absence of a signal confirms the presence of another, without interacting directly with the system,” says Morello. Or, as he puts it rather more memorably, “Sometimes, silence can be loud.” In the actual device, this means the meddling electron only needs to come off the atom once, after which the protocol just listens to the empty states.
The numbers back the metaphor up. Using this adaptive approach on the antimony qudit, the team pushed its readout fidelity from about 98.93 per cent to 99.61 per cent, while cutting the total measurement time to roughly a third. That sounds like a modest bump. It isn’t.
“This value is significant because it puts our system in the range of measurement fidelities necessary to perform successful quantum error correction,” says Vaartjes. Cross that threshold and you are, in principle, in territory where the error-correcting machinery can actually keep up with the errors.
What gives the work legs beyond one chip in Sydney is that the underlying problem is everywhere. Lots of quantum platforms read out their delicate qubits indirectly, by poking a messenger particle and seeing what happens, and in many of them one measurement outcome jostles the system harder than the other. The UNSW group reckons the same stop-at-first-meow logic should carry over to nitrogen-vacancy centres in diamond, to spin qubits in quantum dots, to clusters of donor atoms, even to arrays of neutral atoms held in laser tweezers. “Because many architectures also employ similar hardware, the new protocol can readily be adapted to other platforms that suffer from errors during measurement,” says Morello. Better still, it runs on a humble bit of programmable logic already sitting in most labs; Vaartjes credits the result to “a fast FPGA, a cup of coffee, a dedicated team of clever researchers and a long Friday afternoon of coding”.
None of this means a working quantum computer is around the corner. Readout is only one of several things that all have to clear the fault-tolerance bar at once, and in the antimony system the measurement still takes long enough, milliseconds, that the information can start to fade before you are done.
Which is partly why the team is already eyeing faster cousins of the setup, including a version that reads the nucleus through a nearby quantum dot in a fraction of the time.
Still, there is something quietly satisfying about the core idea. For a century the Schrödinger cat has been shorthand for the impossibility of looking without disturbing. Here is a case where, by choosing very carefully what not to look at, you can have a peek and let the cat live. “We can now extract information about the quantum system just gently enough to keep it intact,” says Morello. Whether that gentleness scales up to the millions of measurements a real machine will demand is the next thing to find out.
Frequently Asked Questions
Why does measuring a quantum computer risk destroying its data?
In quantum systems, the act of looking can itself jolt the thing you are observing into a different state. Error correction needs to check for mistakes over and over while a calculation runs, so each of those checks is a chance to corrupt the very information being protected. The UNSW result matters because it shows you can gather that information while barely touching the system at all.
How does listening to the “empty” boxes actually tell you anything?
Once you have a first guess about where the quantum state is, you probe only the states it supposedly isn’t in. If those stay silent, your confidence in the guess climbs, and because nothing responded, the system’s underlying physics is left undisturbed. It is information drawn from an absence of signal rather than a direct poke, which is why the team calls silence loud.
Is 99.61 per cent really good enough for a quantum computer?
For this particular step, readout, it crosses into the range physicists think is needed for error correction to keep pace with errors, which is why the lead author flagged it as significant. But readout is only one ingredient; gates and initialisation all have to clear similar bars at the same time. So it is a meaningful milestone rather than a finish line.
Could this trick work on other kinds of quantum computers, not just silicon atoms?
The researchers think so, because the basic problem, reading a qubit indirectly through a messenger particle that disturbs one outcome more than the other, crops up across many platforms. They point to nitrogen-vacancy centres in diamond, quantum-dot spin qubits, donor clusters and neutral-atom arrays as likely candidates. And since it runs on programmable logic chips most labs already own, adopting it should be relatively painless.
ScienceBlog.com has no paywalls, no sponsored content, and no agenda beyond getting the science right. Every story here is written to inform, not to impress an advertiser or push a point of view.
Good science journalism takes time — reading the papers, checking the claims, finding researchers who can put findings in context. We do that work because we think it matters.
If you find this site useful, consider supporting it with a donation. Even a few dollars a month helps keep the coverage independent and free for everyone.