In a prototype quantum computer, a magneto-optical trap (cube) cools rubidium atoms to nearly absolute zero. PHOTO: QUERA COMPUTING
Dear Commons Community,
Science has a featured article this morning entitled, “Atomic Explosion” that describes using atoms to develop a quantum computer. Below is an excerpt. Be forewarned, the article gets into the weeds a bit.
“Inside a cramped, windowless laboratory, mirrors, lenses, and beam splitters crowd together on a massive table, forming a dense and seemingly random thicket. In fact, each gizmo is precisely placed to control laser beams that ricochet through the setup before entering a port in a gleaming vacuum chamber jutting up through the table. Within the chamber, the beams suspend individual atoms of rubidium against gravity’s pull, arrange them in patterns, and manipulate their internal quantum states.
Dolev Bluvstein, a physicist at Harvard University, points to a digital camera mounted to peer down into the chamber. Several bright green dots shine on its screen. They are the atoms—hovering, fluorescing, making the atomic scale visible. They are also the quantum bits, or qubits, at the heart of a quantum computer, one that is pushing boundaries and rivaling more established concepts. Speaking over the thrum of pumps, Bluvstein says, “No matter what the advantages of different computational approaches, I think neutral atoms are the most fun because we take pictures of atoms.”
The effort at Harvard represents the leading edge in so-called neutral atom-based quantum computing. For decades, the approach languished as physicists struggled to coax even two atoms to interact, as they must to perform computations. Meanwhile, researchers pursuing other approaches raced ahead, assembling functional, albeit noisy, quantum computers with dozens of qubits consisting of tiny superconducting circuits or individual ions trapped on microchips.
But, thanks to a few key advances, atom-based quantum computing has come roaring back. Physicists can now assemble arrays of thousands of atoms—thousands of potential qubits. Because all the atoms of a particular element and isotope are identical, they should be more reliable and easier to control than manufactured superconducting qubits. “Our qubits don’t need improving,” says Dana Anderson, a physicist at the University of Colorado Boulder and chief technology officer for the startup Infleqtion. “Nature makes them, and we just plug them in.”
“What we have done [could] put this entire field on a completely different slope.”
Even proponents of rival technologies acknowledge that the atomic approach is enjoying a moment. “They will definitely suck the air out of the room for the next few years,” says William Oliver, a physicist at the Massachusetts Institute of Technology who works with superconducting qubits. However, Oliver warns, atom-based quantum computing faces challenges of its own, including the slow pace of herding atoms with lasers. “It’s not clear that it will be fast enough to be practical,” he says.
Nevertheless, many startups are already selling rudimentary atom-based quantum computers, and dozens of academic groups are exploring the approach. The leader of the Harvard effort, Mikhail Lukin, a Russian-born physicist with a boyish mop of hair and piercing blue eyes, says the progress is driving all of quantum computing forward. “I am very hopeful that what we have done will put this entire field on a completely different slope.”
WHETHER IT’S PREPARING your taxes or portraying mayhem in a video game, your computer functions by manipulating binary numbers, long strings of 0s and 1s. It encodes those numbers in tiny electrical switches called bits, flipped on or off to signify a 0 or a 1. In contrast, a quantum computer manipulates qubits that can be set to 0, 1, or 0 and 1 simultaneously—until they are measured, at which point the qubit’s state collapses randomly to 0 or 1.
In principle, those simultaneous states enable a quantum computer to tackle certain problems—perhaps cracking cryptographic codes or modeling molecules—that would overwhelm any conventional computer. Possible solutions to a problem correspond to abstract quantum waves sloshing through the qubits. The waves interfere with one another, like ripples on a pond, so that those signifying incorrect solutions cancel one another and the correct solution pops out.
Such algorithms require special states in which the qubits share a quantum link called entanglement. When two qubits are entangled, their states can be uncertain, but correlated. For example, both can be 0 and, simultaneously, both can be 1. When measured, they will collapse into the same state. In contrast, if the qubits were in 0-and-1 states but not entangled, one might collapse to 0 and the other to 1.
A qubit can be anything with two addressable quantum states that’s not too susceptible to interference from its surroundings and can interact with others of its kind. One popular choice has been an ion—an atom stripped of an electron. The two states, 0 and 1, correspond to the ion spinning up or down. When trapped in a row by electric fields on the surface of a microchip, ions can be manipulated with laser light or microwaves and can interact through vibrations zipping along the row. So far, ions are the most precisely controllable qubits, and companies such as Quantinuum and IonQ have assembled computers with dozens of them.
Another leading type of qubit consists of a tiny circuit made of superconducting metal that resonates with unquenchable current. Controlled by microwaves and much larger than ions and atoms, superconducting qubits are more susceptible to noise and, like any manufactured object, suffer tiny irregularities. Nevertheless, both Google and IBM have developed quantum computers with more than 100 superconducting qubits, and those machines have performed the most complex quantum computations so far.
In the 1990s, when the quest to build a quantum computer began in earnest, atoms seemed like an obvious choice for qubits. Nothing is more pristinely quantum mechanical than an atom, whose behavior quantum theory was developed to explain. An atom possesses a ladder of precisely predictable quantum states whose energy depends on how its electrons whizz around the nucleus and spin.
Physicists had already developed powerful techniques to control atoms. Precisely tuned lasers could ease an atom from one quantum state to another and could cool atoms to nearly absolute zero, a step necessary to protect their delicate quantum states from the scrambling effects of heat. Researchers could also trap single atoms in a spot of laser light, in so-called optical tweezers, and even assemble 2D grids of atoms, the starting point of a quantum computer, without any of the hassles of etching circuits onto chips.
But atom-based quantum computing stumbled out of the gate. Because they’re uncharged, atoms only interact feebly with their environment. That’s good for quantum computing, as it makes atom qubits less susceptible to noise. But atoms also barely interact with one another. That’s bad for quantum computing, as qubits must interact to perform logical operations. Without interactions, qubits don’t compute.
IT WASN’T UNTIL 2000 that physicists solved the problem. Lukin, Peter Zoller, a theorist at the University of Innsbruck, and their colleagues found a way to momentarily strengthen the interaction between two neighboring atoms 1 billion–fold.
An atom with a single electron in its outer shell, such as the rubidium Lukin’s group uses, has a ladder of states determined by the kinetic energy of that electron and the directions in which the electron and nucleus are spinning. In quantum computing, a particular pair of the lowest energy states represents 0 and 1.
If the electron is in its 0 state, laser light of a specific frequency can boost it to a very high energy Rydberg state, tugging the electron away from the nucleus and swelling the atom to 1000 times its normal size. That enables it to exert an electrical effect on a second, nearby atom, shifting the energies of its ladder of states. Now, the same laser no longer has the correct frequency to boost that atom’s electron to its Rydberg state. Exciting the first atom blocks excitation of the second.
Making atoms compute
Individual atoms trapped in beams of laser light can act as a quantum computer’s qubits, which can be set to either 0, 1, or 0 and 1 at the same time. Two of the atom’s lowest energy states, which can differ in how much angular momentum the atom has, serve as 0 and 1. Atoms could make better qubits than alternatives such as ions and superconducting circuits, some scientists say, but an atom-based quantum computer has to overcome several hurdles.
This “Rydberg blockade” makes it possible to entangle two atoms that start in their 0 states. A laser can put the first atom into both 0 and 1 at the same time. Another pulse then boosts it from 0 to its Rydberg state while it also remains in 1. That atom then simultaneously blocks (in its Rydberg state) and doesn’t block (in its 1 state) the effect of other pulses designed to drive the second atom from 0, up to its Ryberg state, and down to 1. So, the second atom ends up in 0 and 1. When a final pulse returns the first atom from its Rydberg state, both atoms are 0 and, at the same time, both are 1. That is, they’re entangled.
It’s as if you’re a teenager hoping your older brother will give you a ride to a party, which he’ll do only if he’s in a good mood. If your brother is quantum mechanical and can be in both good and bad moods at once, you’ll end up both at the party and stuck at home, and in a good mood and a bad mood simultaneously. What’s more, your mood, good or bad, will be the same as your brother’s. Your moods will be entangled.
Called a Rydberg gate, this entangling operation enables all of atom-based quantum computing, Zoller says. “The Rydberg gate is the transistor of this sort of computer.” Making the scheme work took years. In 2010, Mark Saffman, a physicist at the University of Wisconsin–Madison, and colleagues used it to entangle two atoms. But the operation successfully entangled the qubits just 58% of the time. That “fidelity” was low because of laser noise, Saffman says.
Eight years later, Antoine Browaeys, a physicist at the Institute of Optics of France’s national research agency, and colleagues conquered the noise problems. By 2019, Lukin and colleagues had streamlined the scheme so that it required just two laser pulses. “The improvement was spectacular,” says Browaeys, a co-founder of the startup Pasqal. “The fidelity went from 80% to 95% or 98% almost immediately.” Now, several groups have demonstrated two-qubit gates with fidelities exceeding 99.5%, rivaling those of superconducting qubits.
The Ryberg gate only works if two atoms are within a few micrometers of each other. So to exploit it, researchers had to overcome another major challenge, Zoller says: They had to learn how to move atoms. Those skills are now on full display in Lukin’s lab, where lasers shuttle atoms in the vacuum chamber among separate zones for storing the qubits, making them interact, and reading out their final states, 0 or 1, with a laser. Curiously, to perform such legerdemain, physicists rely on off-the-shelf technology.
Standing before the optical table in Lukin’s lab, physicist Alexandra Geim explains the two key pieces of tech. The grid of laser beams that traps the atoms comes from a so-called spatial light modulator (SLM), she says, pointing toward a machined aluminum box lurking amid the chaos. An SLM is essentially a programmable mirror made of liquid crystal, and the devices are ordinarily used in digital projectors. Here an SLM splits one laser beam into an array of many.
To steer the beams and move the atoms trapped in them, physicists rely on a device called an acousto-optical deflector, another aluminum box. It consists of a crystal that vibrates with sound when exposed to radio waves. The sound waves can then redirect a laser beam through a process called diffraction, enabling researchers to move atoms in microseconds. “You put in one laser beam and deflect it into a bunch of different qubits,” Geim says. “It’s very enabling for our parallel control.”
THAT ABILITY TO MOVE atoms could help solve a central problem in quantum computing: error correction. Compared with conventional bits, qubits are easily perturbed and much harder to monitor. To protect the information in an ordinary bit, it can be copied on to others. Comparing the copies reveals which, if any, have flipped. That won’t work with qubits, because it’s impossible to copy the unknown state of one qubit on to another, and measurement crushes two-way states anyway.
Instead, researchers expand the state of one qubit to many through entanglement. They keep watch over these “data” qubits by interleaving them with extra, “ancillary” qubits so that the state of an ancilla reveals whether the neighboring data qubits are in the same or different states. There’s no need to measure the data qubits themselves, and repeatedly measuring the ancillas helps the diffuse “logical” qubit—the data and ancillary qubits combined—hold its state longer any of the individual physical qubits can.
The scheme works, as Google researchers reported in December 2024. They used a chip with 107 superconducting qubits and encoded a single logical qubit onto a subset of them, using an error-correction scheme called the surface code. As the logical qubit grew from nine to 25 to 49 data qubits, its error rate—the rate at which the state of the logical qubit became unclear—fell by half at each step. The largest logical qubit held its state for 291 microseconds, 2.4 times as long as any of the physical qubits.
But the achievement also highlights a looming challenge for the superconducting technology. A full-fledged computer would manipulate logical qubits encoded in as many as 1000 physical qubits. If those physical qubits are fixed in place on a chip, then making two neighboring logical qubits interact requires shuffling information between the adjacent grids. Known as lattice surgery, that maneuver would require thousands of operations among pairs of physical qubits.
Atom qubits can avoid that messy surgery. Suspended in light beams, atoms can be moved around and rearranged at will, unlike ions in a trap or superconducting circuits. So physicists can encode one logical qubit on a grid of atoms, stack it on top of a second grid representing another logical qubit, and apply the necessary laser pulses to make the two logical qubits interact. Afterward, they can move the logical qubits apart again.
Lukin’s group has done just that. In early 2024, the researchers showed they could, for example, encode two logical qubits in surface codes, bring them together to entangle them, and separate them again. In movies the Harvard team makes of their computations, the orderly movements of atoms resemble an antique loom weaving its magic.
Like the Google team, Lukin’s group showed that the error rate fell as the number of physical data qubits in a logical qubit rose. Still, the team hasn’t quite yet matched Google’s feat, Saffman says. They measured their ancillas only once per trial, which isn’t enough to constantly correct for errors, he says. “They’re doing more error detection than error correction.”
Lukin says his team is working on repeated readout. He maintains his team’s work marks a paradigm shift, from just demonstrating a logical qubit to making such things interact. “We are going away from thinking of individual physical qubits to thinking of logical qubits.”
INTEREST IN atom-based quantum computing has exploded. Globally, close to 200 academic groups are working in it, Browaeys estimates. At least a half-dozen companies are trying to market atom-based systems. Most prominent among them is QuEra, the startup Lukin and colleagues founded in 2018, which in February received a $230 million investment from Google and others. It occupies a quiet, secured, mint green building overlooking the Charles River, 3 kilometers southwest of Harvard. “In 2020 this was like an empty room,” says Alex Lukin, a physicist at the company and Mikhail Lukin’s nephew, standing beside one of QuEra’s machines.
QuEra’s machines are much sleeker than the homey setup at Harvard, their custom electronics arranged in neat racks, their optical tables more sparse. In May, the company installed a 256-qubit rig at Japan’s National Institute of Advanced Industrial Science and Technology. The $41 million machine will reside next to a conventional supercomputer at the University of Tokyo, as part of a budding high-performance computing ecosystem. Japan’s interest “goes beyond just use of the machine,” says Takuya Kitagawa, president of QuEra. “They want to grow the user base in the country and get involved with the broader supply chain.”
Similarly, Planqc, a startup spun out of the Max Planck Institute of Quantum Optics (MPQ), is building an atom-based quantum computer for the German Aerospace Center (DLR), a government agency. Sebastian Blatt, a physicist at MPQ and a co-founder of Planqc, says DLR is planning to use the machine for a variety of problems including developing new materials, simulating chemical reactions, and optimization—picking the best solution among many viable ones.
Other companies are trying to take advantage of the variety in nature’s atomic toolbox. Whereas most groups use rubidium or some other alkali metal, whose atoms have a lone outer-shell electron, scientists at the startup Atom Computing are using ytterbium, which has two. The result is a richer palette of quantum states that in principle provides greater control over the atoms, says Ben Bloom, Atom’s co-founder and chief technology officer. “This allows you to do things that you just can’t do with alkali systems.”
Scientists at Infleqtion are working to incorporate two different types of atoms into a single quantum computer. In a logical qubit, one kind of atoms might serve as the data qubits and the other might serve as ancillas, says Saffman, Infleqtion’s chief scientist for quantum information. The two types of atoms would react to different frequencies of light, so the ancillas could be read out without having to first move them away from the data qubits, he explains.
Kenji Ohmori, a physicist with Japan’s National Institutes of Natural Science and founder of the startup Yaqumo, hopes to replace the Rydberg gate with something faster. With laser pulses less than 1 nanosecond long, he and his colleagues force two rubidium atoms into the Rydberg state simultaneously, before the blockade can kick in. The atoms then interact so strongly they become entangled in just 6.5 nanoseconds. “We are almost comparable to the superconductor qubits or even faster,” Ohmori says. But so far the gate’s fidelity is just 71%.
Some experts say atom-based quantum computing has shown more promise than progress. “It still looks a little like a physics experiment,” says Christopher Monroe, an expert on ion quantum computing at Duke University and co-founder of IonQ. “I’m hoping these neutral atom companies will show us an honest to God system that executes an algorithm.” That’s exactly what developers of atom-based systems are striving to do, Blatt says. “The next big 5-year projects worldwide, they’re going to center around putting all these pieces together.”
We will have to wait and see whether this becomes the “big” breakthrough in the development of quantum computing.
Tony
Scott Olson/Getty Images
