In the past five years, neutral atoms have emerged as dark horse candidates in the race to build a quantum computer, a machine that would exploit the laws of quantum physics to solve certain important computational problems far more efficiently than any conventional computer.
In a neutral-atom quantum processor, atoms are suspended in ultrahigh vacuum by arrays of tightly focused laser beams called optical tweezers. Researchers have scaled up to arrays of more than 100 alkali atoms, each of which has one valence electron, and executed quantum algorithms using smaller arrays. Now they’re exploring new quantum information-processing and measurement capabilities in arrays of atoms with two valence electrons, including the alkaline-earth atoms in the second column of the periodic table and a few others with similar properties, such as ytterbium.
Tweezer arrays of alkaline-earth atoms have shown promise in both quantum computing and precision timekeeping—they can encode new kinds of qubits with long coherence times and serve as state-of-the-art atomic clocks. In the future, they may help researchers implement protocols for fault-tolerant quantum error correction and harness quantum entanglement on a large scale to further push the limits of atomic clock performance.
How atoms become qubits
In an optical tweezer trap, the electric field of the focused laser beam induces a small polarization in an atom, pulling it toward the region of maximum intensity at the center of the trap. Researchers can then drag the atom around as they please or hold it in place and zap it with other lasers or microwave pulses to excite specific atomic transitions.
To create a neutral-atom quantum processor, researchers generate an array of optical tweezers by splitting an incident laser beam into many beams that are focused into a glass vacuum cell by a powerful microscope objective lens. Clouds of cold atoms are loaded into the tweezer array and then shuffled to produce a filled array with a single atom in each tweezer. Optical devices outside the vacuum cell enable researchers to control the precise position of each atom, rapidly reconfigure the array, and tailor the trapping potential of each tweezer independently.
Each atomic species has infinitely many discrete energy levels associated with distinct quantum states. Any pair of states can in principle serve as a qubit, the basic building block of a quantum computer. In practice, researchers select a pair of long-lived low-energy states that enable many consecutive quantum logic operations to be performed before quantum information leaks from the qubit into its environment in a process known as decoherence.
Neutral atoms in low-energy states interact weakly with each other and thus can be packed into compact arrays. Proponents see that as a key advantage of neutral-atom quantum computing relative to more mature approaches that use ions trapped in electric fields or superconducting circuits at millikelvin temperatures. Scaling up quantum computers from hundreds of qubits to millions is a challenge for all proposed architectures, but space is not an issue for neutral atoms; a millimeter-scale array could hold as many as a million qubits.
To turn on interactions between qubits, researchers target a pair of adjacent atoms with a laser pulse that excites one of them to a high-energy state called a Rydberg state, in which a valence electron orbits far from the nucleus. The Rydberg atom’s strong electric dipole interactions prevent the laser from also exciting its neighbor, an effect known as a Rydberg blockade, but it’s impossible to know which of the atoms was excited. The result is a single excitation shared between two qubits that can’t be described separately—the characteristic feature of entanglement, the key phenomenon that allows quantum computers to outperform their classical counterparts. Rydberg entanglement fidelity has improved markedly in the past five years but still lags behind that of trapped ions and superconducting qubits.
In April two independent groups, one led by Mikhail Lukin at Harvard University and the other by Mark Saffman at the University of Wisconsin–Madison, reported the first demonstrations of multistep quantum algorithms in arrays of rubidium atoms. Rubidium is an alkali atom that’s long been a workhorse of atomic physics, in part because its single valence electron gives it a simple energy-level structure akin to that of hydrogen. Rubidium-based quantum processors encode qubits in the atoms’ hyperfine states, closely spaced energy levels arising from the interaction of the valence electron’s spin with that of the nucleus. Hyperfine qubits have longer coherence times than qubits encoded in an atom’s electronic transitions, but the unpaired electron spin still makes them susceptible to decoherence from stray magnetic fields and residual interactions with the tweezer light.
Those sources of decoherence have spurred research into quantum computing with tweezer arrays of alkaline-earth atoms, which have a more complex energy-level structure owing to their two valence electrons. With the extra complexity comes new technical challenges but also novel ways to encode and manipulate quantum information. “It’s the natural progression of the field toward more complex atoms as techniques and laser technology have matured,” says Saffman.
Harnessing alkaline earths
One advantage of alkaline-earth atoms is that their valence electrons pair off in the electronic ground state, so there’s no hyperfine interaction. Rather, in alkaline-earth atoms whose nuclei have nonzero spin, the nuclear spin is isolated from environmental disturbances that naturally couple to electron spins. Qubits encoded in the nuclear spin states of alkaline-earth atoms are thus expected to have much longer coherence times than hyperfine qubits.
In May the Berkeley-based startup Atom Computing published a paper describing its first-generation quantum processor, a tweezer array in which qubits are encoded in two nuclear spin levels of strontium-87 atoms. The Atom Computing team reported a coherence time (denoted T2*) longer than 20 seconds, more than three orders of magnitude larger than typical values for alkali hyperfine qubits.
That same month, groups led by Jeff Thompson at Princeton University and Adam Kaufman at the University of Colorado Boulder reported that they’d achieved control over long-lived nuclear spin qubits in arrays of ytterbium-171 atoms. Notably, 171Yb has a simpler nuclear spin structure than any other stable isotope of an alkaline-earth-like atom. It’s a natural two-level system, whereas 87Sr’s 10 distinct nuclear spin states necessitate additional lasers to single out two levels that can encode a qubit.
The rich electronic energy-level structure of alkaline-earth-like atoms may prove useful to quantum computing in other ways. If one compares each atomic species to a Swiss Army knife with a specific set of tools, Thompson says, then alkaline-earth atoms are deluxe models with extra features. “We’ve come up with ways to use other gadgets on this Swiss Army knife that we didn’t even know about when we bought it,” he says, pointing to a technique for improving entanglement fidelity and a new quantum error correction protocol that both rely on a long-lived metastable state characteristic of alkaline-earth atoms that has no counterpart in alkali atoms.
The distinctive metastable state owes its long lifetime—typically tens of seconds and even longer in certain atoms—to the fact that decay to the ground state can proceed only via a strongly suppressed higher-order process. An atom prepared in a superposition of the metastable state and the ground state will oscillate for a long time at a frequency equal to that of the photon emitted in a transition between the two states, known as the clock transition because it provides the frequency reference for the world’s best atomic clocks.
Those clocks—called optical clocks because the clock transition frequency in widely used atoms like strontium, ytterbium, and aluminum falls in or near the visible part of the electromagnetic spectrum—mainly belong to either of two distinct classes that differ primarily in how the atoms are trapped. The tweezer arrays developed for neutral-atom quantum processors have emerged as a third optical-clock architecture with the potential to combine some of the advantages of the more mature platforms.
The two dominant optical-clock architectures have complementary strengths and weaknesses. The most accurate clocks use electric fields to trap single aluminum ions that are extremely well isolated from their environments, but they require long averaging times to achieve high precision. The most precise clocks, called optical lattice clocks, use laser light to trap neutral strontium atoms, but in place of the tightly focused beams used in tweezer arrays they use periodic potentials formed by the interference of two counterpropagating beams to trap as many as 100 000 atoms. Such large ensembles enable lattice clocks to accumulate precision rapidly, but systematic effects like residual interactions between atoms trapped at the same lattice site limit their accuracy.
Tweezer-array optical clocks may help researchers get the best of both worlds, with precise single-particle control comparable to that of ion clocks and a clearer path toward scaling up to large ensembles. Kaufman, who developed the first tweezer clocks concurrently with independent work by Manuel Endres at Caltech, published a paper in 2020 describing a second-generation tweezer clock with precision approaching the lattice clock record at the time. In the future, tweezer clock precision might be further improved by using quantum information protocols to generate many-particle entangled states, which Kaufman calls “a frontier direction in quantum science.”
The rapid progress in quantum science with neutral-atom tweezer arrays over the past five years suggests that those efforts may yield further unexpected advantages. “It gives me new optimism about what we can do in quantum technology as a whole,” Thompson says.