Quantum computers and simulators are of enormous interest because of their potential to shed light on mysteries of physics that are difficult to model using conventional computers. Some physical platforms used in realizing quantum-computing protocols — including trapped ions and several solid-state systems based on superconductors — have received increased attention in the past year. But in this issue, two groups report technical breakthroughs that will aid the development of another platform: trapped neutral atoms. Barredo et al.1 report their use of precision optical-engineering methods to sort atoms into arbitrary 3D patterns, whereas Kumar et al.2construct cubic lattices by revisiting a fanciful thought experiment known as Maxwell’s demon. The ability to organize neutral atoms exactly into planned 3D arrays will be valuable for the development of neutral-atom quantum computers that use a large number of quantum bits (qubits).
Arrays of isolated neutral atoms have long shown promise for quantum computing because neutral-atom qubits are extremely well isolated from environmental noise and are highly controllable, and also because such systems can be scaled up to large numbers of qubits3,4. Given that controlled interactions between atoms are needed to perform quantum-computing operations, neutral-atom quantum computers will need qubits to be precisely arranged in a specified pattern. However, developing methods for sorting atoms into patterns has proved challenging. Neutral-atom qubits require ultracold temperatures and extremely high vacuums to function, and therefore require complicated apparatus; ordering them into arrays using optical techniques adds an extra level of practical complexity. Progress has been made in arranging neutral atoms in one and two dimensions5–8, but 3D stacking will become essential as the number of qubits used approaches the hundreds, or to construct arrangements that have topologies not achievable in two dimensions.
Kumar et al. have extended their previously reported approach9 to assemble cold clouds of caesium atoms into a 3D lattice. The method begins with a randomly populated optical lattice: a trap formed from the interference patterns of counter-propagating lasers, in which atoms can be confined much like eggs in cartons. After imaging and recording the random locations of atoms in the lattice, the authors implement a sorting protocol that involves intricately controlling the polarizations of the lattice lasers, while using additional ‘addressing’ lasers and microwaves to position any given atom within a 5 × 5 × 5 array of lattice sites (Fig. 1). In this way, up to 50 neutral atoms can be precisely ordered into an array that is suitable for use in a quantum computer.
Kumar et al. frame their sorting and preparation protocol in terms of Maxwell’s demon. This thought experiment was proposed by James Clerk Maxwell in 1867, and explores the nature of entropy, a measure of disorder. Maxwell postulated that a reversible sorting mechanism (a sentient demon, although a non-sentient process would also work) could partition gas molecules into two sub-volumes. But this sorting process would lower the entropy of the gas in apparent violation of the second law of thermodynamics, which states that the entropy of any isolated system can only increase. How can this conundrum be explained? The answer turns out to be that the act of sorting inevitably increases the entropy of the Universe. Because the dominant entropy in Kumar and colleagues’ experiments is associated with the physical arrangement of the atoms, their work is a realization of an omniscient Maxwell’s demon, summoned to organize the initial arrangement of a qubit array.
Meanwhile, Barredo et al. extend their previously reported method10 for 2D atom sorting to three dimensions. Their approach to disorder and sorting is different from Kumar and colleagues’ method, but just as effective. They use a holographic technique whereby a laser beam is reflected off a spatial light modulator and then focused to form traps known as optical tweezers. In this way, they generate arrays of traps in arbitrary configurations that can be loaded with up to 72 cold rubidium atoms. To remove disorder and build the desired atomic configuration, the authors use a separate, movable optical tweezer to pluck atoms from ‘wrong’ traps and either move them to correct sites or discard them (Fig. 2). This allows them to build qubit arrays in standard grid patterns, in topologies such as a Möbius strip, and even in the shape of the Eiffel Tower (see Fig. 2 of the paper1).
Barredo and colleagues go on to engineer an interaction between two qubits in a sorted array. To do this, they excite the atoms into ‘Rydberg’ states, which produce atomic electrical dipoles that allow the qubits to sense each other through dipole–dipole interactions. By contrast, atoms in their ground states have vanishingly small dipole–dipole interactions. Rydberg interactions have previously been used to enable quantum-logic operations carried out by small systems of neutral-atom qubits3,4, and could form the basis of both the current groups’ future efforts to develop quantum computers.
The two papers report similar milestones for the assembly of neutral-atom quantum computers, with Barredo and colleagues also reporting a working two-qubit interaction. However, the atoms in Barredo and colleagues’ system are not as cold as they could be, which means that the entropy remaining in their arrays is substantially greater than in Kumar and colleagues’ system. The resulting micrometre-scale motion of the atoms within the traps could limit the performance of future devices based on this system — a restriction that does not apply to Kumar and co-workers’ apparatus. But Barredo and colleagues’ approach does allow qubit arrays of any spatial design to be made, whereas Kumar and co-workers’ apparatus generates only a cubic lattice. These differences might not be important for near-term quantum-computing goals, however. It remains to be seen whether quantum entanglement (a phenomenon that produces stronger correlations between particles than those permitted by classical physics, and which fuels quantum-computing algorithms) can be created for such large numbers of working qubits.
Both papers report technical tours de force, and showcase how far neutral-atom systems have come in terms of stability, reproducibility and technical sophistication. The next step is probably to generate quantum entanglement between arbitrary pairs of atoms in sorted arrays. It will also be interesting to see which exotic quantum states can be simulated using these qubit arrays, especially if some of those states cannot be modelled using existing computational approaches11. Finally, it will be exciting to see whether the potential advantages of neutral atoms will now begin to pay dividends in the race to develop a working quantum computer.
Nature 561, 43-44 (2018)