Physicists in the United States have demonstrated how metasurfaces can be used to create arrays of optical tweezers that could potentially trap hundreds of thousands of neutral atoms (Nature, doi: 10.1038/s41586-025-09961-5). They reckon that the research points the way to quantum computers large enough to carry out the kind of error correction needed to perform useful calculations.

Optical tweezers from metasurfaces

Optical tweezers are tightly focused laser beams that can trap and move tiny particles, including atoms. Thousands of such beams can be generated to form an array by directing the light from a single laser through devices such as spatial light modulators or acousto-optical deflectors. However, the need to combine these devices with optics that have high numerical apertures and other complex components limits their trapping capability to about 10,000 atoms.

In the latest work, the teams of Sebastian Will and Nanfang Yu, led by graduate students Aaron Holman and Yuan Xu, at Columbia University instead create tweezer arrays using metasurfaces. These artificially fabricated materials consist of vast numbers of subwavelength refractive structures laid out in a two-dimensional grid that modifies the phase front of incoming laser beams upon transmission.

The fact that the individual structures on metasurfaces are smaller than the wavelength of radiation they manipulate confers a couple of major advantages compared with the larger feature sizes, or “pixels,” of more traditional light-shaping devices. For one thing, the much higher numerical apertures that they yield enable tighter focusing—so much so that they can trap atoms directly rather than having to rely on additional optics for demagnification. What’s more, a higher density of pixels implies more traps of a given quality from a certain-sized device.

Trapping atoms

The researchers made the metasurfaces from arrays of tiny dielectric pillars, each 750 nm high and less than 200 nm wide. They did so using two different materials: silicon-rich silicon nitride, which is compatible with CMOS manufacturing, and titanium dioxide, which can handle higher powers and shorter wavelengths than its silicon counterpart.

By placing multiple metasurfaces on a single substrate that they moved using a twin-axis translation stage, the researchers were able to test out different sets of tweezer arrays. Exposing the surfaces to the light from a green laser (with a wavelength of 520 nm), they directed the manipulated beam through a series of lenses to a vacuum chamber containing an ultracold cloud of strontium-88 atoms—and in the process trapping those atoms in the desired tweezer array.

They detected the trapped atoms by fluorescence, confirming that they had created the arrays encoded in the metasurfaces. The arrays included, one with 183 atom traps in the shape of the Statue of Liberty, another that featured a little over 1,000 traps in a square lattice, and one in a circular pattern with 16 traps spaced just a fraction over 1 µm apart.

In addition to a more detailed study of their trapping arrays, which found their depth and spacing to be as uniform as those of today’s best arrays, the scientists also showed that their metasurface-based technology should be able to trap far greater numbers of atoms in the future. Specifically, they fabricated a square lattice with 600 tweezer traps on a side that they spaced 2.5 µm apart—yielding a total of 360,000 traps (although atom-less for the time being).

By placing multiple metasurfaces on a single substrate that they moved using a twin-axis translation stage, the researchers were able to test out different sets of tweezer arrays.

Future directions and applications

The researchers argue that such scalability suggests the technology could be ideal for building quantum computers from neutral atoms. The ability to generate quantum bits in superpositions of two states at the same time implies the possibility of massive parallel processing. This means such computers could in principle execute programs using just hundreds of logical qubits that classical computers with billions of bits would find impossible to do in any reasonable amount of time. However, the need to correct errors resulting from the fragility of quantum states requires each such logical qubit to be spread over hundreds of physical qubits, hence the need for large arrays of these entities.

The researchers explain that neutral atoms have the advantage over other types of qubit, such as those made from loops of superconductor or trapped ions, of being extremely plentiful and also identical to one another. One major weak point, however, has been the difficulty of controlling them on large scales.

Last year, scientists at the California Institute of Technology, USA, used spatial light modulators to create a tweezer array holding 6,100 atoms and showed they could maintain the atoms in a superposition state for several seconds even while moving them around. Although they didn’t use the atoms to perform calculations, the researchers plan to soon carry out the key step needed for such computation: entanglement.

The Columbia group has not yet placed the atoms in its arrays in a superposition, but Will and his colleagues say this should be possible given the high uniformity of their tweezer arrays. As for filling the kind of much larger array that they demonstrated, they say this is a question of laser power. Having had only about 1 W available for the recent experiments, they were limited to no more than around 1,000 trapped atoms. Moving to hundreds of thousands of atoms will instead require powers of around 100 W, they say.