Focusing through coupled multimode fibers for miniaturized single-atom tweezer arrays

Abstract

Real-world deployment of quantum technologies requires systems in low SWaP-C packages. Quantum platforms based on ultracold atoms typically trap and control atoms with optical tweezers using light focused through bulky objective lenses, which are costly and sensitive to misalignment. Multimode fibers (MMFs) offer a compact route to focusing light, but their use in forming optical tweezers for cold atoms has been limited to a single fiber. Here we use a 2x1 MMF fused coupler to obtain two independent control paths sharing a common distal facet. We form near-diffraction-limited tweezers through each channel, suppress interference by applying a small frequency offset between the channels, and demonstrate that the tweezer separation is insensitive to misalignment of the input coupling. This work provides a route to reconfigurable single-atom arrays in low SWaP-C environments.

The paper investigates whether a multimode-fiber (MMF) fused coupler can provide independently controlled optical foci at a shared distal facet, as a compact alternative to objective-based atom-tweezer systems. It demonstrates optical focusing and control, but does not trap or image atoms. A commercial $2\times1$ coupler combines two 50-$\mu$m-core, numerical-aperture-0.22 MMF inputs. Separate regions of a digital micromirror device shape 780-nm light entering each channel. Each channel’s transmission matrix is measured using four-step inline holography over 440 probe modes, then used to reconstruct inputs targeting selected output positions approximately $50\,\mu\mathrm{m}$ beyond the common facet. An 80-MHz frequency offset makes interchannel interference average away on anticipated atomic-motion timescales. Camera measurements give focal-intensity FWHMs around $2.0(1)\,\mu\mathrm{m}$ across the accessible field, compared with the quoted $1.8\,\mu\mathrm{m}$ diffraction limit. Measured versus programmed tweezer separation has slope $0.996(3)$. Residual intensity at the other channel’s target averages $-21$ dB, comparable to the speckle background. Deliberately misaligning one input primarily reduces focal intensity; the measured position changes by $1.9(1)$ nm per micrometre of input displacement. The demonstration uses approximately $50\,\mu\mathrm{W}$ per focus; the estimated requirement for a $100\,\mu\mathrm{K}$ potassium trap is $250\,\mu\mathrm{W}$. Larger arrays require greater total power. Single-atom trapping and imaging, atom transport, and microsecond independent switching remain prospective; the latter would require additional channel-specific acousto-optic modulators.