Dual-Element, Two-Dimensional Atom Array with Continuous-Mode Operation

Quantum processing architectures that include multiple qubit modalities offer compelling strategies for high-fidelity operations and readout, quantum error correction, and a path for scaling to large system sizes. Such hybrid architectures have been realized for leading platforms, including superconducting circuits and trapped ions. Recently, a new approach for constructing large, coherent quantum processors has emerged based on arrays of individually trapped neutral atoms. However, these demonstrations have been limited to arrays of a single atomic element where the identical nature of the atoms makes crosstalk-free control and nondemolition readout of a large number of atomic qubits challenging. Here we introduce a dual-element atom array with individual control of single rubidium and cesium atoms. We demonstrate their independent placement in arrays with up to 512 trapping sites and observe negligible crosstalk between the two elements. Furthermore, by continuously reloading one atomic element while maintaining an array of the other, we demonstrate a new continuous operation mode for atom arrays without any off-time. Our results enable avenues for auxiliary-qubit-assisted quantum protocols such as quantum nondemolition measurements and quantum error correction, as well as continuously operating quantum processors and sensors.

Popular Summary

Individual atoms trapped in optical tweezers are a promising platform for building large-scale programmable quantum devices. However, implementations of this architecture have been restricted to arrays composed of a singular atomic element, which suffer from limited control techniques and destructive losses during measurement, since all atomic qubits are in resonance with one another. To address this fundamental challenge and open new avenues for quantum information processing with cold atoms, we introduce for the first time an atom array technology with two atomic elements: rubidium and cesium. In such an architecture, one element can act as the “helping buddy” for the other element, enabling nondestructive measurements and new ways of state preparation and control.

The large frequency separation between the atomic resonances of the two elements allows us to independently cool, trap, image, and position each atom. Leveraging this control, we place the atoms in arbitrary 2D array geometries with up to 512 trapping sites and observe negligible crosstalk between the elements. This enables us to operate the array indefinitely by continuously reloading one atomic element while maintaining an array of the other, a feature not available in single-species arrays.

Our dual-element atom array paves the way for “buddy-assisted” protocols, where only the atoms of one element need to be measured at any time. We can treat one element as “computational” and the other as a helping buddy and transfer information between them like a quantum baton pass. By employing this buddy system and only imaging the helper qubits, we can circumvent measurement losses. Such a system facilitates nondestructive quantum measurements and error-correction schemes with hundreds of atomic qubits.

Figure 1

Trapping and imaging two-dimensional, dual-element arrays of neutral atoms. Optical tweezers are generated using crossed acousto-optic deflectors (AODs) and a spatial light modulator (SLM) at laser wavelengths of 811 nm (red) and 910 nm (dark red), respectively. An optional spatial filter in the AOD path allows us to mask traps selectively and generate desired geometries. The AOD and SLM trapping arrays are then combined via a polarizing beam splitter (PBS). These combined traps propagate along a shared beam path and are focused by a glass-corrected high-numerical-aperture microscope objective with $\mathrm{NA}=0.65$ into a vacuum chamber, thereby creating arbitrary array geometries (shown is a surface-code-inspired geometry with two interleaved arrays). An identical objective images the traps onto a charge-coupled-device (CCD) camera to enable feedback-based intensity homogenization. Atomic fluorescence at 780 nm (blue) for Rb and 852 nm (yellow) for Cs is collected using the first objective and reflected along a shared beam path toward an electron-multiplying CCD (EMCCD) camera using a custom dichroic. The signal to background is improved by separating the fluorescence wavelengths before the EMCCD and performing individual spatial filtering.

Figure 2

A dual-element 512-site atom array. Averaged and single-shot single-atom resolved fluorescence images of the dual-element array with Cs counts in yellow and Rb counts in blue. (a) Averaged fluorescence image of the dual-element array. Scale bar indicates a distance of $20\mu \mathrm{m}$. (b) Single-shot image of the dual-element array. (c) Averaged and single-shot images of the $17\times 16$ Cs array. (d) Averaged and single-shot images of the $16\times 15$ Rb array. See Appendixes pp1 and pp2 for detailed imaging sequence, parameters, and fluorescence histograms.

Figure 3

Homogeneity and loading statistics for the Rb and Cs arrays. (a) Stark shifts across a $17\times 16$ Cs array before (gray) and after (yellow) trap intensity correction via the weighted Gerchberg-Saxton algorithm based on feedback from the atoms. Inset provides an example Stark shift measurement in which the frequency of a pushout beam is swept to determine the trap-induced ac Stark shift. (b) Stark shifts across a $16\times 15$ Rb array before (gray) and after (blue) trap intensity correction via the optimization of the rf tones driving the AODs based on feedback from the atoms. Inset provides an example Stark shift measurement. (c) Loading statistics for the Cs array with and without the presence of the Rb MOT and Rb tweezers. Dashed line indicates the average loading efficiency. Reported loading efficiency errors are the standard deviations of their respective distributions. (d) Loading statistics for the Rb array with and without the presence of the Cs MOT and Cs tweezers. Dashed line indicates the average loading efficiency.

Figure 4

A continuous-mode atom array. (a) The pulse sequence used to reload the Rb atoms and Cs atoms into the atom array. Rb (Cs) atoms are reloaded into the array while the Cs (Rb) atoms are held in their optical tweezers. The shaded regions, yellow for Cs and blue for Rb, indicate the atomic element available for manipulation or computation during the specified time window. By performing the pulse sequence repeatedly, a continuously available atomic array can be maintained. (b) The number of Rb (blue) and Cs (yellow) atoms in each image from a 50-min data run. The dashed lines indicate the average atom counts. The number of atoms available for manipulation as a function of time (green) indicates that the atom array continuously operates with over 115 atoms (red solid line) at any moment in time.

Figure 5

Arbitrary geometries with dual-element arrays. Averaged fluorescence images for arbitrary array shapes with Cs in SLM trap sites (yellow) and Rb in AOD trap sites (blue). Scale bars indicate $10\mu \mathrm{m}$. (a) A Rb-dressed, Cs hexagonal lattice. (b) A bipartite honeycomb lattice. (c) Chicago landmarks: the Sears Tower and The Bean (Cloud Gate).

Figure 6

Experimental sequence. Diagram of the experimental sequence. Cs (Rb) 2D MOT refers to the state (on or off) of the Cs (Rb) 2D MOT cooling beams. We first load a dual-element MOT for approximately 250 ms. After MOT formation, the 2D MOT beams are turned off, and the 3D MOT beam powers and detunings are ramped to perform polarization-gradient cooling. After a short wait time to allow atoms not trapped in the tweezers to disperse, a Cs blowout pulse is performed to remove Cs atoms trapped in weak out-of-plane SLM traps. Two sets of images are performed to measure atom statistics. The dual-wavelength optical tweezer array at 811 and 910 nm remains on through this entire sequence.

Figure 7

Example atomic fluorescence histograms. Sample fluorescence counts for a single Rb atom (top) and a single Cs atom (bottom) collected during a 40-ms imaging time. Thresholds (dashed lines) are placed between the atom and background signals by fitting the fluorescence counts to a bimodal Poisson distribution and extracting the minimum between the modes.

Figure 8

Atom lifetimes in optical tweezers. Typical atom lifetimes for Rb and Cs atoms under continuous laser cooling with the PGC light while trapped in the 811 and 910 tweezers, respectively. Solid lines indicate exponential fits to the data. Error bars are smaller than the size of the markers.