The Signal
Princeton Team Uses Rydberg Blockade to Speed Ytterbium Loading
A Princeton-led team reports a Rydberg-blockade method that prepares single ytterbium atoms in optical tweezers in tens of microseconds. The faster protocol matches conventional loading’s filling fraction, while a metastable-state version r
Quantum.world Team · Written with AI · fact-checked ·
Conventional light-assisted collisions take milliseconds to reduce a randomly loaded optical tweezer to one atom; a Princeton-led team reports doing that in 64.8 microseconds. But the speed advantage comes with a tradeoff: the fastest version retains single atoms at a rate comparable to the conventional method, while a higher-filling version requires extra preparation time and still loses atom pairs.
The results, published in PRX Quantum on October 28, 2026, offer a new way to clear excess atoms from ytterbium tweezers. The study’s central tension is practical, not just numerical: shorten the loading step, or accept additional overhead to keep more single atoms.
Why atom loading takes milliseconds
Optical tweezer arrays begin with atoms loaded stochastically, so an individual trap can contain none, one or several atoms. Quantum operations require controlled occupancy, and conventional light-assisted collisions remove excess atoms through slow collisional dynamics. The paper puts that timescale at several milliseconds.
The new method uses interactions between highly excited Rydberg atoms to make a different selection. Within a multiply occupied tweezer, Rydberg blockade lets the researchers target excess atoms for removal one at a time; autoionization then ejects the excited atom. Repeating the excitation-and-removal cycle reduces multiple occupancy without relying on a single collision event to clear the trap.
64.8 microseconds reaches 1% multiple occupancy
In the first implementation, the team drives a two-photon Rydberg transition from ytterbium’s ground state. After 18 cycles, the probability of finding more than one atom falls to 1%, while 58.2% of tweezers retain a single atom. The authors report that this filling fraction is comparable to light-assisted collisions under the same conditions, achieved more than 100 times faster.
The gain is not cost-free: the paper identifies single-atom loss during repeated excitation as a limit on the final filling fraction. In the two-photon approach, that loss is linked to decoherence from intermediate-state scattering and the finite lifetime of the Rydberg state. The result is a faster route to a similar occupancy level, rather than a higher one.
A higher filling fraction takes extra preparation
A second route starts by transferring atoms into the metastable state ³P₀, then uses single-photon Rydberg excitation. Lower single-atom loss raises the reported filling fraction to 74.8(3)%. The improvement comes with additional time to cool and prepare atoms in ³P₀, so the authors distinguish the short excitation sequence from the full preparation overhead.
That version also exposes a different bottleneck. The team reports an unexplained two-body loss mechanism that limits the final filling fraction; it can remove both atoms rather than just the excess one. Yiyi Li and colleagues write: “The final filling fraction is limited by an unexplained two-body loss mechanism, which, if solved, could enable fast, quasi-deterministic loading.” Yiyi Li et al., Princeton University and California Institute of Technology
The measured tradeoff sets the next test
The demonstrations use ¹⁷¹Yb and rely on Rydberg excitation plus a fast way to remove Rydberg-state atoms. The researchers say the method needs no additional hardware or control beyond what is already used for high-fidelity two-qubit gates, and describe other removal techniques as possible alternatives to autoionization.
For now, the clearest benchmark is the contrast between the two operating regimes: 58.2(2)% single-atom filling in 64.8 microseconds with two-photon excitation, and 74.8(3)% with the metastable-state route, whose preparation adds time. The unresolved two-body loss remains the reported ceiling on that higher figure.