41. Neutral Atoms
41.1Optical tweezers
A tightly focused laser beam, detuned from atomic resonance, polarizes a neutral atom and traps it at the focus. Paint an array of such foci with spatial light modulators or acousto-optic deflectors and you hold hundreds to thousands of single atoms (rubidium, cesium, strontium) in 2D patterns — arrays of 6,100 atoms were demonstrated in 2025. Loading is probabilistic per site (~50–60%), so rearrangement algorithms — grab atoms with movable tweezers and fill the holes — are a signature, software-solvable feature of the platform. The geometry of the array is literally programmable per experiment, which no fixed-lattice platform can offer.
41.2Ultracold atoms
Atoms must be cold to stay trapped: laser cooling plus evaporative cooling brings them to microkelvin temperatures inside an ultra-high-vacuum cell. Being identical atoms, all sites are equivalent — no device-to-device fabrication variation, the same advantage ions have. Qubit states are hyperfine ground states or clock states, with coherence of seconds. Infrastructure is a real differentiator: apart from photonics hardware, the machine is room temperature and tabletop-scale — no dilution refrigerator in the physics. That changes who can build and host these machines, and it is one reason neutral-atom systems scaled in atom count faster than any other platform.
41.3Rydberg excitation
Excite the valence electron to a state with principal quantum number n ≈ 70: a huge, diffuse electron orbit carrying an enormous electric dipole moment. Rydberg atoms interact strongly at micrometer distances — exactly the tweezer spacing — through dipole–dipole coupling, and the interaction strength grows steeply with n. Two-photon laser pulses drive |1⟩ → |r⟩ in tens of nanoseconds. The Rydberg state decays back in ~100 µs and is sensitive to stray electric fields, which sets the platform's noise floor and calibration burden. This single interaction is the transistor of neutral-atom quantum computing: everything else in 41.4–41.6 builds on it.
41.4Rydberg blockade
If atom A is excited to |r⟩, its electric field shifts the neighboring atoms' Rydberg level by more than the excitation laser's linewidth: they can no longer be excited — one excitation per blockade radius of ~5–10 µm. This is a strong nonlinearity assembled from atoms, and it implements controlled-phase gates directly. It also powers the platform's analog mode: instead of digital gates, sweep the laser slowly and let the blockade geometry adiabatically prepare entangled many-body states — the mode in which QuEra's 2023–24 experiments with hundreds of atoms produced sampling results that challenged classical simulation. Digital and analog modes share hardware but demand different compilers.
41.5Gate operations
Digital mode: Rydberg-mediated CZ between neighboring atoms in 100 ns–1 µs, with reported two-qubit fidelities around 99.5% (best ~99.8% in specialized geometries); single-qubit gates ~100 ns via microwave or Raman pulses. A unique option exists on no other platform: atoms can be moved during the computation, so "connectivity" includes transport — gate non-adjacent atoms by moving them together first. Mid-circuit measurement and single-atom reloading have been demonstrated, the primitives fault tolerance needs. The honest summary: speed is competitive with superconductors for some gates, and the gap to close versus ions is fidelity, not scale.
41.6Arrays
Connectivity is defined in software, per run: triangular, square, Kagome, or multi-layer geometries, plus storage and entangling zones. This is architecturally unlike fixed-lattice superconductors — quantum error-correcting codes that need nonlocal connections (certain LDPC codes and surface-code variants) map naturally onto atom arrays, and the geometry can change between circuit executions. The compiler's job changes accordingly: scheduling atom moves becomes part of compilation (Part XII). QuEra and Pasqal expose such arrays through cloud access, and the 2023 Harvard/QuEra work ran error-corrected logical qubits on 200+ atom arrays — the first logical-qubit demonstrations on this platform, three years after Google's.
41.7Scaling
The headline is atom count: thousands of physical qubits already exist, scaling by optically engineering larger arrays — there is no wiring wall like 39.5. The debts are equally concrete: two-qubit fidelity must roughly double its nines to reach ion class; atoms are lost during operation and must be detected and replaced mid-circuit, an unsolved systems problem at speed; laser, imaging, and vacuum systems grow complex with array size; and readout crosstalk across dense arrays needs management. Both landmark events of 2023–24 — analog advantage-class sampling (41.4) and logical-qubit demonstrations (41.6) — happened on this platform first. Its ceiling is genuinely unknown, which cuts both ways.