Neutral-Atom Qubits

A neutral-atom quantum processor stores each qubit inside a single, laser-cooled atom held at the focus of a tightly focused laser beam — an optical tweezer. The qubit lives in the electronic ground manifold, in two hyperfine "clock" states split by a microwave gap: $6.835\,\mathrm{GHz}$ in $^{87}\mathrm{Rb}$ . Because every atom of a given isotope is fundamentally identical, the qubits arrive pre-calibrated — no fabrication spread analogous to the junction disorder that makes each superconducting qubit slightly different. Nature supplies the register; the laboratory only has to hold and address it. Two features make the platform distinctive. First, the traps are movable: acousto-optic deflectors paint hundreds of tweezers at will, a camera reads which are loaded, and feedback shuffles atoms into defect-free chains, grids, or arbitrary graphs — and can transport them mid-circuit to bring any two qubits together. Connectivity is choreographed, not wired. Second, entanglement is switched on and off by the Rydberg blockade: promote an atom to high principal quantum number $n$ and it swells into a balloon of enormous, long-range van der Waals interaction scaling as $n^{11}$ , so when two nearby atoms are driven toward the same Rydberg level the interaction shoves the doubly-excited state out of resonance and at most one can go up. This conditional dynamics is exactly the resource that becomes a native controlled-phase gate. Single-qubit rotations come from resonant microwaves or a focused two-photon Raman drive; readout is state-selective fluorescence, one hyperfine state bright and the other dark, all sites imaged at once. The limits — finite Rydberg lifetime, atom loss, and laser noise — set today's gate-fidelity ceiling and motivate erasure-conversion and array-native error correction. This article assembles the physics that satisfies the DiVincenzo criteria on an optical table cooled by light.

Key takeaways

A neutral-atom qubit is one laser-cooled atom (commonly $^{87}\mathrm{Rb}$ or $^{133}\mathrm{Cs}$ ) in an optical tweezer, encoding $|0\rangle,|1\rangle$ in two long-lived hyperfine ground states — identical atom to atom, no fabrication disorder.
Hundreds of tweezers are painted by acousto-optic deflectors and rearranged atom-by-atom into defect-free patterns; atoms transport mid-circuit, giving reconfigurable, effectively all-to-all connectivity.
Single-qubit gates use microwave or two-photon Raman drives; readout is state-selective fluorescence imaging of the whole array at once.
Two-qubit entanglement uses the Rydberg blockade: exciting an atom to a high- $n$ state shifts its neighbour's resonance by the van der Waals interaction $V=C_6/R^6$ , forbidding double excitation and yielding a native controlled-phase gate.
Typical numbers: hyperfine splitting a few GHz, spacing $\sim3$ – $5\,\mu\mathrm{m}$ , blockade radius $R_b\sim6$ – $10\,\mu\mathrm{m}$ , $n\sim50$ – $100$ , two-qubit gates in $\sim0.1$ – $1\,\mu\mathrm{s}$ at fidelity $\gtrsim99\%$ .
Key idea: Identical atoms are held in a reconfigurable optical-tweezer array, encode a qubit in two hyperfine ground states, and entangle via the Rydberg blockade — the van der Waals shift of a high- $n$ state that forbids simultaneous excitation of neighbouring atoms.

Where this lives today

Neutral-atom arrays are one of the fastest-scaling quantum platforms. Registers of hundreds to over a thousand atoms are now routine, and because atoms transport mid-circuit, the architecture natively supplies the connectivity stabilizer codes crave — a 2023 experiment ran multiple logical qubits and error-corrected circuits on a reconfigurable few-hundred-atom array by shuttling entangled pairs between storage, entangling, and readout zones. Everything here is load-bearing there: coherence rides on the clock states of equation (1); gate speed and its error floor come from the Rydberg Rabi frequency and lifetime of Worked example 2; the entangling operation is the blockade of equations (4)–(7). The same hardware doubles as a programmable simulator — arrange atoms in a chosen geometry, tune the interaction, and the array realizes Ising-type spin models, letting experiments watch phase transitions and quench dynamics of hundreds of spins. The open frontiers are concrete: Rydberg lifetime and laser noise capping fidelity near $99.5\%$ ; atom loss during excitation and imaging; and the signature countermeasure, erasure conversion — an advantage largely unique to this hardware.

The physical picture

Picture a grid of microscopic optical bowls, each cradling one atom in the intensity gradient of a red-detuned laser. The qubit is a tiny internal compass — two hyperfine clock states that barely feel the outside world, which is why coherence is long. Computation is choreography: microwaves rotate the compass, atoms slide to bring partners together, and one is briefly flung up to a Rydberg orbital so vast it feels its neighbour and vetoes that neighbour's excitation. Idle qubits sit in the ground manifold, non-interacting; the coupling exists only during the flash of a gate. The first figure fixes the geometry and the numbers that decide whether two atoms are inside each other's blockade.

Figure

A tweezer register of identical atoms: the qubit is two hyperfine clock states of ⁸⁷Rb split by 6.835 GHz, and a control atom driven to a Rydberg state with Rabi frequency Ω/2π = 1 MHz and C₆/h = 880 GHz·μm⁶ blockades everything inside R_b = (C₆/ħΩ)^(1/6) ≈ 9.8 μm, so a neighbour at 5 μm is deep inside the blockade

The blockade is best read as a two-atom energy diagram. Drive both atoms from $|g\rangle$ toward the same Rydberg level $|r\rangle$ . The state $|gg\rangle$ couples equally to $|gr\rangle$ and $|rg\rangle$ , whose symmetric combination — the bright state $|W\rangle$ — soaks up both paths and oscillates $\sqrt2$ faster than a lone atom. The doubly-excited $|rr\rangle$ would sit at twice the single-excitation energy, but the interaction $V=C_6/R^6$ lifts it far above resonance, so the drive never reaches it. That vetoed rung is the whole mechanism: one collective excitation allowed, two forbidden, and the conditional phase is a CZ gate. The second figure is the punchline.

Figure

The blockade energy diagram: the ground pair couples to the symmetric bright state at the enhanced rate √2 Ω, but the doubly-excited state is pushed up by the interaction V = C₆/R⁶ to V/ħΩ ≈ 56 and is off-resonant, so double excitation is blocked and the residual leakage is (ħΩ/V)² ≈ 3×10⁻⁴

The core equations

The qubit is the two field-insensitive hyperfine clock states; everything else is control on top. Writing the encoding first fixes what $|0\rangle,|1\rangle$ mean throughout.

|\psi\rangle = \alpha\,|0\rangle + \beta\,|1\rangle,\qquad |0\rangle\equiv|F{=}1,m_F{=}0\rangle,\ \ |1\rangle\equiv|F{=}2,m_F{=}0\rangle \qquad (1)

where: $F$ — total hyperfine angular-momentum quantum number; $m_F$ — magnetic sublevel, the $m_F=0$ clock states carrying zero linear Zeeman shift; $\alpha,\beta$ — complex amplitudes with $|\alpha|^2+|\beta|^2=1$ . Reading it: because both states shift identically to first order in field, ambient noise dephases them only at second order — the origin of the long $T_2^*$ . The atom is held by the light itself: a far-detuned beam induces a dipole that couples back to its own field, giving an energy proportional to intensity — the dipole potential.

U(\mathbf r)=-\frac{1}{2}\,\mathrm{Re}[\alpha_{\mathrm{pol}}(\omega)]\,|\mathbf E(\mathbf r)|^2 \;\approx\; -\frac{3\pi c^2}{2\omega_0^3}\,\frac{\Gamma}{\Delta}\,I(\mathbf r) \qquad (2)

where: $\alpha_{\mathrm{pol}}$ — polarizability at trap frequency $\omega$ ; $I(\mathbf r)$ — local beam intensity; $\Delta=\omega-\omega_0$ — detuning from resonance $\omega_0$ ; $\Gamma$ — natural linewidth. Reading it: for red detuning $\Delta<0$ the potential is negative where the intensity is high, pulling the atom to the focus, while the heating scattering falls faster ( $\propto1/\Delta^2$ ) than the depth ( $\propto1/\Delta$ ) — hence far-detuned traps. Single-qubit rotations are resonant two-level drives: a microwave photon bridges the clock states at Rabi frequency $\Omega$ , with the pulse phase setting the rotation axis on the Bloch sphere. Focused optical addressing instead uses a two-photon Raman transition through a virtual excited state, whose coupling and light shift are read off below.

\Omega_{\mathrm{Raman}}=\frac{\Omega_1\Omega_2}{2\Delta},\qquad \delta_{\mathrm{AC}}=\frac{|\Omega_1|^2-|\Omega_2|^2}{4\Delta} \qquad (3)

where: $\Omega_1,\Omega_2$ — single-photon Rabi frequencies of the two Raman beams; $\Delta$ — detuning from the intermediate excited state, taken large to suppress scattering; $\delta_{\mathrm{AC}}$ — differential light shift that must be compensated for resonant driving. Reading it: the effective coupling grows as $1/\Delta$ while scattering falls as $1/\Delta^2$ , the same trade-off that governs the trap. The entangling resource is the interaction between two Rydberg atoms. For $S$ states it is van der Waals, with a coefficient that explodes with principal quantum number.

V_{\mathrm{vdW}}(R)=\frac{C_6}{R^6},\qquad C_6\propto n^{11} \qquad (4)

where: $R$ — interatomic separation; $C_6$ — van der Waals coefficient; $n$ — principal quantum number, typically $50$ – $100$ . Reading it: the $n^{11}$ scaling means a modest increase in $n$ multiplies the interaction enormously — but also its sensitivity to stray fields, the central tension of Rydberg control. Setting the interaction equal to the drive's energy resolution defines the length over which double excitation is forbidden — the blockade radius.

R_b=\left(\frac{C_6}{\hbar\Omega}\right)^{1/6} \qquad (5)

where: $R_b$ — separation below which the doubly-excited state is off-resonant; $\Omega$ — Rydberg-excitation Rabi frequency. Reading it: any pair closer than $R_b$ is blockaded, and the sixth-root makes $R_b$ insensitive to the exact interaction — a factor of ten in $C_6$ moves $R_b$ by only about $1.5$ (Worked example 1). Inside a blockade the pair collapses to a single collective excitation whose drive is enhanced — the fingerprint of full blockade.

|W\rangle=\frac{1}{\sqrt 2}\big(|r g\rangle+|g r\rangle\big),\qquad \Omega_{\mathrm{coll}}=\sqrt 2\,\Omega \qquad (6)

where: $|r\rangle,|g\rangle$ — Rydberg and ground states of one atom; $\Omega_{\mathrm{coll}}$ — collective Rabi frequency of the blockaded pair. Reading it: seeing the pair oscillate at $\sqrt2\,\Omega$ rather than $\Omega$ is the signature that the blockade is complete (Derivation 2). The conditional dynamics of that blockade, run through the pulse sequence, is a controlled-phase gate.

U_{CZ}=\mathrm{diag}\big(e^{i\phi_{00}},\,e^{i\phi_{01}},\,e^{i\phi_{10}},\,e^{i\phi_{11}}\big)\;\sim\;\mathrm{diag}(1,1,1,-1) \qquad (7)

where: $\phi_{ij}$ — accumulated phase for computational input $|ij\rangle$ after the Rydberg pulse sequence. Reading it: the gate is entangling precisely because the four inputs pick up phases that do not factorize — the combination $\phi_{11}-\phi_{01}-\phi_{10}+\phi_{00}=\pi$ is the whole content of the CZ (Derivation 1).

Derivation AControlled-phase gate from the Rydberg blockade

The goal is to show how the conditional dynamics of two blockaded atoms, driven by the original Jaksch–Cirac–Zoller three-pulse sequence, assembles the controlled-phase gate $U_{CZ}=\mathrm{diag}(1,1,1,-1)$ . Encode $|1\rangle$ as the state coupled to $|r\rangle$ by a resonant $\pi$ -pulse and $|0\rangle$ as a spectator; apply $\pi$ on the control, $2\pi$ on the target, $\pi$ on the control, and track the phase each input accumulates when $V\gg\hbar\Omega$ .

Step 1. Case $|00\rangle$ : neither atom couples to $|r\rangle$ , so the pulses are off-resonant spectators and the state returns unchanged, $\phi_{00}=0$ . Why: fixing this reference lets the other three phases be read as physical, gate-relevant quantities.

Step 2. Cases $|01\rangle$ and $|10\rangle$ : exactly one atom reaches $|r\rangle$ . A complete $2\pi$ rotation $|1\rangle\to|r\rangle\to|1\rangle$ returns the atom with a sign flip, $\phi=\pi$ . Why: a $2\pi$ pulse is not the identity but $-\mathbb{1}$ on the driven subspace — that minus sign is the raw material the sequence sculpts into the gate.

Step 3. Case $|11\rangle$ — the blockade: after the control $\pi$ -pulse the control sits in $|r\rangle$ , so the target's $2\pi$ pulse is detuned by $V\gg\hbar\Omega$ and cannot climb to $|r\rangle$ — it is blocked and acquires no $\pi$ . The control then returns. Why: the control in $|r\rangle$ removes the target's minus sign, so the $|11\rangle$ phase differs from the naive sum — the conditionality that makes the gate entangling.

Step 4. Bookkeeping: the differential phase is $\phi_{11}-\phi_{01}-\phi_{10}+\phi_{00}=\pi$ , and single-qubit $Z$ rotations absorb $\phi_{01},\phi_{10}$ , leaving $\mathrm{diag}(1,1,1,-1)$ . Why: only the non-factorizable combination survives local corrections, and a $\pi$ there is by definition a controlled-phase — a universal entangler with the single-qubit gates. *In the ideal full-blockade limit the truth table reproduces CZ exactly; the leading correction is the residual double-excitation leakage $(\hbar\Omega/V)^2$ , as detailed in Saffman, Walker and Mølmer (Rev. Mod. Phys. 82, 2313, 2010).*

Go deeper · Derivation B · 1 minCollective √2 enhancement under full blockade

Derivation BCollective √2 enhancement under full blockade

The goal is to prove that a perfectly blockaded pair, driven from $|gg\rangle$ , oscillates at $\Omega_{\mathrm{coll}}=\sqrt2\,\Omega$ rather than the single-atom $\Omega$ — the cleanest signature that a blockade is complete. Take two atoms each driven from $|g\rangle$ to $|r\rangle$ at Rabi frequency $\Omega$ , and impose full blockade by removing the doubly-excited $|rr\rangle$ from the accessible space (its shift $V\to\infty$ ). The claim: constructive interference of the two excitation paths multiplies the effective coupling by exactly $\sqrt2$ .

Step 1. Restrict the Hilbert space to $\{|gg\rangle,|gr\rangle,|rg\rangle\}$ , dropping $|rr\rangle$ . Why: this is the content of blockade — the forbidden rung is gone, turning a four-state problem into an exactly solvable three-state one.

Step 2. Write the drive on this subspace, $H=\tfrac{\hbar\Omega}{2}\big(|gr\rangle\langle gg|+|rg\rangle\langle gg|+\mathrm{h.c.}\big)$ . Why: each atom's laser couples the ground pair to its own single-excitation state, and both originate from the same $|gg\rangle$ — the geometry that forces interference.

Step 3. Define the symmetric bright state $|W\rangle=\tfrac{1}{\sqrt2}(|gr\rangle+|rg\rangle)$ ; the antisymmetric combination is dark. Why: only the symmetric superposition overlaps the drive out of $|gg\rangle$ , collapsing the dynamics to a genuine two-level system.

Step 4. In this basis $H=\tfrac{\hbar(\sqrt2\,\Omega)}{2}\big(|W\rangle\langle gg|+|gg\rangle\langle W|\big)$ , a two-level drive at frequency $\sqrt2\,\Omega$ . Why: the two amplitudes add, and $\langle W|(|gr\rangle+|rg\rangle)=\sqrt2$ delivers the enhancement — a pair oscillating at $\sqrt2\,\Omega$ shows no population is leaking to $|rr\rangle$ . *This collective-enhancement result was observed between two atoms by Gaëtan et al. and Urban et al. (both Nature Physics, 2009), and generalizes to $\sqrt N\,\Omega$ for $N$ atoms sharing one blockade volume.*

Limiting cases

The blockade physics interpolates between regimes set by the ratio of interaction $V(R)$ to drive strength $\hbar\Omega$ , and by the character of the Rydberg interaction.

Case	Result	Meaning
Full blockade	$V(R)\gg\hbar\Omega$	Doubly-excited state far off-resonance; only one collective Rydberg excitation allowed — the clean regime used for CZ gates, with leakage $(\hbar\Omega/V)^2$ .
No blockade	$V(R)\ll\hbar\Omega$	Atoms excite independently; no conditional phase and entanglement vanishes — the pair is simply too far apart for the given $\Omega$ .
Van der Waals scaling	$V=\dfrac{C_6}{R^6},\ C_6\propto n^{11}$	Interaction and blockade radius grow steeply with $n$ , but so does sensitivity to stray electric fields and blackbody radiation.
Resonant dipole–dipole	$V=\dfrac{C_3}{R^3}$	At a Förster resonance the interaction is dipolar and longer-ranged, enlarging $R_b$ and allowing coherent excitation hopping between atoms.
Many atoms in one blockade	$\Omega_{\mathrm{coll}}=\sqrt N\,\Omega$	$N$ atoms inside a single blockade volume share one excitation, giving multi-qubit gates and a $\sqrt N$ -enhanced collective Rabi frequency.

Common trapThe blockade is not an always-on wire between qubits

It is tempting to read "all-to-all connectivity" as a fixed coupling graph, like wires on a chip. It is nothing of the sort. The strong Rydberg interaction exists only during the microsecond an atom is transiently excited to $|r\rangle$ inside a gate; the instant the pulse ends the atom falls back to the ground manifold, where hyperfine qubits are essentially non-interacting. So two idle neighbours do not slowly entangle, and two distant qubits are coupled not by a standing bond but by moving one atom next to the other and firing the blockade. "Connectivity" is about which atoms you co-locate and drive relative to $R_b$ — not a hardwired topology. The corollary is a real error channel: an atom left in $|r\rangle$ too long, or a neighbour drifting inside $R_b$ during transport, produces unwanted conditional phases, so gate timing and array geometry matter as much as the pulses.

Worked examples

1 · Is a neighbour inside the blockade? (confidence builder)

Two $^{87}\mathrm{Rb}$ atoms sit $R=5\,\mu\mathrm{m}$ apart, driven to a Rydberg state with $C_6/h=880\,\mathrm{GHz}\cdot\mu\mathrm{m}^6$ ( $n\approx70$ ) at $\Omega/2\pi=1\,\mathrm{MHz}$ . Find the blockade radius, confirm the pair is blockaded, and estimate the leakage into $|rr\rangle$ .

Step 1. Blockade radius from equation (5). With $C_6/h$ in $\mathrm{GHz}\cdot\mu\mathrm{m}^6$ and $\Omega/2\pi$ in GHz, $R_b=\big[(C_6/h)/(\Omega/2\pi)\big]^{1/6}=(880/0.001)^{1/6}=(8.8\times10^5)^{1/6}\,\mu\mathrm{m}$ . The sixth root gives $R_b=\mathbf{9.8\,\mu m}$ .

Step 2. Compare to the separation: $R=5\,\mu\mathrm{m}<R_b=9.8\,\mu\mathrm{m}$ , so the pair is comfortably inside the blockade, and the sixth-root makes this robust — a factor of ten in $C_6$ shifts $R_b$ only to $\sim14\,\mu\mathrm{m}$ . The atoms are blockaded.

Step 3. Quantify the strength: $V/\hbar\Omega=(C_6/h)/\big[(\Omega/2\pi)\,R^6\big]=880/(0.001\times5^6)=880/15.625=\mathbf{56}$ . The doubly-excited state is shifted $56$ linewidths off resonance — deep in full blockade.

Step 4. Residual leakage into $|rr\rangle$ scales as $(\hbar\Omega/V)^2=(1/56)^2\approx\mathbf{3\times10^{-4}}$ . A few parts in ten thousand — small, but Worked example 2 shows it is not what limits the gate. Result: $R_b=9.8\,\mu$ m, $V/\hbar\Omega\approx56$ , leakage $\approx3\times10^{-4}$ .

2 · What actually limits the gate: finite Rydberg lifetime

The same drive $\Omega/2\pi=1\,\mathrm{MHz}$ runs the three-pulse CZ, with effective Rydberg lifetime $\tau_{\mathrm{Ryd}}\approx150\,\mu\mathrm{s}$ (radiative plus blackbody). Estimate the gate error from spontaneous decay in $|r\rangle$ , and compare it with the blockade leakage above.

Step 1. Pulse timing: a resonant $\pi$ -pulse takes $t_\pi=\pi/\Omega=\pi/(2\pi\times10^6\,\mathrm{s}^{-1})=\mathbf{0.5\,\mu s}$ , so the $\pi$ – $2\pi$ – $\pi$ sequence lasts $t_{\mathrm{gate}}\approx4\,t_\pi=2\,\mu\mathrm{s}$ . Fast against hyperfine coherence, but not against Rydberg decay.

Step 2. Rydberg-occupancy time: the control is parked in $|r\rangle$ for the middle $2\pi$ window, about $t_r\approx2t_\pi=1\,\mu\mathrm{s}$ ; the blockaded target barely goes up. Roughly $1\,\mu\mathrm{s}$ of Rydberg residence is exposed to decay per gate.

Step 3. Decay probability: $P_{\mathrm{decay}}\approx t_r/\tau_{\mathrm{Ryd}}=1\,\mu\mathrm{s}/150\,\mu\mathrm{s}\approx\mathbf{7\times10^{-3}}$ . One emission event scrambles the phase — a per-gate error near $0.7\%$ .

Step 4. Compare: Rydberg decay $\sim7\times10^{-3}$ beats blockade leakage $\sim3\times10^{-4}$ by over an order of magnitude. This mirrors real experiments — fidelities near $99.5\%$ are set by Rydberg lifetime and laser noise, not imperfect blockade, so higher $n$ (longer $\tau\propto n^3$ ), faster pulses, and erasure conversion are the levers. Result: $t_\pi=0.5\,\mu$ s, $P_{\mathrm{decay}}\approx0.7\%$ , dominating the $\approx0.03\%$ leakage.

Physics of the portrait

The neutral-atom processor is where a 2000-era gate proposal met two decades of tweezer engineering to become a machine that rearranges its own qubits.

Origins · 1 minMikhail Lukin and how this was discovered

Mikhail Lukin

born 1971 · Rydberg blockade and neutral-atom quantum processors

For most of the 1990s neutral atoms were prized for metrology and cold-collision physics, not computing: identical and coherent, but with no fast, controllable way to make two of them interact. The turning point came in 2000, when Dieter Jaksch, Juan Ignacio Cirac, Peter Zoller and co-workers — Mikhail Lukin among them — proposed the Rydberg blockade as a fast two-qubit gate, and in 2001 Lukin and collaborators extended it to the dipole blockade of mesoscopic ensembles, where a whole cloud shares one excitation. The proposal sat ahead of the hardware until 2009, when two groups directly observed the blockade and its $\sqrt2$ collective enhancement between two atoms. Lukin then drove the platform's second act: single-atom tweezers, defect-free array assembly, programmable Rydberg simulators of hundreds of spins, and reconfigurable processors that transport entangled atoms to wire up arbitrary connectivity and run error-corrected logical qubits. What began as a paper mechanism — one atom's Rydberg balloon vetoing its neighbour — became the entangling engine of a leading architecture.

History

2000

— Jaksch, Cirac, Zoller and co-workers propose a fast neutral-atom CZ gate using the Rydberg blockade

2001

— Lukin and co-workers extend the blockade to mesoscopic atomic ensembles

2009

— Gaëtan et al. and Urban et al. observe the Rydberg blockade and √2 collective enhancement between two atoms

2016

— Endres, Barredo and co-workers assemble defect-free tweezer arrays by atom-by-atom rearrangement

2022

— Bluvstein and co-workers run a processor based on coherent transport of entangled atom arrays

2023

— Bluvstein and co-workers demonstrate error-corrected logical qubits on a reconfigurable neutral-atom array

Going further

Texts. Saffman, Walker & Mølmer, "Quantum information with Rydberg atoms", Rev. Mod. Phys. 82, 2313 (2010) · Browaeys & Lahaye, "Many-body physics with individually controlled Rydberg atoms", Nature Physics 16, 132 (2020) · Jaksch et al., "Fast quantum gates for neutral atoms", Phys. Rev. Lett. 85, 2208 (2000) · Levine et al., "Parallel implementation of high-fidelity multiqubit gates with neutral atoms", Phys. Rev. Lett. 123, 170503 (2019) · Bluvstein et al., "A quantum processor based on coherent transport of entangled atom arrays", Nature 604, 451 (2022) · Nielsen & Chuang, Quantum Computation and Quantum Information, Ch. 7 (physical realizations) · Preskill, Lecture Notes for Physics 219, Ch. 7. The $^{87}\mathrm{Rb}$ hyperfine splitting ( $6.835\,\mathrm{GHz}$ ), $V=C_6/R^6$ with $C_6\propto n^{11}$ , and $R_b=(C_6/\hbar\Omega)^{1/6}$ are treated in Saffman–Walker–Mølmer (2010); the optical-dipole potential and its $\Gamma/\Delta$ versus $\Gamma^2/\Delta^2$ scaling follow Grimm–Weidemüller–Ovchinnikov. Key checkable results include: $R_b\approx9.8\,\mu\mathrm{m}$ and $V/\hbar\Omega\approx56$ at $R=5\,\mu\mathrm{m}$ from $C_6/h=880\,\mathrm{GHz}\cdot\mu\mathrm{m}^6$ , $\Omega/2\pi=1\,\mathrm{MHz}$ ; the $\sqrt2$ enhancement from the three-state Hamiltonian; the CZ truth table and $(\hbar\Omega/V)^2$ leakage from the Jaksch protocol; and the decay estimate $P\approx t_r/\tau_{\mathrm{Ryd}}\approx0.7\%$ . Quoted device figures ( $n\sim50$ – $100$ , spacing $\sim3$ – $5\,\mu\mathrm{m}$ , gates $\sim0.1$ – $1\,\mu\mathrm{s}$ , fidelity $\gtrsim99\%$ , $\tau_{\mathrm{Ryd}}\sim150\,\mu\mathrm{s}$ ) are representative values, not fabricated precision.

Prerequisites. The DiVincenzo Criteria · Sideband Cooling · Trapped-Ion Qubits. Leads to. The Surface Code · Erasure Qubits · Quantum Simulation.

Ask the tutor: Want to walk through how the same Rydberg blockade that runs a two-atom CZ generalizes to a native multi-qubit gate when several atoms share one blockade volume?

Why are neutral atoms said to be "identical" qubits, and does it matter?

It matters a great deal. Every $^{87}\mathrm{Rb}$ atom has exactly the same hyperfine splitting and interaction coefficients, set by atomic physics, not fabrication. Contrast a superconducting chip, where each transmon's frequency depends on a junction whose area varies device to device, forcing per-qubit calibration and frequency-collision headaches as chips grow. With neutral atoms a pulse correct for one atom is correct for all, so a single global drive rotates the whole register at once. The disorder that remains is in the environment — trap depth, position jitter, laser noise — a more tractable problem than intrinsic qubit spread.

How can atoms have "all-to-all" connectivity when the Rydberg interaction is short-ranged?

Two ways that stack. First, within a blockade radius of several microns every atom can directly entangle with every other, so a local cluster is fully connected. Second, and more powerful, the traps move: tweezers transport an atom across the array while it carries quantum information, park it beside a distant partner, fire a blockade gate, and carry it back. Connectivity is transport plus local blockade, not a fixed graph. This is why the platform maps so naturally onto codes needing long-range operations — you route the qubits to the interaction, not the interaction to the qubits, which a fixed chip cannot do.

Why excite the atom to such a high Rydberg state rather than a lower level?

Because you need an enormous, switchable interaction, and only Rydberg states provide it. A ground-state atom is nearly inert — its van der Waals coefficient is minuscule, so two atoms microns apart barely feel each other. Promote the valence electron to $n\sim70$ and the orbital swells thousands of times; $C_6\propto n^{11}$ grows by roughly twenty orders of magnitude, so the pair interacts strongly across several microns — exactly the scale of a tweezer array. The price: the same huge polarizability makes Rydberg atoms sensitive to stray fields and gives them finite lifetime, so the excitation must be brief — used only for the microsecond of a gate, then returned to the safe, long-lived ground manifold.

What sets the two-qubit gate fidelity, and why is it not yet as high as trapped ions?

The leading limits are finite Rydberg lifetime, laser phase and intensity noise, and atomic motion. As Worked example 2 shows, an atom spends about a microsecond in $|r\rangle$ per gate, and with lifetime near $150\,\mu\mathrm{s}$ that alone gives a per-gate error of order $10^{-2}$ ; blockade leakage $(\hbar\Omega/V)^2$ is usually an order of magnitude smaller. Trapped ions entangle through phonon modes that do not decay on gate timescales, so their gates exceed $99.9\%$ . Neutral-atom gates have reached roughly $99.5\%$ and are improving through higher $n$ , faster and more coherent lasers, and — distinctively — erasure conversion.

How are atoms read out without disturbing the ones you are not measuring?

Readout is state-selective fluorescence. Light on a cycling transition couples only to one hyperfine state: an atom there scatters thousands of photons and lights up on a camera, while one in the other state — or shelved dark — stays dark. The whole array is imaged in parallel, and the bright-or-dark outcome gives the bit. The subtlety is that scattering heats the atom and can knock it out of the qubit manifold, so imaging is destructive; mid-circuit measurement therefore uses separate readout zones or shelved ancillas so the fluorescence never touches the data qubits.

What is erasure conversion, and why is it special to this platform?

Error correction must find and locate errors, and the "where" is expensive. Erasure conversion, enabled by the atom's rich level structure, makes the location free. Encode the qubit in levels chosen so the dominant errors — atom loss and decay out of the qubit states — leave the atom in a distinguishable "erased" configuration; a fast non-destructive check then reveals which sites failed each cycle. An error of known location is an erasure, and codes tolerate about twice as many erasures as errors of unknown position, so the threshold roughly doubles. Neutral atoms suit this uniquely because their many long-lived internal states let designers route leading error channels into heralded outcomes — a lever sparse-level superconducting qubits lack.