Rydberg Atom Arrays

A Rydberg atom array is a quantum processor assembled one atom at a time. Individual neutral atoms — alkali rubidium and cesium, or alkaline-earth-like strontium and ytterbium — are caught in tightly focused optical tweezers, then shuffled into a defect-free, reconfigurable lattice of hundreds to thousands of sites in one, two, or three dimensions. The qubit lives in a pair of long-lived ground (hyperfine) states, where atoms barely feel one another and the information is well shielded from the world. The entangling resource is switched on by promoting an atom to a Rydberg state — an electronic level with huge principal quantum number $n\sim 50\text{–}100$ , so bloated that its dipole scales as $n^2$ and its pair interaction dwarfs anything between ground-state atoms. Two Rydberg atoms shift each other through $V(r)=C_6/r^6$ with $C_6\propto n^{11}$ ; once that shift exceeds the coupling $\Omega$ , a second atom cannot be excited while the first is. That is the Rydberg blockade, and the distance where interaction balances drive — the blockade radius $R_b$ — sets the length scale of every protocol. The same physics runs in two modes. Digitally, a three-pulse blockade sequence makes a pair's dynamics conditional on its joint state, imprinting a controlled-phase gate, while single-qubit gates are microwave or Raman drives on the hyperfine manifold; because connectivity is literally the geometry of the atoms — which can be moved mid-circuit — the wiring is reconfigurable in a way fixed-coupling chips cannot match. Analogically, one global laser turns the array into a programmable Ising/PXP spin model, exposing quantum magnetism, many-body scars, and native optimization — a dual character that has made neutral-atom arrays a leading route to scalable quantum information processing.

Key takeaways

Qubits are neutral atoms in optical-tweezer arrays; the entangling force is the sharply ranged van der Waals interaction $V(r)=C_6/r^6$ between Rydberg states, with $C_6\propto n^{11}$ .
The Rydberg blockade — one excited atom shifting its neighbours out of resonance — forbids two excitations within a blockade radius $R_b=(C_6/\hbar\Omega)^{1/6}$ and is the engine of the two-qubit gate.
Inside the bubble the drive couples $|gg\rangle$ to the symmetric bright state at the enhanced rate $\sqrt{2}\,\Omega$ ; a $\pi$ – $2\pi$ – $\pi$ sequence makes a controlled-phase (CZ) gate, and a Hadamard turns CZ into CNOT.
One global laser realizes $H=\sum_i\tfrac{\Omega_i}{2}\sigma_i^x-\sum_i\Delta_i n_i+\sum_{i<j}V_{ij}n_i n_j$ , a programmable Ising/PXP model with $\mathbb{Z}_2,\mathbb{Z}_3,\ldots$ antiferromagnetic ground states.
The blockade is never perfect: leakage floors the gate error at $\sim\Omega^2/2V^2$ , but finite Rydberg lifetime, laser phase noise, and Doppler dephasing dominate the real budget; demonstrated two-qubit fidelities now exceed 99.5%.
Key idea: excite atoms to a Rydberg state and the $C_6/r^6$ van der Waals interaction becomes so strong that, within the blockade radius, only one atom can be excited at a time; this conditional excitation, driven by global lasers on reconfigurable tweezer geometries, is simultaneously a universal entangling gate and a programmable spin-model simulator.

Where this lives today

Rydberg arrays are now a leading contender for fault-tolerant quantum computing. In 2024 Bluvstein and co-workers ran a logical processor on reconfigurable arrays of several hundred atoms, shuttling them between storage, entangling, and readout zones to implement surface-code and qLDPC primitives at fidelities above 99.5% — non-local connectivity from mid-circuit atom movement that no fixed-coupling chip shares. On the analog side, programmable geometries realize spin liquids on kagome and ruby lattices, symmetry-protected topological phases, and native encodings of the NP-hard maximum-independent-set problem on unit-disk graphs. Alkaline-earth-like strontium and ytterbium carry an optical clock transition alongside the Rydberg manifold, so the same hardware builds tweezer-array clocks whose Rydberg entanglement pushes metrology toward the Heisenberg limit. The open problems — fidelities past $99.9\%$ , scaling to tens of thousands of atoms, fast low-loss readout — are the levers Example 2 identifies.

The physical picture

Picture two atoms a few microns apart. In their ground states they scarcely notice each other — neutral atoms are almost inert at these distances. Promote one to a Rydberg level, where the electron orbits hundreds of times farther out, and it becomes a giant antenna: its dipole grows by thousands, shifting the second atom's Rydberg transition by megahertz to gigahertz. Tune the laser to the bare transition and the second atom is off-resonant — frozen out, blockaded. The first figure makes this quantitative: a blockade bubble holding both atoms, with the doubly-excited state pushed away while the bright state is driven at $\sqrt{2}\,\Omega$ .

Figure

Two rubidium atoms 4 μm apart inside one blockade bubble of radius R_b = 8.7 μm: the drive couples the ground pair gg to the symmetric bright state W = (gr+rg)/√2 at the enhanced rate √2 Ω, while the doubly-excited pair rr is pushed out of resonance by V/2π = 210 MHz and stays dark

Because the interaction falls as $1/r^6$ it is extraordinarily sharply ranged: halve the distance and the shift grows by $2^6=64$ . Just inside $R_b$ the blockade is near-perfect; just outside, atoms are effectively independent. Connectivity is therefore geometry — you choose which atoms talk by placing the tweezers, and rewire by moving atoms mid-circuit. The second figure is the blockade curve $V(r)$ for a real Rydberg state, plotted against the drive so the blockade radius is where they cross.

Figure

Blockade curve for rubidium 70S atoms: the van der Waals shift V(r) = C₆/r⁶ with C₆/h = 862 GHz·μm⁶ crosses the drive Ω/2π = 2 MHz at the blockade radius R_b = 8.7 μm; at a 4 μm gate spacing V/2π = 210 MHz, so V/Ω ≈ 105 and the leakage floor Ω²/2V² ≈ 4.5×10⁻⁵

With one global laser on every atom, the blockade becomes a many-body Hamiltonian: the occupation $n_i$ plays an Ising spin, $\Delta$ a longitudinal field, $\Omega$ a transverse field. Turn $\Delta$ up so atoms want to be excited while the blockade forbids neighbours, and the array orders into an antiferromagnetic crystal $\ldots rgrg\ldots$ — a quantum phase transition watched site by site.

The core equations

The starting point is the interaction. Two atoms in identical Rydberg $S$ states carry no permanent dipole, so the leading coupling is the second-order van der Waals shift, whose explosive $n$ -dependence lets micron-scale atoms interact at gigahertz strength.

V(r)=\frac{C_6}{r^6},\qquad C_6 \propto n^{11} \qquad (1)

where: $V(r)$ — energy shift of the doubly-excited Rydberg pair at separation $r$ ; $C_6$ — van der Waals coefficient, strongly state-dependent; $n$ — principal quantum number; $r$ — interatomic distance. Reading it: the $1/r^6$ makes the interaction a near-step function of distance, while $n^{11}$ dials its magnitude over orders of magnitude through the level. Setting the interaction shift equal to the laser coupling defines the blockade radius, inside which a second excitation is pushed off resonance — the most useful length scale on the platform.

R_b=\left(\frac{C_6}{\hbar\,\Omega}\right)^{1/6} \qquad (2)

where: $R_b$ — blockade radius; $\Omega$ — single-atom ground-to-Rydberg Rabi frequency; $\hbar$ — reduced Planck constant. Reading it: the sixth root crushes the $\Omega$ dependence, so $R_b$ is set almost entirely by the state ( $C_6$ ) and sits at several microns — larger than a typical tweezer spacing, so neighbours blockade cleanly. Inside the bubble a resonant drive cannot address the atoms independently: it couples the ground pair to the single symmetric superposition, faster than a lone atom by $\sqrt{2}$ — the experimental fingerprint of the blockade.

|gg\rangle \xrightarrow{\ \Omega\ } \tfrac{1}{\sqrt 2}\big(|gr\rangle+|rg\rangle\big),\qquad \Omega_{\mathrm{coll}}=\sqrt{2}\,\Omega \qquad (3)

where: $|g\rangle,|r\rangle$ — single-atom ground and Rydberg states; $\Omega_{\mathrm{coll}}$ — collective Rabi frequency of the pair. Reading it: the pair shares one excitation as a $W$ -state, and the two paths interfere constructively to speed the oscillation by $\sqrt{2}$ — measured directly in experiments. Promoting to a full array with a site-resolved drive gives the platform's master Hamiltonian — a transverse-field Ising model with long-range $1/r^6$ coupling, of which every experiment, digital or analog, is a special case.

H=\sum_i \frac{\Omega_i}{2}\,\sigma_i^x-\sum_i \Delta_i\,n_i+\sum_{i<j}V_{ij}\,n_i n_j \qquad (4)

where: $\sigma_i^x=|g\rangle_i\langle r|+|r\rangle_i\langle g|$ — drive (transverse-field) operator; $n_i=|r\rangle_i\langle r|$ — Rydberg occupation, the Ising spin; $\Delta_i$ — detuning, a longitudinal field; $V_{ij}=C_6/r_{ij}^6$ — van der Waals coupling. Reading it: $\Omega$ flips spins, $\Delta$ pays for excitations, $V_{ij}$ penalizes excited neighbours — three knobs that place the array anywhere in its phase diagram. In the hard-blockade limit the $V_{ij}$ term becomes an exact constraint — adjacent excitations forbidden — and the Hamiltonian projects onto the constrained subspace as the PXP model.

H_{\mathrm{PXP}}=\Omega\sum_i P_{i-1}\,\sigma_i^x\,P_{i+1},\qquad P=|g\rangle\langle g| \qquad (5)

where: $P_{i\pm1}$ — projector onto a ground-state neighbour (hard blockade); $\Omega$ — resonant drive at $\Delta=0$ . Reading it: atom $i$ may flip only if both neighbours are in $|g\rangle$ ; this innocent kinetic constraint produces quantum many-body scars — eigenstates that refuse to thermalize. Finally, the intrinsic gate error. Because $V$ is finite the doubly-excited state is only detuned, so a small residual population leaks in — a floor no pulse-shaping removes, and the reason experiments push $V$ high while fighting the decay it costs.

P_{\mathrm{double}}\approx \frac{\Omega^2}{2V^2},\qquad \bar F_{\mathrm{gate}}\gtrsim 1-a\,\frac{\Omega^2}{V^2}-b\,\frac{\Gamma}{\Omega} \qquad (6)

where: $P_{\mathrm{double}}$ — residual double-excitation probability; $V$ — interaction shift at the chosen spacing; $\bar F$ — average two-qubit fidelity; $a,b$ — order-one constants; $\Gamma=\Gamma_0 n^{-3}+\Gamma_{\mathrm{bb}}n^{-2}$ — Rydberg decay (radiative plus blackbody). Reading it: the two terms oppose — raising $V$ shrinks leakage but demands higher $n$ or tighter spacing, both raising $\Gamma/\Omega$ , so real gates sit at the balance point.

Derivation ACollective Rabi enhancement in the Rydberg blockade

The goal is to show, from the bare two-atom Hamiltonian, why a pair inside the blockade radius behaves not as two independent driven atoms but as a single effective two-level system oscillating between $|gg\rangle$ and one symmetric bright state at $\sqrt{2}\,\Omega$ — and why the antisymmetric partner never participates. This $\sqrt{2}$ is the cleanest signature that the blockade is real, so deriving it from first principles fixes the physics and conventions we reuse in the gate.

Step 1. Work in the two-atom basis $\{|gg\rangle,|gr\rangle,|rg\rangle,|rr\rangle\}$ with a resonant drive $\tfrac{\Omega}{2}(\sigma_1^x+\sigma_2^x)$ on each atom and the van der Waals shift $V\,|rr\rangle\langle rr|$ . Why: the minimal Hamiltonian containing both the drive that creates excitations and the interaction that penalizes having two.

Step 2. Rotate the singly-excited sector into symmetric and antisymmetric combinations, $|W\rangle=\tfrac{1}{\sqrt2}(|gr\rangle+|rg\rangle)$ and $|D\rangle=\tfrac{1}{\sqrt2}(|gr\rangle-|rg\rangle)$ . Why: the drive is symmetric under exchanging the atoms, so it connects only states of the same symmetry.

Step 3. The ground pair connects only to the bright state with $\langle W|H|gg\rangle=\Omega/\sqrt2$ , which connects onward to $|rr\rangle$ with the same $\Omega/\sqrt2$ , while $\langle D|H|gg\rangle=0$ exactly. Why: the paths $gg\to gr$ and $gg\to rg$ add in phase for $|W\rangle$ and cancel for $|D\rangle$ — the interference behind both the $\sqrt2$ and the dark state.

Step 4. In the strong-blockade limit $V\gg\Omega$ the state $|rr\rangle$ is detuned far beyond the coupling, so it is only virtually populated and can be adiabatically eliminated. Why: with $|rr\rangle$ and $|D\rangle$ removed, the dynamics collapse onto $\{|gg\rangle,|W\rangle\}$ .

Step 5. The pair therefore executes Rabi oscillations $|gg\rangle\leftrightarrow|W\rangle$ at $\Omega_{\mathrm{coll}}=\sqrt2\,\Omega$ , an effective two-level system. Why: the enhancement seen in the 2009 experiments, and the ingredient the gate exploits to make the collective response state-dependent. Verified against Saffman–Walker–Mølmer (Rev. Mod. Phys. 82, 2313, 2010): diagonalizing the $4\times4$ Hamiltonian as $V\to\infty$ gives the $\sqrt2$ splitting.

Go deeper · Derivation B · 2 minThe blockade controlled-phase gate truth table

Derivation BThe blockade controlled-phase gate truth table

The goal is to trace how a fixed three-pulse laser sequence produces a two-qubit controlled-phase gate by following each of the four inputs through the pulses and showing that exactly one is protected by the blockade, leaving a diagonal unitary with a single conditional sign. This is the original Jaksch–Cirac–Zoller–Côté–Lukin proposal of 2000, from which the whole neutral-atom gate set descends.

Step 1. Encode the qubit in ground states $\{|0\rangle,|1\rangle\}$ , with only $|1\rangle$ coupled to $|r\rangle$ by $\Omega$ , and apply the sequence: a $\pi$ pulse on the control, a $2\pi$ on the target, then a $\pi$ on the control. Why: a $\pi$ pulse swaps $|1\rangle\leftrightarrow|r\rangle$ , while a full $2\pi$ returns an isolated atom to $|1\rangle$ with a sign flip — the two rotations the gate plays off against each other.

Step 2. Input $|00\rangle$ is untouched — neither atom couples to the laser — so it returns as $+|00\rangle$ . Why: it sets the zero against which the conditional phase is measured.

Step 5. Input $|11\rangle$ : the control $\pi$ excites it to $|r\rangle$ ; the blockade now shifts the target out of resonance so its $2\pi$ pulse cannot complete — the target picks up no phase — while the control $\pi$ returns with a single $-1$ , giving $-|11\rangle$ . Why: the one input where the blockade acts, suppressing the target rotation that would have added a second sign flip.

Step 6. Collecting the outputs gives $U=\mathrm{diag}(1,-1,-1,-1)$ , which single-qubit $Z$ rotations convert to the canonical $\mathrm{diag}(1,1,1,-1)$ CZ. Why: CZ plus single-qubit rotations is universal, and CZ between target Hadamards yields CNOT — a complete gate set. Verified against Jaksch et al. (Phys. Rev. Lett. 85, 2208, 2000); the residual error inherits the $\Omega^2/V^2$ floor of equation (6).

Go deeper · Derivation C · 1 minAntiferromagnetic crystals from the detuned Hamiltonian

Derivation CAntiferromagnetic crystals from the detuned Hamiltonian

The goal is to show how the array Hamiltonian, driven globally and slowly ramped in detuning, undergoes a quantum phase transition from a trivial paramagnet to a $\mathbb{Z}_2$ -ordered antiferromagnetic crystal, and how the van der Waals tail generates a staircase of higher-period orders — the analog face of the platform, realized in Bernien's 51-atom chain.

Step 1. Take the array Hamiltonian (4) with uniform $\Omega,\Delta$ on a chain, keeping only the dominant nearest-neighbour coupling $V\gg\Omega,\Delta$ , so $H=\tfrac{\Omega}{2}\sum_i\sigma_i^x-\Delta\sum_i n_i+V\sum_i n_i n_{i+1}$ . Why: near-neighbour blockade is the essential constraint; longer-range tails refine, not change, the picture.

Step 2. In the classical limit $\Omega\to0$ the energy is $E[\{n_i\}]=-\Delta\sum_i n_i+V\sum_i n_i n_{i+1}$ — detuning rewards excitations, interaction punishes excited neighbours. Why: with the drive off the ground state is a definite bit string, found by energy minimization.

Step 3. For $\Delta<0$ excitations cost energy, so every atom stays in $|g\rangle$ and $\langle n_i\rangle=0$ — the disordered paramagnet. Why: the trivial phase, adiabatically connected to all atoms in the ground state.

Step 4. For $0<\Delta<V$ it pays to excite atoms but never two in a row, so the ground state is the half-filled pattern $n_i=\tfrac12[1+(-1)^i]$ , an antiferromagnet $\ldots rgrg\ldots$ breaking $\mathbb{Z}_2$ two ways. Why: this broken-symmetry crystal is the ordered phase, its two degenerate patterns the hallmark of $\mathbb{Z}_2$ order.

Step 5. Restoring $\Omega$ smears the boundary near $\Delta_c\sim\Omega$ into a quantum critical point of the Ising universality class; the $1/r^6$ tail then adds period-3 and period-4 orders, a devil's-staircase of $\mathbb{Z}_3,\mathbb{Z}_4,\ldots$ crystals. Why: what the experiments scan — the observed plateaus and critical scaling match Fendley–Sengupta–Sachdev theory. Verified against Bernien et al. (Nature 551, 579, 2017).

Limiting cases

The competition between drive $\Omega$ , interaction $V(r)=C_6/r^6$ , and detuning $\Delta$ carves the platform into a handful of operating regimes.

Case	Result	Meaning
Strong blockade	$V(r)\gg\Omega\ \Rightarrow\ r\ll R_b$	Only one excitation per volume; the pair is an effective two-level system with collective rate $\sqrt2\,\Omega$ — the gate regime.
Weak / no blockade	$V(r)\ll\Omega\ \Rightarrow\ r\gg R_b$	Atoms drive independently and both can be excited; good for parallel single-qubit control, useless for entanglement.
Antiblockade (facilitation)	$V(r)=\Delta$	A detuned drive is resonant only once a neighbour is excited, so one excitation seeds the next — avalanche and epidemic-like spreading.
Förster resonance	$V_{\mathrm{dd}}(r)=C_3/r^3,\ \|sp\rangle\leftrightarrow\|ps\rangle$	Interaction turns dipolar ( $1/r^3$ ) and off-diagonal, realizing an XY (flip-flop) spin model with excitation transport instead of Ising order.
Perfect blockade	$V\to\infty$	The array collapses onto the constrained PXP model — frozen excited neighbours and non-thermal quantum scars.

Common trapthe misconception that a strong blockade means a perfect gate

The blockade is never perfect. Because $V$ is finite the doubly-excited state is only detuned, so a residual double-excitation probability $\sim(\Omega/V)^2$ leaves an intrinsic leakage floor on any blockade gate — you cannot pulse-shape it away, only push $V$ higher. And even a flawless blockade would not give a flawless gate: the qubit lives in ground states, but the gate transiently parks population in $|r\rangle$ , whose finite lifetime (radiative $\propto n^{-3}$ plus blackbody $\propto n^{-2}$ , $\tau\sim100\ \mu\mathrm{s}$ at $n\sim70$ ), together with laser phase noise, stray fields, and Doppler dephasing, sets the true error budget. The blockade fixes what the gate can do in principle; the Rydberg state's fragility fixes what it does in the lab.

Worked examples

1 · Blockade radius and leakage floor (fully worked)

Rubidium atoms are excited to the $70S$ state ( $C_6/h=862\ \mathrm{GHz\cdot\mu m^6}$ ) and driven at $\Omega/2\pi=2\ \mathrm{MHz}$ ; the gate pair sits $r=4\ \mu\mathrm{m}$ apart. Find the blockade radius, confirm strong blockade, and estimate the leakage floor.

Step 1. Blockade radius. In frequency units $R_b=\big[(C_6/h)/(\Omega/2\pi)\big]^{1/6}=\big[862000/2\big]^{1/6}=(4.31\times10^5)^{1/6}\ \mu\mathrm{m}$ , so $R_b=\mathbf{8.7\ \mu m}$ .

Step 2. Inside the bubble? The spacing $r=4\ \mu\mathrm{m}$ is well under $R_b=8.7\ \mu\mathrm{m}$ , so the pair is strongly blockaded. The shift there is $V/2\pi=(C_6/h)/r^6=862000/4^6=\mathbf{210\ MHz}$ , so $V/\Omega\approx\mathbf{105}$ — two orders of magnitude.

Step 3. Leakage floor. The residual double-excitation probability is $P_{\mathrm{double}}\approx\Omega^2/2V^2=1/(2\times105^2)\approx\mathbf{4.5\times10^{-5}}$ . So at $n=70$ , $4\ \mu\mathrm{m}$ spacing, and a $2\ \mathrm{MHz}$ drive the blockade is essentially perfect — intrinsic leakage infidelity below $10^{-4}$ , so something else must dominate the real error.

2 · The real gate error budget

Continue Example 1. A $\pi$ pulse takes $t_\pi=\pi/\Omega=250\ \mathrm{ns}$ , and the full CZ sequence parks population in $|r\rangle$ for roughly $t_r\approx0.5\ \mu\mathrm{s}$ . With Rydberg lifetime $\tau=100\ \mu\mathrm{s}$ at $n=70$ , estimate the decay error and the fidelity, and identify the dominant term.

Step 1. Decay error. The probability of a spontaneous or blackbody decay while the atom sits in $|r\rangle$ is $P_{\mathrm{decay}}\approx t_r/\tau=0.5/100=\mathbf{5\times10^{-3}}$ — a hundred times the leakage floor of Example 1.

Step 2. Compare. Leakage $\sim4.5\times10^{-5}$ ; decay $\sim5\times10^{-3}$ . The Rydberg lifetime, not the blockade quality, dominates the error budget — the point of the warning box.

Step 3. Fidelity. Adding the two, $1-\bar F\approx a\,\Omega^2/V^2+b\,\Gamma/\Omega\sim5\times10^{-3}$ with order-one $a,b$ , so $\bar F\approx\mathbf{0.995}$ . This matches demonstrated fidelities above 99.5% and names the levers: colder atoms and faster (higher- $\Omega$ ) gates cut the decay fraction, and a cryogenic apparatus suppresses the blackbody term.

Physics of the portrait

The Rydberg blockade sits where nineteenth-century spectroscopy meets twenty-first-century quantum engineering: a resource named for the states of Johannes Rydberg, turned into a gate a century later.

Origins · 1 minRydberg blockade and how this was discovered

Rydberg blockade

conditional single-atom excitation, proposed 2000 · neutral-atom quantum information

The blockade is named for the highly excited states Johannes Rydberg catalogued in 1888, when he found the alkali spectral lines fit a simple formula built from one constant and integer quantum numbers — the large- $n$ states that now bear his name, for a century a spectroscopist's curiosity remarkable only for their size and sensitivity. The blockade idea arrived in 2000, when Dieter Jaksch, Juan Ignacio Cirac, Peter Zoller, Robin Côté and Mikhail Lukin proposed that the huge dipole interaction between two Rydberg atoms could shift a second excitation out of resonance, giving a two-qubit gate whose speed is nearly independent of the interatomic distance — a decisive advantage over motional gates. The prediction waited nearly a decade: in 2009 the groups of Antoine Browaeys and Mark Saffman observed the blockade directly between two trapped atoms, including the tell-tale $\sqrt2$ collective Rabi enhancement. Defect-free arrays followed in 2016, Bernien's 51-atom simulator with its quantum scars in 2017, and by 2024 reconfigurable arrays were running logical, error-corrected circuits. What began as a formula for alkali spectra is now the organizing principle of an entire quantum-computing platform.

History

1888

— Johannes Rydberg fits the alkali spectral series with his constant, defining the high-n states

2000

— Jaksch, Cirac, Zoller, Côté and Lukin propose the dipole-blockade two-qubit gate

2009

— the Browaeys and Saffman groups observe the blockade and the √2 collective Rabi rate between two atoms

2016

— real-time atom rearrangement produces the first defect-free tweezer arrays

2017

— Bernien and co-workers run a 51-atom simulator, revealing Z2 order and quantum many-body scars

Going further

References. M. Saffman, T. G. Walker, K. Mølmer, "Quantum information with Rydberg atoms", Rev. Mod. Phys. 82, 2313 (2010) · D. Jaksch et al., "Fast quantum gates for neutral atoms", Phys. Rev. Lett. 85, 2208 (2000) · A. Browaeys, T. Lahaye, "Many-body physics with individually controlled Rydberg atoms", Nat. Phys. 16, 132 (2020) · H. Bernien et al., "Probing many-body dynamics on a 51-atom quantum simulator", Nature 551, 579 (2017) · D. Bluvstein et al., "Logical quantum processor based on reconfigurable atom arrays", Nature 626, 58 (2024) · Nielsen & Chuang, Quantum Computation and Quantum Information, Ch. 7 · Preskill, Lecture Notes on Quantum Computation, Ch. 7.

Provenance. The scalings ( $C_6\propto n^{11}$ , $R_b=(C_6/\hbar\Omega)^{1/6}$ , the $\sqrt2$ enhancement, $P_{\mathrm{double}}\approx\Omega^2/2V^2$ , the $n^{-3}/n^{-2}$ decay split) trace to Saffman–Walker–Mølmer (2010) and Browaeys–Lahaye (2020); the blockade gate to Jaksch et al. (2000); the phase diagram and scars to Bernien et al. (2017) and Fendley–Sengupta–Sachdev. The worked numbers use a representative $70S$ rubidium $C_6/h=862\ \mathrm{GHz\cdot\mu m^6}$ ; a verifier should confirm $R_b=8.7\ \mu\mathrm{m}$ , $V/2\pi=210\ \mathrm{MHz}$ at $4\ \mu\mathrm{m}$ , and $P_{\mathrm{double}}\approx4.5\times10^{-5}$ . The lifetime and fidelity floor are representative demonstrated values, not constants.

Builds on: DiVincenzo Criteria · Two-level atom and Rabi oscillations · van der Waals and dipole interactions · Pauli operators and the Ising model. Leads to: Mølmer–Sørensen Gate · Trapped-Ion Qubits · Quantum simulation of spin models · Quantum error correction with neutral atoms.

Ask the tutor: Want to see how changing $n$ or the tweezer spacing reshapes the blockade radius, the gate error budget, or the antiferromagnetic phase diagram? Ask and we will work a case together.

What makes a Rydberg state so special that it can entangle atoms microns apart?

The valence electron in a Rydberg state of principal quantum number $n$ orbits at radius $\propto n^2$ , so at $n\sim70$ the atom is a thousand times larger than its ground state and carries a dipole matrix element scaling as $n^2$ . The pair interaction, second order in that coupling, scales as $n^{11}$ — which is why micron-separated atoms, inert in their ground states, shift each other by hundreds of megahertz once excited. That switchable interaction is the whole entangling resource: excite with a laser and it appears; de-excite and the atoms are isolated memory qubits again.

Why store the qubit in ground states rather than directly in Rydberg levels?

Because the two jobs want opposite properties. A memory qubit must be long-lived and non-interacting — what a pair of hyperfine ground states offers: coherence times from seconds to minutes and negligible field sensitivity. An entangling state must interact strongly, which is what makes the Rydberg level fragile — it lives only $\sim100\ \mu\mathrm{s}$ and couples to every passing field. The platform gets both by keeping the qubit in the ground manifold and only briefly promoting it to $|r\rangle$ during a gate.

Is the blockade a real energy barrier, or just a detuning?

It is a detuning, not a wall. Exciting the second atom is not forbidden by any conservation law — it is simply pushed off resonance by the shift $V$ , so a laser on the bare transition drives it only weakly. That is why the blockade is imperfect: the residual double-excitation probability is $\sim(\Omega/V)^2$ , finite for any finite $V$ . "Blockade" is shorthand for "the second excitation is so far detuned that, to the accuracy you care about, it does not happen". Larger $V$ sharpens it but never makes it exact.

Why does one global laser turn the array into an Ising model?

In the array Hamiltonian $H=\sum_i\tfrac{\Omega_i}{2}\sigma_i^x-\sum_i\Delta_i n_i+\sum_{i<j}V_{ij}n_i n_j$ , identify the Rydberg occupation $n_i$ with an Ising spin: $\Omega$ is a transverse field that flips it, $\Delta$ a longitudinal field that biases it, and $V_{ij}$ a spin-spin coupling penalizing excited neighbours. A single global beam sets uniform $\Omega$ and $\Delta$ , so the only site-dependence left is $V_{ij}$ — which you program by placing the atoms. The result is a transverse-field Ising model whose graph is drawn by your tweezers, with antiferromagnetic crystal ground states.

How is a Rydberg array's connectivity different from a superconducting chip's?

On a superconducting processor the couplings are fixed at fabrication: a qubit talks only to its wired neighbours, and distant gates route through chains of swaps. On a Rydberg array the coupling graph is the physical arrangement of atoms, and an atom can be picked up by a moving tweezer and set down elsewhere mid-circuit. Connectivity is therefore reconfigurable through motion: bring any two qubits together to entangle them, then transport them to readout. This is what makes neutral atoms attractive for surface and qLDPC codes, which benefit from long-range, rearrangeable connectivity a static lattice cannot provide.

Can the same machine really both compute and simulate?

Yes, and that dual character is a main attraction. Digital, gate-based computing uses local addressing and the blockade CZ gate to run arbitrary circuits, with atoms shuttled between zones. Analog simulation uses a single global laser to evolve the array under the programmable Ising/PXP Hamiltonian, reading out many-body order directly. The hardware — tweezers, atoms, Rydberg lasers — is identical; only the control changes, and groups run both modes, sometimes on one apparatus.