Rydberg Blockade

A Rydberg atom is an ordinary atom with one valence electron promoted to a state of large principal quantum number $n$ , orbiting at a radius that swells as $n^2 a_0$ . Everything about such an atom is exaggerated: the polarizability, the transition dipole moments, and — decisively for quantum information — the interaction between two of them. Two ground-state alkali atoms a few microns apart barely feel each other; excite both to a Rydberg level and their van der Waals interaction $V(R)=C_6/R^6$ , with $C_6\propto n^{11}$ , can leap to hundreds of megahertz or more. The Rydberg blockade lives in that contrast: atoms are inert and storable in the ground state, yet acquire a huge, laser-switchable interaction the instant they are coupled toward $|r\rangle$ . The mechanism is starkly simple. Drive two atoms on the transition $|g\rangle\to|r\rangle$ with Rabi frequency $\Omega$ . Exciting one atom is resonant and cheap. Exciting both would land the pair in $|rr\rangle$ , whose energy is pushed up by $V(R)$ ; when $V(R)\gg\hbar\Omega$ that doubly-excited state is detuned far out of reach and simply cannot be populated. The first excitation blockades the second. What remains is an effective two-level system between $|gg\rangle$ and the symmetric singly-excited state $|W\rangle$ , driven at the collectively enhanced rate $\sqrt2\,\Omega$ — and $\sqrt N\,\Omega$ for $N$ atoms sharing one blockade volume, which then admit exactly one excitation among them. Proposed by Jaksch, Cirac, Zoller and collaborators in 2000 as the engine of neutral-atom logic and observed for single atoms in 2009, the blockade is now the beating heart of one of the leading routes to scalable quantum computing. It is strong and fast (gates in hundreds of nanoseconds), switchable, micron-ranged so atoms can be individually addressed, and native to large reconfigurable tweezer arrays. It is also more than a gate: its constraint — no two excitations within a blockade radius — is exactly the Rydberg-Ising model behind neutral-atom quantum simulation, complementary to the trapped-ion and superconducting platforms.

Key takeaways

The Rydberg blockade is the suppression of simultaneous laser excitation of two or more nearby atoms to a high-lying Rydberg state, caused by the strong interaction $V(R)$ shifting the doubly-excited level out of resonance.
Inside the blockade radius $R_b$ , where $V(R_b)\simeq\hbar\Omega$ , only one collective excitation is admitted; the symmetric bright state is driven at $\sqrt N\,\Omega$ for $N$ atoms, and at $\sqrt2\,\Omega$ for a pair.
Van der Waals scaling $V=C_6/R^6$ with $C_6\propto n^{11}$ makes the interaction enormous at micron range for $n\sim50$ – $100$ , giving blockade radii of several microns and switchable, conditional two-qubit interactions.
The blockade supplies the state-dependent interaction for entangling gates in neutral-atom processors — the interaction part of the DiVincenzo criteria — through the Jaksch π–2π–π sequence or modern single-pulse protocols.
Fidelity is bounded below by finite-blockade leakage $P_{rr}\sim(\Omega/V)^2$ , Rydberg decay $\Gamma t_{\mathrm{gate}}$ , laser noise, and atomic motion, giving an intrinsic error floor $\sim(\Gamma/V)^{2/3}$ .
Key idea: the first Rydberg excitation shifts the doubly-excited state by $V(R)$ ; when $V(R)\gg\hbar\Omega$ a second excitation is energetically forbidden, so atoms inside the blockade radius share exactly one collective excitation and are driven at $\sqrt N\,\Omega$ .

The core insight

The first Rydberg excitation shifts the energy of the doubly-excited state by the interaction $V(R)$ ; when $V(R)\gg\hbar\Omega$ a second excitation is energetically forbidden, so an ensemble inside the blockade radius shares exactly one collective excitation and is driven at $\sqrt N\,\Omega$ .

Where this lives today

The blockade is no longer a proposal — it is the physics inside a fast-growing class of quantum machines. Companies and academic groups build processors from hundreds to thousands of neutral atoms in optical-tweezer arrays, with entangling gates driven directly by the blockade and two-qubit fidelities above 99.5%. Because the interaction is laser-switched and the atoms are physically movable, connectivity is reconfigurable — atoms are shuttled together to interact and apart to idle — and the same platform has demonstrated dozens of logical qubits and real-time error correction. In a parallel line, the same constraint powers analog quantum simulation: a detuned array realizes the Rydberg-Ising/PXP model, whose rule forbidding two excitations within $R_b$ produces crystalline ordered phases, frustrated magnetism, and quantum many-body scars on hundreds of atoms. Ensemble blockade, meanwhile, gives deterministic single-photon sources and Rydberg-mediated photon–photon gates for networking. The live frontiers are pushing gate fidelity past $99.9\%$ against the Rydberg-lifetime floor, and wiring arrays together — through atom transport and photonic links — into fault-tolerant machines.

The physical picture

Picture two atoms as two switches you try to flip up — to $|r\rangle$ — with the same key, the laser. Flipping one is easy. But the first flipped switch instantly raises a wall in front of the second: the interaction energy $V$ . If that wall towers over the strength of your key turn $\hbar\Omega$ , the second switch cannot move. The pair therefore admits exactly one excitation, and because the laser cannot tell which atom should carry it, it drives the symmetric superposition of both — faster, by $\sqrt2$ . The first figure is that story as an energy ladder: the doubly-excited rung floats up out of resonance and the dynamics close on a clean, collectively-enhanced qubit.

Figure

Two-atom energy ladder under blockade: the drive couples the ground state gg to the symmetric bright state W at the enhanced rate √2 Ω, but the doubly-excited state rr is pushed up by the interaction V and, when V ≫ ħΩ, is detuned out of reach so it stays empty; the antisymmetric dark state D is decoupled by symmetry

The interaction is exquisitely tunable. Because $C_6\propto n^{11}$ , a modest change of principal quantum number swings $V$ by orders of magnitude, so one can set the blockade radius precisely between intended gate partners and unintended neighbours. Far from any resonance the interaction is van der Waals, $C_6/R^6$ ; tuned onto a Förster resonance it becomes resonant dipole–dipole, $C_3/R^3$ , longer-ranged and angle-dependent. The single number that decides whether two atoms are blockaded is the blockade radius $R_b$ , where the interaction crosses the drive energy — and the second figure plots that crossing for a concrete, laboratory-scale case.

Figure

The van der Waals interaction V(R) = C₆/R⁶ for a typical ⁸⁷Rb n ≈ 70 state with C₆/h ≈ 900 GHz·μm⁶, on a log axis, crossing the drive energy ħΩ for Ω = 2π × 2 MHz at the blockade radius R_b ≈ 8.8 μm; atoms closer than R_b are blockaded, and at a working spacing of 5 μm the shift V is about 29 ħΩ

The core equations

The whole phenomenon rests on how two Rydberg atoms interact. Far from any Förster resonance the leading term is a second-order (van der Waals) energy that falls as the sixth power of separation, with a coefficient that explodes with principal quantum number — this steep $n$ -dependence is what makes the interaction switchable across orders of magnitude.

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

where: $V(R)$ — interaction energy of two atoms both in the Rydberg state, separated by $R$ ; $C_6$ — van der Waals coefficient, steeply $n$ -dependent; $R$ — interatomic separation; $n$ — principal quantum number. Read it as: doubling $n$ multiplies the interaction by $2^{11}\approx2000$ , so a few steps in $n$ reposition the blockade entirely. The blockade radius is the single geometric number the platform runs on: the separation at which the interaction equals the drive energy. Closer than $R_b$ the pair is blockaded; farther, it is essentially independent.

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

where: $R_b$ — blockade radius, where $V(R_b)=\hbar\Omega$ ; $\Omega$ — single-atom Rabi frequency of the $|g\rangle\to|r\rangle$ drive. The sixth-root makes $R_b$ remarkably insensitive to $\Omega$ : a hundredfold change in drive strength moves the blockade radius by only about a factor of two. The full many-atom dynamics follow from a driven Hamiltonian with three pieces — coherent driving, laser detuning, and the pairwise Rydberg interaction that acts only when both atoms occupy $|r\rangle$ . This is also the Hamiltonian of neutral-atom quantum simulation.

H=\sum_{i}\frac{\hbar\Omega}{2}\big(|g\rangle_i\langle r|+|r\rangle_i\langle g|\big)-\hbar\Delta\sum_i|r\rangle_i\langle r|+\sum_{i<j}V_{ij}\,|r\rangle_i\langle r|\,|r\rangle_j\langle r| \qquad (3)

where: $\Omega$ — Rabi frequency; $\Delta$ — laser detuning from the $|g\rangle\to|r\rangle$ resonance; $V_{ij}$ — pairwise interaction between atoms $i,j$ ; $|g\rangle,|r\rangle$ — ground (qubit) and Rydberg states. The interaction term costs energy only for doubly-occupied Rydberg pairs — precisely the term the blockade suppresses. Under strong blockade the drive can promote only one shared excitation, and it does so into the symmetric bright state with a matrix element boosted by the square root of the atom number — the collective enhancement, reducing to $\sqrt2$ for a pair.

\Omega_{\mathrm{coll}}=\sqrt{N}\,\Omega,\qquad |gg\rangle\;\xleftrightarrow{\;\sqrt2\,\Omega\;}\;|W\rangle \qquad (4)

where: $\Omega_{\mathrm{coll}}$ — effective drive of the symmetric singly-excited bright state; $N$ — number of atoms inside one blockade volume; $|W\rangle$ — symmetric single-excitation state. The enhancement is a coherence effect: the $N$ ways to place one excitation add in amplitude, not probability. The blockade is never perfect. A finite $V/\Omega$ leaves a small off-resonant admixture of $|rr\rangle$ , and its residual population is the intrinsic error the mechanism cannot escape.

P_{rr}\simeq\frac{\Omega^2}{2V^2} \qquad (5)

where: $P_{rr}$ — probability of leakage into the doubly-excited state; $\Omega$ — drive; $V$ — interaction shift. The order-unity prefactor depends on pulse shape, but the scaling $(\Omega/V)^2$ is universal and sets the blockade-error floor. The conditional interaction becomes logic as a controlled-phase (CZ) gate: the pair accumulates a phase on the $|11\rangle$ input that no product of single-qubit phases can reproduce, the signature of a maximally entangling operation.

U_{\mathrm{CZ}}=\mathrm{diag}\big(1,\,e^{i\phi_{01}},\,e^{i\phi_{10}},\,e^{i\phi_{11}}\big),\qquad \phi_{11}-\phi_{01}-\phi_{10}=\pi \qquad (6)

where: $\phi_{ij}$ — accumulated phase on the computational basis state $|ij\rangle$ . The nonseparable combination $\phi_{11}-\phi_{01}-\phi_{10}=\pi$ is exactly what the blockade supplies and single-qubit gates cannot. Near a Förster resonance the interaction changes character from van der Waals to resonant dipole–dipole: it falls only as $1/R^3$ , reaches farther, and depends on the angle to the quantization axis.

V_{\mathrm{dd}}(R,\theta)=\frac{C_3}{R^3}\,(1-3\cos^2\theta),\qquad C_3\propto n^4 \qquad (7)

where: $C_3$ — dipole–dipole coefficient; $\theta$ — angle between the interatomic axis and the quantization axis. The factor $1-3\cos^2\theta$ vanishes at the magic angle $\theta\approx54.7^\circ$ , where the resonant interaction (and the blockade) can be switched off geometrically. Finally, the resource that caps everything is the Rydberg lifetime. Radiative decay and black-body-driven transfer to nearby levels scale differently with $n$ , and their sum sets the time budget for any gate.

\tau_{\mathrm{eff}}^{-1}=\Gamma_0+\Gamma_{\mathrm{bb}},\qquad \Gamma_0\propto n^{-3},\quad \Gamma_{\mathrm{bb}}\propto n^{-2} \qquad (8)

where: $\tau_{\mathrm{eff}}$ — effective Rydberg lifetime; $\Gamma_0$ — spontaneous (radiative) rate; $\Gamma_{\mathrm{bb}}$ — black-body-induced rate. Higher $n$ lengthens the radiative lifetime but black-body decay falls more slowly, so at room temperature the effective lifetime saturates at tens to hundreds of microseconds.

Derivation ABlockade reduces two atoms to an effective two-level system

The goal is to prove that a strong interaction on the doubly-excited state collapses the four-dimensional two-atom Hilbert space onto a clean qubit — the transition $|gg\rangle\leftrightarrow|W\rangle$ driven at $\sqrt2\,\Omega$ — while $|rr\rangle$ is frozen out and the antisymmetric state $|D\rangle$ never participates. We build the symmetric-state basis, compute the drive matrix elements, locate the interaction shift, and adiabatically eliminate the far-detuned rung, so that the workhorse of the entangling gate falls out as the $N=2$ instance of collective enhancement.

Step 1. Work in the symmetric/antisymmetric basis $\{|gg\rangle,\ |W\rangle=\tfrac{1}{\sqrt2}(|gr\rangle+|rg\rangle),\ |D\rangle=\tfrac{1}{\sqrt2}(|gr\rangle-|rg\rangle),\ |rr\rangle\}$ for two atoms driven identically on resonance ( $\Delta=0$ ). Why: a symmetric drive can only connect symmetric states, so $|D\rangle$ is decoupled by symmetry and the problem lives in the three-state ladder $|gg\rangle$ – $|W\rangle$ – $|rr\rangle$ .

Step 2. Compute the drive matrix elements. Acting with $H_{\mathrm{drive}}=\tfrac{\hbar\Omega}{2}\sum_i(\sigma_+^{(i)}+\sigma_-^{(i)})$ gives $\langle W|H|gg\rangle=\tfrac{\hbar}{2}\sqrt2\,\Omega$ and $\langle rr|H|W\rangle=\tfrac{\hbar}{2}\sqrt2\,\Omega$ : both links inherit the $\sqrt2$ because two atoms can each supply the excitation. Why: this $\sqrt2$ is the two-atom collective enhancement, the same coherent amplitude addition that becomes $\sqrt N$ for larger ensembles.

Step 3. Add the interaction. Only $|rr\rangle$ carries it, $\langle rr|H|rr\rangle=V$ , while $|gg\rangle$ and $|W\rangle$ sit at zero energy. So the ladder is two resonantly-coupled rungs plus a third rung floated up by $V$ . Why: the interaction acts as a state-selective detuning that singles out $|rr\rangle$ for suppression without touching the qubit subspace.

Step 4. Adiabatically eliminate $|rr\rangle$ . For $V\gg\hbar\Omega$ the top rung is populated only to order $(\Omega/V)^2$ and can be integrated out, leaving $H_{\mathrm{eff}}=\tfrac{\hbar\sqrt2\,\Omega}{2}\big(|gg\rangle\langle W|+\mathrm{h.c.}\big)$ — a two-level system oscillating at $\sqrt2\,\Omega$ . Why: this is the blockade made precise; the residual $(\Omega/V)^2$ is exactly the leakage $P_{rr}$ that limits gate fidelity. *Two quick checks: setting $N=2$ in $\sqrt N\,\Omega$ reproduces $\sqrt2\,\Omega$ , as expected for a pair; taking $V\to0$ restores the full uncoupled three-rung ladder, confirming the reduction to a two-level system is a strong-blockade limit, not an exact identity.*

Go deeper · Derivation B · 1 minThe Jaksch π–2π–π controlled-phase gate

Derivation BThe Jaksch π–2π–π controlled-phase gate

The goal is to construct a controlled- $Z$ from three pulses using nothing but the blockade, and to show that only the $|11\rangle$ input escapes a $-1$ phase — the nonseparable difference that makes the gate entangling. The qubit is encoded in $|0\rangle$ (dark, uncoupled to $|r\rangle$ ) and $|1\rangle$ (bright, coupled to $|r\rangle$ by a resonant $\pi$ pulse). Pulses hit control, then target, then control, and we track each computational input through the sequence.

Step 1. Apply a $\pi$ pulse on the control's $|1\rangle\to|r\rangle$ transition, $|1\rangle_c\to -i|r\rangle_c$ (a control in $|0\rangle$ is untouched). Why: this conditionally parks a Rydberg excitation on the control — the presence or absence of which will switch the blockade on the target.

Step 2. Apply a full $2\pi$ pulse on the target with the control dark ( $|0\rangle$ ). With no blockade the target completes $|1\rangle_t\to|r\rangle_t\to|1\rangle_t$ and picks up the $-1$ phase of a $2\pi$ rotation: $|01\rangle\to-|01\rangle$ , $|00\rangle\to|00\rangle$ . Why: an unobstructed $2\pi$ Rabi cycle is a $2\pi$ rotation of a spin-½, whose universal signature is a sign flip.

Step 3. Repeat the target $2\pi$ pulse but with the control bright ( $|1\rangle$ , now in $|r\rangle$ ). The control's Rydberg excitation shifts the target's $|r\rangle$ by $V$ , blockading it: the $2\pi$ pulse is off-resonant, does essentially nothing, and no $-1$ is acquired, $|1\rangle_t\to+|1\rangle_t$ . Why: this is the conditional step — the same target pulse acts completely differently depending on the control, which is the whole content of a controlled gate.

Step 4. Apply the final $\pi$ pulse to the control, $|r\rangle_c\to -i|1\rangle_c$ , returning it to the computational space; the two control $\pi$ pulses compose to a common sign absorbed into single-qubit corrections. Collecting phases gives $|00\rangle\to|00\rangle$ , $|01\rangle\to-|01\rangle$ , $|10\rangle\to-|10\rangle$ , $|11\rangle\to+|11\rangle$ — a CZ up to single-qubit $Z$ rotations. Why: only $|11\rangle$ is spared the $-1$ because only there does the blockade intervene, producing the nonseparable phase pattern of an entangling gate. *The resulting truth table satisfies $\phi_{11}-\phi_{01}-\phi_{10}+\phi_{00}=\pi$ (mod $2\pi$ ), the defining condition of a maximally entangling controlled-phase gate.*

Go deeper · Derivation C · 1 minThe blockade-error floor and the optimal drive

Derivation CThe blockade-error floor and the optimal drive

The goal is to find how small the intrinsic gate error can be made, by balancing the two errors that pull in opposite directions: finite-blockade leakage, which wants a slow, weak drive, and Rydberg decay, which wants a fast one. We write each error in terms of the drive $\Omega$ , minimize their sum, and read off the optimal drive and the resulting floor. The answer, $(\Gamma/V)^{2/3}$ , is the design rule of every high-fidelity Rydberg gate: push the interaction $V$ up and the decay rate $\Gamma$ down.

Step 1. The leakage error is the residual double-excitation, $E_{\mathrm{leak}}\sim(\Omega/V)^2$ , from the off-resonant $|rr\rangle$ admixture of equation (5). Why: this is smaller for a weaker drive, because a gentle drive is more easily detuned out of the blockaded state.

Step 2. The decay error is the chance the atom spontaneously leaves $|r\rangle$ during the gate, $E_{\mathrm{decay}}\sim\Gamma\,t_{\mathrm{gate}}\sim\Gamma/\Omega$ , since the gate time scales as $1/\Omega$ . Why: this is smaller for a stronger drive, because a fast gate spends less time in the fragile Rydberg state — exactly opposite to the leakage.

Step 3. Add them, $E(\Omega)\sim\Omega^2/V^2+\Gamma/\Omega$ , and minimize: $\mathrm{d}E/\mathrm{d}\Omega=2\Omega/V^2-\Gamma/\Omega^2=0$ gives $\Omega_{\mathrm{opt}}\sim(\Gamma V^2)^{1/3}$ . Why: the optimum sits where the two competing errors are comparable, the generic signature of a well-balanced trade-off.

Step 4. Substitute back: $E_{\min}\sim(\Gamma/V)^{2/3}$ . Why: the floor depends only on the ratio of decay rate to interaction, so improving the gate is a race to make $V/\Gamma$ large — high $n$ raises $V$ , and cold, low-field, low-black-body conditions lower $\Gamma$ . *Two consistency checks: $E_{\min}$ is dimensionless, as it should be, since $\Gamma$ and $V/\hbar$ are both rates; and the required ratio $V/\Gamma\sim10^3$ – $10^4$ for errors at the $10^{-3}$ level is consistent with the Saffman–Walker–Mølmer error-floor analysis.*

Limiting cases

The blockade sorts into clean regimes set by the ratio of interaction to drive and by the distance law of $V(R)$ .

Case	Result	Meaning
Perfect blockade	$V\gg\hbar\Omega$	$\vert rr\rangle$ is fully detuned; the pair is an ideal two-level system $\{\vert gg\rangle,\vert W\rangle\}$ driven at $\sqrt2\,\Omega$ , with double-excitation probability $\to0$ .
Finite-blockade leakage	$P_{rr}\sim(\Omega/V)^2$	Residual $\vert rr\rangle$ population caps fidelity; high-fidelity logic needs $V/\Omega\gtrsim10$ – $100$ (here $\approx29$ at $5\,\mu$ m).
Anti-blockade / facilitation	$\Delta=V(R)$	Detune the laser by exactly the interaction shift and the double excitation becomes resonant — used deliberately for facilitated excitation and dissipative state preparation.
van der Waals vs dipole–dipole	$C_6/R^6$ vs $C_3/R^3$	Far from a Förster resonance the interaction is $1/R^6$ ( $C_6\propto n^{11}$ ); on resonance it is $1/R^3$ ( $C_3\propto n^4$ ), longer-ranged and angle-dependent.
Collective / mesoscopic	$\Omega_{\mathrm{coll}}=\sqrt N\,\Omega$	An ensemble in one blockade volume admits a single shared excitation driven at $\sqrt N\,\Omega$ — a superatom, the basis of ensemble memories and single-photon sources.

Common trapThe blockade is a hard wall

It is tempting to treat the blockade as an on/off switch that perfectly forbids double excitation, but it is a finite energy barrier, not an infinite one. Leakage into $|rr\rangle$ scales as $(\Omega/V)^2$ , so any real $V/\Omega$ leaves a residual double-excitation probability that sets a floor on gate fidelity — the reason blockade gates are engineered with $V/\Omega\sim10$ – $100$ , not merely $V>\Omega$ . A second, subtler trap: the collective $\sqrt N$ enhancement belongs only to the symmetric bright state $|W\rangle$ . Positional disorder, differential light shifts, and decoherence leak population into the $N-1$ dark collective states, which the laser does not drive and which quietly degrade both the enhancement and the qubit. "Perfect blockade" and "collective enhancement" are idealizations that a fidelity budget must treat as approximate.

Worked examples

1 · Blockade radius and leakage floor for a rubidium array (fully worked)

Two ⁸⁷Rb atoms are driven to a Rydberg state near $n\approx70$ , for which $C_6/h\approx900\ \mathrm{GHz\,\mu m^6}$ , with Rabi frequency $\Omega=2\pi\times2\ \mathrm{MHz}$ . Find the blockade radius, and the residual double-excitation probability when the atoms are placed at a typical spacing of $5\ \mu$ m.

Step 1. Blockade radius. The condition $V(R_b)=\hbar\Omega$ reads $(C_6/h)/R_b^6=\Omega/2\pi$ , so

R_b=\left(\frac{C_6/h}{\Omega/2\pi}\right)^{1/6}=\left(\frac{900\ \mathrm{GHz\,\mu m^6}}{0.002\ \mathrm{GHz}}\right)^{1/6}=\big(4.5\times10^{5}\big)^{1/6}\ \mu\mathrm{m}=\mathbf{8.8\ \mu m}.

The sixth-root at work: even this enormous $C_6$ gives a blockade radius of only a few microns.

Step 2. Interaction at the working spacing. At $R=5\ \mu$ m,

\frac{V}{h}=\frac{900\ \mathrm{GHz\,\mu m^6}}{(5\ \mu\mathrm{m})^6}=\frac{900{,}000\ \mathrm{MHz\,\mu m^6}}{15{,}625\ \mu\mathrm{m^6}}=\mathbf{57.6\ MHz},

so $V/\hbar\Omega=57.6/2=\mathbf{28.8}$ — the pair sits comfortably inside the blockade radius ( $5\ \mu\mathrm{m}<8.8\ \mu\mathrm{m}$ ), deep in the strong-blockade regime.

Step 3. Leakage floor. From equation (5),

P_{rr}\simeq\frac{1}{2}\left(\frac{\Omega}{V}\right)^2=\frac{1}{2\,(28.8)^2}=\mathbf{6.0\times10^{-4}}.

The blockade alone therefore limits the two-qubit gate to an error of order $10^{-3}$ — already at the threshold-relevant level, before decay and technical noise are added. R_b ≈ 8.8 μm, V ≈ 29 ħΩ at 5 μm, P_rr ≈ 6 × 10⁻⁴.

2 · Gate time and the decay-versus-leakage budget

For the same parameters, estimate the entangling-gate time and compare the leakage error against Rydberg decay, taking an effective lifetime $\tau_{\mathrm{eff}}\approx100\ \mu$ s.

Step 1. Gate time. The pair is driven at the collective rate $\sqrt2\,\Omega=2\pi\times2.83\ \mathrm{MHz}$ ; a population-transfer $\pi$ pulse on $|gg\rangle\leftrightarrow|W\rangle$ takes $t_\pi=\pi/(\sqrt2\,\Omega)=1/(2\sqrt2\times2\ \mathrm{MHz})\approx\mathbf{0.18\ \mu s}$ . A full CZ, a few such pulses, completes in well under a microsecond — hundreds of nanoseconds, as advertised.

Step 2. Decay error. With $\Gamma=1/\tau_{\mathrm{eff}}=10^4\ \mathrm{s^{-1}}$ and roughly $1\ \mu$ s of accumulated Rydberg population over the gate, the spontaneous-decay probability is $E_{\mathrm{decay}}\sim\Gamma\,t\approx10^4\times10^{-6}=\mathbf{10^{-2}}$ . Decay, not blockade leakage ( $6\times10^{-4}$ ), is the dominant intrinsic error here — the practical message of the platform.

Step 3. Intrinsic floor. The balanced optimum of Derivation 3 gives $E_{\min}\sim(\Gamma/V)^{2/3}$ ; with $\Gamma=10^4\ \mathrm{s^{-1}}$ and $V/\hbar=2\pi\times57.6\ \mathrm{MHz}=3.6\times10^{8}\ \mathrm{s^{-1}}$ , $E_{\min}\sim(2.8\times10^{-5})^{2/3}\approx\mathbf{9\times10^{-4}}$ — an error floor near $10^{-3}$ , consistent with demonstrated two-qubit fidelities above 0.99. t_gate ≈ 0.2 μs, decay error ≈ 1%, intrinsic floor ≈ 10⁻³.

Physics of the portrait

The Rydberg blockade is the work of many hands, but one physicist turned it from a two-atom gate proposal into a named mechanism and a scalable technology for both computing and simulation.

Origins · 1 minMikhail Lukin and how this was discovered

Mikhail Lukin

1971– · Harvard, co-director Harvard–MIT CUA · dipole blockade, Rydberg quantum optics, neutral-atom arrays

Mikhail Lukin has done more than almost anyone to make the Rydberg blockade a working tool of quantum science. In 2001, with collaborators, he extended the two-atom blockade of Jaksch, Cirac and Zoller to mesoscopic ensembles — the dipole blockade — showing that a whole cloud inside one blockade volume behaves as a single "superatom" admitting one collective excitation driven at $\sqrt N\,\Omega$ , the idea behind ensemble memories and single-photon sources. His group went on to realize Rydberg quantum optics, in which single photons interact through atomic blockade, and, from around 2016, to build programmable optical-tweezer arrays of individually controlled atoms. Those arrays produced landmark quantum simulations — a 51-atom chain revealing symmetry-broken ordered phases and long-lived "quantum scars" in 2017, later two-dimensional simulators of frustrated magnetism — and, more recently, high-fidelity gates and demonstrations of logical qubits and error correction on neutral-atom hardware. The through-line of his career is the conviction, borne out repeatedly, that the same strong, switchable interaction can serve equally as a gate, a photonic nonlinearity, and a many-body constraint. The modern picture of the blockade — a resource that scales from two atoms to hundreds — is very much the field he built.

History

1888

— Rydberg's spectral formula names the highly excited atomic states

2000

— Jaksch, Cirac, Zoller and collaborators propose the blockade entangling gate for neutral atoms

2001

— Lukin and collaborators generalize the blockade to mesoscopic ensembles (the dipole blockade)

2009

— the blockade and √2 collective enhancement are observed for two individual atoms

2023

— optical-tweezer blockade gates surpass 99.5% two-qubit fidelity

Going further

Texts and papers. Jaksch, Cirac, Zoller, Côté, Lukin, "Fast Quantum Gates for Neutral Atoms," Phys. Rev. Lett. 85, 2208 (2000) · Lukin, Fleischhauer, Côté, Duan, Jaksch, Cirac, Zoller, "Dipole Blockade and Quantum Information Processing in Mesoscopic Atomic Ensembles," PRL 87, 037901 (2001) · Saffman, Walker, Mølmer, "Quantum information with Rydberg atoms," Rev. Mod. Phys. 82, 2313 (2010) · Levine et al., "Parallel Implementation of High-Fidelity Multiqubit Gates with Neutral Atoms," PRL 123, 170503 (2019) · Browaeys & Lahaye, "Many-body physics with individually controlled Rydberg atoms," Nat. Phys. 16, 132 (2020) · Gallagher, Rydberg Atoms, Cambridge (1994) · Nielsen & Chuang, Quantum Computation and Quantum Information, Ch. 7. The key scalings are $C_6\propto n^{11}$ , $C_3\propto n^4$ , and $R_b=(C_6/\hbar\Omega)^{1/6}$ ; the $\sqrt N$ enhancement follows from direct matrix-element computation in the $\{|gg\rangle,|W\rangle,|rr\rangle\}$ basis; the $(\Omega/V)^2$ leakage and $(\Gamma/V)^{2/3}$ error floor come from the balanced error-minimisation of Derivation 3. The anchor numbers — with $C_6/h=900\ \mathrm{GHz\,\mu m^6}$ and $\Omega=2\pi\times2$ MHz — are $R_b=8.8\ \mu$ m, $V/h=57.6$ MHz and $V/\hbar\Omega=28.8$ at $5\ \mu$ m, $P_{rr}=6.0\times10^{-4}$ , $t_\pi=0.18\ \mu$ s, and $E_{\min}\approx9\times10^{-4}$ . The $C_6$ value is representative of an $n\approx70$ ⁸⁷Rb state; quoted fidelities (above 0.99, above 0.995) are standard published values.

Builds on: DiVincenzo Criteria · Mølmer–Sørensen Gate · Sideband Cooling. Leads to: Trapped-Ion Qubits · Cirac–Zoller Gate.

Ask the tutor: Want to walk through why the π–2π–π sequence flips only the |11⟩ phase, or how the √N superatom turns a whole atomic cloud into a single-photon source?

Is the Rydberg blockade an electrostatic or a quantum-mechanical effect?

Both, and cleanly so. The underlying force is electrostatic: two Rydberg atoms have enormous dipoles, and their van der Waals (or resonant dipole–dipole) interaction shifts the energy of the doubly-excited state $|rr\rangle$ by $V(R)$ . But the blockade — the fact that a second excitation is forbidden — is quantum-mechanical: it is an energy-conservation statement about a coherently driven transition. When the interaction shift $V$ far exceeds the drive coupling $\hbar\Omega$ , the laser cannot bridge the gap to $|rr\rangle$ on resonance, so the amplitude to double-excite is suppressed to order $(\Omega/V)^2$ . The classical interaction sets the barrier; the quantum drive's inability to climb it is the blockade.

Why does the collective Rabi frequency scale as √N and not N?

Because amplitudes add, not probabilities. Under blockade the laser couples the all-ground state $|G\rangle=|g\cdots g\rangle$ only to the symmetric single-excitation state $|W\rangle$ , which is a superposition of the $N$ ways to place one excitation. Acting with the symmetric drive on $|G\rangle$ produces a sum of $N$ equal terms; that sum equals $\sqrt N$ times the normalized $|W\rangle$ , because normalizing an equal superposition of $N$ terms divides by $\sqrt N$ . So the effective coupling is $\sqrt N\,\Omega$ . If probabilities had added you would get $N$ , but quantum mechanics adds the amplitudes coherently, and the geometric mean of " $N$ terms" and "norm $\sqrt N$ " leaves the $\sqrt N$ enhancement — the same collective factor seen in superradiance.

How is the Rydberg blockade different from the Coulomb interaction in trapped ions?

Both use a strong interaction to make qubits talk, but the mechanisms are opposite in character. Trapped ions carry a net charge and interact through the always-on Coulomb force, mediated by shared quantized motion (phonons). Neutral Rydberg atoms have no net charge and essentially no interaction in the ground state — the coupling is switched on only while the atoms are laser-excited to $|r\rangle$ , and it is a direct state-dependent energy shift, not a phonon exchange. The blockade needs no shared vibrational bus and no ground-state cooling of motion, which is why neutral-atom arrays scale by simply adding tweezers, at the cost of the finite Rydberg lifetime that ions do not suffer.

What exactly sets the blockade radius, and why is it so insensitive to the laser?

The blockade radius $R_b$ is the separation at which the interaction equals the drive energy, $V(R_b)=\hbar\Omega$ , giving $R_b=(C_6/\hbar\Omega)^{1/6}$ . The striking feature is the sixth-root: because $V$ falls as $1/R^6$ , the crossing point moves very slowly with the drive. Increasing $\Omega$ by a factor of 100 shrinks $R_b$ by only $100^{1/6}\approx2.15$ . In practice this means $R_b$ is set almost entirely by the atomic physics — the principal quantum number $n$ through $C_6\propto n^{11}$ — and only weakly by how hard you drive. To move the blockade radius substantially you change the Rydberg state, not the laser intensity.

What is anti-blockade, and why would anyone want it?

Anti-blockade, or facilitation, is the deliberate opposite of the blockade. Instead of driving on bare resonance, you detune the laser by exactly the interaction shift, $\Delta=V(R)$ . The doubly-excited state, which the blockade normally pushes out of resonance, is then brought back onto resonance — but only for atoms at the right separation — so a first excitation facilitates rather than forbids a nearby second one, seeding avalanches along a chain. This is a resource, not a bug: facilitation dynamics probe non-equilibrium phase transitions, epidemic-like spreading, and dissipative many-body physics, and engineer correlated states the plain blockade cannot produce.

If the Rydberg state decays, how can these gates ever be high-fidelity?

The trick is to spend as little time as possible in $|r\rangle$ . A Rydberg state lives only tens to hundreds of microseconds, but a blockade gate completes in hundreds of nanoseconds, so the population spends perhaps a microsecond in the fragile state — a decay probability of order $10^{-2}$ , which modern echoed and time-optimal pulse designs push lower by minimizing the integrated Rydberg population. The intrinsic floor, from balancing leakage against decay, is $(\Gamma/V)^{2/3}$ : making the interaction $V$ large (high $n$ ) while keeping the decay rate $\Gamma$ small (cold atoms, low stray fields, controlled black-body environment) drives the error toward $10^{-3}$ and below. Demonstrated fidelities above 0.995 show the balance is winnable, though beating $99.9\%$ against the lifetime remains the central challenge.

Does the blockade only work for two atoms, or for many at once?

It works for any number of atoms sharing one blockade volume. If $N$ atoms all sit within $R_b$ of each other, the blockade forbids any second excitation, so the whole group admits one shared excitation and behaves as a single "superatom" driven at $\sqrt N\,\Omega$ — the basis of ensemble memories and single-photon sources. Spread the atoms so each has only some neighbours within $R_b$ , and the same constraint becomes a many-body interaction — no two excitations closer than $R_b$ — which is exactly the Rydberg-Ising model behind neutral-atom quantum simulation. The two-atom gate is just the smallest instance of a broadly scalable effect.