Simple Algorithm
Quantum Computing · Rydberg State · 28 min

The Rydberg Blockade

Giant atoms that forbid their neighbors from exciting.

A rubidium atom excited to n≈70n \approx 70 is a few tenths of a micrometer across, and it interacts with a similar neighbor more strongly than ground-state atoms do by many orders of magnitude. Within several micrometers, one excited atom shifts every other one out of resonance. This blockade is the entangling mechanism of neutral-atom quantum computers.

What you’ll learn

  • Derive the scaling laws of Rydberg atoms and the origin of the C6/r6C_6/r^6 interaction in dipole–dipole coupling.
  • State and prove the blockade theorem for NN atoms, including the N\sqrt N collective enhancement.
  • Prove when a diagonal two-qubit phase gate equals CZ up to local rotations, and apply it to Jaksch-type and global-pulse protocols.
  • Derive the intrinsic error budget of a blockade gate and its optimum, Emin⁡∝(Bτ)−2/3E_{\min}\propto(B\tau)^{-2/3}.

Rydberg atoms by the numbers

Rydberg states are hydrogen-like, with an effective principal quantum number n∗=n−δℓn^* = n - \delta_\ell, where the quantum defect δℓ\delta_\ell accounts for the penetrable ionic core. Their properties follow hydrogenic scalings.

PropertyScalingConsequence
Binding energy∝n∗−2\propto n^{*-2}Weakly bound and easily perturbed
Orbital radius∝n∗2\propto n^{*2}A few tenths of a μm at n≈70n \approx 70
Dipole moment to neighboring levels∝n∗2\propto n^{*2}Huge transition dipoles between Rydberg states
Spacing of neighboring levels∝n∗−3\propto n^{*-3}Microwave-frequency transitions
Radiative lifetime∝n∗3\propto n^{*3} at zero temperatureOf order 100 μs at room temperature for n≈70n\approx70, where blackbody radiation shortens it
Polarizability∝n∗7\propto n^{*7}Strong sensitivity to stray electric fields
Van der Waals C6C_6∝n∗11\propto n^{*11}MHz-scale interactions at several μm
Standard scalings; see T. F. Gallagher, Rydberg Atoms (Cambridge, 1994), and Saffman, Walker & Mølmer (2010).
DerivationWhere C₆/r⁶ comes from
  1. Vdd=14πϵ0r3[d1⋅d2−3(d1⋅r^)(d2⋅r^)]V_{\rm dd} = \frac{1}{4\pi\epsilon_0 r^3}\Big[\mathbf d_1\cdot\mathbf d_2 - 3(\mathbf d_1\cdot\hat{\mathbf r})(\mathbf d_2\cdot\hat{\mathbf r})\Big]

    Two atoms in the same Rydberg state ∣rr⟩\ket{rr} couple through the dipole–dipole interaction to pair states ∣r′r′′⟩\ket{r'r''} of opposite parity.

  2. Hpair=(0C3/r3C3/r3ΔF)H_{\rm pair} = \begin{pmatrix}0 & C_3/r^3\\ C_3/r^3 & \Delta_F\end{pmatrix}

    Keep the most strongly coupled pair state; ΔF\Delta_F is its energy offset (the Förster defect), and C3∝d2∝n∗4C_3 \propto d^2 \propto n^{*4}.

  3. E±=ΔF2±ΔF24+C32r6E_\pm = \frac{\Delta_F}{2} \pm \sqrt{\frac{\Delta_F^2}{4} + \frac{C_3^2}{r^6}}

    Exact eigenvalues of the two-level pair model.

  4. r≫Rc=∣C3/ΔF∣1/3:Err≈−C32ΔF r6≡C6r6,C6=−C32ΔF∝n∗8n∗−3=n∗11r \gg R_c = \abs{C_3/\Delta_F}^{1/3}:\qquad E_{rr} \approx -\frac{C_3^2}{\Delta_F\,r^6} \equiv \frac{C_6}{r^6},\qquad C_6 = -\frac{C_3^2}{\Delta_F} \propto \frac{n^{*8}}{n^{*-3}} = n^{*11}

    At long range the shift of the branch connected to ∣rr⟩\ket{rr} is second order: van der Waals, with a sign set by ΔF\Delta_F. At short range, r≪Rcr \ll R_c, the states mix fully and the interaction becomes resonant dipole–dipole, ±C3/r3\pm C_3/r^3. Electric fields can tune ΔF\Delta_F to zero (a Förster resonance), giving 1/r31/r^3 interactions at all distances.

The blockade theorem

The 2\sqrt2 speed-up for two atoms was observed by Gaëtan et al. (2009) and is now a standard calibration of the blockade. A blockaded ensemble behaves as one "superatom", the basis of Rydberg quantum optics. For computing, the key consequence is conditional dynamics: whether an atom can be excited depends on whether its neighbor already is.

From blockade to a CZ gate

Jaksch et al. (2000). A π-pulse on the control, a 2π2\pi-pulse on the target and another π-pulse on the control give φ01=φ10=φ11=π\varphi_{01} = \varphi_{10} = \varphi_{11} = \pi, so the invariant is −π≡π-\pi \equiv \pi: a CZ. The blockade enters only through φ11\varphi_{11}, since the target's 2π2\pi-pulse is suppressed while the control sits in ∣r⟩\ket r. Two features make the scheme robust: the exact value of VV is irrelevant as long as V≫ℏΩV \gg \hbar\Omega, and the atoms need not sit at precisely defined distances.

Global protocols. Levine et al. (2019) drove both atoms with the same two pulses, separated by a jump of the laser phase. The inputs ∣01⟩\ket{01} and ∣10⟩\ket{10} then undergo single-atom dynamics at Ω\Omega, while ∣11⟩\ket{11} undergoes blockaded dynamics between ∣11⟩\ket{11} and (∣1r⟩+∣r1⟩)/2(\ket{1r}+\ket{r1})/\sqrt2 at 2 Ω\sqrt2\,\Omega. Pulse length, detuning and phase jump are chosen so that both return to the qubit subspace with φ11=2φ01+π\varphi_{11} = 2\varphi_{01} + \pi: a CZ up to single-qubit phases, by the theorem. Evered et al. (2023) used smooth, time-optimal phase-modulated pulses (Jandura & Pupillo, 2022) and reported CZ fidelities of about 99.5%, applied in parallel to up to 60 atoms.

Check your understanding

A calibrated Rydberg gate produces the phases φ01=φ10=0.3π\varphi_{01} = \varphi_{10} = 0.3\pi and φ11=1.6π\varphi_{11} = 1.6\pi. Up to single-qubit ZZ rotations, it is

  1. the identity: no entanglement
  2. a controlled phase of 0.6π0.6\pi
  3. a CZ
  4. a controlled phase of 1.6π1.6\pi
Show the answer

C. a CZ

The invariant is 1.6π−0.3π−0.3π=π1.6\pi - 0.3\pi - 0.3\pi = \pi, so the gate is a CZ dressed with local ZZ rotations, exactly the structure of the global-pulse protocols.

The error budget of a blockade gate

Even with perfect lasers, a blockade gate has two intrinsic errors that pull in opposite directions. Time spent in ∣r⟩\ket r risks decay (lifetime τ\tau), which argues for a strong, fast drive; but a strong drive leaks into the doubly excited state, shifted by the blockade frequency B=V/ℏB = V/\hbar, which argues for a weak one. Saffman and Walker (2005) analyzed this trade-off for the Jaksch protocol; here are its two leading terms.

DerivationOptimizing the Rabi frequency
  1. tˉr=14(0+πΩ+3πΩ+3πΩ)=7π4Ω\bar t_r = \frac14\Big(0 + \frac{\pi}{\Omega} + \frac{3\pi}{\Omega} + \frac{3\pi}{\Omega}\Big) = \frac{7\pi}{4\Omega}

    Average time spent in ∣r⟩\ket r over the four inputs. ∣00⟩\ket{00}: never. ∣01⟩\ket{01}: half of a 2π2\pi-pulse, π/Ω\pi/\Omega. ∣10⟩\ket{10} and ∣11⟩\ket{11}: half of each control π-pulse (π/2Ω\pi/2\Omega twice) plus the full 2π/Ω2\pi/\Omega in between, 3π/Ω3\pi/\Omega.

  2. Edecay≈tˉrτ=7π4ΩτE_{\rm decay} \approx \frac{\bar t_r}{\tau} = \frac{7\pi}{4\Omega\tau}

    Decay probability, to first order.

  3. Eblock≈Ω28B2E_{\rm block} \approx \frac{\Omega^2}{8B^2}

    During the target's 2π2\pi-pulse with the control in ∣r⟩\ket r, off-resonant excitation of ∣rr⟩\ket{rr} (detuning BB) leaves a time-averaged population of about Ω2/2B2\Omega^2/2B^2, and only one of the four inputs is affected. (A further term, from off-resonant excitation of the other qubit level, depends on the qubit frequency and is omitted.)

  4. E(Ω)=7π4Ωτ+Ω28B2  ⇒  Ω∗=(7πB2τ)1/3E(\Omega) = \frac{7\pi}{4\Omega\tau} + \frac{\Omega^2}{8B^2}\;\Rightarrow\;\Omega_* = \Big(\frac{7\pi B^2}{\tau}\Big)^{1/3}

    Minimize over Ω\Omega by setting dE/dΩ=0dE/d\Omega = 0.

  5. Emin⁡=21π8 Ω∗τ=21π8 (7π)1/3 (Bτ)−2/3≈2.9 (Bτ)−2/3E_{\min} = \frac{21\pi}{8\,\Omega_*\tau} = \frac{21\pi}{8\,(7\pi)^{1/3}}\,(B\tau)^{-2/3} \approx 2.9\,(B\tau)^{-2/3}

    At the optimum the decay term is twice the blockade term. The intrinsic error falls only as the 2/32/3 power of the figure of merit BτB\tau.

Predict before you look

With a blockade shift B/2π=50B/2\pi = 50 MHz and a Rydberg lifetime τ=100\tau = 100 μs, estimate the minimum intrinsic error Emin⁡≈2.9 (Bτ)−2/3E_{\min}\approx2.9\,(B\tau)^{-2/3} of a Jaksch gate, in percent.

Show the answer

0.3 %

Bτ=2π×50×106×10−4≈3.1×104B\tau = 2\pi\times50\times10^6\times10^{-4} \approx 3.1\times10^4 and (3.1×104)2/3≈1.0×103(3.1\times10^4)^{2/3} \approx 1.0\times10^3, so Emin⁡≈3×10−3E_{\min} \approx 3\times10^{-3}, about 0.3%, at an optimal Ω∗/2π≈4.4\Omega_*/2\pi \approx 4.4 MHz. An intrinsic floor at the 10−310^{-3} level is why reaching 99.9% needs better protocols as well as longer-lived, more strongly interacting states.

Technical errorMechanismMitigation
Doppler dephasingThermal velocity vv gives the ground–Rydberg coherence a random phase keffvtk_{\rm eff}vt; for two-photon excitation at ~10 μK, T2∗T_2^* is a few μsColder atoms; counter-propagating beams to minimize keffk_{\rm eff}; short gates
Laser phase noiseNoise at Fourier frequencies near Ω\Omega drives errors directlySpectral filtering of the excitation lasers with optical cavities
Intermediate-state scatteringTwo-photon excitation through a short-lived PP state scatters photonsLarge intermediate detuning, at the cost of laser power
Intensity and position fluctuationsFluctuating Ω\Omega changes rotation angles; positions matter little deep in the blockade regimeBeam shaping and stabilization; colder atoms
LeakageDecay or blackbody transfer to other Rydberg states, or to levels outside the qubitLeakage detection; erasure conversion (Wu, Kolkowitz, Puri & Thompson, 2022)

Key takeaways

  • Rydberg properties follow hydrogenic scalings; C6=−C32/ΔF∝n∗11C_6 = -C_3^2/\Delta_F \propto n^{*11} arises at second order from dipole–dipole coupling, crossing over to C3/r3C_3/r^3 at short range.
  • Perfect blockade confines NN atoms to {∣G⟩,∣W⟩}\{\ket G, \ket W\}, with collective Rabi frequency N Ω\sqrt N\,\Omega.
  • A diagonal gate is CZ up to local ZZ rotations iff φ11−φ01−φ10≡π\varphi_{11}-\varphi_{01}-\varphi_{10}\equiv\pi; Jaksch and global-pulse protocols satisfy this without needing the exact value of VV.
  • Decay and blockade leakage trade off, giving Emin⁡≈2.9 (Bτ)−2/3E_{\min}\approx2.9\,(B\tau)^{-2/3}; the best reported CZ fidelities are about 99.5% as of the mid-2020s.

Check yourself

In the interactive lesson these are scored with your confidence, and what you miss comes back for review.

Question 1

Moving from n=50n = 50 to n=70n = 70 at a fixed Rabi frequency increases the blockade radius by about (ignore quantum defects)

  1. 1.4×
  2. 1.85×
  3. 40×
  4. It does not change
Show the answer

B. 1.85×

Rb∝C61/6∝n11/6R_b \propto C_6^{1/6} \propto n^{11/6}, so 1.411/6≈1.851.4^{11/6} \approx 1.85. C6C_6 itself grows by 1.411≈401.4^{11} \approx 40.

Question 2

Why is the Jaksch gate insensitive to the exact distance between the two atoms?

  1. Rydberg interactions do not depend on distance
  2. Distance errors are removed by the single-qubit phase corrections after the gate
  3. The atoms are always exactly one lattice constant apart
  4. The ideal phases only require the doubly excited state to be far off resonance (V≫ℏΩV \gg \hbar\Omega); its exact shift does not enter
Show the answer

D. The ideal phases only require the doubly excited state to be far off resonance (V≫ℏΩV \gg \hbar\Omega); its exact shift does not enter

In the blockade limit the ∣11⟩\ket{11} dynamics is "nothing happens to the target", whatever the value of VV. Residual errors scale as (ℏΩ/V)2(\hbar\Omega/V)^2.

Question 3

A better Rydberg state doubles both the blockade shift BB and the lifetime τ\tau. The optimal intrinsic error Emin⁡E_{\min} of a Jaksch gate is multiplied by about

  1. 0.400.40
  2. 1/41/4
  3. 0.630.63
  4. 1/21/2
Show the answer

A. 0.400.40

Emin⁡∝(Bτ)−2/3E_{\min}\propto(B\tau)^{-2/3} and BτB\tau grows fourfold: 4−2/3≈0.404^{-2/3}\approx0.40.

Question 4

At short distances, r≪Rc=∣C3/ΔF∣1/3r \ll R_c = \abs{C_3/\Delta_F}^{1/3}, the interaction between two Rydberg atoms

  1. vanishes
  2. still scales as 1/r61/r^6
  3. scales as 1/r31/r^3 (resonant dipole–dipole)
  4. scales as 1/r121/r^{12}
Show the answer

C. scales as 1/r31/r^3 (resonant dipole–dipole)

When C3/r3≫∣ΔF∣C_3/r^3 \gg \abs{\Delta_F} the pair states mix completely and E±≈±C3/r3E_\pm \approx \pm C_3/r^3.

Further reading

  • D. Jaksch, J. I. Cirac, P. Zoller, S. L. Rolston, R. Côté & M. D. Lukin, "Fast quantum gates for neutral atoms", Physical Review Letters 85, 2208 (2000).
  • M. Saffman & T. G. Walker, "Analysis of a quantum logic device based on dipole-dipole interactions of optically trapped Rydberg atoms", Physical Review A (2005).
  • H. Levine et al., "Parallel implementation of high-fidelity multiqubit gates with neutral atoms", Physical Review Letters (2019).
  • S. J. Evered et al., "High-fidelity parallel entangling gates on a neutral-atom quantum computer", Nature (2023).
  • M. Saffman, T. G. Walker & K. Mølmer, "Quantum information with Rydberg atoms", Reviews of Modern Physics 82, 2313 (2010).