Simple Algorithm
Quantum Computing · Excited State · 22 min

Neutral Atoms & Rydberg Arrays

Hundreds of atoms in optical tweezers, entangled via Rydberg states.

Trap an atom with nothing but a focused laser beam, then make hundreds of copies with a hologram. Neutral-atom processors hold large arrays of identical qubits that barely interact, until a laser lifts two of them into giant Rydberg states that interact strongly across micrometers.

What you’ll learn

  • Derive the optical dipole potential and explain why red-detuned tweezers trap atoms at intensity maxima.
  • Explain stochastic loading and the atom-by-atom assembly of defect-free arrays.
  • Describe hyperfine and nuclear-spin qubits in tweezers, their main dephasing mechanism and how they are read out.
  • Compute the Rydberg blockade radius and derive the 2\sqrt2-enhanced collective Rabi oscillation.
  • Analyze the Jaksch et al. blockade gate and show that it is a CZ up to single-qubit phases.

Trapping atoms with light

A neutral atom has no net charge, so the Paul-trap trick is unavailable. Instead, light induces an oscillating electric dipole in the atom, and the energy of that dipole depends on the light's intensity. Below resonance (red detuning) the induced dipole oscillates in phase with the field and the atom is pulled toward high intensity; above resonance it is pushed away.

Focus a red-detuned beam to a spot about a micrometer across and you get an optical tweezer, a trap a few hundred μK to about a mK deep. Atoms are first gathered and cooled to around 10 μK in a magneto-optical trap and optical molasses, then caught in the tweezers.

DerivationThe dipole potential
  1. H=−ℏΔ ∣e⟩ ⁣⟨e∣+ℏΩ2(∣e⟩ ⁣⟨g∣+∣g⟩ ⁣⟨e∣),Δ=ωL−ω0H = -\hbar\Delta\,\dyad{e} + \frac{\hbar\Omega}{2}\big(\ketbra{e}{g} + \ketbra{g}{e}\big),\qquad \Delta = \omega_L - \omega_0

    A two-level atom in the frame rotating at the laser frequency ωL\omega_L, with Rabi frequency Ω\Omega and detuning Δ\Delta from the resonance ω0\omega_0.

  2. ΔEg=∣ℏΩ/2∣2Eg−Ee=ℏΩ24Δ\Delta E_g = \frac{\abs{\hbar\Omega/2}^2}{E_g - E_e} = \frac{\hbar\Omega^2}{4\Delta}

    Second-order perturbation theory for ∣Δ∣≫Ω\abs\Delta \gg \Omega, with Eg=0E_g = 0 and Ee=−ℏΔE_e = -\hbar\Delta: the light shift of the ground state.

  3. U(r)=ℏ Ω(r)24Δ∝I(r)ΔU(\mathbf r) = \frac{\hbar\,\Omega(\mathbf r)^2}{4\Delta} \propto \frac{I(\mathbf r)}{\Delta}

    Since Ω2∝I\Omega^2 \propto I, red detuning (Δ<0\Delta < 0) makes a potential well at the intensity maximum: the focus.

  4. Γsc≈Γ Ω24Δ2  ⇒  ℏΓsc=Γ∣Δ∣ ∣U∣\Gamma_{\rm sc} \approx \Gamma\,\frac{\Omega^2}{4\Delta^2}\;\Rightarrow\;\hbar\Gamma_{\rm sc} = \frac{\Gamma}{\abs\Delta}\,\abs U

    The atom also scatters photons, at a rate set by its small excited-state admixture. Depth falls like 1/Δ1/\Delta but scattering like 1/Δ21/\Delta^2, so tweezers are detuned far from resonance (tens of THz for typical alkali tweezers), and scattering becomes a minor source of heating.

Check your understanding

Why does a tweezer tuned below the atomic resonance attract atoms to its focus?

  1. Radiation pressure from absorbed photons pushes atoms toward the focus
  2. The induced dipole oscillates in phase with the field, which lowers the atom's energy where the intensity is highest
  3. The light slightly ionizes the atom, and the ion is trapped electrostatically
  4. The laser heats the surrounding gas, which pushes atoms inward
Show the answer

B. The induced dipole oscillates in phase with the field, which lowers the atom's energy where the intensity is highest

The light shift U=ℏΩ2/4ΔU = \hbar\Omega^2/4\Delta is negative for Δ<0\Delta < 0 and proportional to intensity, so the energy is lowest at the focus. The force is the gradient of this potential.

Defect-free arrays, atom by atom

Many tweezers at once come from a spatial light modulator (a programmable hologram) or from acousto-optic deflectors (AODs) driven with many radio-frequency tones. Arrays of hundreds to thousands of sites are routine as of the mid-2020s. But with a loading probability p≈12p \approx \tfrac12 per site, the chance that all NN sites are filled is pNp^N: about 10−3010^{-30} for N=100N = 100.

The fix is rearrangement. Image the array (atoms fluoresce under resonant light), compute which loaded atoms should go where, drag them with a mobile tweezer steered by an AOD into the target pattern, and image again to check. Endres et al. (2016) and Barredo et al. (2016) demonstrated defect-free one- and two-dimensional arrays this way.

Qubits in tweezers

Two families of long-lived qubits are common:

  • Hyperfine qubits in alkali atoms, such as 87Rb^{87}\mathrm{Rb} ∣F=1,mF=0⟩↔∣F=2,mF=0⟩\ket{F{=}1,m_F{=}0}\leftrightarrow\ket{F{=}2,m_F{=}0} at 6.8 GHz or 133Cs^{133}\mathrm{Cs} at 9.2 GHz: clock states, as for ions.
  • Nuclear-spin qubits in alkaline-earth(-like) atoms such as 171Yb^{171}\mathrm{Yb} or 87Sr^{87}\mathrm{Sr}. Their electronic ground state has zero electronic angular momentum, so the qubit lives in the nuclear spin and barely notices magnetic fields or the tweezer light.

Single-qubit gates use global microwaves or Raman beams, plus focused beams that shift selected atoms out of resonance for local addressing.

Readout is fluorescence imaging on a camera. To make it state-selective, a resonant "push-out" pulse first ejects the atoms in one hyperfine state, and the survivors are imaged. This is destructive and takes milliseconds, far slower than gates. Mid-circuit readout therefore uses separate readout zones, or a second atomic species whose fluorescence does not disturb the data qubits.

Rydberg atoms and the blockade

Ground-state atoms a few micrometers apart hardly interact: ideal for storing qubits, useless for gates. The trick is to excite atoms briefly to a Rydberg state with principal quantum number n∼50n \sim 50–100100. Such an atom is enormous (its radius grows as n2n^2, reaching a few tenths of a micrometer at n=70n = 70), and two Rydberg atoms at distance rr interact through the van der Waals potential

V(r)=C6r6,C6∝n11.V(r) = \frac{C_6}{r^6}, \qquad C_6 \propto n^{11}.
Rb=(C6ℏΩ)1/6R_b = \Big(\frac{C_6}{\hbar\Omega}\Big)^{1/6}
Within the blockade radius the interaction exceeds the excitation linewidth, set by the Rabi frequency Ω\Omega, and two atoms cannot both be excited.
Predict before you look

For rubidium in the 70S70S Rydberg state, C6/h≈860C_6/h \approx 860 GHz μm⁶. With a Rydberg Rabi frequency Ω/2π=2\Omega/2\pi = 2 MHz, estimate the blockade radius RbR_b.

Show the answer

8.7 μm

Rb6=C6/ℏΩ=(860×109/2×106) μm6=4.3×105 μm6R_b^6 = C_6/\hbar\Omega = (860\times10^{9}/2\times10^{6})\ \mu\text{m}^6 = 4.3\times10^5\ \mu\text{m}^6, so Rb≈8.7R_b \approx 8.7 μm. The steep r−6r^{-6} law makes RbR_b insensitive to Ω\Omega: doubling Ω\Omega shrinks it by only 21/6≈1.122^{1/6} \approx 1.12. At 5 μm the interaction is h×55h\times55 MHz; at 10 μm it is below 1 MHz.

DerivationTwo atoms, one excitation
  1. H=ℏΩ2∑i=1,2(∣r⟩ ⁣⟨g∣i+∣g⟩ ⁣⟨r∣i)+V ∣rr⟩ ⁣⟨rr∣H = \frac{\hbar\Omega}{2}\sum_{i=1,2}\big(\ketbra{r}{g}_i + \ketbra{g}{r}_i\big) + V\,\dyad{rr}

    Both atoms are driven resonantly from ∣g⟩\ket g to ∣r⟩\ket r, with interaction energy VV when both are excited.

  2. H∣gg⟩=ℏ2 Ω2∣W⟩,∣W⟩=∣gr⟩+∣rg⟩2H\ket{gg} = \frac{\hbar\sqrt2\,\Omega}{2}\ket{W},\qquad \ket W = \frac{\ket{gr}+\ket{rg}}{\sqrt2}

    The drive couples ∣gg⟩\ket{gg} only to the symmetric singly excited state, with a matrix element enhanced by 2\sqrt2. The antisymmetric state (∣gr⟩−∣rg⟩)/2(\ket{gr}-\ket{rg})/\sqrt2 is never populated.

  3. ⟨rr|H|W⟩=ℏ2 Ω2,Err=V≫ℏΩ\mel{rr}{H}{W} = \frac{\hbar\sqrt2\,\Omega}{2},\qquad E_{rr} = V \gg \hbar\Omega

    ∣W⟩\ket W couples onward to ∣rr⟩\ket{rr}, but that state is shifted by VV, so the transition is far off resonance.

  4. ∣gg⟩ ↔ ∣W⟩  at Rabi frequency 2 Ω,Prr∼(ℏΩV)2\ket{gg}\ \leftrightarrow\ \ket{W}\ \ \text{at Rabi frequency}\ \sqrt2\,\Omega,\qquad P_{rr} \sim \Big(\frac{\hbar\Omega}{V}\Big)^2

    In the blockade regime the pair acts as a single two-level "superatom" that oscillates 2\sqrt2 times faster, and double excitation is suppressed as (ℏΩ/V)2(\hbar\Omega/V)^2.

A blockade gate

Jaksch et al. turned the blockade into a gate. Only the qubit state ∣1⟩\ket1 is laser-coupled to ∣r⟩\ket r. The sequence is: a π-pulse on the control, a 2π2\pi-pulse on the target, and another π-pulse on the control.

DerivationPhase bookkeeping
  1. ∣00⟩ → ∣00⟩\ket{00}\ \to\ \ket{00}

    Neither atom is coupled to ∣r⟩\ket r.

  2. ∣01⟩ → −∣01⟩\ket{01}\ \to\ -\ket{01}

    The target makes a full Rabi cycle ∣1⟩→∣r⟩→∣1⟩\ket1\to\ket r\to\ket1. A resonant 2π2\pi rotation of a two-level system multiplies it by −1-1.

  3. ∣10⟩ → (−i)2∣10⟩=−∣10⟩\ket{10}\ \to\ (-i)^2\ket{10} = -\ket{10}

    The control goes to ∣r⟩\ket r and back with two π-pulses, each contributing a factor −i-i.

  4. ∣11⟩ → −∣11⟩\ket{11}\ \to\ -\ket{11}

    The control is excited, so the target's 2π2\pi-pulse is blockaded and does nothing; the control's round trip contributes −1-1.

  5. U=diag⁡(1,−1,−1,−1)=(Z⊗Z)⋅CZU = \diag(1,-1,-1,-1) = (Z\otimes Z)\cdot\mathrm{CZ}

    Up to single-qubit ZZ gates this is a controlled-ZZ. The exact value of VV does not matter, as long as V≫ℏΩV \gg \hbar\Omega.

Check your understanding

In the Jaksch protocol, which input state ends with a different phase from the other three?

  1. ∣00⟩\ket{00}
  2. ∣01⟩\ket{01}
  3. ∣10⟩\ket{10}
  4. ∣11⟩\ket{11}
Show the answer

A. ∣00⟩\ket{00}

∣00⟩\ket{00} is untouched (phase +1+1), while ∣01⟩\ket{01}, ∣10⟩\ket{10} and ∣11⟩\ket{11} all pick up −1-1, by three different routes. The pattern (1,−1,−1,−1)(1,-1,-1,-1) is a CZ followed by Z⊗ZZ\otimes Z.

The native entangler is therefore a CZ, and a CNOT is a CZ with Hadamards on the target. Practice compiling with CZ alone: make a three-atom GHZ state.

Modern protocols drive both atoms at once with global, shaped pulses. Levine et al. (2019) used two global pulses separated by a laser phase jump (fidelity about 97%), and Evered et al. (2023) reached about 99.5% for CZ gates applied in parallel across an array, using smooth phase-modulated pulses. Gate times are a fraction of a microsecond. Errors come from Rydberg-state decay, laser phase and intensity noise, Doppler dephasing from atomic motion and imperfect blockade.

Key takeaways

  • Red-detuned optical tweezers trap single atoms through the light shift U=ℏΩ2/4ΔU = \hbar\Omega^2/4\Delta, with scattering suppressed by an extra factor Γ/∣Δ∣\Gamma/\abs\Delta.
  • Collisional blockade gives each tweezer 0 or 1 atom, so arrays are assembled atom by atom with mobile tweezers.
  • Hyperfine and nuclear-spin qubits store information for up to seconds; readout is fluorescence imaging, slow and usually destructive.
  • Rydberg atoms interact as C6/r6C_6/r^6; within the blockade radius Rb=(C6/ℏΩ)1/6R_b = (C_6/\hbar\Omega)^{1/6} only one excitation is allowed, which yields a CZ gate insensitive to the exact interaction strength.

Check yourself

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

Question 1

You double the detuning of a far-detuned tweezer and double its power to keep the trap depth fixed. The photon scattering rate

  1. doubles
  2. stays the same
  3. halves
  4. drops fourfold
Show the answer

C. halves

ℏΓsc=(Γ/∣Δ∣)∣U∣\hbar\Gamma_{\rm sc} = (\Gamma/\abs\Delta)\abs U: at fixed depth UU, scattering falls as 1/Δ1/\Delta.

Question 2

With everything else fixed, you double the Rydberg Rabi frequency Ω\Omega. The blockade radius

  1. halves
  2. shrinks by about 11%
  3. stays the same
  4. doubles
Show the answer

B. shrinks by about 11%

Rb=(C6/ℏΩ)1/6∝Ω−1/6R_b = (C_6/\hbar\Omega)^{1/6} \propto \Omega^{-1/6}, and 2−1/6≈0.892^{-1/6} \approx 0.89.

Question 3

Why do neutral-atom experiments rearrange atoms after loading?

  1. Each tweezer holds 0 or 1 atom with probability about ½, so an N-site array is completely filled with probability only about 2−N2^{-N}
  2. Atoms load at random positions within each tweezer
  3. Rearrangement cools the atoms to the motional ground state
  4. Tweezers can only be loaded one at a time
Show the answer

A. Each tweezer holds 0 or 1 atom with probability about ½, so an N-site array is completely filled with probability only about 2−N2^{-N}

Collisional blockade makes occupation binary and random. Imaging plus targeted moves turns a random half-filled array into a defect-free one.

Question 4

Three atoms lie within each other's blockade radius and are driven together from ∣ggg⟩\ket{ggg}. In the perfect-blockade limit they

  1. oscillate independently, each at Ω\Omega
  2. cannot be excited at all
  3. oscillate between ∣ggg⟩\ket{ggg} and ∣rrr⟩\ket{rrr} at 3Ω3\Omega
  4. oscillate between ∣ggg⟩\ket{ggg} and the symmetric single-excitation state at 3 Ω\sqrt3\,\Omega
Show the answer

D. oscillate between ∣ggg⟩\ket{ggg} and the symmetric single-excitation state at 3 Ω\sqrt3\,\Omega

The drive couples ∣ggg⟩\ket{ggg} to (∣rgg⟩+∣grg⟩+∣ggr⟩)/3(\ket{rgg}+\ket{grg}+\ket{ggr})/\sqrt3 with matrix element 3 ℏΩ/2\sqrt3\,\hbar\Omega/2, and all multiply excited states are blockaded.

Question 5

Why is standard state-selective readout of neutral atoms destructive?

  1. Imaging ionizes the atoms
  2. The camera absorbs the atoms
  3. Atoms in one hyperfine state are pushed out of their traps before imaging, so "absent" encodes that state
  4. Imaging requires switching off the tweezers permanently
Show the answer

C. Atoms in one hyperfine state are pushed out of their traps before imaging, so "absent" encodes that state

The push-out converts the qubit state into presence or absence, which a fluorescence image detects reliably, at the price of losing the atoms in one state.

Go deeper

Further reading

  • M. Saffman, T. G. Walker & K. Mølmer, "Quantum information with Rydberg atoms", Reviews of Modern Physics 82, 2313 (2010).
  • A. Browaeys & T. Lahaye, "Many-body physics with individually controlled Rydberg atoms", Nature Physics 16, 132 (2020).
  • 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).
  • R. Grimm, M. Weidemüller & Y. B. Ovchinnikov, "Optical dipole traps for neutral atoms", Advances in Atomic, Molecular, and Optical Physics 42, 95 (2000).