Neutral Atoms

Neutral-atom platforms use optical lattices, optical tweezers, magnetic traps, quantum gas microscopes, and Rydberg excitation to prepare and measure controllable atomic systems. For open-system physics, they are a rich middle ground: atoms are well isolated from charge noise, but they can be lost, heated by light, dephased by differential shifts, and driven into short-lived highly excited states.

We connect photon scattering, fluorescence imaging, atom loss, and Rydberg decay to explicit open-system models. Rydberg Atoms develops the atomic structure, while Neutral-Atom and Rydberg Qubits connects these processes to gates and array operation.

Trapped Motion and Internal States

A neutral atom in an optical trap has external motion and internal levels. Near the bottom of a trap, one often approximates the potential as harmonic:

$$V(\mathbf r) \approx V_0 + \frac{m}{2} \sum_i \omega_i^2 x_i^2 .$$

Internal states may be hyperfine, Zeeman, clock, or optical states. A minimal qubit Hamiltonian is

$$\frac{H_q}{\hbar} = \frac{\omega_q}{2}\sigma_z,$$

with controls supplied by microwave, Raman, optical, or Rydberg lasers.

The important open-system feature is that internal and external degrees of freedom are not automatically independent. Trap light shifts internal transitions, photon scattering heats motion and can flip internal states, and imaging light can both reveal and disturb atom occupation.

Optical Traps and Photon Scattering

A far-detuned optical dipole trap uses the AC Stark shift. In a simple two-level estimate, the trap depth scales as

$$U \approx \frac{\hbar\Omega^2}{4\Delta},$$

while the off-resonant photon-scattering rate scales as

$$\Gamma_{\mathrm{sc}} \approx \Gamma \frac{\Omega^2}{4\Delta^2} \approx \frac{\Gamma}{|\Delta|} \frac{|U|}{\hbar}.$$

Here $\Omega$ is the resonant Rabi frequency, $\Delta=\omega_L-\omega_0$ is laser minus atomic transition frequency, and $\Gamma$ is the natural linewidth of the optical transition. The estimate hides level structure and polarization, but the tradeoff is robust: stronger confinement generally costs more scattering unless detuning, wavelength, or atomic structure are chosen carefully.

Scattering can produce:

• motional heating from recoil;
• spin flips or leakage to unwanted hyperfine states;
• dephasing from which-state information carried by photons;
• atom loss if heating exceeds trap depth.

Thus an optical trap is both a conservative potential and a weakly dissipative environment.

Imaging and Occupation Measurement

Neutral-atom imaging usually detects fluorescence. In optical tweezer arrays and quantum gas microscopes, the measurement often asks whether a site is occupied. A simple count model is

$$P(n\mid 1) = e^{-\mu_1} \frac{\mu_1^n}{n!}, \qquad P(n\mid 0) = e^{-\mu_0} \frac{\mu_0^n}{n!},$$

where $1$ means occupied, $0$ means empty, and $\mu_1\gg\mu_0$ for good collection and low background.

Imaging is a strong measurement of atom presence. The same scattered photons that produce a camera signal also impart recoil. High-fidelity imaging therefore requires cooling during detection, short exposure, deep traps, or a combination of all three.

In dense gases or lattices, imaging can also be affected by light-assisted collisions. Depending on species and protocol, two atoms on one site may be lost during imaging, turning the measurement into a parity-sensitive detection rather than a direct atom-number measurement.

Atom Loss

Atom loss is not merely dephasing. It transfers the state out of the occupied-atom subspace, so the full description must include an empty-site sector. For a single occupied state $|1\rangle$ and a lost state $|\varnothing\rangle$, a simple collapse operator is

$$L_{\mathrm{loss}} = \sqrt{\Gamma_{\mathrm{loss}}} |\varnothing\rangle\langle 1|.$$

If loss is the only process, the survival probability is

$$P_1(t) = e^{-\Gamma_{\mathrm{loss}}t}.$$

In a many-body lattice or tweezer array, local one-body loss is often modeled by

$$\dot\rho = \sum_j \Gamma_j \mathcal D[c_j]\rho,$$

where $c_j$ removes an atom from site $j$. This changes particle number and can turn a coherent many-body state into a mixture over loss locations.

For an internal qubit with equal state-independent loss, separate channels $L_0=\sqrt{\Gamma}|\varnothing\rangle\langle0|$ and $L_1=\sqrt{\Gamma}|\varnothing\rangle\langle1|$ give survival $e^{-\Gamma t}$. Combining these into one coherent jump can create an artificial dark superposition; it represents a different environment. A detected loss is an erasure only when an accessible record flags it. Common loss mechanisms include background-gas collisions, heating out of finite-depth traps, photoionization, inelastic collisions, failed rearrangement moves, and Rydberg-state decay to untrapped states.

Dephasing

Neutral-atom qubits often dephase because their transition frequency fluctuates:

$$\frac{H_{\mathrm{noise}}(t)}{\hbar} = \frac{\delta\omega(t)}{2}\sigma_z .$$

Sources include:

• differential AC Stark shifts from trap intensity noise;
• magnetic-field noise;
• laser phase noise;
• Doppler shifts and finite temperature;
• inhomogeneous trap frequencies across an array;
• interactions with other atoms;
• imperfect cancellation at magic wavelengths or clock settings.

A Markovian pure-dephasing model is

$$\dot\rho = \frac{\Gamma_\phi}{2} \mathcal D[\sigma_z]\rho .$$

Not all dephasing is Markovian. Quasi-static inhomogeneity can often be refocused by spin echo, while fast noise produces irreversible decay on the experiment timescale. A quoted coherence time is therefore meaningful only together with the pulse sequence and noise environment.

Rydberg Excitation and Decay

Rydberg states have large electric dipoles and strong interactions, which make them valuable for gates, blockade physics, and quantum simulation. A common driven Rydberg-array Hamiltonian is

$$\frac{H}{\hbar} = \sum_i \left[ \frac{\Omega_i}{2} \left( |r_i\rangle\langle g_i| + |g_i\rangle\langle r_i| \right) - \Delta_i n_i \right] + \sum_{i\lt j} \frac{V_{ij}}{\hbar}n_i n_j,$$

where

$$n_i=|r_i\rangle\langle r_i|.$$

Here $V_{ij}$ and $C_6/R_{ij}^6$ are energies, whereas $\Omega_i$ and $\Delta_i$ are angular frequencies. For van der Waals interactions,

$$V_{ij} \approx \frac{C_6} {R_{ij}^6}.$$

The simplest drive-defined blockade radius is estimated by setting the interaction shift comparable to the Rabi frequency:

$$|V(R_b)| \approx \hbar\Omega, \qquad R_b \approx \left( \frac{|C_6|} {\hbar\Omega} \right)^{1/6}.$$

This is only a crossover convention. Rydberg Blockade explains the finite-interaction dynamics, collective enhancement, and gate errors. Here the purpose of the formula is to identify the open-system channels acting on that driven manifold.

Open-system channels for Rydberg physics include:

• radiative decay to lower states;
• blackbody-induced transitions between Rydberg levels;
• photoionization;
• motional forces from state-dependent potentials;
• dephasing from laser phase noise and Doppler shifts;
• leakage to states outside the intended two-level manifold.

A schematic decay channel is

$$L_{r\to g} = \sqrt{\Gamma_{rg}} |g\rangle\langle r|,$$

but realistic modeling often needs multiple final states and loss channels.

Imaging Backaction and Cooling

Imaging neutral atoms while keeping them trapped is a cooling problem as much as a detection problem. Scattered photons heat the atom by recoil; cooling removes that energy while preserving site occupation or internal-state information.

Cooling and loss must act on compatible state spaces. Extend the trapped motional levels by an empty-site state $|\varnothing\rangle$, and let $a$ vanish on that state. One model is

$$\dot\rho=\Gamma_-\mathcal D[a]\rho +\Gamma_+\mathcal D[a^\dagger]\rho +\sum_n\mathcal D[L_n]\rho, \qquad L_n=\sqrt{\Gamma_n}|\varnothing\rangle\langle n|.$$

The rates $\Gamma_n$ describe loss from trapped level $n$; no extra rate multiplies a dissipator whose jump already contains its square root. A finite-depth trap often needs energy-dependent loss rather than one constant rate. The full state remains normalized, while the probability of remaining trapped can decrease.

For quantum gas microscopes, Raman cooling, electromagnetically induced transparency cooling, or polarization-gradient cooling may be interleaved with fluorescence collection. For tweezer arrays, imaging and rearrangement errors can dominate the preparation fidelity of a large array even when single-atom physics is clean.

Noise in Arrays

Neutral-atom arrays add spatial structure. Each site can have different trap depth, intensity noise, pointing noise, temperature, and optical phase. In a many-atom experiment, the relevant noise may be:

• local, affecting one site independently;
• global, affecting many sites coherently;
• slowly varying across the array;
• correlated with rearrangement, imaging, or laser addressing.

These distinctions matter. Global phase noise can sometimes be refocused or tracked; independent atom loss cannot. Site-dependent detuning disorder may look like dephasing in one experiment and like a static Hamiltonian term in another.

Exercises

Loss Survival Probability

For a single atom with collapse operator

$$L_{\mathrm{loss}} = \sqrt{\Gamma} |\varnothing\rangle\langle 1|,$$

show that the occupation probability obeys

$$P_1(t) = e^{-\Gamma t}.$$

Solution

The loss jump transfers population from $|1\rangle$ to $|\varnothing\rangle$. The occupied-state population obeys

$$\frac{dP_1}{dt} = - \Gamma P_1 .$$

Solving this differential equation gives

$$P_1(t) = P_1(0)e^{-\Gamma t}.$$

For an initially occupied site, $P_1(0)=1$, so $P_1(t)=e^{-\Gamma t}$.

Trap Depth and Scattering

Using the far-detuned two-level estimates

$$U \approx \frac{\hbar\Omega^2}{4\Delta}, \qquad \Gamma_{\mathrm{sc}} \approx \Gamma \frac{\Omega^2}{4\Delta^2},$$

show that

$$\Gamma_{\mathrm{sc}} \approx \frac{\Gamma}{|\Delta|} \frac{|U|}{\hbar}.$$

Solution

Taking magnitudes in the trap-depth expression gives

$$|U| \approx \frac{\hbar\Omega^2}{4|\Delta|}.$$

Therefore

$$\frac{\Omega^2}{4\Delta^2} = \frac{1}{|\Delta|} \frac{\Omega^2}{4|\Delta|} \approx \frac{1}{|\Delta|} \frac{|U|}{\hbar}.$$

Multiplying by $\Gamma$ gives

$$\Gamma_{\mathrm{sc}} \approx \frac{\Gamma}{|\Delta|} \frac{|U|}{\hbar}.$$

The result is an approximate scaling relation; real atoms require sums over levels, polarizations, and tensor light shifts.

Blockade Radius

Assume a van der Waals interaction

$$|V(R)| = \frac{|C_6|}{R^6}.$$

Derive the blockade-radius estimate from $|V(R_b)|=\hbar\Omega$.

Solution

Set

$$\frac{|C_6|} {R_b^6} = \hbar\Omega .$$

Solving for $R_b$ gives

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

The blockade picture is approximate; finite detuning, laser linewidth, motion, and many-body geometry all modify practical performance.

Dephasing Versus Loss

Why can atom loss not be modeled as pure dephasing of a qubit that remains in the same two-dimensional Hilbert space?

Solution

Pure dephasing preserves populations in the qubit basis and only suppresses coherences. Atom loss removes the atom from the trapped or computational subspace. The possible outcomes now include a lost state or vacuum sector, and particle number changes.

A loss model therefore requires an enlarged Hilbert space, an erasure flag, or a conditional description with postselection. Treating loss as pure dephasing would miss both the changed normalization of occupied-site probabilities and the information contained in detecting an empty site. See Erasure and Loss Channels for the channel-level distinctions.

References

• W. S. Bakr, J. I. Gillen, A. Peng, S. Fölling, and M. Greiner, “A quantum gas microscope for detecting single atoms in a Hubbard-regime optical lattice,” Nature 462, 74-77 (2009).
• I. Bloch, J. Dalibard, and W. Zwerger, “Many-body physics with ultracold gases,” Reviews of Modern Physics 80, 885-964 (2008).
• A. Browaeys and T. Lahaye, “Many-body physics with individually controlled Rydberg atoms,” Nature Physics 16, 132-142 (2020).
• R. Grimm, M. Weidemüller, and Y. B. Ovchinnikov, “Optical dipole traps for neutral atoms,” Advances in Atomic, Molecular, and Optical Physics 42, 95-170 (2000).
• A. M. Kaufman and K.-K. Ni, “Quantum science with optical tweezer arrays of ultracold atoms and molecules,” Nature Physics 17, 1324-1333 (2021).
• M. Saffman, T. G. Walker, and K. Mølmer, “Quantum information with Rydberg atoms,” Reviews of Modern Physics 82, 2313-2363 (2010).
• J. F. Sherson, C. Weitenberg, M. Endres, M. Cheneau, I. Bloch, and S. Kuhr, “Single-atom-resolved fluorescence imaging of an atomic Mott insulator,” Nature 467, 68-72 (2010).