AMO Platforms and Quantum Control

An AMO platform is an experimentally controlled quantum system built from atomic, molecular, optical, or closely analogous degrees of freedom. A platform is more than a species, trap, or qubit. It combines a way to prepare states, a controllable interaction, and a measurement. The same atom can serve as a clock, a qubit, or a probe of many-body physics, depending on how these parts are arranged.

These distinctions matter in practice. An atom can possess an exceptionally narrow transition and still make a poor sensor if the atoms cannot be loaded reproducibly. A qubit can have a long coherence time and still make a poor processor if its gates leak outside the computational subspace. A trap can confine particles for minutes and still be unsuitable for many-body physics if its temperature, density, or interaction parameters are not known. An experiment therefore needs both coherent dynamics and reliable preparation and readout.

AMO experiments are unusually clean because their microscopic constituents, couplings, and measurement channels can often be identified one by one. They are not clean because noise has disappeared. Laser phase noise, magnetic-field drift, blackbody radiation, photon scattering, motional heating, collisions, imperfect vacuum, detector errors, and calibration drift remain part of the physical model.

Helpful background. Density Operators describes imperfectly prepared states, Rabi Oscillations explains driven rotations, and AC Stark Shifts supplies the interaction behind optical trapping.

Enter this chapter

Prepare and confine atoms. Begin with Laser Cooling, Doppler Cooling, and Sub-Doppler Cooling. Then compare Magneto-Optical Traps, Optical Dipole Traps, and Optical Tweezers. Evaporative Cooling and Ultracold Atoms lead to Bose–Einstein Condensates, Degenerate Fermi Gases, and Optical Lattices.

Couple internal states through motion or strong interactions. Follow Ion Traps into Trapped-Ion Control, or Rydberg Atoms into Rydberg Blockade.

Explore other degrees of freedom. Cold Molecules, Cavity QED Platforms, Circuit QED, and Atom Interferometry illustrate different interactions and measurements. Quantum Control in AMO brings together pulse design, calibration, and noise.

A Common Mathematical Model

Many platforms can be described, over a declared operating regime, by

$$H(t;\boldsymbol{\theta}) = H_0(\boldsymbol{\theta}) + \sum_k u_k(t)H_k,$$

where:

• $H_0$ contains uncontrolled or slowly tuned system parameters $\boldsymbol{\theta}$;
• $u_k(t)$ are classical control waveforms such as laser amplitudes, phases, frequencies, magnetic fields, electrode voltages, or microwave envelopes;
• $H_k$ are the operators through which those controls act.

For an open system, a common Markovian working model is

$$\dot{\rho} = -\frac{i}{\hbar}[H(t),\rho] + \sum_\mu \gamma_\mu \mathcal D[L_\mu]\rho,$$

with

$$\mathcal D[L]\rho = L\rho L^\dagger -\frac{1}{2} \left\{ L^\dagger L,\rho \right\}.$$

This equation is not a declaration that every laboratory environment is Markovian. It is a compact engineering model whose validity must be tested over the bandwidth and duration of the experiment. Colored laser noise, quasistatic field drift, trap-loss events, collision histories, and feedback latency can require stochastic Hamiltonians, non-Markovian models, or explicit classical state variables.

A measurement with outcome $y$ is represented at the probability level by positive operators $E_y$:

$$p(y|\rho,\boldsymbol{\theta}) = \operatorname{Tr}(E_y\rho), \qquad \sum_y E_y=I.$$

The physical state update requires a quantum instrument, not only the POVM. In many AMO experiments, however, the immediate inference problem is already substantial: $E_y$ must include finite photon collection, dark counts, thresholding, state leakage, atom loss, optical pumping during detection, and spatially varying response.

From preparation to readout

A long coherence time, a strong interaction, and accurate readout are different achievements. The following questions help distinguish them.

Preparation

What density operator enters the experiment?

Preparation may include:

• slowing and capturing a thermal beam;
• laser cooling or sympathetic cooling;
• evaporative cooling;
• optical pumping into a Zeeman or hyperfine state;
• resolved-sideband cooling of a motional mode;
• deterministic rearrangement of a neutral-atom array;
• heralding a photon or cavity state;
• initializing a superconducting circuit by relaxation, measurement, or active reset.

A quoted temperature does not completely specify the state. One may also need the particle-number distribution, internal-state purity, motional-mode occupations, spatial correlations, condensate fraction, or residual entanglement with preparation apparatus.

Control

Which terms in $H(t)$ are independently programmable, with what bandwidth, dynamic range, spatial resolution, and calibration uncertainty?

A nominal Rabi frequency $\Omega$ is not yet a control specification. Its amplitude and phase noise, detuning, spatial profile, polarization, turn-on transient, off-resonant couplings, and correlations with other controls can all matter. In an array, the response is generally $\Omega_j(t)$ rather than one global number.

Evolution

Which degrees of freedom are retained in the model, and which are treated as an environment?

Examples include:

• an internal clock doublet coupled to external motion;
• spins coupled through phonon modes in an ion chain;
• atoms tunneling and interacting in optical-lattice bands;
• Rydberg states coupled by van der Waals or resonant dipole interactions;
• a two-level emitter coupled to a lossy cavity mode;
• a superconducting artificial atom coupled to resonators and transmission lines.

Model reduction is part of platform design. It must be supported by scale separation, not merely by the desire for a simple Hamiltonian.

Measurement

Which observable is inferred, from which raw detector record?

State-dependent fluorescence, absorption images, phase-contrast images, ionization, cavity transmission, homodyne current, dispersive microwave readout, and time-tagged photon counts implement different quantum instruments. The final binary label in a data file may conceal a continuous photon record and a nontrivial classifier.

Inference and feedback

How are uncertain parameters estimated and controls updated?

Feedback occurs on several time scales:

• fast analog or digital feedback within one experimental shot;
• adaptive selection of a later pulse based on an earlier measurement;
• shot-to-shot correction of laser, magnetic-field, or trap drift;
• periodic recalibration of amplitudes, detunings, imaging thresholds, and transfer functions;
• long-term maintenance of vacuum, alignment, and reference standards.

The Laser Stabilization page gives a frequency-domain treatment of a representative laboratory feedback loop.

A Scale Hierarchy for Control

No single figure of merit ranks every AMO platform, but a useful experiment usually establishes several inequalities.

For a coherent operation at scale $\Omega$,

$$\Omega \gg \Gamma_{\mathrm{decoh}}, \Gamma_{\mathrm{loss}}, \Gamma_{\mathrm{heat}},$$

where each rate must be interpreted for the relevant state and mode. The rough number of coherent radians available is $\Omega/\Gamma$, but this ratio is not a gate fidelity. Coherent calibration errors, leakage, crosstalk, and nonstationary noise do not reduce to one exponential rate.

Spectral selectivity often requires

$$\Omega \ll \Delta_{\mathrm{unwanted}},$$

where $\Delta_{\mathrm{unwanted}}$ is the detuning to a leakage transition, motional sideband, neighboring site, or higher band. Fast control and spectral selectivity therefore compete.

To resolve motional sidebands, the trap frequency must exceed the effective transition linewidth and the excitation bandwidth, including power broadening. A large one-photon optical linewidth does not preclude narrow Raman sidebands between long-lived states. Spatial localization, discussed below, is a separate requirement.

For an adiabatic ramp, a useful local diagnostic for each coupled pair of instantaneous eigenstates is

$$\frac{\hbar|\langle m(t)|\dot H(t)|n(t)\rangle|}{|E_m(t)-E_n(t)|^2}\ll1.$$

The changing matrix elements and squared energy gaps both matter; an inverse-gap time alone is insufficient. Endpoint behavior and accumulated transitions also require checking. The ramp must remain short enough to limit decoherence and drift.

The right hierarchy is task dependent:

• cooling compares friction and diffusion;
• trapping compares depth, temperature, oscillation frequencies, and loss;
• coherent gates compare interaction speed, leakage gaps, and noise spectra;
• sensing compares interrogation time, contrast, particle number, cycle time, and systematic shifts;
• simulation compares Hamiltonian scales, temperature, preparation error, and observable resolution.

Comparing physical platforms

The comparison depends on the task and operating conditions. These are characteristic resources and constraints, not a ranking.

Platform	Useful interaction or degree of freedom	Common constraints
Neutral atoms	Collisions, lattice tunneling, Rydberg interactions, optical control of geometry	Temperature, vacancies, scattering, site-dependent shifts
Trapped ions	Internal states coupled through collective Coulomb motion	Heating, micromotion, spectator modes, optical phase stability
Cold molecules	Rotation, parity structure, electric dipoles, internal-state-dependent chemistry	Dense spectra, branching into dark states, reactive loss
Cavity systems	Repeated emitter–field coupling and photon-mediated interactions	Photon loss, mode matching, spatially varying coupling
Traveling photons	Path, polarization, time, and frequency modes	Loss and weak deterministic interactions between photons
Superconducting circuits	Engineered nonlinear circuits coupled to microwave resonators	Materials, cryogenic wiring, parameter variations, multilevel leakage

Atoms of one isotope have reproducible internal structure, but their surrounding fields need not be uniform. Molecules add rich rotational and vibrational structure along with additional preparation and detection challenges. Circuits illustrate many of the same Hamiltonians with fabricated parameters and different dominant noise sources.

For sensing with moving atoms, Atom Interferometry explains how pulses split and recombine matter-wave amplitudes. Its familiar acceleration phase $\mathbf k_{\mathrm{eff}}\cdot\mathbf aT^2$ assumes a specified pulse geometry; laser phase, vibration, rotation, wavefronts, and finite pulse duration also enter the measured phase.

Cooling and Trapping

Cooling and trapping answer different questions.

• Cooling narrows or reshapes a momentum or energy distribution and exports entropy to another degree of freedom.
• Trapping creates a restoring potential or dynamical confinement that keeps particles in a chosen region of phase space.

A conservative trap does not cool by itself. Loading a hot distribution into a static potential may select low-energy particles, but selection and thermalization must be distinguished from dissipative cooling.

Temperature and phase-space density

For a nondegenerate gas in thermal equilibrium, temperature characterizes a kinetic-energy distribution. The thermal de Broglie wavelength may be written

$$\lambda_{\mathrm{dB}} = \frac{h}{\sqrt{2\pi m k_{\mathrm B}T}},$$

and the phase-space density is

$$\mathcal D = n\lambda_{\mathrm{dB}}^3.$$

Reducing $T$ raises $\mathcal D$, but atom loss can lower $n$. Successful evaporative cooling depends on rethermalization and on increasing phase-space density despite deliberate particle removal. A colder cloud is not automatically closer to degeneracy if its density has collapsed.

For a nondegenerate gas in a harmonic trap, with $k_{\mathrm B}T\gg\hbar\omega_x$ so that the motion along this axis is classical, the variance about the trap center is

$$\langle x^2\rangle = \frac{k_{\mathrm B}T}{m\omega_x^2}.$$

This relation is useful for thermometry only when the harmonic and thermal assumptions are justified and imaging resolution is included.

Dissipation by scattered photons

Laser cooling uses momentum-selective absorption and spontaneous emission to transfer entropy from atomic motion to the electromagnetic environment. A single absorption changes momentum by approximately $\hbar\mathbf k$; spontaneous emission randomizes recoil directions and produces momentum diffusion.

The mean force can be expanded near small velocity as

$$\mathbf F(\mathbf v) \simeq \mathbf F_0 -\boldsymbol{\alpha}\mathbf v,$$

where a positive symmetric part of $\boldsymbol\alpha$ makes the frictional contribution to $d(mv^2/2)/dt$ negative. In the isotropic case this reduces to a positive scalar friction coefficient. The final temperature follows from the competition between friction and diffusion, not from friction alone. Doppler, polarization-gradient, Raman-sideband, and dark-state cooling use different internal-state and motional structures, so they have different limits and validity conditions.

Conservative optical confinement

For a polarizable particle in a far-detuned monochromatic field, a common cycle-averaged form of the dipole potential is

$$U(\mathbf r) \simeq -\frac{1}{4} \operatorname{Re}\alpha(\omega) |\mathbf E(\mathbf r)|^2,$$

Here $\mathbf E$ is the peak complex field amplitude in $\mathbf E_{\mathrm{phys}}(t)=\operatorname{Re}[\mathbf E e^{-i\omega t}]$, and $\alpha$ is the polarizability for the chosen state and polarization. Red-detuned light usually attracts a ground-state atom toward high intensity; blue-detuned light usually repels it. Near an isolated transition, take laser-minus-transition detuning $\Delta=\omega-\omega_0$ with linewidth and Rabi frequency much smaller than $|\Delta|$, but $|\Delta|\ll\omega_0$. In that regime the useful scaling is

$$U\propto\frac{I}{\Delta}, \qquad \Gamma_{\mathrm{sc}} \propto \frac{I}{\Delta^2}.$$

At fixed trap depth, increasing $|\Delta|$ while increasing intensity reduces the scattering rate approximately as $1/|\Delta|$. Real atoms require sums over transitions, scalar/vector/tensor polarizabilities, and careful polarization conventions.

Magnetic and electrodynamic confinement

Magnetic traps use spatially varying Zeeman energy. They are naturally state-selective and can suffer loss near field zeros. Magneto-optical traps combine velocity-dependent radiation pressure with a position-dependent Zeeman shift; they are dissipative capture devices, not conservative equilibrium traps.

In a charge-free region, a static electrostatic potential obeys Laplace’s equation and has no strict local maximum or minimum. A point charge therefore cannot be stably trapped in all three directions by electrostatic forces alone. Radio-frequency Paul traps use time-dependent quadrupole fields and an effective pseudopotential, while Penning traps combine static electric and magnetic fields. Stability, secular motion, and micromotion must be distinguished.

Entropy has to go somewhere

Every cooling method identifies an entropy sink:

• spontaneous photons in laser cooling;
• selectively removed particles in evaporative cooling;
• a colder auxiliary species in sympathetic cooling;
• a lossy cavity field in cavity-assisted cooling;
• a resettable internal state in algorithmic or sideband cooling.

Claims of cooling should state what is cooled, how its distribution is measured, and where entropy is exported.

Coherent Control

Internal-state rotations

For a driven two-level subspace in a rotating frame, one useful convention is

$$H_{\mathrm{rot}}(t) = \frac{\hbar}{2} \left[ -\delta(t)\sigma_z + \Omega_x(t)\sigma_x + \Omega_y(t)\sigma_y \right].$$

Writing

$$\Omega_x+i\Omega_y = \Omega e^{i\phi}$$

shows that pulse amplitude sets a rotation rate while optical or microwave phase sets an equatorial rotation axis. On resonance, a pulse with fixed phase has rotation angle

$$\Theta = \int \Omega(t)\,dt.$$

This ideal reduction omits leakage levels, counter-rotating terms, inhomogeneity, motion, differential light shifts, and source noise. Those omissions are acceptable only when bounded.

Raman and multiphoton control

Two optical fields can couple long-lived states through an off-resonant intermediate level. In a simple large-detuning limit,

$$\Omega_{\mathrm{eff}} \sim \frac{\Omega_1\Omega_2^*}{2\Delta},$$

up to convention-dependent factors and multilevel sums. Large detuning suppresses intermediate-state population but requires more optical power for a fixed effective rate. Differential AC Stark shifts and spontaneous scattering remain coupled design constraints.

Motion as a resource

For a harmonic mode of frequency $\omega_t$,

$$x = x_0(a+a^\dagger), \qquad x_0 = \sqrt{\frac{\hbar}{2m\omega_t}}.$$

A laser with effective wavevector difference $\Delta k$ couples to motion through the Lamb–Dicke parameter

$$\eta = \Delta k\,x_0.$$

The Lamb–Dicke condition $|\eta|\sqrt{2n+1}\ll1$ for significantly occupied motional levels permits expansion of the spatial phase. Resolving carrier and sideband frequencies additionally requires sufficiently narrow excitation; the Lamb–Dicke condition alone does not ensure spectral resolution. This enables ground-state cooling and motional entangling gates in trapped ions and neutral atoms. It also makes laser geometry and motional occupation part of the internal-state control problem.

Engineered interactions

Different platforms create interactions in different ways:

Resource	Representative scale	What it enables	What must be controlled
Contact collisions	$g_{\mathrm{3D}}=4\pi\hbar^2a_s/m$	thermalization, nonlinear dynamics, Hubbard interactions	scattering length, density, loss, dimensionality
Dipole interactions	$C_3(1-3\cos^2\theta)/R^3$	anisotropic spin and motional models	orientation, short-range loss, state purity
Rydberg van der Waals interaction	$C_6/R^6$	blockade, gates, long-range spin models	laser detuning, lifetime, motion, pair resonances
Ion phonon mediation	mode-dependent spin–motion coupling	entangling gates, effective spin models	mode spectrum, heating, detuning, spectator modes
Cavity mediation	functions of $g$, $\kappa$, $\gamma$, and detuning	collective coupling, nondestructive readout, networks	photon loss, mode profile, cooperativity

An effective interaction should be accompanied by its derivation regime. Eliminating an excited state, motional mode, or cavity field also transforms noise and measurement channels.

Open-loop, closed-loop, and robust control

• Open-loop control applies a precomputed waveform.
• Feedback control updates a waveform using a measurement record.
• Feedforward applies a correction based on a measured disturbance or earlier outcome without comparing the final output to a setpoint.
• Robust control designs acceptable performance across an uncertainty set.
• Optimal control minimizes a declared cost functional subject to a model and constraints.

An optimized pulse is only as trustworthy as its model, transfer-function calibration, and out-of-sample validation. Further derivations are given in Optimal Control, Dynamical Decoupling, and Control Limits and Noise.

Measurement

Fluorescence and shelving

State-dependent fluorescence maps an internal state onto a bright or dark photon-count distribution. If $n$ photons are detected in a time window, a threshold rule may classify the state. The two distributions overlap because of photon shot noise, background counts, imperfect cycling, state decay, and off-resonant pumping.

The resulting binary POVM can be parameterized by assignment probabilities:

$$E_{\mathrm B} = p(\mathrm B|0)|0\rangle\langle0| + p(\mathrm B|1)|1\rangle\langle1|,$$ $$E_{\mathrm D} = I-E_{\mathrm B},$$

when coherences do not affect the detector. This is a detector model, not necessarily the state-update rule.

Ensemble imaging

Absorption imaging infers optical depth from incident and transmitted intensity. Fluorescence imaging detects scattered light. Dispersive imaging uses a phase or polarization shift and can reduce absorption, although it still causes backaction.

Quantitative imaging requires:

• saturation and detuning corrections;
• an effective absorption cross section;
• polarization and magnetic-field control;
• optical resolution and depth of field;
• camera gain, offset, nonlinearity, and noise;
• multiple-scattering and high-optical-depth corrections when relevant.

Destructive time-of-flight imaging gives momentum information after expansion but does not directly photograph the in-trap momentum operator. The inference depends on interactions and expansion dynamics.

Cavity and continuous measurement

A cavity converts an atomic dispersive shift or absorption into changes in an output field. Homodyne or heterodyne detection retains field-quadrature information. The measurement rate, quantum efficiency, cavity bandwidth, and backaction determine what can be inferred in real time.

Continuous records should not be interpreted as noiseless trajectories of a pre-existing classical state. Their conditioned quantum evolution is treated by Stochastic Master Equations.

Calibration hierarchy

A mature measurement claim separates:

1. raw record: camera counts, arrival times, voltages, or digitizer samples;
2. detector model: likelihood $p(y|x,\boldsymbol{\eta})$ with nuisance parameters $\boldsymbol{\eta}$;
3. state or parameter estimator: the rule mapping records to an estimate;
4. uncertainty: statistical and systematic components;
5. validation: held-out calibrations, null experiments, injected signals, and consistency checks.

Correcting an assignment matrix can reduce bias but amplify variance and calibration error. It cannot recover information erased by a singular detector response.

Quantum Simulation

A quantum simulator uses a controlled quantum system to learn about another Hamiltonian, dynamical process, or universality class.

Analog simulation

An analog simulator engineers a Hamiltonian whose low-energy or dynamical behavior approximates a target model. Neutral atoms in optical lattices can realize Hubbard-type models; trapped ions can realize tunable spin interactions; Rydberg arrays can realize constrained spin models; molecules can realize anisotropic dipolar models.

For a single-band optical lattice, a reduction may lead to

$$H = -\sum_{\langle i,j\rangle,\sigma} \left(t_{ij}c_{i\sigma}^\dagger c_{j\sigma}+t_{ij}^*c_{j\sigma}^\dagger c_{i\sigma}\right) + U\sum_i n_{i\uparrow}n_{i\downarrow} + \sum_i V_i n_i.$$

Each unordered neighboring pair is counted once, and the second hopping term is the Hermitian conjugate of the first. The Hubbard Model develops this Hamiltonian. A platform implementation must additionally establish:

• occupation of the intended lattice band;
• values and inhomogeneity of $t_{ij}$, $U$, and $V_i$;
• temperature or entropy;
• preparation history;
• detection fidelity and finite-size boundaries;
• residual long-range interactions, loss, and heating.

Digital simulation

Digital simulation decomposes evolution into calibrated gates. Its errors include gate infidelity, leakage, crosstalk, idle evolution, measurement error, and approximation error from product formulas or other algorithms.

The analog/digital distinction is not absolute. Floquet engineering, stroboscopic analog blocks, measurement-based protocols, and variational circuits interpolate between them.

Validation

A simulator is not validated merely because its data look physically plausible. Useful checks include:

• exactly solvable limits;
• independent observables tied by conservation laws or sum rules;
• convergence with Trotter step, lattice depth, system size, or bond dimension where applicable;
• comparison with controlled classical calculations in overlapping regimes;
• randomized or blinded calibration tests;
• forward prediction of data not used for fitting;
• explicit sensitivity to uncertain Hamiltonian parameters.

Quantum advantage and physical insight are separate claims. A platform can teach valuable physics before it exceeds every classical computation.

For quantum-information tasks, specify the computational subspace, leakage states, and the allowed preparation, gate, and readout operations. A long idle coherence time does not determine a gate fidelity: coherent calibration errors, loss, crosstalk, and correlations can dominate during an operation. Quantum Control in AMO develops these distinctions.

For sensing, the observable may be a phase $\phi(T)=\int_0^T\Delta\omega(t)\,dt$. The Precision Measurement and Metrology guide explains how phase sensitivity, averaging time, systematic shifts, and accuracy differ. Neither a narrow atomic line nor a large atom number alone determines the final uncertainty.

Worked Architecture: A Neutral-Atom Experiment

Consider a programmable neutral-atom sequence using alkali atoms in optical tweezers. The purpose is not to prescribe one apparatus but to show how the layers fit.

1. Source, slowing, and capture

An atomic source feeds a slowed beam or vapor-cell loading region. A magneto-optical trap captures atoms using radiation-pressure forces made position dependent by Zeeman shifts in a magnetic-field gradient. Relevant observables include atom number, temperature, cloud size, loading rate, and loss rate.

2. Transfer to tweezers

Optical tweezers provide focused conservative microtraps. Loading is probabilistic in a simple collisional-blockade regime, so an image identifies occupied sites and moving tweezers rearrange atoms into a target geometry.

The transfer error budget includes:

• trap-depth and waist calibration;
• differential light shifts;
• photon scattering;
• heating during motion;
• atom loss;
• occupancy-classification errors.

3. Internal-state preparation

Optical pumping prepares a hyperfine or Zeeman state. A bias field defines a quantization axis. Microwave or Raman spectroscopy calibrates the transition frequency and inhomogeneous detuning.

The prepared density operator is not exactly $|0\rangle\langle0|^{\otimes N}$. Residual population, thermal motion, vacancies, and correlated laser errors should be represented or bounded.

4. Coherent operation

Single-particle rotations use microwave or Raman fields. Entangling dynamics may transiently excite a Rydberg state. If the pair interaction shift $V(R)$ greatly exceeds the relevant excitation scale, double excitation is suppressed: the blockade regime.

The inequality $|V|\gg\hbar|\Omega|$ is necessary but not sufficient for a high-fidelity operation. Finite Rydberg lifetime, Doppler shifts, intermediate state scattering, laser phase noise, position fluctuations, and neighboring atoms all contribute.

5. Readout

State-dependent removal or fluorescence maps internal states to site occupancy or photon counts. The analysis must distinguish:

• a qubit in the dark state;
• an atom physically lost before detection;
• an atom removed intentionally;
• a bright state misclassified as dark;
• a dark state pumped bright during readout.

A ternary model can be more honest than forcing every outcome into one bit.

6. Validation

Single-atom Rabi and Ramsey data calibrate local controls. Pair experiments calibrate interaction shifts. Randomized sequences probe accumulated error. Independent imaging calibrations estimate assignment and loss matrices. Exactly solvable small arrays and symmetry checks validate many-body observables.

This architecture illustrates a general lesson: every headline capability rests on preparation, calibration, and measurement layers that are themselves quantum experiments.

Exercises

1. Coherent-operation budget

A control has $\Omega/2\pi=100\ \mathrm{kHz}$. During the operation, phase-flip jumps $\sigma_z$ occur as a Poisson process of rate $r_\phi=200\ \mathrm{s}^{-1}$, while an independent Poisson process removes the atom at rate $r_\ell=50\ \mathrm{s}^{-1}$. The corresponding free coherence-decay rate from these phase flips alone is $2r_\phi$, not $r_\phi$.

1. Estimate the duration of an ideal resonant $\pi$ pulse.
2. Estimate the probability of at least one dephasing-or-loss event during the pulse by using $(r_\phi+r_\ell)t_\pi\ll1$.
3. Explain why this estimate is not a gate-fidelity prediction.

Solution

The angular Rabi frequency is

$$\Omega = 2\pi(100\times10^3)\ \mathrm{s}^{-1}.$$

An ideal $\pi$ pulse takes

$$t_\pi = \frac{\pi}{\Omega} = \frac{1}{2(100\times10^3)} = 5.0\ \mu\mathrm{s}.$$

The small-event-probability estimate is

$$p_{\mathrm{event}} \simeq (200+50)(5.0\times10^{-6}) = 1.25\times10^{-3}.$$

This is only a rough decoherence-and-loss contribution. It omits coherent overrotation, detuning, pulse-shape distortion, leakage, crosstalk, state dependence of the rates, and the precise way a dephasing event affects the chosen fidelity measure. The ratio $\Omega/\Gamma$ is a scale diagnostic, not a complete gate benchmark.

2. Thermal width in a harmonic trap

Noninteracting atoms of mass $m$ obey Maxwell–Boltzmann statistics and have $k_{\mathrm B}T\gg\hbar\omega_x$. They are in thermal equilibrium in $V(x)=m\omega_x^2x^2/2$.

1. Derive $\langle x^2\rangle=k_{\mathrm B}T/(m\omega_x^2)$.
2. If an imaging system has an independent Gaussian point-spread width $\sigma_{\mathrm{PSF}}$, what width is measured?

Solution

The classical position distribution is

$$p(x) \propto \exp\left( -\frac{m\omega_x^2x^2}{2k_{\mathrm B}T} \right).$$

Comparison with a Gaussian $\exp[-x^2/(2\sigma_x^2)]$ gives

$$\sigma_x^2 = \langle x^2\rangle = \frac{k_{\mathrm B}T}{m\omega_x^2}.$$

Convolution of independent Gaussian widths adds variances:

$$\sigma_{\mathrm{meas}}^2 = \sigma_x^2+\sigma_{\mathrm{PSF}}^2.$$

Ignoring the point-spread function biases the inferred temperature upward. The derivation also fails when the trap is anharmonic, the gas is degenerate, the motional spacing is not small relative to $k_{\mathrm B}T$, or the state is not in thermal equilibrium.

3. Detuning error in a nominal pi pulse

A rectangular pulse is designed as a resonant $\pi$ pulse with duration $t_\pi=\pi/\Omega$, but the actual detuning is a constant $\delta$.

1. Write the excited-state probability.
2. Expand the final transfer error through order $(\delta/\Omega)^2$.

Solution

For initial ground state, the generalized Rabi frequency is

$$\Omega_R = \sqrt{\Omega^2+\delta^2},$$

and

$$P_e(t) = \frac{\Omega^2}{\Omega_R^2} \sin^2\left(\frac{\Omega_Rt}{2}\right).$$

At $t=t_\pi=\pi/\Omega$, let $\epsilon=\delta/\Omega$. Then

$$P_e(t_\pi) = \frac{1}{1+\epsilon^2} \sin^2\left( \frac{\pi}{2}\sqrt{1+\epsilon^2} \right).$$

The pulse-area argument changes only at order $\epsilon^2$, and its effect on the sine squared begins at order $\epsilon^4$. The prefactor gives

$$P_e(t_\pi) = 1-\epsilon^2+O(\epsilon^4).$$

Therefore the leading transfer error is

$$1-P_e \simeq \left(\frac{\delta}{\Omega}\right)^2.$$

4. Correcting a binary readout

Let the true populations be $\mathbf p=(p_0,p_1)^{\mathsf T}$ and the observed frequencies be $\mathbf f=(f_0,f_1)^{\mathsf T}$. The assignment matrix is

$$A = \begin{pmatrix} 0.98 & 0.07\\ 0.02 & 0.93 \end{pmatrix}, \qquad \mathbf f=A\mathbf p.$$

For $\mathbf f=(0.61,0.39)^{\mathsf T}$, estimate $\mathbf p$. Why should this correction not be applied without uncertainty propagation?

Solution

The determinant is

$$\det A = (0.98)(0.93)-(0.07)(0.02) = 0.9100.$$

Thus

$$A^{-1} = \frac{1}{0.9100} \begin{pmatrix} 0.93 & -0.07\\ -0.02 & 0.98 \end{pmatrix}.$$

Applying it,

$$p_0 = \frac{0.93(0.61)-0.07(0.39)}{0.9100} \simeq 0.593,$$ $$p_1 = \frac{-0.02(0.61)+0.98(0.39)}{0.9100} \simeq 0.407.$$

The inverse amplifies shot noise and inherits uncertainty from every calibrated element of $A$. If the detector response drifts or if atom loss is not represented by this two-outcome model, the corrected result can remain biased. A likelihood model with calibration nuisance parameters is usually more reliable than treating $A$ as exact.

References

• A. Blais, A. L. Grimsmo, S. M. Girvin, and A. Wallraff, “Circuit quantum electrodynamics,” Reviews of Modern Physics 93, 025005 (2021), doi:10.1103/RevModPhys.93.025005.
• I. Bloch, J. Dalibard, and W. Zwerger, “Many-body physics with ultracold gases,” Reviews of Modern Physics 80, 885–964 (2008), doi:10.1103/RevModPhys.80.885.
• C. Brif, R. Chakrabarti, and H. Rabitz, “Control of quantum phenomena: past, present and future,” New Journal of Physics 12, 075008 (2010), doi:10.1088/1367-2630/12/7/075008.
• M. Brownnutt, M. Kumph, P. Rabl, and R. Blatt, “Ion-trap measurements of electric-field noise near surfaces,” Reviews of Modern Physics 87, 1419–1482 (2015), doi:10.1103/RevModPhys.87.1419.
• L. D. Carr, D. DeMille, R. V. Krems, and J. Ye, “Cold and ultracold molecules: science, technology and applications,” New Journal of Physics 11, 055049 (2009), doi:10.1088/1367-2630/11/5/055049.
• A. D. Cronin, J. Schmiedmayer, and D. E. Pritchard, “Optics and interferometry with atoms and molecules,” Reviews of Modern Physics 81, 1051–1129 (2009), doi:10.1103/RevModPhys.81.1051.
• C. J. Foot, Atomic Physics (Oxford University Press, 2005).
• S. Giorgini, L. P. Pitaevskii, and S. Stringari, “Theory of ultracold atomic Fermi gases,” Reviews of Modern Physics 80, 1215–1274 (2008), doi:10.1103/RevModPhys.80.1215.
• 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), doi:10.1016/S1049-250X(08)60186-X.
• S. Haroche and J.-M. Raimond, Exploring the Quantum: Atoms, Cavities, and Photons (Oxford University Press, 2006).
• D. Leibfried, R. Blatt, C. Monroe, and D. Wineland, “Quantum dynamics of single trapped ions,” Reviews of Modern Physics 75, 281–324 (2003), doi:10.1103/RevModPhys.75.281.
• A. D. Ludlow, M. M. Boyd, J. Ye, E. Peik, and P. O. Schmidt, “Optical atomic clocks,” Reviews of Modern Physics 87, 637–701 (2015), doi:10.1103/RevModPhys.87.637.
• H. J. Metcalf and P. van der Straten, Laser Cooling and Trapping (Springer, 1999).
• M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, 10th anniversary ed. (Cambridge University Press, 2010).
• W. D. Phillips, “Laser cooling and trapping of neutral atoms,” Reviews of Modern Physics 70, 721–741 (1998), doi:10.1103/RevModPhys.70.721.
• M. Saffman, T. G. Walker, and K. Mølmer, “Quantum information with Rydberg atoms,” Reviews of Modern Physics 82, 2313–2363 (2010), doi:10.1103/RevModPhys.82.2313.