01 How the course works
The course follows the site's three stories in order: build a lab, run a first experiment with single atoms, then turn the atoms into qubits and link them with light. Students meet each piece of physics first in a lecture and then in a page they can push on.
Senior undergraduates and new graduate students
Prerequisites: undergraduate quantum mechanics and electromagnetism, and enough Python to read a short script. No lab experience is assumed. A first-year student joining an AMO group can also follow it alone, one week at a time.
Two lectures, the pages, one demonstration
Lecture 1 sets up the physics; Lecture 2 turns it into the engineering decision a lab actually makes. The listed pages are the demonstrations and the homework. The "in class" line suggests a live demonstration that works on a projector in a few minutes.
Three kinds of problem
Use the tool find a number or a behaviour with the page. By hand derive or check it from a formula, so the tool is never a black box. Design make a decision a lab would make and defend it.
Core equations, scales, and limiting regimes.
Hardware trade-offs, noise budgets, and species choices.
Push sliders live on the toolkit page in front of the class.
Tool exploration, analytical check, and lab design defense.
02 The fourteen weeks
Weeks 1–4 build the apparatus, weeks 5–10 trap, image and cool single atoms, and weeks 11–14 turn them into qubits, gates and networks. Week 7 is a natural midterm; week 14 ends with project presentations.
Atoms, light and the two-level system
- Lecture 1
- Two-level atom, Rabi oscillations, the Bloch sphere, detuning and the generalised Rabi frequency.
- Lecture 2
- Choosing an atom: linewidths, recoil, hyperfine structure, bosons and fermions, what each species is good for.
- In class
- Drive Rabi oscillations live while changing the detuning; watch the Bloch vector tilt and the contrast drop.
Lasers and polarisation
- Lecture 1
- Which lasers an experiment needs and why: cooling, repump, imaging, trap and Rydberg light; AOM and fibre schemes.
- Lecture 2
- Jones vectors, wave plates, Stokes parameters; circular light and the quantisation axis.
- In class
- Rotate a quarter-wave plate in the polarimetry simulator and follow the state round the Poincaré sphere.
Locking lasers to atoms and cavities
- Lecture 1
- Saturated absorption, error signals, PID servos and what limits them.
- Lecture 2
- Offset locks with beat notes, Pound–Drever–Hall locking to a cavity, ultrastable references.
- In class
- Perturb a free-running laser, then close each lock in turn and compare how it recovers.
Vacuum and the lab bench
- Lecture 1
- Why ultra-high vacuum: background collisions, pumping chains, conductance, bake-out.
- Lecture 2
- Everyday optics and RF: fibre coupling, AOMs, dBm budgets, Zeeman shifts, Clebsch–Gordan coefficients.
- In class
- Work a real fibre-coupling and AOM double-pass layout with the calculators open on the projector.
Laser cooling and the MOT
- Lecture 1
- Radiation pressure, optical molasses, the Doppler limit; sub-Doppler and gray-molasses cooling.
- Lecture 2
- The magneto-optical trap: capture velocity, detuning and gradient trade-offs, species differences.
- In class
- Sweep the MOT detuning and show the capture velocity and temperature pulling in opposite directions.
Imaging clouds and measuring temperature
- Lecture 1
- Absorption imaging: Beer–Lambert, optical depth, saturation and atom-number calibration.
- Lecture 2
- Time of flight: ballistic expansion, fitting a temperature, the recoil and Doppler scales.
- In class
- Fit a temperature from simulated time-of-flight points entered by the class.
Controlling the experiment
- Lecture 1
- Hardware timing, TTL/DAC/ADC/DDS/AWG signals, real device delays, pulse shapes and their spectra.
- Lecture 2
- ARTIQ, labscript and NI systems; writing a shot in Python; git for a shared experiment. Midterm review.
- In class
- Toggle real versus compensated delays in the shot diagram; then compare square and Blackman pulse spectra.
Optical tweezers and wavefronts
- Lecture 1
- Dipole potentials, trap depth and frequencies, photon scattering from trap light, array geometry.
- Lecture 2
- Aberrations and Zernike modes, Strehl ratio, SLM holograms and the Gerchberg–Saxton algorithm.
- In class
- Change the objective NA and watch depth, trap frequencies and the Lamb–Dicke parameter move together.
Seeing and measuring single atoms
- Lecture 1
- Fluorescence imaging of one atom: scattering rate, collection efficiency, camera noise, detection fidelity.
- Lecture 2
- Release and recapture: measuring a single atom's temperature from a survival curve.
- In class
- Move the threshold across the bright and dark histograms and watch the two error rates trade.
Ground-state cooling and the full cycle
- Lecture 1
- Lamb–Dicke regime, resolved and unresolved sideband cooling, Raman and gray-molasses schemes.
- Lecture 2
- A whole experimental cycle: loading, imaging, rearrangement and the exponential wall of large arrays.
- In class
- Run 1,000 virtual shots and sweep the target array size to show the success probability collapse.
Coherence and dynamical decoupling
- Lecture 1
- T₁, T₂* and T₂; dephasing versus dissipation; how each is measured.
- Lecture 2
- Spin echo, CPMG and XY sequences; filter functions and noise spectroscopy.
- In class
- Compare a Ramsey decay and an echo decay for the same noise, then add CPMG pulses.
Rydberg atoms and two-qubit gates
- Lecture 1
- Rydberg scaling laws, van der Waals interactions, the blockade radius and array geometry.
- Lecture 2
- CZ gates and their error budget: decay, Doppler dephasing, laser noise, blockade leakage, SPAM.
- In class
- Raise n and the Rabi frequency in turn and watch which line of the error budget dominates.
Benchmarking, processors and automation
- Lecture 1
- Randomized and interleaved benchmarking; zoned architectures and atom transport.
- Lecture 2
- Running a lab as a system: feedback and feedforward, calibration loops, optimisers, machine learning and AI agents.
- In class
- Push the power-servo bandwidth until the loop oscillates, then add feedforward.
Light–matter interfaces, networks and the field
- Lecture 1
- Cavity QED: coupling rates, cooperativity, Purcell enhancement; atom–photon entanglement and remote links.
- Lecture 2
- Where the field is going: reading the literature, platforms and companies. Final project presentations.
- In class
- Fire 2,000 photons in the cavity-QED race in each coupling regime and compare the escape fractions.
03 Problem sets
One set per week, with problems for every page that week uses. Most take 20–40 minutes each. Filter by problem style, search by keyword, check off problems as you solve them, or expand all to print a full packet.
Week 01 Atoms, light and the two-level system 6 problems
🔵 Quantum Computing Learning Path
Open the page →- Use the tool In the Rabi oscillation explorer set Ω/2π = 1 MHz and Δ = 0 and read off the π-pulse time. Then set Δ/2π = 1 MHz. How do the maximum excited-state population and the oscillation frequency change?
- By hand Derive Pmax = Ω²/(Ω² + Δ²) and the generalised Rabi frequency Ω′ = √(Ω² + Δ²) and check both against your readings in problem 1.
- Design With optimal angles the CHSH game is won with probability cos²(π/8) ≈ 0.854, against 0.75 for any classical strategy. Play 1,000 rounds. How many rounds would you need to separate the quantum and classical win rates by five standard deviations?
⚛️ Atomic Species Selector
Open the page →- Use the tool Use “Compare two” for ⁸⁷Rb and ¹³³Cs. Tabulate the cooling wavelength, linewidth, Doppler limit and recoil temperature for each.
- By hand Check one Doppler limit with TD = ħΓ/2kB and one recoil temperature with Tr = ħ²k²/(2m kB) (the library's single-photon convention; some texts quote twice this). Explain from the formulas why lithium has the highest recoil temperature of the alkalis.
- Design Choose a species for (a) a tweezer quantum computer with a nuclear-spin qubit, (b) an optical clock and (c) a dipolar quantum gas. Justify each choice with two numbers from the library.
Week 02 Lasers and polarisation 6 problems
💡 Laser System Planner
Open the page →- Use the tool Generate the laser plan for ⁸⁷Rb. List every beam the planner proposes with its wavelength and the role it plays in the experiment.
- By hand The repumper drives F = 1 → F′ = 2 while the cooler drives F = 2 → F′ = 3. Using the ground and excited hyperfine splittings, compute the frequency offset between them and the EOM frequency needed to make the repumper as a sideband of the cooling laser.
- Design Generate the plan for ¹⁷¹Yb and compare it with ⁸⁷Rb: number of lasers, wavelengths that need frequency doubling, and narrow-line requirements. Which system would you rather build on a fixed budget, and why?
🔭 Polarimetry Simulator
Open the page →- Use the tool Send horizontal light through a quarter-wave plate with its fast axis at 45°. Predict the Stokes vector, then check it in the simulator.
- By hand Show with Jones matrices that a half-wave plate at angle θ rotates linear polarisation by 2θ. Confirm with the simulator at θ = 22.5°.
- Design Optical pumping into a stretched state needs pure σ⁺ light along the quantisation axis. Describe how you would measure and optimise circular purity with a polarising beam splitter, a quarter-wave plate and a power meter, and what value of S₃ you would aim for.
Week 03 Locking lasers to atoms and cavities 3 problems
🔐 Laser Locking Guide
Open the page →- Use the tool In the saturated-absorption tab, raise the vapour-cell temperature from 300 K to 330 K. Describe what happens to the Doppler background and the sub-Doppler features, and explain both.
- By hand Two beams of 1 mW each fall on a photodiode with responsivity 0.5 A/W. Compute the amplitude of the beat-note photocurrent, 2R√(P₁P₂), and compare with the beat-note calculator.
- Design A cavity is 100 mm long with finesse 10,000 at 780 nm. Compute its free spectral range and linewidth, choose a PDH modulation frequency, and justify the choice. Check your numbers with the PDH tab.
Week 04 Vacuum and the lab bench 9 problems
⚗️ Vacuum Systems Guide
Open the page →- Use the tool With a 100 L/s pump behind a 38 mm bore, 150 mm long tube, find the effective pumping speed at the chamber. Now double the tube length. What limits the speed, the pump or the tube?
- By hand From Section 03, list the pumping chain from atmosphere to ultra-high vacuum with the pressure range of each stage. Explain why an ion pump cannot be started at atmosphere.
- Design Your tweezer lifetime drops from 20 s to 5 s overnight. Using Section 06, describe how you would use the ion-pump current, a residual-gas analyser and the atoms themselves to tell a real leak from outgassing or a virtual leak.
🔬 AMO Lab Operations Handbook
Open the page →- Use the tool Use the fibre-coupling calculator for 780 nm light, a 1 mm input beam and a 5 µm mode-field diameter. What focal length does it suggest?
- By hand Derive the same focal length from f = πwinwf/λ with wf = MFD/2, and estimate how much coupling you lose if you only have a lens 30% longer.
- Design Open the debugging drawer and pick one pitfall you have met (or could meet) in a real lab. Write the five-step check you would follow, in the order you would follow it.
🧮 Quick Lab Console
Open the page →- Use the tool Use the Zeeman calculator for gF = 1/2, mF = 2 at 1 G, then at 10 G.
- By hand Check problem 1 with ΔE/h = gFmFμBB/h, using μB/h ≈ 1.4 MHz/G. Then plan an RF chain: a 10 dBm driver must deliver 2 W (33 dBm) to an AOM through a 2 dB cable loss. How much amplifier gain is needed? Check it with the dBm calculator.
- Design Using the Clebsch–Gordan table, compare the σ⁺ transition strengths from mF = 0 and mF = 2 on the ⁸⁷Rb D2 line (F = 2 → F′ = 3). Explain why experiments cycle on the stretched transition for imaging.
Week 05 Laser cooling and the MOT 6 problems
❄️ Laser Cooling Simulator
Open the page →- Use the tool In the Doppler tab at s₀ = 0.1, vary δ/Γ and find where the damping (the slope of F(v) at v = 0) is largest. Show that it peaks near |δ| = (Γ/2)√((1 + s₀)/3), which tends to Γ/2√3 at low intensity, not at the δ = −Γ/2 that minimises the Doppler temperature, and explain why the two differ.
- Use the tool In the sideband tab with η = 0.15 and Γ/ω = 0.05, cool from ⟨n⟩ = 20. Then raise Γ/ω to 0.5 and repeat. Explain why the final ⟨n⟩ gets worse.
- Design Lithium's excited-state hyperfine structure is unresolved, so ordinary sub-Doppler cooling works poorly. Use the comparison tab to choose a cooling route for ⁶Li and for ¹³³Cs, and explain the difference.
🧲 MOT Designer
Open the page →- Use the tool For ⁸⁷Rb with the default beams, step the detuning from 1 Γ to 3 Γ. Record the capture velocity and the Doppler temperature at each step and describe the trade-off.
- By hand Check one temperature with T = (ħΓ/4kB)(1 + s + 4δ²/Γ²)/(2|δ|/Γ) and confirm that it is minimised at |δ| = Γ/2 when s ≪ 1.
- Design Use Section 03 to compare a lithium and a cesium MOT. Which parameters must change, and why does a lithium MOT usually need a separate cooling stage afterwards?
Week 06 Imaging clouds and measuring temperature 6 problems
🌑 Absorption Imaging Lab
Open the page →- Use the tool For ⁸⁷Rb, run the imaging calculator at I/Isat = 0.1 and 1.0 with the other settings fixed. How do the effective cross-section, the OD noise and the minimum detectable atom number change, and why?
- By hand With the resonant cross-section σ₀ = 3λ²/2π, compute the peak optical depth of a Gaussian cloud of 10⁶ atoms with an rms radius of 100 µm.
- By hand Each scattered photon gives the atom a recoil velocity vr = ħk/m. How many photons does it take to Doppler-shift a ⁸⁷Rb atom by Γ/2, and what does that imply for the probe pulse length?
🌡️ MOT Temperature
Open the page →- Use the tool Set T = 10 µK and σ₀ = 100 µm, add five synthetic data points between 2 and 15 ms, and fit the temperature. How close is the fit to the input?
- By hand Using σ²(t) = σ₀² + kBTt²/m, find the time at which the cloud size has doubled for your parameters.
- Design A thermal cloud released from an elongated trap becomes round, while a condensate inverts its aspect ratio. Use the aspect-ratio control to explain why, and say what this means for choosing the time of flight.
Week 07 Controlling the experiment 3 problems
🎛️ Experiment Control Systems
Open the page →- Use the tool In the shot diagram, switch from “Real delays” to “Delays compensated”. Which channels moved, by how much, and why does the mechanical shutter need the largest correction?
- Use the tool With T = 10 µs and a neighbouring transition 200 kHz away, compare square and Blackman pulses. Find the shortest Blackman pulse that keeps the drive at the neighbour below −40 dB.
- By hand A 16-bit DAC with a ±10 V range drives a coil with 1 G/V. What is the smallest field step? Is it fine enough for a 1 mG field-noise budget? Check with the signal-budget calculator.
Week 08 Optical tweezers and wavefronts 6 problems
🔦 Tweezer Array Design Lab
Open the page →- Use the tool For ⁸⁷Rb at the default wavelength and power, record the waist, trap depth, radial and axial frequencies and the Lamb–Dicke parameter. Then lower the NA by 30% and record them again.
- By hand Using U ∝ P/w₀², ωr ∝ √U/w₀ and ωz ∝ √U/zR, predict the changes you saw in problem 1 from the change in waist alone.
- Design Design a 10 × 10 array: choose the spacing, check the cross-talk criterion, and compute the total power you need. Is your design limited by the laser or by the optics?
🌊 Zernike Wavefront Lab
Open the page →- Use the tool Build a wavefront with 0.5 λ of coma and read the Strehl ratio. Compare it with the Maréchal estimate S ≈ exp[−(2πσ)²], where σ is the rms wavefront error in waves.
- Use the tool Add an OAM order of l = 1 to a blazed grating and compute the far field. Explain the shape you see.
- Design Run the Gerchberg–Saxton algorithm for a target pattern with 10, 30 and 100 iterations. How does the uniformity change, and how would you check uniformity on the atoms rather than on a camera?
Week 09 Seeing and measuring single atoms 6 problems
📷 Single-atom Imaging
Open the page →- Use the tool With the defaults (NA 0.5, 5 ms exposure), record the scattering rate, the collected photons and the detection fidelity.
- By hand Check the collected fraction with (1 − cos θ)/2, where sin θ = NA, and multiply through by the transmission and quantum efficiency.
- Design Switch to the “Marginal (SNR ≈ 1.6)” preset. Change one thing at a time (exposure, NA, camera type) to reach 99.5% fidelity, and say which change you would make first in a real lab and why.
🎯 Single-atom Temperature
Open the page →- Use the tool Simulate survival curves for 5, 10 and 20 µK. Record the release time at which survival falls to 50% for each.
- By hand Explain with a simple ballistic argument why the 50% release time scales roughly as 1/√T, and check whether your data from problem 1 follow it.
- Design Lower the Strehl ratio from 1 to 0.6. Why does the inferred temperature change if you keep fitting with the ideal trap? What does that mean for a real measurement?
Week 10 Ground-state cooling and the full cycle 6 problems
🧊 Single-atom Cooling
Open the page →- Use the tool Compute the Lamb–Dicke parameter for ⁸⁷Rb in a 100 kHz trap with 780 nm light.
- By hand Check problem 1 with η = k√(ħ/2mω) and explain why a lighter atom in the same trap is harder to cool to the ground state.
- Design Use the technique comparison to choose a ground-state cooling method for ⁶Li and for ¹³³Cs in 100 kHz tweezers, and justify each with the relevant linewidth and Lamb–Dicke parameter.
🧩 Virtual Tweezer Lab
Open the page →- Use the tool Run 1,000 shots and sweep the target block size. At what size does the defect-free success probability fall below 50%?
- By hand If each atom survives the sequence with probability p, a block of N atoms succeeds with probability ≈ pN. Use your data to estimate p, and predict the success probability for N = 100.
- Design You need a 100-atom defect-free block with at least 50% probability per shot. Which settings (loss per move, survival of one image, vacuum lifetime) would you improve first, and by how much?
Week 11 Coherence and dynamical decoupling 6 problems
🌀 Decoherence Lab
Open the page →- Use the tool Watch Sections 03–05. Sketch the Bloch-vector trajectory and the σz signal for pure T₁ decay, pure dephasing and both together, and mark where the vector ends up in each case.
- By hand Show that T₂ ≤ 2T₁ follows from 1/T₂ = 1/2T₁ + 1/Tφ.
- Design A Ramsey signal decays as a Gaussian in time, while the echo signal hardly decays. What does this say about the noise, and which measurement would you do next?
🛡️ Dynamical Decoupling
Open the page →- Use the tool Use the enhancement-factor calculator for α = 1 and α = 2. Fit the slope of T₂(N)/T₂* against N on a log scale.
- By hand Compare your fitted exponents with Nα/(α+1), the scaling expected for S(ω) ∝ ω−α noise.
- Design In the filter-function viewer, where is the passband of CPMG with N pulses in total time T? Design a noise-spectroscopy measurement that maps S(ω) between 1 and 100 kHz.
Week 12 Rydberg atoms and two-qubit gates 6 problems
🔮 Rydberg Blockade Lab
Open the page →- Use the tool For ⁸⁷Rb, compare n = 50 and n = 70: record the C₆ coefficient, the lifetime and the blockade radius at Ω/2π = 1 MHz.
- By hand Using C₆ ∝ n¹¹ and Rb = (C₆/ħΩ)1/6, predict the ratio of blockade radii and compare with problem 1.
- Design Choose an atom spacing and lattice geometry (square or triangular) so that nearest neighbours are blockaded but next-nearest neighbours are not. Check it in the tool.
📊 Rydberg Gate Error Budget
Open the page →- Use the tool With the default settings, identify the largest error source. Double the Rabi frequency and record what changes.
- By hand The Doppler dephasing error grows with temperature. Halve the atom temperature and check that the Doppler term changes as you would predict from the velocity spread.
- Design Starting from the defaults, reach a 99.5% CZ fidelity by changing at most three parameters. Defend each change as achievable in a real lab.
Week 13 Benchmarking, processors and automation 9 problems
📈 Randomized Benchmarking
Open the page →- Use the tool Run standard RB with rC = 0.01 and the default SPAM fidelity. Read off the decay constant p.
- By hand Show that SPAM errors change the amplitude A and offset B of F(m) = Apm + B but not p, and convert p to an average error per Clifford for one qubit.
- Design Use interleaved RB to estimate the error of one gate from rC,ref and rC,int. What assumptions make this estimate unreliable?
🎛️ Zoned Atom Processor
Open the page →- Use the tool Run the same circuit with “In order” and “Optimised” placement. Compare the total transport time and the circuit success probability.
- By hand Estimate the circuit success probability by hand as the product of gate, move and readout fidelities, and compare with 1,000 simulated shots.
- Design Why do zoned architectures separate storage, entangling and readout zones? Give one advantage and one cost of each zone.
🤖 Automating the Experiment
Open the page →- Use the tool In the power-servo simulator with an integral (I) servo, find the bandwidth at which the loop starts to oscillate for loop delays of 0.5, 1 and 2 µs.
- By hand Compare your answers in problem 1 with f ≈ 1/4τ, where τ is the total delay including the AOM response. Explain the phase-margin argument behind it.
- Design Follow the nine-step workflow in Section 06 to write an optimisation plan for one stage of an experiment you know: the figure of merit, its noise, the parameters and bounds, and the optimiser you would choose.
Week 14 Light–matter interfaces, networks and the field 8 problems
💎 Cavity QED Coupling Lab
Open the page →- Use the tool Enter a 10 mm cavity with finesse 10,000 and a 20 µm waist for ⁸⁷Rb at 780 nm. Record g, κ, γ, the cooperativity C and the regime.
- By hand Check g₀ with the page's formula g₀ = √(3πΓc³/2ω²V), V = πw₀²L/4, and verify C = g₀²/κγ from your values, taking care with the half-width conventions for κ and γ.
- Design In the photon race, compare the measured fraction of photons through the output mirror with C/(1 + C)·κ/(κ + γ). What finesse and waist would you need for 90% collection?
🔗 Remote Entanglement
Open the page →- Use the tool In the rate calculator, increase the NA from 0.5 to 0.7. By what factor does the two-photon heralded rate change?
- By hand Explain why the two-photon scheme scales as η² while the single-photon scheme scales as η, and what the single-photon scheme gives up in fidelity.
- Design For a 10 km link, find the memory coherence time needed for a three-segment repeater chain to beat direct transmission, using the repeater tool.
📚 AMO Paper Roadmap
Open the page →- Design Pick one paper from the stage of the roadmap that matches this week. Write a one-page summary: the question, the method, the key figure, and which AMO Toolkit page lets you reproduce a number from it.
- Design Using one of the platform comparison pages, write a 500-word memo recommending a hardware platform for a specific application (for example, a 100-qubit simulator or a quantum network node). Cite at least three sources.
04 Assessment and final projects
A suggested weighting, which works well when the problem sets are short and the project carries the real design work. Pick a preset or tune the weights to match your syllabus — the Markdown syllabus export reflects your weights automatically.
Design a lab
Choose an atom and an application. Use the Build pages in order to produce a design document: lasers, vacuum, MOT, control system, tweezer array and an error budget, with every number traced to a page or a source.
Reproduce a figure
Pick a figure from a paper in the roadmap and reproduce it, first with the matching page and then in your own Python code. Explain every place where the two disagree.
Build a calculator
Write a short, tested Python module that reproduces one calculator on this site, with a notebook that checks it against two published numbers.
Error budget for a real result
Take a published two-qubit gate fidelity and rebuild its error budget with the fidelity-budget page. Which terms do the authors' numbers imply, and what is left unexplained?
Optimise a virtual stage
Use the virtual lab or the automation page's simulators as a stand-in experiment. Run a full optimisation workflow, with noise estimates, reference points and a validated result.
Teach a topic
Prepare a 15-minute lecture on one week's topic using the pages as live demonstrations, with three problems of your own and their solutions.
05 Python notebooks
19 Jupyter notebooks reproduce the site's calculators in Python with the same constants and formulas, add a scan or a plot the page does not show, and end by checking their numbers against the live page. They suit the "By hand" problems, the "Build a calculator" project, and students who would rather work in code. Each needs only numpy, scipy and matplotlib.
06 Solutions, adaptations and feedback
Solutions are available on request
Worked solutions are not posted, so the problems stay usable as homework. If you teach a course, write with your name, institution, course and the weeks you need, and the solutions will be sent to you.
- Use any week, problem or project in your own course; please credit AMO Toolkit and link to this page.
- Found an error, or have a problem that worked well in your class? Use the "Found a problem?" panel at the foot of any page.
- Students working alone: check your answers against the "By hand" problems, which are built to agree with the tools.