Use this as your laboratory notebook. Read a chapter, calculate its checkpoint on paper, run the linked lab, then change one parameter and explain the result. Each session can take 45 to 90 minutes; the full course needs several sessions.
Start with notation and units, the paper map, and the Python setup. Commands below run from the repository root. Lab numbers preserve the original repository order; the recommended learning order is deliberately different.
| Chapter | By the end you should be able to… | Run and investigate |
|---|---|---|
| 01 | Explain why nonlinearity and refrigeration solve different problems | python hands-on/12-microwave-chain/chain.py; compare 10 and 100 mK |
| 02 | Quantize an LC mode and calculate its zero-point scales | python tutorial/figures/make_figures.py; inspect the computed chapter-02 figures |
| 03 | Build the charge-basis matrix from the cosine | python hands-on/09-transmon-spectrum/transmon.py; compare charge cutoffs |
| 04 | Distinguish exact spectrum, quartic approximation, and asymptotic dispersion | Lab 09 again; sweep EJ at fixed EC |
| 05 | Connect a spectral splitting to an excitation-swap time | python hands-on/10-jaynes-cummings/jaynes_cummings.py |
| 06 | Explain pointer separation, ring-up, and receiver noise | python hands-on/04-dispersive-readout/dispersive.py; vary drive and cutoff |
| 07 | Turn a pulse into a rotation and distinguish leakage from gate error | Run bloch.py, rabi.py, and drag.py via python hands-on/run_labs.py 01 02 05 |
| 08 | Separate local phases, conditional phase, leakage, and unwanted exchange | python hands-on/06-cz-gate/cz_gate.py; inspect full CZ overlap |
| 09 | Recognize exponential versus nonexponential decay and explain echo | python hands-on/run_labs.py 03 11 |
| 10 | Calculate a staged photon and amplifier-noise budget | python hands-on/12-microwave-chain/chain.py |
| 11 | Extract error per Clifford and state the RB assumptions | python hands-on/07-randomized-benchmarking/rb.py |
| 12 | Explain which errors repetition corrects and what it cannot protect | python hands-on/08-repetition-code/repetition.py |
- Initialization: Is a 5 GHz mode at 100 mK effectively empty? Distinguish oscillator photon occupation from the excited-state population of a thermal qubit.
-
LC: For
$L=10$ nH and$C=100$ fF, calculate$f$ , impedance, and$\Phi_{\rm zpf}Q_{\rm zpf}$ . -
Josephson: Why is the charge-matrix off-diagonal
$-E_J/2$ , and how many rows does cutoff 20 give? -
Transmon: At
$E_C/h=250$ MHz and$E_J/E_C=50$ , which predicted quantity has the larger relative quartic-approximation error: frequency or anharmonicity? -
JC: If
$g/2\pi=100$ MHz, what are the avoided-crossing gap and swap time? -
Readout: For
$\kappa/2\pi=5$ MHz, calculate the photon and field decay times. Does twice the field separation necessarily imply twice the assignment fidelity? -
Control: At
$\Omega/2\pi=20$ MHz, how long is a resonant pi pulse? Does final$P_2=10^{-9}$ prove that the pulse is a high-fidelity X gate? -
CZ: Why can virtual Z rotations correct local phases but not
$|01\rangle \leftrightarrow|10\rangle$ exchange or population in$|02\rangle$ ? -
Noise: Find
$T_2$ for$T_1=30$ us and$T_\phi=40$ us. Can ideal echo undo Markovian energy relaxation? - Chain: At fixed total attenuation, why does moving an attenuator to a warmer stage change the photon budget?
-
RB: A single-qubit fit gives
$p=0.98$ . What is error per Clifford? Why is the default lab's sequence scatter essentially zero? -
QEC: Evaluate the three-bit failure probability at
$p=0.01$ . What does observing zero failures in two million shots establish?
Answers and interpretation
-
$\bar n\approx0.100$ is not negligible. A thermal two-level qubit instead has$p_e=1/(e^{hf/k_BT}+1)\approx0.083$ ; in equilibrium$p_e=\bar n/(2\bar n+1)$ . -
$f\approx5.03$ GHz,$Z\approx316$ ohms, and the product is$\hbar/2$ . -
$\cos\varphi=(e^{i\varphi}+e^{-i\varphi})/2$ shifts charge by one; integers -20 through 20 give 41 rows. - Exact numerics give 4.73548 GHz and -287.31 MHz. Leading estimates are 4.75 GHz and -250 MHz: about 0.31% versus 13.0% relative error.
- 200 MHz and 2.5 ns. The gap is
$2g$ ; the full swap takes$\pi/(2g)$ . -
$1/\kappa\approx31.8$ ns and$2/\kappa\approx63.7$ ns. Fidelity depends nonlinearly on overlap, noise, priors, decay, and the discriminator. - 25 ns. No: residual population error and phase error can remain even when leakage is tiny. Test all input states or the full channel.
- A local Z correction is diagonal within the computational basis; it cannot reverse population transfer between basis states or recover lost population.
- 24 us. Echo rejects slow phase noise; it does not reverse irreversible relaxation.
- Each attenuator replaces absorbed photons with its own thermal emission.
-
$r=(1-p)/2=0.01$ . Identical depolarizing channels commute with unitary gates; each sequence has survival$1/2+p^{m+1}/2$ . The lab uses probabilities, not finite-shot samples. -
$3p^2-2p^3=2.98\times10^{-4}$ . Zero failures only bounds the rate: approximately$3/N=1.5\times10^{-6}$ at 95% confidence.
Choose target