These notes cover the complete ECE 405A curriculum as taught by Brad Hauer at the University of Waterloo. They follow the structure of the course textbook — Majidy, Wilson, and Laflamme, Building Quantum Computers: A Practical Introduction — while providing expanded mathematical derivations, physical motivation, and worked examples drawn from primary literature and standard references in quantum information.

---

Sources and References

The following publicly available references are recommended for deeper study:

• Nielsen, M. A. and Chuang, I. L. Quantum Computation and Quantum Information. Cambridge University Press, 10th anniversary edition, 2010. The standard textbook for quantum information theory and algorithms.
• Wilde, M. M. Quantum Information Theory. 2nd ed. Cambridge University Press, 2017. Free arXiv preprint: arXiv:1106.1445. Rigorous treatment of quantum Shannon theory.
• Preskill, J. Lecture Notes on Quantum Computation. Open access at preskill.caltech.edu/ph229. Authoritative notes on quantum error correction and fault tolerance.
• Majidy, S., Wilson, C., and Laflamme, R. Building Quantum Computers: A Practical Introduction. Cambridge University Press. Primary course textbook — organized by hardware platform with the same chapter structure as this course.
• Krantz, P. et al. “A quantum engineer’s guide to superconducting qubits.” Applied Physics Reviews 6, 021318 (2019). Comprehensive review of superconducting qubit physics.
• Bruzewicz, C. D. et al. “Trapped-ion quantum computing: Progress and challenges.” Applied Physics Reviews 6, 021314 (2019). Comprehensive review of trapped-ion systems.
• O’Brien, J. L. “Optical quantum computing.” Science 318, 1567 (2007). Overview of photonic quantum computing.
• IBM Quantum Learning. Open at learn.qiskit.org. Interactive quantum circuits and algorithm tutorials.

---

Chapter 1: Introduction to Quantum Computing

1.1 Historical Motivation and the Case for Quantum Computation

Classical computers have achieved extraordinary things by manipulating bits — binary digits that are always definitively 0 or 1. The transistor, invented in 1947, and its subsequent miniaturization through Moore’s law drove half a century of exponential improvement in computational power. Yet there are problems — integer factorization, quantum simulation, unstructured database search — for which no efficient classical algorithm is known, and for which there are profound theoretical reasons to believe that no efficient classical algorithm can exist.

The idea that quantum mechanics might offer a fundamentally different model of computation was recognized independently by Richard Feynman (1982) and Yuri Manin (1980). Feynman’s observation was simple and physical: simulating a quantum system of \(n\) particles on a classical computer requires tracking \(2^n\) complex amplitudes. Even for \(n = 50\) spins, the state space has more than \(10^{15}\) components — already beyond practical reach. A quantum computer, built from quantum systems, could simulate quantum systems efficiently, requiring only \(n\) qubits to represent what a classical computer needs \(2^n\) bits to track.

David Deutsch formalized the quantum circuit model in 1985, showing that a quantum Turing machine could efficiently simulate any physical process. Peter Shor’s 1994 algorithm for integer factorization in polynomial time provided the first killer application: if built, a large-scale quantum computer would break the RSA cryptosystem on which most of today’s internet security rests. Lov Grover’s 1996 search algorithm demonstrated a provable, unconditional quadratic speedup over any classical algorithm for searching an unstructured database. These discoveries transformed quantum computing from a theoretical curiosity into an engineering imperative.

1.2 The DiVincenzo Criteria

For a physical system to function as a quantum computer, David DiVincenzo identified five necessary (and two sufficient for quantum communication) criteria in 2000. These criteria are not merely a checklist — they define the engineering challenge that each hardware platform must address.

Every hardware platform studied in this course — NMR, superconducting circuits, photonic systems, and trapped ions — represents a different engineering trade-off among these criteria. No platform currently satisfies all five with the fidelity and scale required for fault-tolerant computation, but each illuminates different physics and engineering constraints.

1.3 Quantum Circuits and Computational Complexity

The quantum circuit model provides the standard computational framework. A computation begins with \(n\) qubits in the state \(|0\rangle^{\otimes n}\), applies a sequence of quantum gates drawn from a universal set, and concludes with a measurement. The depth of a circuit is the number of sequential gate layers; its size is the total number of gates.

Computational complexity classes relevant to quantum computing include:

• P: problems solvable by a classical computer in polynomial time.
• NP: problems whose solutions can be verified classically in polynomial time.
• BQP (Bounded-error Quantum Polynomial time): problems solvable by a quantum computer in polynomial time with error probability at most 1/3. BQP contains P and is believed to contain integer factorization.
• QMA (Quantum Merlin Arthur): the quantum analogue of NP — problems for which a quantum proof can be verified in BQP.

The central open question — whether P = NP — remains unsolved. The equally important question of whether BQP is strictly larger than P is also open, though Shor’s algorithm provides strong evidence that it is.

1.4 Introduction to Quantum Error Correction

Physical qubits decohere. Errors accumulate. Left uncorrected, quantum information is destroyed faster than a useful computation can complete. Quantum error correction (QEC) seems paradoxical at first: the no-cloning theorem forbids copying quantum states, and measurement collapses superpositions. How can errors be detected and corrected?

The resolution lies in encoding logical information redundantly across many physical qubits, and measuring only error syndromes — observables that commute with the logical operators and thus reveal error information without revealing (and collapsing) the logical state.

The simplest example is the three-qubit bit-flip repetition code. The logical states are:

\[ |0_L\rangle = |000\rangle, \qquad |1_L\rangle = |111\rangle \]

A single bit-flip error on any qubit can be detected by measuring the parity operators \(Z_1 Z_2\) and \(Z_2 Z_3\) without learning whether the logical state is \(|0_L\rangle\) or \(|1_L\rangle\). The syndrome \((+1, +1)\) indicates no error; \((-1, +1)\) indicates an error on qubit 1; \((-1, -1)\) indicates an error on qubit 2; \((+1, -1)\) indicates an error on qubit 3. The appropriate correction restores the code word. Chapter 7 develops the full theory of QEC including the stabilizer formalism and the threshold theorem.

---

Chapter 2: Review of Quantum Mechanics

2.1 Pure States and Operators

2.1.1 The Hilbert Space Postulate

Quantum mechanics is formulated in the language of Hilbert spaces. A Hilbert space \(\mathcal{H}\) is a complex vector space equipped with an inner product \(\langle \cdot | \cdot \rangle\) that is complete with respect to the induced norm. For finite-dimensional quantum systems (which includes all qubits), \(\mathcal{H} \cong \mathbb{C}^d\) for some dimension \(d\).

Dirac notation provides a powerful and compact language. For a qubit (\(d = 2\)), the standard orthonormal basis \(\{|0\rangle, |1\rangle\}\), called the computational basis, corresponds to column vectors:

\[ |0\rangle = \begin{pmatrix} 1 \\ 0 \end{pmatrix}, \qquad |1\rangle = \begin{pmatrix} 0 \\ 1 \end{pmatrix} \]

A general pure single-qubit state is:

\[ |\psi\rangle = \alpha|0\rangle + \beta|1\rangle, \qquad \alpha, \beta \in \mathbb{C}, \quad |\alpha|^2 + |\beta|^2 = 1 \]

The outer product \(|i\rangle\langle j|\) is a \(d \times d\) matrix with a 1 in position \((i,j)\) and zeros elsewhere. The identity resolves as \(\mathbf{I} = \sum_i |i\rangle\langle i|\).

2.1.2 Observables and Hermitian Operators

Since \(A\) is Hermitian, its eigenvalues are real and its eigenvectors form a complete orthonormal basis (the spectral theorem). Writing \(A = \sum_a a |a\rangle\langle a|\) in its eigenbasis, the spectral decomposition is immediate.

The four most important single-qubit operators are the identity and the three Pauli matrices:

\[ \mathbf{I} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}, \quad \sigma_x = X = \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix}, \quad \sigma_y = Y = \begin{pmatrix} 0 & -i \\ i & 0 \end{pmatrix}, \quad \sigma_z = Z = \begin{pmatrix} 1 & 0 \\ 0 & -1 \end{pmatrix} \]

The Pauli matrices satisfy the commutation relations \([X, Y] = 2iZ\), \([Y, Z] = 2iX\), \([Z, X] = 2iY\), which encode the algebra \(\mathfrak{su}(2)\). They are both Hermitian and unitary, with eigenvalues \(\pm 1\).

2.1.3 The Bloch Sphere

The most elegant visualization of a single-qubit pure state is provided by the Bloch sphere. Every pure state can be written, up to a global phase (which is physically unobservable):

\[ |\psi\rangle = \cos\!\left(\frac{\theta}{2}\right)|0\rangle + e^{i\phi}\sin\!\left(\frac{\theta}{2}\right)|1\rangle \]

where \(\theta \in [0, \pi]\) is the polar angle and \(\phi \in [0, 2\pi)\) is the azimuthal angle. This parameterizes a point on the unit sphere \(S^2\). The north pole \((\theta = 0)\) corresponds to \(|0\rangle\); the south pole \((\theta = \pi)\) corresponds to \(|1\rangle\). The equatorial states \(|\pm\rangle = (|0\rangle \pm |1\rangle)/\sqrt{2}\) lie on the \(x\)-axis, and \(|\pm i\rangle = (|0\rangle \pm i|1\rangle)/\sqrt{2}\) on the \(y\)-axis.

The Bloch vector \(\vec{r} = (\langle X \rangle, \langle Y \rangle, \langle Z \rangle)\) lives on \(|\vec{r}| = 1\) for pure states and in the interior \(|\vec{r}| < 1\) for mixed states. The density matrix of any single-qubit state is:

\[ \rho = \frac{1}{2}(\mathbf{I} + \vec{r} \cdot \vec{\sigma}) = \frac{1}{2}\begin{pmatrix} 1 + r_z & r_x - ir_y \\ r_x + ir_y & 1 - r_z \end{pmatrix} \]

Single-qubit gates are rotations of the Bloch vector. This geometric picture makes the effect of noise immediately visible: decoherence shrinks the Bloch vector toward the origin.

2.2 Mixed States and the Density Matrix

2.2.1 Classical vs. Quantum Uncertainty

A pure state \(|\psi\rangle\) represents maximal knowledge about a quantum system. Often we have less than maximal knowledge — either because the system is entangled with an environment we cannot observe, or because we prepared the system from a statistical ensemble. In both cases, the appropriate description is the density matrix.

A mixed state arises from a classical ensemble: if the system is in state \(|\psi_i\rangle\) with probability \(p_i\) (with \(\sum_i p_i = 1\)), then:

\[ \rho = \sum_i p_i |\psi_i\rangle\langle\psi_i| \]

The purity \(\mu = \text{tr}(\rho^2)\) satisfies \(1/d \leq \mu \leq 1\), with equality on the left for the maximally mixed state \(\rho = \mathbf{I}/d\) and on the right for pure states.

2.2.2 Expectation Values and Measurement

For a qubit in state \(\rho\) and observable \(Z\): outcomes are \(+1\) with probability \(\text{tr}(|0\rangle\langle 0| \rho) = \rho_{00}\) and \(-1\) with probability \(\rho_{11}\). After measuring \(+1\), the state collapses to \(|0\rangle\langle 0|\).

The generalized measurement formalism replaces projectors with positive operator-valued measures (POVMs). A POVM is a set \(\{E_m\}\) with \(E_m \geq 0\) and \(\sum_m E_m = \mathbf{I}\). Outcome \(m\) occurs with probability \(\text{tr}(E_m \rho)\). POVMs are more general than projective measurements and are the appropriate description for imperfect detectors and indirect readout schemes such as dispersive readout in circuit QED.

2.2.3 Partial Trace and Reduced Density Matrix

For a composite system \(\mathcal{H}_A \otimes \mathcal{H}_B\), the reduced density matrix for subsystem \(A\) is obtained by tracing out \(B\):

\[ \rho_A = \text{tr}_B(\rho_{AB}) = \sum_j (\mathbf{I}_A \otimes \langle j|_B)\, \rho_{AB}\, (\mathbf{I}_A \otimes |j\rangle_B) \]

where \(\{|j\rangle_B\}\) is any orthonormal basis of \(\mathcal{H}_B\). The partial trace is the unique operation on \(\rho_{AB}\) that correctly predicts measurement statistics for all observables on \(A\) alone.

If the global state is pure and \(\rho_A\) is mixed, the system \(A\) is entangled with system \(B\). Entanglement entropy is \(S(\rho_A) = -\text{tr}(\rho_A \log_2 \rho_A)\).

2.3 Dynamics and Quantum Gates

2.3.1 The Schrödinger Equation

\[ i\hbar \frac{d}{dt}|\psi(t)\rangle = H(t)|\psi(t)\rangle \]

where \(H = H^\dagger\) is the Hamiltonian operator. For time-independent \(H\), the solution is:

\[ |\psi(t)\rangle = e^{-iHt/\hbar}|\psi(0)\rangle \equiv U(t)|\psi(0)\rangle \]

The operator \(U(t) = e^{-iHt/\hbar}\) is unitary: \(U^\dagger U = \mathbf{I}\). Unitary operators preserve inner products and hence probabilities. Conversely, any unitary can be written \(U = e^{-iA}\) for some Hermitian \(A\), so every unitary can in principle be implemented by a suitable Hamiltonian.

For density matrices, Schrödinger evolution becomes the von Neumann equation:

\[ \frac{d\rho}{dt} = -\frac{i}{\hbar}[H, \rho] \]

2.3.2 Single-Qubit Gates

A single-qubit gate is a \(2\times 2\) unitary matrix. The Pauli matrices generate rotations on the Bloch sphere:

\[ R_x(\theta) = e^{-i\theta X/2} = \cos\!\left(\frac{\theta}{2}\right)\mathbf{I} - i\sin\!\left(\frac{\theta}{2}\right)X = \begin{pmatrix} \cos(\theta/2) & -i\sin(\theta/2) \\ -i\sin(\theta/2) & \cos(\theta/2) \end{pmatrix} \]\[ R_y(\theta) = e^{-i\theta Y/2} = \begin{pmatrix} \cos(\theta/2) & -\sin(\theta/2) \\ \sin(\theta/2) & \cos(\theta/2) \end{pmatrix} \]\[ R_z(\theta) = e^{-i\theta Z/2} = \begin{pmatrix} e^{-i\theta/2} & 0 \\ 0 & e^{i\theta/2} \end{pmatrix} \]

These are rotations by angle \(\theta\) around the \(x\)-, \(y\)-, and \(z\)-axes of the Bloch sphere, respectively. The Euler decomposition states that any single-qubit unitary can be written as \(U = e^{i\alpha}R_z(\beta)R_y(\gamma)R_z(\delta)\) for some angles \(\alpha, \beta, \gamma, \delta\).

The Hadamard gate \(H = (X + Z)/\sqrt{2}\) is one of the most important single-qubit operations:

\[ H = \frac{1}{\sqrt{2}}\begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix} \]

It maps \(|0\rangle \mapsto |+\rangle\) and \(|1\rangle \mapsto |-\rangle\), transforming between the \(Z\)- and \(X\)-eigenbases. The phase gates \(S = \text{diag}(1, i)\) and \(T = \text{diag}(1, e^{i\pi/4})\) complete the standard single-qubit gate set.

2.3.3 Two-Qubit Gates and Entanglement

The CNOT (controlled-NOT) gate acts on two qubits — a control and a target. It flips the target if and only if the control is \(|1\rangle\):

\[ \text{CNOT} = |0\rangle\langle 0| \otimes \mathbf{I} + |1\rangle\langle 1| \otimes X = \begin{pmatrix} 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 1 & 0 \end{pmatrix} \]

in the basis \(\{|00\rangle, |01\rangle, |10\rangle, |11\rangle\}\). The CNOT combined with single-qubit gates is universal for quantum computation.

The four Bell states are maximally entangled two-qubit states:

\[ |\Phi^\pm\rangle = \frac{1}{\sqrt{2}}(|00\rangle \pm |11\rangle), \qquad |\Psi^\pm\rangle = \frac{1}{\sqrt{2}}(|01\rangle \pm |10\rangle) \]

They are produced from product states by applying \(H\) to the first qubit followed by CNOT. The Bell states form an orthonormal basis for the two-qubit Hilbert space \(\mathbb{C}^2 \otimes \mathbb{C}^2\).

2.4 Open Quantum Systems

2.4.1 Quantum Channels

Real quantum systems interact with an environment. The combined system-environment state \(\rho_{SE}\) evolves unitarily, but tracing out the environment yields non-unitary dynamics for the system alone. The most general trace-preserving positive map is described by a quantum channel:

2.4.2 The Lindblad Master Equation

For Markovian (memoryless) system-environment coupling, the density matrix evolves according to the Lindblad master equation, the most general time-local completely positive map in continuous time:

\[ \frac{d\rho}{dt} = -\frac{i}{\hbar}[H, \rho] + \sum_k \gamma_k \left( L_k \rho L_k^\dagger - \frac{1}{2}L_k^\dagger L_k \rho - \frac{1}{2}\rho L_k^\dagger L_k \right) \]

The first term is coherent evolution under Hamiltonian \(H\). The second term is the dissipator, where \(L_k\) are jump operators with rates \(\gamma_k\). The characteristic sandwiching structure \(L\rho L^\dagger - \{L^\dagger L, \rho\}/2\) (using anticommutator notation for brevity) ensures trace preservation and complete positivity.

For a qubit subject to energy relaxation (rate \(\gamma_1 = 1/T_1\)) and pure dephasing (rate \(\gamma_\phi\)):

\[ \frac{d\rho}{dt} = -\frac{i}{\hbar}[H,\rho] + \frac{\gamma_1}{2}(2\sigma^- \rho \sigma^+ - \sigma^+\sigma^- \rho - \rho \sigma^+\sigma^-) + \frac{\gamma_\phi}{2}(Z\rho Z - \rho) \]

where \(\sigma^+ = |1\rangle\langle 0|\) and \(\sigma^- = |0\rangle\langle 1|\) are the raising and lowering operators. The dephasing time \(T_2\) satisfies \(1/T_2 = 1/(2T_1) + 1/T_\phi\), where \(T_\phi = 1/\gamma_\phi\).

2.5 Quantum Information Measures

2.5.1 Von Neumann Entropy

The von Neumann entropy generalizes classical Shannon entropy to quantum states:

\[ S(\rho) = -\text{tr}(\rho\log_2\rho) = -\sum_i \lambda_i \log_2 \lambda_i \]

where \(\{\lambda_i\}\) are the eigenvalues of \(\rho\). For a pure state, \(S = 0\). For the maximally mixed state \(\rho = \mathbf{I}/d\), \(S = \log_2 d\). Von Neumann entropy is subadditive: \(S(AB) \leq S(A) + S(B)\), with equality if and only if \(\rho_{AB} = \rho_A \otimes \rho_B\) (no correlations).

The quantum mutual information quantifies correlations between subsystems \(A\) and \(B\):

\[ I(A:B) = S(\rho_A) + S(\rho_B) - S(\rho_{AB}) \geq 0 \]

For a pure bipartite state, \(I(A:B) = 2S(\rho_A)\), twice the entanglement entropy.

2.5.2 Relative Entropy and Quantum Channel Capacity

The quantum relative entropy \(D(\rho\|\sigma) = \text{tr}(\rho\log_2\rho) - \text{tr}(\rho\log_2\sigma) \geq 0\) (Klein’s inequality) measures the “distance” from \(\sigma\) to \(\rho\), with equality iff \(\rho = \sigma\). It is not a metric (not symmetric) but serves as a contraction under quantum channels: \(D(\mathcal{E}(\rho)\|\mathcal{E}(\sigma)) \leq D(\rho\|\sigma)\). This data processing inequality underlies the impossibility of perfectly distinguishing two quantum states after processing through a noisy channel.

The quantum channel capacity — the maximum rate at which quantum information can be transmitted reliably through a noisy channel — is bounded by the quantum capacity \(Q(\mathcal{E})\) given by the regularized coherent information:

\[ Q(\mathcal{E}) = \lim_{n\to\infty}\frac{1}{n}\max_\rho I_c(\rho, \mathcal{E}^{\otimes n}) \]

where \(I_c(\rho, \mathcal{E}) = S(\mathcal{E}(\rho)) - S_e(\rho, \mathcal{E})\) and \(S_e\) is the entropy exchange. For the depolarizing channel with error probability \(p\), \(Q > 0\) iff \(p < 0.5\) (Hashing bound).

2.6 Measurement Revisited: Generalized and Repeated Measurement

2.6.1 Projective Measurement in Arbitrary Basis

A projective measurement in the \(Z\)-basis has outcomes \(|0\rangle\) and \(|1\rangle\). Measuring in an arbitrary basis defined by unitary \(U\) is equivalent to applying \(U^\dagger\) first, then measuring in \(Z\). For example, measuring in the \(X\)-basis requires applying \(H\) (which maps \(|+\rangle \to |0\rangle\) and \(|-\rangle \to |1\rangle\)) before the \(Z\)-measurement.

2.6.2 Quantum State Tomography

To determine an unknown density matrix \(\rho\), one measures a tomographically complete set of observables: for a single qubit, measuring \(X\), \(Y\), and \(Z\) on many identically prepared copies determines \(\rho\) uniquely via:

\[ \rho = \frac{1}{2}(\mathbf{I} + \langle X\rangle X + \langle Y\rangle Y + \langle Z\rangle Z) \]

For \(n\) qubits, a complete Pauli tomography requires \(4^n - 1\) expectation values, scaling exponentially. Compressed sensing and neural-network-based tomography offer more efficient alternatives at the cost of assumptions about state structure.

2.7 The Quantum Harmonic Oscillator

The quantum harmonic oscillator (QHO) is the workhorse of quantum optics and circuit QED. Its Hamiltonian is:

\[ H = \hbar\omega \left(a^\dagger a + \frac{1}{2}\right) \]

where the bosonic creation and annihilation operators satisfy \([a, a^\dagger] = 1\). The energy eigenstates (Fock states) are \(|n\rangle\), \(n = 0, 1, 2, \ldots\), with \(H|n\rangle = \hbar\omega(n + 1/2)|n\rangle\). Ladder operator actions:

\[ a|n\rangle = \sqrt{n}\,|n-1\rangle, \qquad a^\dagger|n\rangle = \sqrt{n+1}\,|n+1\rangle, \qquad \hat{n}|n\rangle = n|n\rangle \]

Position and momentum quadratures are \(x = \sqrt{\hbar/(2m\omega)}(a + a^\dagger)\) and \(p = i\sqrt{m\omega\hbar/2}(a^\dagger - a)\), satisfying \([x, p] = i\hbar\).

Coherent states \(|\alpha\rangle\) (eigenstates of \(a\): \(a|\alpha\rangle = \alpha|\alpha\rangle\)) are displaced vacuum states and represent the closest quantum analogue of a classical oscillating field. They have Poissonian photon-number statistics: \(p_n = e^{-|\alpha|^2}|\alpha|^{2n}/n!\).

The displacement operator \(D(\alpha) = \exp(\alpha a^\dagger - \alpha^* a)\) maps the vacuum to \(|\alpha\rangle\). Squeezed states reduce uncertainty in one quadrature below the vacuum level \(\sqrt{\hbar/2}\) at the cost of increased uncertainty in the conjugate quadrature — important for continuous-variable quantum computing and high-sensitivity readout.

---

Chapter 3: Spins and Nuclear Magnetic Resonance

3.1 Spin-\(\tfrac{1}{2}\) Quantum Mechanics

A spin-\(1/2\) particle has a two-dimensional Hilbert space \(\mathbb{C}^2\). The spin angular momentum operators are \(\vec{S} = (\hbar/2)\vec{\sigma}\), where \(\vec{\sigma} = (X, Y, Z)\) are the Pauli matrices. The spin eigenstates of \(S_z\) are \(|\uparrow\rangle = |0\rangle\) (spin up, \(m_s = +1/2\)) and \(|\downarrow\rangle = |1\rangle\) (spin down, \(m_s = -1/2\)).

The magnetic dipole moment of a spin-\(1/2\) particle is \(\vec{\mu} = \gamma \vec{S} = (\gamma\hbar/2)\vec{\sigma}\), where \(\gamma\) is the gyromagnetic ratio. For a proton, \(\gamma_p = 2\pi \times 42.577\) MHz/T; for an electron, \(\gamma_e \approx 2\pi \times 28.0\) GHz/T.

3.2 Nuclear Magnetic Resonance as a Qubit Platform

In NMR quantum computing, nuclear spins in a strong magnetic field \(B_0\hat{z}\) provide the qubit. The static Hamiltonian is:

\[ H_0 = -\gamma\hbar B_0 \frac{Z}{2} = -\frac{\hbar\omega_0}{2}Z \]

where the Larmor frequency \(\omega_0 = \gamma B_0\) sets the qubit splitting. For a proton in a 10 T field, \(\omega_0 \approx 2\pi \times 425\) MHz. The ground state (\(|0\rangle = |\uparrow\rangle\)) is the lower-energy state.

3.2.1 Larmor Precession

In the Heisenberg picture, the spin vector precesses about \(\hat{z}\) at the Larmor frequency:

\[ \frac{d\langle\vec{S}\rangle}{dt} = \gamma \langle\vec{S}\rangle \times \vec{B} \]

This is classical Bloch precession: the Bloch vector rotates around the \(z\)-axis. In the rotating frame — obtained by changing to a frame rotating at \(\omega_0\) — the Hamiltonian vanishes to lowest order, and small frequency offsets \(\Delta\omega = \omega_0 - \omega_\text{rf}\) appear as a static effective field.

3.2.2 Rabi Oscillations and the Rotating Wave Approximation

An oscillating RF field \(\vec{B}_1(t) = B_1\cos(\omega_\text{rf}t)\hat{x}\) drives transitions between \(|\uparrow\rangle\) and \(|\downarrow\rangle\). The full Hamiltonian in the lab frame is:

\[ H = -\frac{\hbar\omega_0}{2}Z - \frac{\hbar\Omega_1}{2}(e^{i\omega_\text{rf}t} + e^{-i\omega_\text{rf}t})X \]

where \(\Omega_1 = \gamma B_1\) is the Rabi frequency. Transforming to the rotating frame with \(U_0 = e^{i\omega_\text{rf}tZ/2}\):

\[ H_\text{rot} = -\frac{\hbar\Delta}{2}Z - \frac{\hbar\Omega_1}{2}X + \text{(counter-rotating terms)} \]

where \(\Delta = \omega_0 - \omega_\text{rf}\). The rotating wave approximation (RWA) drops the counter-rotating terms oscillating at \(2\omega_\text{rf}\), valid when \(\Omega_1 \ll \omega_\text{rf}\):

\[ H_\text{RWA} = -\frac{\hbar\Delta}{2}Z - \frac{\hbar\Omega_1}{2}X \]

On resonance (\(\Delta = 0\)), the Hamiltonian is simply \(-(\hbar\Omega_1/2)X\), driving Rabi oscillations: the population oscillates between \(|0\rangle\) and \(|1\rangle\) with frequency \(\Omega_1\). A pulse of duration \(t_\pi = \pi/\Omega_1\) performs a NOT gate (\(\pi\)-pulse); a pulse of duration \(t_{\pi/2} = \pi/(2\Omega_1\)) creates the superposition \((|0\rangle + |1\rangle)/\sqrt{2}\).

3.2.3 Pulsed NMR Operations

Single-qubit rotations are implemented as pulsed RF fields in NMR. A pulse with phase \(\phi_\text{rf}\) in the rotating frame drives rotations about the axis \(\cos\phi_\text{rf}\,\hat{x} + \sin\phi_\text{rf}\,\hat{y}\). An \(x\)-pulse: \(R_x(\theta)\); a \(y\)-pulse: \(R_y(\theta)\). Combining pulses with different phases and durations generates any \(SU(2)\) operation.

3.3 J-Coupling and Two-Qubit Gates in NMR

Multi-qubit operations in NMR exploit the scalar (J) coupling between nuclear spins, mediated by the electron cloud:

\[ H_J = \pi\hbar J\, \vec{\sigma}_1 \cdot \vec{\sigma}_2 = \pi\hbar J(X_1X_2 + Y_1Y_2 + Z_1Z_2) \]

Under the secular approximation (valid when \(|\omega_1 - \omega_2| \gg J\)), this reduces to:

\[ H_J \approx \pi\hbar J Z_1 Z_2 \]

Evolving under \(H_J\) for time \(t = 1/(2J)\) yields the unitary \(e^{-i\pi Z_1 Z_2/2}\), which combined with single-qubit rotations implements the CNOT gate. J-coupling is weak (typical \(J \sim 10\text{–}200\) Hz), making NMR gates slow and susceptible to decoherence during operation.

3.4 Ensemble Measurement and Pseudo-Pure States

NMR works with large ensembles of \(\sim 10^{23}\) identical molecules in thermal equilibrium. The thermal state at room temperature with \(k_BT \gg \hbar\omega_0\) is nearly the maximally mixed state:

\[ \rho_\text{thermal} \approx \frac{\mathbf{I}}{2^n} + \epsilon \rho_\text{pps} \]

where \(\epsilon = \hbar\omega_0/(2k_BT) \sim 10^{-5}\) is the tiny polarization. The signal comes from the traceless deviation from the maximally mixed background. To simulate quantum algorithms, the state is manipulated into a pseudo-pure state:

\[ \rho_\text{pps} = (1-\epsilon)\frac{\mathbf{I}}{2^n} + \epsilon|00\cdots0\rangle\langle00\cdots0| \]

which behaves like a pure state \(|00\cdots0\rangle\) for all practical purposes, since the maximally mixed part contributes only a constant scaling to all observables. Pseudo-pure state preparation requires specialized pulse sequences (e.g., spatial averaging).

3.5 State Tomography in NMR

Reading out the full density matrix requires state tomography: applying rotations to map different operator components into the \(Z\)-basis (which NMR measures via the free-induction decay signal) followed by measurement and Fourier analysis. Full tomography of \(n\) qubits scales as \(4^n\) measurements.

3.5.1 The Ramsey Experiment and T₂ Measurement

The Ramsey experiment is the canonical method for measuring dephasing time. The pulse sequence is:

1. Apply \(\pi/2\) pulse (\(R_x(\pi/2)\)): prepares \((|0\rangle + |1\rangle)/\sqrt{2}\).
2. Free evolution for time \(\tau\) under \(H = -\hbar\Delta Z/2\): the state evolves to \((|0\rangle + e^{i\Delta\tau}|1\rangle)/\sqrt{2}\), accumulating a relative phase.
3. Apply a second \(\pi/2\) pulse: maps the relative phase into population.
4. Measure \(\langle Z\rangle\).

In the presence of dephasing, the measured signal decays as:

\[ \langle Z\rangle(\tau) = \cos(\Delta\tau)\,e^{-\tau/T_2^*} \]

where \(T_2^* = 1/(\pi\Delta\omega_\text{inhom})\) is the inhomogeneous dephasing time. For Gaussian distributed frequency fluctuations with variance \(\sigma_\omega^2\), the decay envelope is Gaussian: \(e^{-\sigma_\omega^2\tau^2/2}\). Adding a \(\pi\)-pulse at \(\tau/2\) (Hahn echo) refocuses static frequency offsets, recovering the longer homogeneous \(T_2\).

3.6–3.8 Initialization, Relaxation, and Silicon Spins

Initialization in NMR uses thermalization (slow) or algorithmic cooling protocols. Decoherence in NMR is characterized by two time scales:

• \(T_1\) (longitudinal, spin-lattice): energy relaxation time, recovery of \(\langle Z \rangle\) to thermal equilibrium. In solution NMR, \(T_1 \sim 1\)–10 s.
• \(T_2\) (transverse, spin-spin): decay of \(\langle X \rangle\) and \(\langle Y \rangle\) due to dephasing from inhomogeneous fields and spin-spin interactions. In solution NMR, \(T_2^* \lesssim T_2 \leq 2T_1\).

The Hahn echo technique refocuses static inhomogeneous broadening, recovering the “true” \(T_2\) from the apparent \(T_2^*\). Dynamical decoupling (CPMG sequences) extends coherence further.

Silicon-based spin qubits exploit the long coherence times of \(^{28}\)Si (spin-zero nucleus, minimal decoherence). Electron spin or \(^{31}\)P nuclear spin qubits in silicon have achieved \(T_2 > 30\) ms, with gate fidelities exceeding 99.9%. These systems bridge NMR physics with scalable semiconductor fabrication.

---

Chapter 4: Photonic Quantum Computing

4.1 Quantization of the Electromagnetic Field

The electromagnetic field in a cavity of volume \(V\) can be decomposed into normal modes, each equivalent to a harmonic oscillator. For mode \(k\) with wavevector \(\vec{k}\) and polarization \(\lambda\):

\[ \vec{A}(\vec{r}, t) = \sum_{k,\lambda} \sqrt{\frac{\hbar}{2\epsilon_0\omega_k V}}\,\hat{\epsilon}_{k\lambda}\left(a_{k\lambda} e^{i\vec{k}\cdot\vec{r} - i\omega_k t} + a_{k\lambda}^\dagger e^{-i\vec{k}\cdot\vec{r} + i\omega_k t}\right) \]

The quantized electric field is \(\vec{E} = -\partial\vec{A}/\partial t\). The vacuum state \(|0\rangle\) has zero photons but nonzero energy \(\sum_k \hbar\omega_k/2\) (zero-point energy). A Fock state \(|n\rangle\) contains exactly \(n\) photons in the mode.

4.1.1 Dual-Rail Encoding

A single photonic qubit is typically encoded in the dual-rail representation: the logical \(|0\rangle\) is a photon in spatial mode 0 (or horizontal polarization), and logical \(|1\rangle\) is a photon in spatial mode 1 (or vertical polarization):

\[ |0_L\rangle = |1\rangle_a|0\rangle_b, \qquad |1_L\rangle = |0\rangle_a|1\rangle_b \]

This encoding is robust to photon loss (which takes both logical states out of the qubit subspace, making errors detectable) and is compatible with linear optical elements.

4.2 Single-Photon Sources and Detectors

4.2.1 Sources

Deterministic single-photon sources are required for photonic quantum computing. The primary platforms are:

• Spontaneous parametric down-conversion (SPDC): a nonlinear crystal converts a pump photon into a correlated pair at half the frequency. Heralded single photons are produced by detecting one photon of the pair. The source is probabilistic; the probability of a single pair per pulse is low (\(\sim 10^{-2}\)), and higher-order terms introduce multi-photon errors.
• Quantum dots in photonic structures: a semiconductor quantum dot in a photonic crystal cavity or micropillar can emit single photons on demand with high purity (\(>98\%\)) and indistinguishability (\(>96\%\)).
• Color centers in diamond: nitrogen-vacancy (NV) centers or silicon-vacancy (SiV) centers emit single photons at room temperature with high purity.

4.2.2 Detectors

• Avalanche photodiodes (APDs): photon-induced avalanche in reverse-biased semiconductor junction. Efficiency \(\sim 50\%\) at 700 nm, dark count rates \(\sim 100\)/s.
• Superconducting nanowire single-photon detectors (SNSPDs): a superconducting wire driven near the critical current. A photon breaks Cooper pairs and creates a resistive hotspot. Efficiencies \(> 95\%\), timing jitter \(< 20\) ps, dark counts \(< 1\)/s.
• Transition edge sensors (TES): bolometric detectors with photon-number resolution. Efficiency \(> 98\%\) but require cooling to \(\sim 100\) mK.

4.3 Linear Optical Elements

Linear optical quantum computing (LOQC) manipulates photons using beam splitters, phase shifters, and wave plates — passive elements that preserve photon number but create superpositions.

A beam splitter (BS) with reflectivity \(r\) and transmissivity \(t\) (\(|r|^2 + |t|^2 = 1\)) performs the transformation:

\[ \begin{pmatrix} a_\text{out} \\ b_\text{out} \end{pmatrix} = \begin{pmatrix} t & ir \\ ir & t \end{pmatrix}\begin{pmatrix} a_\text{in} \\ b_\text{in} \end{pmatrix} \]

A 50:50 BS with \(t = r = 1/\sqrt{2}\) implements the Hadamard gate on the dual-rail qubit. A phase shifter \(e^{i\phi a^\dagger a}\) applies a phase to one mode: it implements the \(R_z(\phi)\) gate.

4.3.1 The KLM Protocol

Knill, Laflamme, and Milburn (KLM, 2001) showed that universal quantum computing with linear optics is possible using only linear optical elements, single-photon sources, and photon-number-resolving detectors — if one accepts a probabilistic scheme with postselection.

The key insight is that nonlinear interactions between photons can be induced measurement-effectively: by coupling ancilla photons and performing a measurement, a nonlinear (entangling) gate can be implemented on the signal photons, with a success probability that can be made close to 1 by using sufficiently large ancilla registers (quantum teleportation-based gates).

The basic probabilistic CZ gate in KLM succeeds with probability \(1/4\). Using teleportation-based schemes, near-deterministic gates can be built at the cost of large photon overhead (\(\sim 10^4\) ancilla photons per gate).

4.4–4.5 Photonic Gates and BB84 Quantum Key Distribution

4.4 Photonic Qubit Gates

In the polarization encoding, single-qubit gates are implemented by:

• Wave plates: half-wave plate (HWP) performs a rotation by \(2\theta\) around the axis at angle \(\theta\) to the horizontal. It implements \(R_z(2\theta)\) in the \(H/V\) basis.
• Quarter-wave plate (QWP): introduces a \(\pi/2\) phase between \(H\) and \(V\), implementing \(S\)-like gates.
• Electro-optic modulators (EOMs): voltage-controlled birefringence for fast, tunable phase shifts.

4.5 Quantum Key Distribution: BB84

The BB84 protocol (Bennett and Brassard, 1984) is the first and most widely deployed QKD protocol. It exploits the no-cloning theorem and the measurement disturbance principle to distribute cryptographic keys with information-theoretic security.

Security follows from the uncertainty principle: Eve cannot simultaneously measure in both \(+\) and \(\times\) bases without disturbing the transmitted qubits. Any interception and re-transmission introduces a QBER of 25% (for the intercept-resend attack in the sifted key), detectable by Alice and Bob.

4.6–4.9 Measurement-Based and Continuous-Variable Quantum Computing

Measurement-based quantum computing (MBQC) uses a highly entangled resource state — the cluster state — and performs computation by sequential adaptive single-qubit measurements. The computation is encoded in the choice of measurement bases; the classical feed-forward (using previous measurement results to adjust future bases) implements the effective unitary.

A 2D cluster state is an entangled state of qubits on a lattice, with CZ gates applied to all neighboring pairs starting from \(|+\rangle^{\otimes N}\):

\[ |\text{cluster}\rangle = \prod_{\langle i,j\rangle} CZ_{ij}\, |+\rangle^{\otimes N} \]

Measurement in the \(XY\)-plane at angle \(\phi\) on a chain of qubits implements \(R_z(\phi)\); the feed-forward corrects the Pauli byproduct operators. MBQC requires the same gate complexity as the circuit model but can be more naturally suited to photonic architectures.

Continuous-variable (CV) quantum computing encodes information in the quadratures \(x\) and \(p\) of harmonic oscillators (optical field modes). Gaussian states (coherent, squeezed) and Gaussian operations (displacement, squeezing, beam-splitter) can be implemented efficiently but are not universal. Adding a non-Gaussian element (such as photon counting) achieves universality. CV approaches naturally suit optical fiber telecommunications wavelengths and allow room-temperature operations.

---

Chapter 5: Trapped-Ion Quantum Computing

5.1 Ion Traps: Paul and Penning Traps

Charged ions can be confined in electromagnetic traps. The two main types are:

Paul trap (RF trap): uses oscillating RF electric fields in a quadrupole geometry to create a pseudopotential that confines ions in 3D. The Earnshaw theorem forbids confinement by static electric fields alone; the RF field creates an effective attractive potential by exploiting the frequency-dependent response of the ion. For a linear Paul trap, ions form a 1D Coulomb crystal along the trap axis, with motional frequencies \(\omega_\text{ax} \ll \omega_\text{rad}\). Typical \(\omega_\text{ax} \sim 2\pi \times 1\) MHz, \(\omega_\text{rad} \sim 2\pi \times 5\) MHz.

Penning trap: uses a combination of static magnetic field (along \(z\)) and electrostatic potential for confinement. Less commonly used for quantum computing due to complexity of manipulating ions in the rotating frame.

5.2 Internal States and Optical/Microwave Transitions

Qubit states in trapped ions are encoded in two internal electronic states of the ion, typically:

• Optical qubits: transitions between a ground state and a metastable excited state (\(\sim\) ms lifetime) at optical frequencies (\(\sim 10^{14}\) Hz). Example: \(^{40}\)Ca\(^+\), \(S_{1/2}\leftrightarrow D_{5/2}\) at 729 nm. High \(\omega_{01}\) means \(k_BT \ll \hbar\omega_{01}\) even at room temperature, allowing reliable initialization by optical pumping.
• Hyperfine qubits: transitions between two hyperfine ground-state levels via microwave or Raman laser transitions. Example: \(^{171}\)Yb\(^+\) (\(F=0\) vs \(F=1\) at 12.6 GHz), or \(^9\)Be\(^+\). Insensitive to magnetic field fluctuations at “magic” field points; coherence times \(T_2 > 10\) min have been demonstrated.

Initialization: optical pumping by a resonant laser depopulates one state, achieving \(|0\rangle\) with fidelity \(> 99.9\%\) in \(\sim 1\) µs.

5.3 Trapped Ion Motion and the Lamb-Dicke Regime

The motion of trapped ions along the trap axis is quantized into phonon modes. The Hamiltonian for a single ion in a harmonic trap with internal qubit coupling to a laser is:

\[ H = \frac{\hbar\omega_0}{2}Z + \hbar\nu a^\dagger a + \frac{\hbar\Omega}{2}\left(\sigma^+ e^{i(\eta(a+a^\dagger) - \delta t)} + \text{h.c.}\right) \]

where \(\nu\) is the trap frequency, \(\delta = \omega_L - \omega_0\) is the laser-qubit detuning, and \(\eta = k_L x_\text{zpf}/\hbar\) is the Lamb-Dicke parameter — the ratio of the laser wavelength to the zero-point motion \(x_\text{zpf} = \sqrt{\hbar/(2m\nu)}\).

In the Lamb-Dicke regime (\(\eta\sqrt{\langle(a+a^\dagger)^2\rangle} \ll 1\)), expanding the exponential to first order:

\[ H_\text{LD} \approx \frac{\hbar\Omega}{2}\left(\sigma^+ e^{-i\delta t}(1 + i\eta(ae^{-i\nu t} + a^\dagger e^{i\nu t})) + \text{h.c.}\right) \]

Three resonance conditions emerge:

• Carrier (\(\delta = 0\)): drives internal qubit transitions without affecting motion.
• Red sideband (\(\delta = -\nu\)): couples \(|g, n\rangle \leftrightarrow |e, n-1\rangle\) — simultaneously flips the qubit and removes a phonon.
• Blue sideband (\(\delta = +\nu\)): couples \(|g, n\rangle \leftrightarrow |e, n+1\rangle\) — simultaneously flips the qubit and adds a phonon.

5.3.1 Sideband Cooling

Repeated red sideband pulses followed by optical pumping back to \(|g\rangle\) remove phonons one by one, cooling the ion toward the motional ground state \(|n=0\rangle\). Starting from a thermal distribution, the average phonon number decreases exponentially. Final temperatures \(\langle n\rangle < 0.1\) phonons are routinely achieved.

5.4–5.5 Ion-Laser Interactions and Two-Qubit Gates

5.4 Cirac-Zoller Gate

The Cirac-Zoller (CZ) gate (1995) uses the shared motional mode of a 1D ion chain as a “quantum bus.” The gate sequence:

1. A \(\pi\)-pulse on the red sideband of ion 1 maps \(|1,0\rangle_{q,m} \to |0,1\rangle_{q,m}\): the qubit state is coherently transferred to the motional mode.
2. A \(2\pi\)-pulse on the red sideband of ion 2 conditioned on \(|1\rangle_m\): \(|1,1\rangle \to e^{i\pi}|1,1\rangle = -|1,1\rangle\) (geometric phase).
3. A reverse \(\pi\)-pulse on ion 1: transfers the motional state back.

The net effect is a conditional phase \(e^{i\pi}\) on \(|11\rangle\), implementing a CZ gate up to single-qubit rotations. Requires sequential addressing of individual ions; gate time \(\sim 1/\Omega_\text{sb} \sim 100\) µs for typical sideband Rabi frequencies.

5.5 Mølmer-Sørensen Gate

The Mølmer-Sørensen (MS) gate drives both blue and red sidebands simultaneously with a bichromatic field, creating an entangling interaction that does not require cooling to the motional ground state:

\[ H_\text{MS} = \frac{\hbar\Omega\eta}{2}(ae^{-i\delta t} + a^\dagger e^{i\delta t})(e^{-i\phi}\sigma^+ + e^{i\phi}\sigma^-) \]

By choosing the detuning \(\delta\) off-resonance from the sideband and using a spin-echo-like sequence, the motional state is disentangled at the end of the gate while the spin-spin interaction accumulates:

\[ U_\text{MS}(\theta) = \exp\!\left(-i\frac{\theta}{4}(X_1\cos\phi + Y_1\sin\phi)(X_2\cos\phi + Y_2\sin\phi)\right) \]

For \(\theta = \pi/2\) and \(\phi = 0\): \(U_\text{MS} = \exp(-i\pi X_1 X_2/4)\). Together with single-qubit rotations, this is universal. Gate times \(\sim 50\text{–}200\) µs; two-qubit fidelities \(> 99.9\%\) have been achieved in research settings.

5.6–5.9 Measurement, Noise, and Atomic Qubits

5.6 Fluorescence Readout

State-selective fluorescence: a laser resonant with a cycling transition (\(S_{1/2} \to P_{1/2}\)) excites \(|1\rangle\) ions to scatter many photons (detected by CCD or PMT) while \(|0\rangle\) ions remain dark. Typical discrimination: \(> 100\) photons per ms for \(|1\rangle\), \(< 1\) for \(|0\rangle\). State-detection fidelity \(> 99.9\%\).

5.7 Decoherence in Trapped Ions

Dominant decoherence mechanisms:

• Motional heating: electric field noise at the trap-frequency drives phonon excitation. Heating rates scale as \(\propto d^{-4}\) with electrode distance (anomalous heating from patch potentials). Cryogenic traps at 4 K reduce heating rates by \(\sim 100\times\).
• Dephasing from laser phase noise and magnetic field fluctuations: limits \(T_2^*\). Spin-echo extends to \(T_2\). For hyperfine qubits at “magic” magnetic fields, \(T_2 \sim\) seconds.
• Off-resonant photon scattering: Raman laser beams cause spontaneous emission at rate \(\propto \Omega^2/\Delta^2\) (detuning \(\Delta\) from intermediate state). Large detuning reduces scattering at cost of slower gates.

5.8 Scalability and Quantum Charge-Coupled Device (QCCD)

The primary scalability challenge for trapped ions is that all ions in a single trap share the same motional modes, limiting gate speed as the chain grows and increasing crosstalk. The QCCD architecture addresses this by shuttling ions between small trapping zones — a “quantum CCD” — using electric field gradients to move individual ions, while performing gates in dedicated interaction zones. Ion shuttling preserves coherence (validated by IonQ, Quantinuum) and allows modular scaling.

---

Chapter 6: Superconducting Quantum Circuits

6.1 Superconductivity and Flux Quantization

At temperatures below the critical temperature \(T_c\), certain metals undergo a phase transition to the superconducting state, characterized by zero DC resistance and the Meissner effect (expulsion of magnetic flux). The underlying physics, described by the BCS theory (Bardeen, Cooper, Schrieffer, 1957), involves the condensation of electrons into Cooper pairs — bound states of two electrons mediated by phonons with opposite momenta and spins.

6.1.1 Cooper Pairs and the BCS Ground State

Cooper pairs have charge \(2e\) and are bosonic. Below \(T_c\), they condense into a macroscopic quantum state described by an order parameter \(\Psi(\vec{r}) = |\Psi|e^{i\phi(\vec{r})}\), where \(\phi\) is the macroscopic phase. The supercurrent density is:

\[ \vec{J}_s = \frac{2e n_s \hbar}{2m}\left(\nabla\phi - \frac{2e}{\hbar c}\vec{A}\right) \]

where \(n_s\) is the Cooper pair density. The London equations follow, with penetration depth \(\lambda_L = \sqrt{m/(4\mu_0 n_s e^2)}\).

6.1.2 Magnetic Flux Quantization

In a superconducting loop, the phase \(\phi\) must be single-valued modulo \(2\pi\). This forces the enclosed magnetic flux to be quantized in units of the flux quantum:

\[ \Phi_0 = \frac{h}{2e} \approx 2.068 \times 10^{-15} \text{ Wb} \]

This quantization is macroscopic and observable at the level of a single loop.

6.2 The Josephson Junction

A Josephson junction (JJ) consists of two superconductors separated by a thin insulating barrier. Cooper pairs can tunnel through coherently. The Josephson relations:

\[ I = I_c \sin\delta, \qquad \frac{d\delta}{dt} = \frac{2eV}{\hbar} = \frac{2\pi V}{\Phi_0} \]

where \(\delta = \phi_1 - \phi_2\) is the gauge-invariant phase difference across the junction and \(I_c\) is the critical current. The first is the DC Josephson effect (supercurrent at zero voltage); the second is the AC Josephson effect (voltage drives phase precession at \(2eV/\hbar\)).

The energy stored in the junction is:

\[ E_J = \frac{\hbar I_c}{2e}(1 - \cos\delta) = E_J^0(1 - \cos\delta) \]

where \(E_J^0 = \Phi_0 I_c / (2\pi)\) is the Josephson energy. The junction also has a capacitance \(C_J\), with charging energy \(E_C = e^2/(2C_J)\).

6.3 Circuit Quantization: The Lagrangian Approach

To systematically quantize superconducting circuits, one derives the circuit Lagrangian in terms of node fluxes and then performs canonical quantization. For a node \(i\) connected to ground, the node flux is:

\[ \Phi_i(t) = \int_{-\infty}^t V_i(t')\,dt' \]

The kinetic and potential energies for circuit elements are:

• Capacitor \(C\): \(T = C\dot{\Phi}^2/2\), \(U = 0\) (or equivalently stored energy \(Q^2/(2C) = C\dot\Phi^2/2\)).
• Inductor \(L\): \(T = 0\), \(U = \Phi^2/(2L)\).
• Josephson junction: \(T = C_J\dot{\Phi}^2/2\) (junction capacitance), \(U = -E_J\cos(2\pi\Phi/\Phi_0)\).

The Lagrangian \(\mathcal{L} = T - U\) leads via the Euler-Lagrange equations to the equation of motion. Canonical quantization promotes the node flux to an operator \(\hat{\Phi}\) and its conjugate charge \(\hat{Q} = \partial\mathcal{L}/\partial\dot\Phi\) to an operator satisfying \([\hat\Phi, \hat Q] = i\hbar\).

In terms of dimensionless phase \(\hat\phi = 2\pi\hat\Phi/\Phi_0\) and Cooper pair number \(\hat n = \hat Q/(2e)\):

\[ [\hat\phi, \hat n] = i \]

This is the fundamental commutation relation for superconducting qubits, analogous to \([x,p] = i\hbar\) for mechanical oscillators. The basis states of \(\hat n\) are charge states \(|n\rangle\), and \(\hat\phi\) generates displacements in charge.

6.3.1 LC Circuits and the Anharmonic Oscillator

A superconducting LC resonator has Hamiltonian:

\[ H = \frac{Q^2}{2C} + \frac{\Phi^2}{2L} \]

where \(Q = \hat{n} \cdot 2e\) (charge, discrete) and \(\Phi = \hat{\phi} \cdot \Phi_0/(2\pi)\) (flux, conjugate to charge: \([\hat{\phi}, \hat{n}] = i\)). This is exactly a harmonic oscillator with frequency \(\omega_r = 1/\sqrt{LC}\). The equally spaced spectrum means that any RF drive populates all levels — it cannot be used as a qubit.

A superconducting LC resonator has Hamiltonian \(H = Q^2/(2C) + \Phi^2/(2L)\). Replacing the inductor with a Josephson junction introduces the \(-E_J\cos\phi\) anharmonicity. The Hamiltonian becomes:

\[ H = 4E_C(\hat{n} - n_g)^2 - E_J\cos\hat{\phi} \]

where \(n_g = C_g V_g/(2e)\) is the dimensionless gate charge controlled by a gate voltage \(V_g\). The two energy scales \(E_C\) and \(E_J\) determine the physics of the qubit.

6.3 Charge Qubits and the Transmon

6.3.1 Cooper Pair Box (Charge Qubit)

In the charge qubit (Cooper pair box) regime \(E_J \ll E_C\), the eigenstates are nearly charge states \(|n\rangle\) (number of Cooper pairs). At the “sweet spot” \(n_g = 1/2\), the two lowest charge states \(|0\rangle\) and \(|1\rangle\) are degenerate and form the qubit. First-order charge noise sensitivity vanishes at this point (\(d\omega_{01}/dn_g = 0\)), but the qubit remains sensitive to higher-order noise and the range of \(n_g\) where noise is suppressed is narrow.

The transition frequency near the sweet spot:

\[ \hbar\omega_{01} \approx \sqrt{8E_C E_J} - E_C \]

6.3.2 The Transmon Qubit

The transmon (transmission-line shunted plasma oscillation qubit), introduced by Koch et al. (2007), operates in the regime \(E_J \gg E_C\) by shunting the junction with a large capacitor. The Hamiltonian \(H = 4E_C(\hat{n}-n_g)^2 - E_J\cos\hat\phi\) is equivalent to a particle of mass \(\propto 1/E_C\) in the cosine potential \(-E_J\cos\phi\). In the large \(E_J/E_C\) limit, the potential wells are deep and the low-lying energy levels can be found by expanding the cosine about its minimum: \(-E_J\cos\phi \approx -E_J + E_J\phi^2/2 - E_J\phi^4/24 + \cdots\). The leading harmonic approximation gives plasma frequency \(\omega_p = \sqrt{8E_C E_J}/\hbar\). The quartic correction shifts the energy levels by \(-E_C/\hbar\) per level, introducing anharmonicity \(\alpha = \omega_{12} - \omega_{01} = -E_C/\hbar\). In this limit:

\[ \omega_{01} \approx \sqrt{8E_C E_J}/\hbar - E_C/\hbar \]

The charge dispersion — the variation of \(\omega_{01}\) with gate charge — falls exponentially with \(E_J/E_C\):

\[ \epsilon_m \propto e^{-\sqrt{8E_J/E_C}} \]

For \(E_J/E_C \approx 50\), charge noise is negligible. The cost is reduced anharmonicity: the anharmonicity \(\alpha = \omega_{12} - \omega_{01} \approx -E_C/\hbar\). Typical values: \(E_C/h \approx 200\text{–}300\) MHz, \(E_J/h \approx 15\text{–}25\) GHz, \(\omega_{01}/2\pi \approx 4\text{–}6\) GHz. The negative anharmonicity allows frequency-selective driving.

6.4 Circuit Quantum Electrodynamics (cQED)

Circuit QED couples a transmon qubit to a microwave resonator (coplanar waveguide resonator or 3D cavity) via capacitive coupling. The combined Hamiltonian is the Jaynes-Cummings model:

\[ H = \hbar\omega_r a^\dagger a + \frac{\hbar\omega_q}{2}Z + \hbar g(a\sigma^+ + a^\dagger\sigma^-) \]

where \(g\) is the qubit-resonator coupling strength (typically \(g/2\pi \sim 50\text{–}200\) MHz for transmons). In the rotating wave approximation, this Hamiltonian has an exact solution in terms of dressed states (polaritons).

6.4.1 Dressed States and the Jaynes-Cummings Ladder

The Jaynes-Cummings Hamiltonian has the important property that it preserves the excitation number \(\hat N = a^\dagger a + \sigma^+\sigma^-\). Working in the basis \(\{|n, g\rangle, |n-1, e\rangle\}\) for each manifold with \(N = n\), the Hamiltonian is block diagonal:

\[ H_n = \hbar\omega_r n\,\mathbf{I} + \frac{\hbar\Delta}{2}Z_n + \hbar g\sqrt{n}\,X_n \]

where \(\Delta = \omega_q - \omega_r\) and we have defined Pauli operators within the two-dimensional manifold \(\{|n,g\rangle, |n-1,e\rangle\}\). The eigenvalues are:

\[ E_{n,\pm} = \hbar\omega_r n \pm \frac{\hbar}{2}\sqrt{\Delta^2 + 4g^2 n} \]

The dressed state splitting (Rabi splitting) at resonance (\(\Delta = 0\)) is \(2g\sqrt{n}\), demonstrating that the coupling is enhanced for higher Fock states. The \(n = 0\) manifold is the bare ground state \(|0, g\rangle\) with energy 0 (plus zero-point resonator energy); the \(n = 1\) manifold gives the vacuum Rabi doublet split by \(2g\).

6.4.2 Strong Coupling Regime

Strong coupling is achieved when \(g > \kappa, \gamma\) — the coupling exceeds both the cavity decay rate \(\kappa\) and the qubit decay rate \(\gamma\). This is routinely achieved in circuit QED, with \(g/\kappa \sim 10^3\text{–}10^4\). In this regime, the vacuum Rabi splitting \(2g\) is clearly resolved in the transmission spectrum of the resonator.

6.4.3 Dispersive Regime and Qubit Readout

In the dispersive regime \(\Delta = \omega_q - \omega_r \gg g\), the Jaynes-Cummings Hamiltonian is perturbatively diagonalized to second order:

\[ H_\text{disp} = \hbar(\omega_r + \chi Z) a^\dagger a + \frac{\hbar\tilde{\omega}_q}{2}Z \]

where the dispersive shift is \(\chi = g^2/\Delta\). The resonator frequency is shifted by \(\pm\chi\) depending on the qubit state. This is the basis for quantum non-demolition (QND) qubit readout: by measuring the resonator transmission phase or amplitude, the qubit state is inferred without directly driving the qubit transition.

The dispersive readout fidelity is limited by qubit mixing (Purcell decay rate \(\kappa g^2/\Delta^2\)) and photon shot noise. Parametric amplifiers (JPAs, TWPAs) are used to amplify the weak microwave signal near the quantum noise limit.

6.5 Superconducting Gates

6.5.1 Single-Qubit Gates

Microwave pulses at the qubit frequency \(\omega_{01}\) drive Rabi oscillations on the \(|0\rangle\leftrightarrow|1\rangle\) transition. A pulse of amplitude \(\Omega\) and phase \(\phi\) drives:

\[ H_d = \hbar\Omega(t)\cos(\omega_d t + \phi)\cdot X \]

In the rotating frame at \(\omega_d\), this becomes a DC drive along \(\cos\phi\,X + \sin\phi\,Y\). Pulse shaping (Gaussian, DRAG — Derivative Removal via Adiabatic Gate) suppresses leakage to \(|2\rangle\) due to the finite anharmonicity. Single-qubit gate fidelities \(> 99.9\%\) are routinely achieved.

6.5.2 Two-Qubit Gates

Cross-resonance (CR) gate: Drive qubit 1 at the frequency of qubit 2 (\(\omega_{q2}\)). The coupling (via bus resonator or direct capacitor) creates an effective \(ZX\) interaction:

\[ H_\text{CR} \approx \hbar\frac{\Omega_\text{CR}}{2}(Z_1\otimes X_2) \]

The CR gate is the basis of IBM’s native CNOT and is an all-microwave, fixed-frequency-qubit gate.

Parametric gates: Modulating the qubit frequency at the frequency difference \(|\omega_1 - \omega_2|\) activates resonant exchange coupling, implementing iSWAP-family gates.

Controlled-phase (CZ) gate: Exploiting the \(|11\rangle \to |02\rangle\) transition via the nonlinearity, a conditional phase is accumulated. Used by Google and others.

6.6 Flux Qubits and SQUIDs

A flux qubit consists of a superconducting loop with one (or three) small Josephson junctions. The potential landscape has two minima at opposite fluxes through the loop; tunneling between them splits the two-well potential into symmetric and antisymmetric states, forming the qubit. The tunneling amplitude (and hence \(\omega_{01}\) and anharmonicity) is controlled by an external flux.

A SQUID (Superconducting QUantum Interference Device) consists of two Josephson junctions in a superconducting loop. The critical current is:

\[ I_c^\text{eff} = 2I_c\left|\cos\!\left(\frac{\pi\Phi_\text{ext}}{\Phi_0}\right)\right| \]

A DC SQUID thus acts as a flux-tunable Josephson junction, enabling in situ control of \(E_J\) (and hence qubit frequency) by varying the external flux. This is used in flux-tunable transmons (split transmons) for two-qubit parametric gates.

6.7–6.8 Circuit QED Measurement and Markovian Noise

6.7 Quantum Non-Demolition Readout

In the dispersive limit, homodyne or heterodyne detection of the resonator output field determines the qubit state. The measurement duration \(T_m\) trades off signal-to-noise ratio (SNR \(\propto T_m\)) against qubit coherence (fidelity decreases as \(e^{-T_m/T_1}\)). Optimal readout minimizes \(T_m\) while maintaining high fidelity; typical times \(T_m \sim 100\text{–}500\) ns with fidelities \(> 99\%\).

6.8 Noise in Superconducting Qubits

Primary decoherence mechanisms:

• Dielectric loss (\(1/f\) TLS noise): two-level systems (TLS) in amorphous oxides at surfaces and interfaces absorb microwave energy. Loss tangent \(\tan\delta \sim 10^{-6}\) in high-quality substrates (Si, sapphire). Dominated by surfaces.
• Flux noise (\(1/f\) spectrum): random fluctuations in flux through the qubit loop due to surface spins or magnetic impurities. Relevant for flux-sensitive qubits; transmons are much less sensitive.
• Charge noise: gate-charge fluctuations. Negligible for transmons due to exponentially suppressed charge dispersion.
• Quasiparticle tunneling: non-equilibrium quasiparticles (broken Cooper pairs) tunnel through the junction, causing relaxation. Mitigation: quasiparticle traps (normal-metal islands), filters.
• Radiative decay (Purcell effect): coupling to the resonator enables decay channel \(\gamma_P = (g/\Delta)^2 \kappa\). Mitigated by Purcell filters.

State-of-the-art transmon \(T_1 \sim 100\text{–}500\) µs, \(T_2 \sim 50\text{–}300\) µs. Best reported values (2024): \(T_1 \sim 1\) ms.

6.9 Bosonic (Continuous Variable) Qubits

Rather than encoding a qubit in two levels of a transmon, bosonic qubit schemes encode logical information in the infinite-dimensional Hilbert space of a harmonic oscillator (microwave cavity mode), protected by the symmetries of the encoding.

Cat qubits: logical states are superpositions of coherent states \(|\alpha\rangle\) and \(|-\alpha\rangle\) with opposite amplitudes. The Schrödinger cat states \(|\mathcal{C}_\pm\rangle \propto |\alpha\rangle \pm |-\alpha\rangle\) have approximate Z eigenvalues for large \(|\alpha|\). Two-photon driven dissipation stabilizes the qubit manifold and biases errors toward phase flips (easier to correct).

GKP (Gottesman-Kitaev-Preskill) code: logical states are superpositions of evenly-spaced coherent states forming a comb in phase space. GKP qubits can correct single-phonon loss errors and have been experimentally demonstrated in trapped-ion motion and superconducting cavities.

Binomial and dual-rail codes: exploit the photon-number degree of freedom to encode logical information redundantly against specific loss patterns.

---

Chapter 7: Quantum Error Correction

7.1 The Necessity of Error Correction

Physical qubits decohere. For any circuit of depth \(D\) with error rate \(p\) per gate, the total error probability scales as \(\sim Dp\). Without error correction, a useful computation (say, Shor’s algorithm for a 2048-bit number, requiring \(\sim 10^{10}\) gates) would succeed with probability essentially zero for any realistic \(p\).

Quantum error correction (QEC) achieves fault-tolerance: by encoding logical qubits redundantly and periodically correcting errors, the logical error rate can be made arbitrarily small provided the physical error rate falls below a threshold \(p_\text{th}\).

7.2 Fundamental Limits: No-Cloning and the QEC Conditions

Two results shape quantum error correction: the no-cloning theorem and the Knill-Laflamme conditions.

Despite the no-cloning theorem, QEC is possible because the syndrome measurement reveals error information without revealing the logical state. The Knill-Laflamme (KL) conditions precisely characterize when this is possible:

The condition has a clear interpretation: the errors \(E_a\) and \(E_b\) must be “distinguishable” without revealing whether the logical state is \(|0_L\rangle\) or \(|1_L\rangle\). If \(C_{ab}\) were to depend on \(i, j\), learning the syndrome would also reveal the logical state and collapse it.

7.3 Quantum Error Correction Codes

7.3.1 Three-Qubit Codes

The three-qubit bit-flip code corrects arbitrary single-qubit \(X\)-errors (but not \(Z\)-errors):

\[ |0_L\rangle = |000\rangle, \quad |1_L\rangle = |111\rangle \]

Syndrome measurement: measure \(Z_1Z_2\) and \(Z_2Z_3\). The syndromes \((Z_1Z_2, Z_2Z_3)\) take values \((\pm1, \pm1)\):

• \((+1,+1)\): no error
• \((-1,+1)\): \(X_1\)
• \((-1,-1)\): \(X_2\)
• \((+1,-1)\): \(X_3\)

The three-qubit phase-flip code corrects \(Z\)-errors by conjugating with Hadamards: \(|0_L\rangle = |+++\rangle\), \(|1_L\rangle = |---\rangle\).

The nine-qubit Shor code combines both to correct any single-qubit error (including arbitrary rotations, which are discretized by the syndrome measurement into \(X\), \(Y\), or \(Z\) errors):

\[ |0_L\rangle = \frac{1}{2\sqrt{2}}(|000\rangle + |111\rangle)^{\otimes 3}, \quad |1_L\rangle = \frac{1}{2\sqrt{2}}(|000\rangle - |111\rangle)^{\otimes 3} \]

7.4 Stabilizer Formalism

The stabilizer formalism (Gottesman, 1997) provides a compact description of many important QEC codes. A stabilizer code \([[n, k, d]]\) encodes \(k\) logical qubits into \(n\) physical qubits with distance \(d\) (can correct \(\lfloor(d-1)/2\rfloor\) errors).

Error detection: an error \(E\) is detectable if and only if there exists \(S \in \mathcal{S}\) such that \(\{S, E\} = 0\) (anticommutes). Measuring all generators of \(\mathcal{S}\) produces the error syndrome — a binary string of \(\pm 1\) eigenvalues that identifies the error up to code degeneracy.

7.4.1 The Surface Code

The surface code is the leading candidate for fault-tolerant quantum computing. It is a 2D stabilizer code on a square lattice of \(n = d^2\) qubits (data qubits at vertices; syndrome qubits at plaquette centers and vertex stars):

• Plaquette operators (\(Z\)-type): \(B_p = \prod_{i \in p} Z_i\) — detect \(X\)-errors.
• Star operators (\(X\)-type): \(A_s = \prod_{i \in s} X_i\) — detect \(Z\)-errors.

The surface code has parameters \([[d^2, 1, d]]\): one logical qubit, distance \(d\). The logical error rate is:

\[ p_L \approx A\left(\frac{p}{p_\text{th}}\right)^{(d+1)/2} \]

for physical error rate \(p < p_\text{th} \approx 1\%\). To achieve logical error rate \(10^{-10}\) requires \(d \approx 17\text{–}23\) for \(p = 10^{-3}\), demanding \(\sim 300\text{–}500\) physical qubits per logical qubit.

Error correction is performed by decoding the syndrome via minimum-weight perfect matching (MWPM) or neural-network decoders, which identify the most likely error chain consistent with the observed syndrome.

7.5 The Threshold Theorem

The proof proceeds by showing that concatenated error correction codes reduce the logical error rate super-exponentially: after \(l\) levels of concatenation, the logical error rate is:

\[ p_L^{(l)} \approx p_\text{th}\left(\frac{p}{p_\text{th}}\right)^{c^l} \]

for some constant \(c > 1\) depending on the code. For \(p < p_\text{th}\), \(p_L^{(l)} \to 0\) as \(l \to \infty\).

---

Chapter 8: Quantum Algorithms

8.1 The Quantum Circuit Model and Universality

In the circuit model, \(n\) qubits are initialized in \(|0\rangle^{\otimes n}\), a sequence of gates from a universal set \(\mathcal{G}\) is applied, and measurements are performed.

The set \(\{H, T\}\) generates a dense subgroup of \(SU(2)\) — the Clifford + T gate set. Combined with CNOT, it is universal: any \(n\)-qubit unitary can be approximated to within \(\epsilon\) using \(O(\text{poly}(n, 1/\epsilon))\) gates. The Clifford group (generated by \(\{H, S, CNOT\}\)) is not universal but is efficiently classically simulable (Gottesman-Knill theorem). The T gate is the “magic” that makes the full set universal.

The stabilizer formalism underlies the proof: Clifford gates map Pauli operators to Pauli operators by conjugation, so the stabilizer tableau (tracking \(n\) stabilizer generators, each an \(n\)-qubit Pauli string) evolves efficiently. This theorem explains why quantum speedup requires non-Clifford resources — in particular T gates or other non-stabilizer states.

Magic state distillation converts many noisy non-Clifford states into fewer high-fidelity T states, enabling fault-tolerant universal computation from Clifford gates (which can be implemented transversally in the surface code) plus distilled T states. The resource cost is steep: each logical T gate currently requires \(\sim 10^3\) physical gate operations, making T count minimization a central optimization problem.

8.2 Quantum Parallelism and Interference

The power of quantum computation rests on two mechanisms: quantum parallelism (a state \(H^{\otimes n}|0\rangle^{\otimes n} = 2^{-n/2}\sum_{x=0}^{2^n-1}|x\rangle\) encodes all \(2^n\) inputs simultaneously) and quantum interference (phases on amplitude branches can be made to interfere destructively for wrong answers and constructively for right answers).

8.2.1 Deutsch-Jozsa Algorithm

Given a function \(f: \{0,1\}^n \to \{0,1\}\) promised to be either constant (same output for all inputs) or balanced (equal numbers of 0 and 1 outputs), determine which. Classically, worst case requires \(2^{n-1}+1\) queries. The Deutsch-Jozsa algorithm determines this in a single query.

8.3 Grover’s Search Algorithm

Given an unstructured database of \(N = 2^n\) items with exactly one “marked” item (oracle \(f(x^*) = 1\), \(f(x) = 0\) otherwise), find \(x^*\). Classically requires \(O(N)\) queries on average. Grover’s algorithm requires \(O(\sqrt{N})\).

The algorithm works by iterating the Grover operator \(G = (2|\psi\rangle\langle\psi| - \mathbf{I})O_f\) approximately \(\pi\sqrt{N}/4\) times, where \(|\psi\rangle = H^{\otimes n}|0\rangle^{\otimes n}\) is the uniform superposition and \(O_f\) marks the target: \(O_f|x\rangle = (-1)^{f(x)}|x\rangle\).

8.4 Quantum Walks and Amplitude Amplification

Grover’s algorithm is a special case of amplitude amplification (Brassard, Høyer, Mosca, Tapp 2000). Given any quantum algorithm \(\mathcal{A}\) that produces the marked state with probability \(p\), amplitude amplification boosts the success probability to near 1 in \(O(1/\sqrt{p})\) applications of \(\mathcal{A}\) and its inverse — without knowing \(p\) in advance (via the fixed-point or oblivious variants). The key operator is the reflection about the initial state:

\[ Q = -\mathcal{A} S_0 \mathcal{A}^\dagger S_\chi \]

where \(S_0 = \mathbf{I} - 2|0\rangle\langle 0|\) reflects about \(|0\rangle\) and \(S_\chi = \mathbf{I} - 2P_\text{good}\) marks the good states. In the 2D subspace, \(Q\) acts as a rotation by \(2\arcsin(\sqrt{p})\), exactly as in Grover’s analysis.

Quantum walks (the quantum analogue of random walks) provide an alternative framework. The coined quantum walk on a graph \(G\) has a Hilbert space \(\mathcal{H}_V \otimes \mathcal{H}_C\) (vertex \(\otimes\) coin). Each step applies a coin operation \(C\) then a conditional shift \(S\). Unlike classical random walks, quantum walks can explore graphs quadratically faster (hitting time improvement) and achieve exponential speedups on certain oracular problems.

8.5 Shor’s Factoring Algorithm

Shor’s algorithm factors an \(L\)-bit integer \(N\) in polynomial time \(O(L^3)\), exponentially faster than the best known classical algorithm (number field sieve, sub-exponential). The algorithm reduces factoring to order-finding via a classical reduction, and uses quantum phase estimation (QPE) to find the order efficiently.

8.5.1 Reduction to Order-Finding

Choose a random \(a\) with \(1 < a < N\) and \(\gcd(a, N) = 1\) (otherwise we already found a factor). The order \(r\) of \(a\) modulo \(N\) is the smallest positive integer such that \(a^r \equiv 1 \pmod{N}\). If \(r\) is even and \(a^{r/2} \not\equiv -1 \pmod{N}\), then:

\[ \gcd(a^{r/2} \pm 1, N) \in \{1, N\}^c \]

yields a non-trivial factor of \(N\). Classical number theory shows this succeeds with probability \(\geq 1/2\) for a random \(a\).

8.5.2 Quantum Phase Estimation

Given a unitary \(U\) with eigenstates \(|u_j\rangle\) and eigenvalues \(e^{2\pi i\phi_j}\), QPE estimates \(\phi_j\) to \(t\) bits of precision using \(t\) ancilla qubits and one controlled-\(U^{2^k}\) gate for each:

\[ |\text{QPE output}\rangle \approx \sum_j c_j |\tilde{\phi}_j\rangle|u_j\rangle \]

where \(\tilde{\phi}_j\) is a \(t\)-bit approximation to \(\phi_j\). The key circuit: prepare ancillas in \(|0\rangle^{\otimes t}\), apply \(H^{\otimes t}\), apply controlled-\(U^{2^k}\) for \(k = 0, \ldots, t-1\), apply the inverse QFT, measure.

8.5.3 Quantum Fourier Transform

The QFT over \(\mathbb{Z}_{2^n}\) maps:

\[ \text{QFT}|j\rangle = \frac{1}{\sqrt{2^n}}\sum_{k=0}^{2^n-1} e^{2\pi i jk/2^n}|k\rangle \]

The classical DFT on \(N = 2^n\) points requires \(O(N\log N)\) operations. The QFT requires only \(O(n^2)\) quantum gates using the Cooley-Tukey decomposition into Hadamard gates and controlled-phase gates \(R_k = \text{diag}(1, e^{2\pi i/2^k})\):

\[ |j_1\cdots j_n\rangle \xrightarrow{\text{QFT}} \frac{1}{2^{n/2}}\bigotimes_{l=1}^n \left(|0\rangle + e^{2\pi i j/2^l}|1\rangle\right) \]

This product form reveals that the QFT can be built from \(O(n)\) layers of gates. The exponential speedup over classical FFT does not by itself confer a computational advantage (measurement only gives one sample from the QFT), but it becomes powerful when used inside QPE or hidden subgroup algorithms.

8.5.4 Order-Finding via QPE

To find the order \(r\) of \(a \pmod{N}\), define the modular exponentiation unitary \(U_a|x\rangle = |ax \pmod{N}\rangle\). Its eigenstates are:

\[ |u_s\rangle = \frac{1}{\sqrt{r}}\sum_{j=0}^{r-1}e^{-2\pi i sj/r}|a^j \pmod{N}\rangle, \quad s = 0, \ldots, r-1 \]

with eigenvalues \(e^{2\pi i s/r}\). QPE estimates \(s/r\) as a fraction, and the continued fractions algorithm recovers \(r\) from \(s/r\) with high probability. The total complexity is \(O(L^3)\) for an \(L\)-bit \(N\), dominated by the modular exponentiation circuit.

---

Chapter 9: Quantum Communication and Entanglement

9.1 Quantum Teleportation

Quantum teleportation (Bennett et al., 1993) transmits an unknown qubit state from Alice to Bob using a pre-shared Bell pair and two classical bits.

Teleportation does not violate no-signaling: the classical communication of \(m\) is required before Bob can apply his correction. Without it, his qubit is in the maximally mixed state. Teleportation is a fundamental primitive in quantum networks and measurement-based quantum computing.

9.1.1 No-Communication Theorem

Quantum teleportation requires classical communication — Alice’s 2 bits cannot be replaced by quantum correlations alone. The no-communication theorem guarantees this: for any bipartite state \(\rho_{AB}\) and any measurement \(\{M_m^A\}\) by Alice, the reduced state of Bob is:

\[ \rho_B = \text{tr}_A(\rho_{AB}) = \text{const (independent of Alice's measurement choice)} \]

This follows from the linearity of the partial trace and unitarity of Bob’s operations. No information propagates from Alice to Bob faster than light, even when entanglement is shared. The 2 classical bits transmitted in teleportation carry the “correction” information — without them, Bob’s qubit is in the maximally mixed state regardless of \(|\psi\rangle\).

9.1.2 Security of BB84: Information-Theoretic Proof Sketch

The security of BB84 against the most general collective attacks follows from the decoupling theorem. The key steps:

1. Phase error rate bounds information leakage. The QBER in the \(X\)-basis (phase error rate \(e_\phi\)) bounds Eve’s information on the \(Z\)-basis (bit error rate \(e_b\)) via the Sauer-Shelah lemma. For a collective attack, \(e_\phi = e_b\) by symmetry.
2. Privacy amplification. Hashing the sifted key reduces Eve’s information to zero exponentially fast. If Eve holds system \(E\) correlated with Alice’s \(n\)-bit sifted key \(K\), after privacy amplification by a random hash function \(f\) of rate \(R\), Eve’s residual information is bounded by \(2^{-n(H(K|E) - R)}\). Choosing \(R = H(K|E) - \epsilon\) makes the residual exponentially small.
3. Key rate. The asymptotic secure key rate per transmitted bit is:

\[ r = 1 - H(e_b) - H(e_\phi) \geq 1 - 2H(e_b) \]

where \(H(e) = -e\log_2 e - (1-e)\log_2(1-e)\) is the binary entropy. The protocol is secure for \(e_b < 11\%\) (the threshold where \(r > 0\)).

9.2 Entanglement and the CHSH Inequality

Bell inequalities test whether correlations in quantum mechanics can be explained by local hidden variable (LHV) theories. The CHSH inequality (Clauser, Horne, Shimony, Holt, 1969) is the most experimentally tested:

For measurements on a bipartite system with dichotomic outcomes \(\pm 1\), define:

\[ \mathcal{B}_\text{CHSH} = \langle A_1 B_1 \rangle + \langle A_1 B_2 \rangle + \langle A_2 B_1 \rangle - \langle A_2 B_2 \rangle \]

where \(A_1, A_2\) are Alice’s measurement settings and \(B_1, B_2\) are Bob’s. Any LHV theory satisfies \(|\mathcal{B}_\text{CHSH}| \leq 2\). Quantum mechanics predicts the maximum (Tsirelson’s bound) \(|\mathcal{B}_\text{CHSH}| \leq 2\sqrt{2} \approx 2.828\).

For the Bell state \(|\Phi^+\rangle\) with optimal measurement settings (\(A_1 = Z\), \(A_2 = X\), \(B_1 = (Z+X)/\sqrt{2}\), \(B_2 = (Z-X)/\sqrt{2}\)):

\[ \mathcal{B}_\text{CHSH} = 2\sqrt{2} \]

which violates the classical bound by \(\sqrt{2}\). Loophole-free Bell test experiments (Hensen et al. 2015; Giustina et al. 2015; Shalm et al. 2015) have conclusively demonstrated this violation, ruling out all local realistic theories.

9.3 The E91 Protocol

Ekert’s E91 protocol (1991) uses entangled pairs rather than prepared qubits. A source distributes one half of a Bell pair to Alice and one to Bob. They choose random measurement settings; when they choose compatible settings, they get perfectly correlated results — the raw key. The security proof is based on the CHSH inequality: any eavesdropping disturbs the entanglement and reduces the CHSH value below \(2\sqrt{2}\), detectable by Alice and Bob on a sample of their data.

---

Chapter 10: Synthesis — Cross-Platform Comparison and the Path to Fault Tolerance

10.0 Gate Fidelity Metrics and Randomized Benchmarking

Quantifying gate performance requires measuring gate fidelity in a way that is robust to state preparation and measurement (SPAM) errors. Randomized benchmarking (RB) achieves this by:

1. Applying a random sequence of \(m\) Clifford gates \(C_1, \ldots, C_m\), followed by the inverse \(C_m^\dagger\cdots C_1^\dagger\).
2. Measuring survival probability \(F(m)\) (probability of returning to the initial state).
3. Fitting \(F(m) = A p^m + B\), where the decay parameter \(p\) gives the average Clifford gate error rate \(\epsilon_\text{Clifford} = (1-p)(d-1)/d\) for a \(d\)-dimensional system.

The advantage of RB: SPAM errors enter only in \(A\) and \(B\), not in \(p\). The average single-qubit Clifford gate fidelity is \(1 - \epsilon\), related to the average gate fidelity over the Haar measure.

Interleaved RB characterizes a specific gate by interleaving it between random Clifford gates and comparing the decay rate to the reference RB curve:

\[ \epsilon_\text{gate} \approx \frac{(d-1)}{d}\left(1 - \frac{p_\text{interleaved}}{p_\text{reference}}\right) \]

Gate set tomography (GST) provides complete characterization of the gate set as a superoperator (process matrix) using linear inversion or maximum-likelihood estimation, at the cost of exponentially more experiments.

10.1 DiVincenzo Scorecard

Having studied four hardware platforms in depth, we can now systematically apply the DiVincenzo criteria:

Criterion	NMR	Superconducting	Photonic	Trapped Ion
Well-characterized qubits	✓	✓	✓	✓
Initialization	✗ (pseudo-pure)	✓ (>99%)	✓ (single photon)	✓ (>99.9%)
Long coherence vs. gate time	Moderate (\(T_2/t_g \sim 10^4\))	Moderate (\(T_2/t_g \sim 10^3\))	Excellent (photons don’t decohere)	Best (\(T_2/t_g \sim 10^8\))
Universal gate set	✓	✓	✓ (probabilistic CNOT)	✓
Qubit-selective measurement	Partial	✓	✓ (photon detection)	✓
Scalability	✗	✓ (2D arrays)	Partial (routing difficult)	Partial (QCCD)

10.1.1 Fault-Tolerant Syndrome Extraction

A critical but often overlooked aspect of error correction is that the syndrome measurement circuits themselves must be fault-tolerant — errors in ancilla qubits or two-qubit gates during syndrome extraction must not spread catastrophically to data qubits. The key principle is the fault-tolerance criterion: a fault-tolerant gadget for a distance-\(d\) code must not cause more than \(\lfloor(d-1)/2\rfloor\) errors in the data for any single fault location.

For the surface code, syndrome extraction circuits use weight-4 stabilizer measurements (four CNOT gates per plaquette/star). With a single ancilla per stabilizer:

\[ |\text{anc}\rangle = |0\rangle \xrightarrow{H} |{+}\rangle \xrightarrow{\text{CNOTs}} \frac{1}{\sqrt{2}}(|0\rangle + (-1)^s|1\rangle) \xrightarrow{H} |s\rangle \]

where \(s \in \{0,1\}\) is the syndrome bit (\(Z\)-type syndromes use CNOT with the ancilla as control; \(X\)-type syndromes use it as the target or swap the role with Hadamards). A single gate error in this circuit can produce a data qubit error, but since the surface code has distance \(d\), a pattern of at most \(\lfloor(d-1)/2\rfloor\) such errors is correctable.

In practice, syndromes are measured repeatedly (typically \(d\) times per error correction round) to distinguish genuine errors from measurement errors. The 3D space-time syndrome graph (2D space + time dimension) is decoded simultaneously, accounting for both spatial and temporal error correlations.

10.2 Current State of the Art

As of 2024–2025, the leading platforms in terms of qubit count and error rates are:

Superconducting circuits: Google’s Willow processor (105 qubits) demonstrated below-threshold error rates for the surface code — logical qubit error rates decreasing as the code distance increases, the first demonstration of the surface code threshold in a 2D system (2024). IBM’s Condor processor (1121 qubits, 2023) demonstrates scaling, though not yet fault-tolerant.

Trapped ions: IonQ and Quantinuum achieve the highest two-qubit gate fidelities (\(> 99.9\%\)). Quantinuum’s H2 system (56 qubits in a 32-qubit fully-connected QCCD) has demonstrated logical qubit operations with post-selected fidelities \(> 99.99\%\).

Photonics: PsiQuantum is pursuing fault-tolerant photonic computing with silicon photonics; the approach requires \(\sim 10^6\) physical qubits per logical qubit and is long-term.

10.3 Quantum Advantage and the Road Ahead

Quantum advantage — the demonstration that a quantum computer performs a specific task faster than any classical computer — was claimed by Google in 2019 (random circuit sampling), but the classical hardness of this task has been subsequently debated. A provable, practically useful quantum advantage beyond simulation remains the central open challenge.

The path to fault-tolerant universal quantum computing requires:

1. Physical error rates below the fault-tolerance threshold (\(p < 10^{-3}\) for the surface code with depolarizing noise).
2. \(\sim 10^3\text{–}10^4\) physical qubits per logical qubit (for the surface code at practical distances).
3. Fast enough classical control to decode syndromes in real time (decoding latency \(\lesssim 1\) µs for superconducting qubits).
4. Efficient magic state distillation for non-Clifford gates (T gates), which cost \(\sim 10^3\) physical operations per logical T gate at current estimates.

The combined resource requirement for a fault-tolerant Shor calculation on a 2048-bit RSA key is estimated at \(\sim 10^6\) physical qubits with millisecond coherence times — a target that motivates the engineering across all platforms studied in this course.