The Bloch Sphere

A general single-qubit pure state is a unit vector in $\mathbb{C}^2$:

with $\alpha,\beta \in \mathbb{C}$. Naively this state is specified by four real numbers (two real and two imaginary parts). The normalization constraint removes one real degree of freedom, leaving three. But quantum mechanics postulates that two state vectors differing only by an overall complex phase, $|\psi\rangle$ and $e^{i\gamma}|\psi\rangle$ for real $\gamma$, are physically indistinguishable: every measurement probability, computed via $|\langle\phi|\psi\rangle|^2$, is unchanged by such a factor. This is the global phase, and it removes one further real degree of freedom, leaving exactly two real parameters. Two real parameters are precisely what is needed to label a point on the surface of a sphere: a polar angle $\theta \in [0,\pi]$ and an azimuthal angle $\varphi \in [0,2\pi)$, exactly as for latitude and longitude on a globe. Writing the amplitudes in this parametrization,

is called the Bloch sphere representation. The half-angle $\theta/2$ is a deliberate convention: it makes $\theta$ range over the full $[0,\pi]$ while $|0\rangle$ and $|1\rangle$, which are physically orthogonal (perfectly distinguishable), end up diametrically opposite on the sphere ($\theta = 0$ and $\theta = \pi$) rather than at right angles. This matches physical intuition better than the raw amplitude angle would. The key conceptual split is between global phase and relative phase. Global phase multiplies the entire state vector, $e^{i\gamma}|\psi\rangle$, and is unobservable — it never appears in the Bloch angles at all. Relative phase is the phase difference between the $|1\rangle$ amplitude and the $|0\rangle$ amplitude; after factoring out global phase to make the $|0\rangle$ coefficient real and non-negative, this relative phase survives as exactly the angle $\varphi$. It is directly observable, for instance by measuring in the $X$ or $Y$ basis, and it is what distinguishes states like $|+\rangle = \frac{1}{\sqrt2}(|0\rangle+|1\rangle)$ from $|-\rangle = \frac{1}{\sqrt2}(|0\rangle - |1\rangle)$, which have the same $\theta$ but different $\varphi$.

Step 1 — start from amplitudes. Given $|\psi\rangle = \alpha|0\rangle + \beta|1\rangle$ with $|\alpha|^2+|\beta|^2=1$, write $\alpha = |\alpha|e^{i\gamma_\alpha}$ and $\beta = |\beta|e^{i\gamma_\beta}$ in polar form.

Step 2 — strip global phase. Multiply the whole state by $e^{-i\gamma_\alpha}$. This does not change the physical state, but now the $|0\rangle$ coefficient is real and non-negative: $\alpha' = |\alpha|$, and $\beta' = |\beta|e^{i(\gamma_\beta - \gamma_\alpha)}$. Define $\varphi = \gamma_\beta - \gamma_\alpha \pmod {2\pi}$ — this is the relative phase.

Step 3 — read off angles. Since $|\alpha|^2 + |\beta|^2 = 1$, there is a unique $\theta \in [0,\pi]$ with

Concretely, $\theta = 2\arccos(|\alpha|)$ (equivalently $\theta = 2\arcsin(|\beta|)$, with the branch fixed by $\theta\in[0,\pi]$).

Step 4 — plot the point. $(\theta,\varphi)$ are spherical coordinates on a unit sphere. The corresponding Cartesian (Bloch vector) coordinates are:

These are exactly the expectation values of the Pauli operators: $x = \langle\psi|X|\psi\rangle$, $y=\langle\psi|Y|\psi\rangle$, $z=\langle\psi|Z|\psi\rangle$, which is why $(x,y,z)$ is called the Bloch vector.

Step 5 (inverse) — angles back to amplitudes. Given $(\theta,\varphi)$, reconstruct

which is unique up to the (physically irrelevant) global phase you discarded in Step 2. ``` FUNCTION StateToBlochAngles(alpha, beta):

# alpha, beta are complex numbers, |alpha|^2 + |beta|^2 = 1 ASSERT |alpha|^2 + |beta|^2 ≈ 1 # normalization check r_alpha = |alpha| r_beta = |beta| theta = 2 arccos(clamp(r_alpha, -1, 1)) # theta in [0, pi] IF r_alpha < EPS: # alpha ≈ 0 (pure |1>) phi = arg(beta) # any global phase convention ELSE IF r_beta < EPS: # beta ≈ 0 (pure |0>) phi = 0 # azimuth undefined at pole; pick 0 ELSE: phi = arg(beta) - arg(alpha) # relative phase phi = phi mod 2pi RETURN (theta, phi) FUNCTION BlochAnglesToState(theta, phi): alpha = cos(theta / 2) beta = exp(i phi) sin(theta / 2) RETURN (alpha, beta) # canonical representative,

# alpha real and ≥ 0 ``` The pseudocode's edge-case handling matters: at the poles ($\theta=0$ or $\theta=\pi$) one of $\alpha,\beta$ is exactly zero, so the relative phase $\varphi$ is geometrically meaningless (longitude is undefined at the poles of a globe too) — the function returns a conventional value rather than dividing by zero or taking `arg` of `0`.

"Complexity" here is representational, not computational. Converting between amplitudes $(\alpha,\beta)$ and angles $(\theta,\varphi)$ costs a constant number of arithmetic operations: one `arccos`, one or two `arg` (complex argument / `atan2`) calls, and a subtraction. There are no loops, no data structures, and no dependence on any input size — it is $O(1)$ time and $O(1)$ space per qubit, because a single qubit's pure state is described by a fixed, small number of real parameters regardless of context.

Correctness — this is a bijection (up to global phase). The argument has three parts. Existence. Given any valid amplitudes $\alpha,\beta$ with $|\alpha|^2+|\beta|^2=1$, Step 2–3 always produce a valid $\theta \in [0,\pi]$: this follows because $|\alpha| \in [0,1]$ by normalization, and $\arccos$ is defined and single-valued on $[-1,1]$, mapping onto $[0,\pi]$. The value $\varphi$ always exists modulo $2\pi$ as the difference of two arguments (except at the measure-zero poles, handled by convention). Uniqueness up to global phase. Suppose two amplitude pairs $(\alpha,\beta)$ and $(\alpha',\beta')$ give the same $(\theta,\varphi)$ via the canonical form (real non-negative $\alpha$). Then $\alpha=\alpha'=\cos(\theta/2)$ and $\beta=\beta'=e^{i\varphi}\sin(\theta/2)$ exactly, since the canonicalization removes all phase freedom. Conversely, if $(\alpha,\beta)$ and $(\alpha',\beta')$ describe the same physical state, they differ by a global phase $e^{i\gamma}$; Step 2 strips exactly this factor by construction (it multiplies by $e^{-i\gamma_\alpha}$, whatever that phase is), so both canonicalize to the identical $(\theta,\varphi)$. Hence the map from physical states (equivalence classes under global phase) to points $(\theta,\varphi)$ is a genuine bijection. Surjectivity onto the sphere. Every $(\theta,\varphi) \in [0,\pi]\times[0,2\pi)$ arises from some valid normalized state, namely the one produced by `BlochAnglesToState`, and one checks directly that $|\cos(\theta/2)|^2 + |\sin(\theta/2)|^2 = 1$ for all $\theta$ — so the map covers the whole sphere, not just part of it.

Why Pauli expectation values match spherical coordinates. The identity $x=\langle X\rangle, y=\langle Y\rangle, z=\langle Z\rangle$ is not a coincidence; it is a direct computation. For $|\psi\rangle=\cos(\theta/2)|0\rangle+e^{i\varphi}\sin(\theta/2)|1\rangle$,

using $Z|0\rangle=|0\rangle, Z|1\rangle=-|1\rangle$, and similarly expanding $X=\begin{pmatrix}0&1\\1&0\end{pmatrix}$ and $Y=\begin{pmatrix}0&-i\\i&0\end{pmatrix}$ against the same state gives $\langle X\rangle=\sin\theta\cos\varphi$ and $\langle Y\rangle=\sin\theta\sin\varphi$. This is exactly why a unitary gate, which conjugates the density matrix and therefore linearly transforms $(\langle X\rangle,\langle Y\rangle,\langle Z\rangle)$, acts as a rotation of the Bloch vector — the geometric intuition used throughout single-qubit gate theory is not a pedagogical shortcut, it is a provable consequence of the Pauli algebra.

Honest limitations. The Bloch sphere is a surface, and it represents only pure states — states with certainty of preparation, no classical uncertainty mixed in. A statistical mixture of pure states (a mixed state, described by a density matrix rather than a state vector) corresponds to a point strictly inside the sphere, with the fully mixed state (maximal uncertainty) at the exact center. The representation also does not extend simply to multiple qubits: an $n$-qubit pure state has $2^{n+1}-2$ real parameters (after normalization and global phase), which for $n\ge2$ already exceeds what any single low-dimensional sphere can encode, and entangled states have no faithful "per-qubit Bloch vector" picture at all — tracing out the rest of the system to get a single-qubit Bloch vector for one qubit of an entangled pair necessarily loses information and lands strictly inside the sphere.

Take the state produced by applying a Hadamard-like superposition with a phase, a common intermediate state in phase-estimation-style circuits:

Sanity check going backward: plug $(\theta,\varphi)=(\pi/2,\pi/2)$ into `BlochAnglesToState`: $\alpha=\cos(\pi/4)=1/\sqrt2$, $\beta=e^{i\pi/2}\sin(\pi/4) = i/\sqrt2$ — matches the original state exactly, confirming the round trip. Now contrast with a state that has both a nontrivial global phase and a nontrivial relative phase, to show the two effects don't get confused:

Here $\alpha = \tfrac{\sqrt3}{2}e^{i\pi/3}$ and $\beta = \tfrac12 e^{i\pi/3}e^{i\pi/4} = \tfrac12 e^{i7\pi/12}$.

• $|\alpha| = \sqrt3/2 \approx 0.866$, so $\theta = 2\arccos(0.866) = 2\cdot(\pi/6) = \pi/3$.
• $\arg(\alpha) = \pi/3$, $\arg(\beta)=7\pi/12$, so $\varphi = 7\pi/12 - \pi/3 = 7\pi/12 - 4\pi/12 = \pi/4$.

Notice: the leading factor $e^{i\pi/3}$ (global phase) never appears in $\theta$ or $\varphi$ — it cancels exactly in the subtraction $\arg(\beta)-\arg(\alpha)$. Only the relative phase $\pi/4$ between the two terms survives, exactly as the theory predicts. Both $|\phi\rangle$ and $\tfrac{\sqrt3}{2}|0\rangle+\tfrac12 e^{i\pi/4}|1\rangle$ map to the identical point $(\theta,\varphi)=(\pi/3,\pi/4)$ on the sphere, confirming they are the same physical state.

The Bloch sphere is named for Felix Bloch, who introduced the geometric picture in the context of nuclear magnetic resonance, where a spin-1/2 particle's state precesses under a magnetic field in a way that is naturally visualized as a vector on a sphere. The same geometry was later recognized as exactly the right tool for describing any two-level quantum system, including the qubit of quantum computing.

history

• 1946 — Felix Bloch publishes the equations describing nuclear magnetic resonance, introducing the vector picture of spin precession that becomes the Bloch sphere.
• 1957 — The Bloch sphere is adopted broadly in the physics of two-level systems (masers, lasers) beyond its original NMR context.
• 1980s — As quantum information theory develops, the two-level system is formalized as the abstract "qubit," and the Bloch sphere becomes its standard visualization.
• 1990s — Textbooks and courses on quantum computing (following the rise of Dirac Notation and the State Vector as a teaching foundation) adopt the Bloch sphere as the canonical picture for single-qubit gates.
• 2000s–present — The Bloch sphere becomes a standard diagnostic tool in experimental quantum hardware labs, used to visualize measured qubit states and gate errors on real superconducting and trapped-ion devices.

Every major single-qubit gate — $X$, $Y$, $Z$, Hadamard, phase gates, and general rotation gates — is most easily understood and remembered as a specific rotation of the Bloch sphere: $X$ is a $\pi$ rotation about the $x$-axis, $Z$ a $\pi$ rotation about the $z$-axis, and a general rotation gate $R_{\hat n}(\theta)$ rotates the Bloch vector by angle $\theta$ about axis $\hat n$. This geometric fact is exactly what makes the transition to Single-Qubit Gates: X, Y, Z, H, Rotation Gates and Arbitrary Single-Qubit Unitaries intuitive rather than purely algebraic. In experimental physics, quantum hardware teams (superconducting circuits, trapped ions, photonic qubits) routinely plot measured qubit states on Bloch spheres to visualize calibration errors, decoherence (the Bloch vector shrinking toward the center as the state becomes mixed over time), and the fidelity of applied pulses. Quantum optimal control — designing microwave pulses that implement a target gate — is frequently framed as steering a point along a desired path on the sphere. Software toolkits for quantum computing (simulators and visualizers used alongside frameworks like Qiskit, Cirq, and similar platforms) include Bloch sphere plotting as a core diagnostic, because it is the fastest way for a human to sanity-check a single-qubit state or debug an unexpected phase. The concept also underlies how textbooks introduce the geometric meaning of The Qubit and Superposition, giving beginners a concrete picture before the abstract linear-algebra machinery of multi-qubit systems and entanglement is introduced.

Ask the tutor: How do the X, Y, and Z gates each correspond to a specific rotation axis on the Bloch sphere, and why does that fully explain their action on any input state?