Bloch Sphere Representation of Gates

A single-qubit pure state, up to an unobservable global phase, is one unit arrow from the centre of a ball to its surface — a point on the Bloch sphere. Write $|\psi\rangle=\cos(\theta/2)|0\rangle+e^{i\varphi}\sin(\theta/2)|1\rangle$ and the arrow has polar angle $\theta$ and azimuth $\varphi$ ; the north pole is $|0\rangle$ , the south pole $|1\rangle$ , and the equator carries the equal superpositions. A gate cannot lengthen the arrow (it preserves purity) nor bend angles between arrows (it preserves inner products), so the only thing left is to turn the whole ball rigidly. Every single-qubit unitary is a rotation of the Bloch sphere — an axis and an angle, nothing more. This is not an analogy but an exact theorem: the surjective two-to-one Lie-group homomorphism $SU(2)\to SO(3)$ . A special-unitary $U$ acts on Bloch vectors by conjugation, $U(\hat r\cdot\vec\sigma)U^{\dagger}=(R\hat r)\cdot\vec\sigma$ , where $R$ is a genuine $3\times3$ rotation and $\vec\sigma=(X,Y,Z)$ are the Pauli matrices. The kernel is exactly $\{+I,-I\}$ — the freedom of global phase — so $U$ and $-U$ paint the same picture. The generator is $R_{\hat n}(\theta)=\exp(-i\tfrac{\theta}{2}\,\hat n\cdot\vec\sigma)=\cos(\theta/2)I-i\sin(\theta/2)\,\hat n\cdot\vec\sigma$ , and the factor of two in $\theta/2$ is the spinor double cover: a full $2\pi$ turn of the sphere sends the state vector to minus itself. From this one formula you read the axis and angle of every named gate, compile arbitrary gates into hardware rotations, and prove circuit identities by moving axes instead of multiplying matrices.

Key takeaways

A single-qubit pure state is a point on the unit Bloch sphere, $|\psi\rangle=\cos(\theta/2)|0\rangle+e^{i\varphi}\sin(\theta/2)|1\rangle$ ↔ $\hat r=(\sin\theta\cos\varphi,\sin\theta\sin\varphi,\cos\theta)$ ; mixed states fill the interior.
Up to global phase, every single-qubit unitary is a rotation $R_{\hat n}(\theta)=\exp(-i\theta\,\hat n\cdot\vec\sigma/2)=\cos(\theta/2)I-i\sin(\theta/2)\,\hat n\cdot\vec\sigma$ of the sphere about axis $\hat n$ by angle $\theta$ .
The gate-to-rotation map is the 2-to-1 homomorphism $SU(2)\to SO(3)$ with kernel $\{+I,-I\}$ , so $\pm U$ give the same rotation; global phase is invisible.
$X,Y,Z$ are $\pi$ -turns about $\hat x,\hat y,\hat z$ ; $S,T$ are $\hat z$ -turns of $\pi/2,\pi/4$ ; Hadamard is a $\pi$ -turn about the diagonal $(\hat x+\hat z)/\sqrt2$ .
Rotations do not commute; a product's axis and angle follow the $SU(2)$ half-angle law, and $R_{\hat n}(2\pi)=-I$ is the spinor sign flip.
Key idea: up to global phase a single-qubit gate is exactly a rotation $R_{\hat n}(\theta)=\exp(-i\theta\,\hat n\cdot\vec\sigma/2)$ of the Bloch sphere; the gate's eigenvectors are the $\pm\hat n$ poles and its eigenphase difference is the angle $\theta$ .

The core insight

Up to global phase, a single-qubit unitary is exactly a rotation $R_{\hat n}(\theta)=\exp(-i\theta\,\hat n\cdot\vec\sigma/2)$ of the Bloch sphere about axis $\hat n$ by angle $\theta$ ; the gate's eigenvectors sit at the $\pm\hat n$ poles and its eigenphase difference is $\theta$ .

Where this lives today

The gate-as-rotation picture is the working language of quantum control and compilation. On real hardware a gate is a pulse: a resonant microwave or laser drive implements $R_{\hat n}(\theta)$ with the angle $\theta$ set by the pulse area and the axis $\hat n$ set by the drive phase, so changing a phase in software literally rotates the axis on the sphere. Native gate sets are chosen in these terms — IBM's superconducting qubits expose $R_z$ (a virtual phase, zero pulse time and essentially exact) with the fixed $\sqrt X$ half-turn of worked example 2, and a compiler builds any single-qubit gate from an $R_z$ – $\sqrt X$ – $R_z$ – $\sqrt X$ – $R_z$ sandwich, the Z-Y decomposition read on the sphere. Composite-pulse and dynamical-decoupling sequences (spin echoes, CPMG, BB1) are geometric paths that return the arrow to its target even when the drive is miscalibrated. And the universality behind small gate sets — the Solovay–Kitaev theorem — is at heart that two rotations about non-parallel axes through angles incommensurate with $\pi$ densely fill $SO(3)$ , so a Hadamard and a $T$ approximate any single-qubit rotation to arbitrary precision with only logarithmic overhead.

The physical picture

Think of the qubit as an arrow from the centre of a unit ball to a point on its surface. A unitary preserves every vector's length and the angle between any two, so on the sphere it leaves one freedom: turn the whole ball rigidly about a line through the centre. That line is the axis $\hat n$ and the amount of turning is the angle $\theta$ — together they name the gate completely. The axis is the gate's eigenbasis lifted onto the sphere: the two eigenstates sit at the poles of $\hat n$ , left untouched except for a rephasing that between them is exactly $\theta$ . So "diagonalise the gate" and "find its Bloch axis and angle" are one task.

Figure

The Bloch sphere: a pure state is a unit arrow at polar angle θ from ẑ and azimuth φ, so cos(θ/2) sets the amplitude of state 0 and φ the relative phase; the vector r̂ = (sinθ cosφ, sinθ sinφ, cosθ) has length 1 for pure states and 0 at the maximally mixed centre, north pole 0 and south pole 1

Because the state vector lives in $SU(2)$ while the observable arrow lives in $SO(3)$ , there is a built-in factor of two: turning the sphere by $\theta$ advances the state vector through $\theta/2$ in the exponential. Run the sphere around by $2\pi$ and every arrow returns home, yet the state vector picks up a minus sign — the spinor sign flip — undetectable because global phase has no image on the sphere. Conjugation carries an equally clean meaning: $UVU^{\dagger}$ is $V$ with its axis rotated by $U$ — why $HZH=X$ (Hadamard swings $\hat z$ onto $\hat x$ ) and why Clifford sandwiches merely permute the Pauli axes.

Figure

Gate as rotation: X, Y, Z are π turns about x̂, ŷ, ẑ; S and T are ẑ turns of π/2 and π/4; Hadamard is a π turn about the diagonal (x̂+ẑ)/√2 that carries the north pole 0 onto the equatorial + state; and composing R_x(π/2) after R_z(π/2) gives a 2π/3 turn about the body diagonal (x̂−ŷ+ẑ)/√3

The core equations

The starting point is the dictionary between a qubit state and a point in space. After the global phase is dropped a pure state keeps two real parameters, exactly the two spherical angles; a general mixed state is the same expression with a Bloch vector of length at most one.

|\psi\rangle=\cos\!\tfrac{\theta}{2}\,|0\rangle+e^{i\varphi}\sin\!\tfrac{\theta}{2}\,|1\rangle\;\longleftrightarrow\;\hat r=(\sin\theta\cos\varphi,\sin\theta\sin\varphi,\cos\theta) \qquad (1)

where: $\theta$ — polar angle, $0\le\theta\le\pi$ ; $\varphi$ — azimuthal angle, $0\le\varphi<2\pi$ ; $\hat r$ — unit Bloch vector. Reading it: the poles ( $\theta=0,\pi$ ) are $|0\rangle$ and $|1\rangle$ , and the half-angle $\theta/2$ is why antipodal points are orthogonal states. Packaging the same data as a density operator makes the whole ball visible. Every one-qubit state is the identity plus a Bloch vector contracted with the Paulis.

\rho=\tfrac12\left(I+\hat r\cdot\vec\sigma\right),\qquad |\hat r|\le 1 \qquad (2)

where: $\rho$ — density operator; $\vec\sigma=(X,Y,Z)$ — Pauli matrices; $\hat r$ — Bloch vector, $|\hat r|=1$ pure, $<1$ mixed, $0$ maximally mixed. Reading it: purity is $\mathrm{tr}\,\rho^2=\tfrac12(1+|\hat r|^2)$ , so length on the sphere measures how pure the state is. A rotation by angle $\theta$ about axis $\hat n$ is the exponential of the corresponding Pauli combination; because that combination squares to the identity, the exponential collapses to a cosine plus a sine — the workhorse formula.

R_{\hat n}(\theta)=\exp\!\left(-i\tfrac{\theta}{2}\,\hat n\cdot\vec\sigma\right)=\cos\!\tfrac{\theta}{2}\,I-i\sin\!\tfrac{\theta}{2}\,(\hat n\cdot\vec\sigma) \qquad (3)

where: $\hat n=(n_x,n_y,n_z)$ — unit axis; $\theta$ — rotation angle; $\hat n\cdot\vec\sigma=n_xX+n_yY+n_zZ$ — a traceless Hermitian involution. Reading it: $\hat n$ is fixed and the poles are eigenvectors with eigenvalues $e^{\mp i\theta/2}$ , whose phase difference is $\theta$ . That a gate rotates the sphere at all is the statement that conjugation of a Bloch vector by a unitary is a genuine three-dimensional rotation — the $SU(2)\to SO(3)$ homomorphism, derived below.

U\,(\hat r\cdot\vec\sigma)\,U^{\dagger}=(R\hat r)\cdot\vec\sigma,\qquad U\in SU(2),\;R\in SO(3) \qquad (4)

where: $U$ — special-unitary representative; $R$ — the $3\times3$ real rotation it induces. Reading it: $U\mapsto R$ is a homomorphism, so a product of gates rotates by the product of rotations. The map is two-to-one, and the factor of two shows up as the spinor sign: a single full turn is the identity on every arrow but minus the identity on the state vector.

R_{\hat n}(2\pi)=-I,\qquad R_{\hat n}(4\pi)=+I \qquad (5)

where: $-I$ — unobservable global phase, the same Bloch rotation as $+I$ . Reading it: the kernel is $\{+I,-I\}$ , so two turns of the state vector make one turn of the sphere look complete. Going backwards — matrix to axis and angle — is trace arithmetic once the global phase is stripped: the trace gives the angle, the Pauli projections the axis.

V=e^{-i\alpha}U\in SU(2),\quad \cos\tfrac{\theta}{2}=\big|\tfrac12\,\mathrm{tr}\,V\big|,\quad n_j=\frac{-\,\mathrm{Im}\,\mathrm{tr}(V\sigma_j)}{2\sin(\theta/2)} \qquad (6)

where: $\alpha=\tfrac12\arg\det U$ — global phase carried by $U$ ; $\theta$ — recovered angle; $n_j$ — axis components, defined when $\sin(\theta/2)\neq0$ . Reading it: if $\theta=0$ then $V=I$ and the axis is undefined, as a do-nothing rotation should be. Products of rotations obey a half-angle law — $SU(2)$ multiplication in axis–angle form — the cross-product term encoding non-commutativity.

\cos\tfrac{\gamma}{2}=\cos\tfrac{\beta}{2}\cos\tfrac{\alpha}{2}-(\hat m\cdot\hat n)\sin\tfrac{\beta}{2}\sin\tfrac{\alpha}{2} \qquad (7)

where: $R_{\hat m}(\beta)R_{\hat n}(\alpha)=R_{\hat k}(\gamma)$ ; $\gamma$ — net angle; $\hat m\cdot\hat n$ — cosine between the axes. Reading it: for parallel axes the last term vanishes and angles add (why all $z$ -rotations commute); for perpendicular axes the net axis tilts out of both.

Derivation AWhy the rotation generator collapses to cosine plus sine

The goal is to show that the operator exponential $\exp(-i\tfrac{\theta}{2}\,\hat n\cdot\vec\sigma)$ — a formally infinite power series of $2\times2$ matrices — sums exactly to the closed form $\cos(\theta/2)I-i\sin(\theta/2)\,\hat n\cdot\vec\sigma$ , and that the only property doing the work is that $\hat n\cdot\vec\sigma$ squares to the identity. This one algebraic fact stands behind every named-gate identity in the article.

Step 1. Square the generator. Using $\sigma_i\sigma_j=\delta_{ij}I+i\epsilon_{ijk}\sigma_k$ , $(\hat n\cdot\vec\sigma)^2=\sum_{i,j}n_in_j\sigma_i\sigma_j=|\hat n|^2\,I=I$ , since the symmetric part gives $|\hat n|^2$ and the antisymmetric $\epsilon_{ijk}$ part vanishes against symmetric $n_in_j$ . Why: an operator whose square is $I$ behaves in an exponential exactly like the imaginary unit in $e^{i\phi}$ , forcing a cosine–sine answer.

Step 2. Split by parity. With $A=\hat n\cdot\vec\sigma$ , the series $\sum_k\tfrac{1}{k!}(-i\tfrac{\theta}{2})^kA^k$ has even powers $A^{2m}=I$ and odd powers $A^{2m+1}=A$ , separating into an $I$ -part and an $A$ -part. Why: only $A^2=I$ makes every even power the identity and every odd power $A$ , so no other matrix survives.

Step 3. Resum. The even series is $\cos(\theta/2)$ and the odd series is $-i\sin(\theta/2)$ , giving $R_{\hat n}(\theta)=\cos(\theta/2)I-i\sin(\theta/2)A$ . Why: these are the Taylor series of cosine and sine, with $I$ and $A$ standing in for the real and imaginary units.

Go deeper · Derivation B · 1 minThe SU(2) to SO(3) homomorphism and its kernel

Derivation BThe SU(2) to SO(3) homomorphism and its kernel

The goal is to prove that conjugating a Bloch vector by a single-qubit unitary genuinely rotates it — that the induced map $R$ on $\mathbb R^3$ is orthogonal with determinant one — and that the only unitaries acting trivially are $\pm I$ , so the gate-to-rotation correspondence is exactly two-to-one. Encoding the Bloch vector as a traceless Hermitian matrix turns the question into linear algebra on a three-dimensional real space.

Step 1. Identify the arena. Set $M=\hat r\cdot\vec\sigma$ ; it is traceless and Hermitian, and the three Paulis are a real basis of such matrices, so $\hat r\mapsto M$ is a linear isomorphism of $\mathbb R^3$ with the traceless Hermitian $2\times2$ matrices. Why: putting the geometry inside a matrix space lets unitary conjugation act on vectors, which is where the rotation comes from.

Step 2. Show conjugation stays in the arena. For any $U\in SU(2)$ , $\mathrm{tr}(UMU^{\dagger})=\mathrm{tr}\,M=0$ and $UMU^{\dagger}$ is Hermitian, so it equals $\hat r'\cdot\vec\sigma$ for some real $\hat r'$ ; defining $R$ by $\hat r'=R\hat r$ gives a linear map on $\mathbb R^3$ . Why: preserving trace and Hermiticity keeps the output a legitimate Bloch vector, so $R$ is well defined.

Step 3. Show $R$ is a proper rotation. Since $\det(\hat r\cdot\vec\sigma)=-|\hat r|^2$ is conjugation-invariant, $|\hat r'|=|\hat r|$ , so $R$ is orthogonal; and connecting $U$ to $I$ by a path in $SU(2)$ along which $\det R\in\{+1,-1\}$ starts at $+1$ forces $\det R=+1$ , giving $R\in SO(3)$ . Why: length preservation forces orthogonality, connectedness rules out reflections.

Step 4. Find the kernel. If $R=I$ then $UMU^{\dagger}=M$ for all traceless Hermitian $M$ , so $U$ commutes with all three Paulis and is a scalar $e^{i\alpha}I$ ; $\det U=1$ forces $U=\pm I$ . Why: only multiples of the identity commute with a full operator basis, and special-unitarity cuts the phase to the two-element kernel — the freedom of global phase.

Go deeper · Derivation C · 1 minRecovering the axis and angle of any single-qubit gate

Derivation CRecovering the axis and angle of any single-qubit gate

The goal is a practical recipe: given an arbitrary $2\times2$ unitary $U$ , strip its unobservable global phase, then read its rotation angle from one trace and its axis from three more. This inverts the axis–angle formula and is exactly what a compiler does turning an abstract gate into a hardware pulse, so every step ends in a number.

Step 1. Remove the global phase. Set $\alpha=\tfrac12\arg\det U$ and $V=e^{-i\alpha}U$ , so $\det V=1$ and $V\in SU(2)$ while $U$ and $V$ paint the identical rotation. Why: only the special-unitary representative has the clean cosine–sine form, and $\alpha$ is invisible on the sphere anyway.

Step 2. Angle from the trace. Since $V=\cos(\theta/2)I-i\sin(\theta/2)\,\hat n\cdot\vec\sigma$ and Paulis are traceless, $\mathrm{tr}\,V=2\cos(\theta/2)$ , so $\theta=2\arccos\big|\tfrac12\mathrm{tr}\,V\big|$ . Why: the axis part contributes nothing to the trace, so it isolates the angle.

Step 3. Axis from the Pauli projections. Using $\mathrm{tr}(\sigma_j\sigma_k)=2\delta_{jk}$ , $\mathrm{tr}(V\sigma_j)=-2i\sin(\theta/2)\,n_j$ , so $n_j=-\,\mathrm{Im}\,\mathrm{tr}(V\sigma_j)/(2\sin(\theta/2))$ when $\sin(\theta/2)\neq0$ ; if $\theta=0$ then $V=I$ and the axis is undefined. Why: projecting $V$ onto each Pauli extracts that component, and the sine fails exactly when there is no rotation to orient.

Step 4. Test on $S$ . Writing $S=\mathrm{diag}(1,i)=e^{i\pi/4}\mathrm{diag}(e^{-i\pi/4},e^{i\pi/4})$ gives $\alpha=\pi/4$ , $\mathrm{tr}\,V=2\cos(\pi/4)$ so $\theta=\pi/2$ , and only $\mathrm{tr}(V\sigma_z)\neq0$ , so $\hat n=\hat z$ — a quarter-turn about $\hat z$ . Why: recovering $S=e^{i\pi/4}R_{\hat z}(\pi/2)$ confirms the recipe end to end.

Limiting cases

Special values of the axis–angle formula recover every named gate, each up to a global phase that leaves no trace on the sphere:

Case	Result	Meaning
Pauli $X$	$R_{\hat x}(\pi)=-iX$	$\pi$ -turn about $\hat x$ ; swaps the $\hat z$ poles, exchanging $\|0\rangle$ and $\|1\rangle$ . The $-i$ is global phase.
Pauli $Y$	$R_{\hat y}(\pi)=-iY$	$\pi$ -turn about $\hat y$ ; sends $\|0\rangle\to\|1\rangle$ up to phase while fixing $\|{+}i\rangle,\|{-}i\rangle$ as poles.
Pauli $Z$	$R_{\hat z}(\pi)=-iZ$	$\pi$ -turn about $\hat z$ ; maps $\|+\rangle\leftrightarrow\|-\rangle$ while fixing $\|0\rangle,\|1\rangle$ .
Phase gate $S$	$S=e^{i\pi/4}R_{\hat z}(\pi/2)$	quarter-turn ( $\pi/2$ ) about $\hat z$ ; rotates the equatorial $x$ – $y$ plane by $90°$ .
$T$ gate	$T=e^{i\pi/8}R_{\hat z}(\pi/4)$	eighth-turn ( $\pi/4$ ) about $\hat z$ ; the irrational-angle generator behind Clifford+ $T$ universality.
Hadamard $H$	$H=e^{i\pi/2}R_{\hat n}(\pi),\ \hat n=\tfrac{1}{\sqrt2}(\hat x+\hat z)$	$\pi$ -turn about the $x$ – $z$ diagonal; swaps $\hat x$ and $\hat z$ , hence the computational and Hadamard bases.
Identity	$R_{\hat n}(0)=I$	no rotation; axis undefined, the $\theta\to0$ boundary.

Common trapThe factor-of-two trap, and reading phase off the sphere

The most common error is dropping the factor of two: a rotation by $\theta$ is generated by $\exp(-i\theta\,\hat n\cdot\vec\sigma/2)$ , not $\exp(-i\theta\,\hat n\cdot\vec\sigma)$ . Pauli $Z=\mathrm{diag}(1,-1)$ looks like a "half" operation, yet rotates the sphere by a full $\pi$ because its eigenphases differ by $\pi$ . The mirror mistake is looking for a gate's global phase on the sphere: $U$ and $e^{i\alpha}U$ are the identical rotation, so it is not there. Do not confuse global with relative phase — the relative phase between branches of a superposition is the azimuth $\varphi$ and is very much visible; only the overall phase vanishes. And a $2\pi$ turn is not "back to start" for the state vector: it returns every arrow but flips the ket's sign.

Worked examples

1 · Composing two quarter-turns into a body-diagonal rotation (fully worked)

Apply $R_z(\pi/2)$ then $R_x(\pi/2)$ ; the circuit unitary is $R_x(\pi/2)R_z(\pi/2)$ (right-most acts first). Find its axis and angle.

Step 1. In cosine–sine form, with $\cos(\pi/4)=\sin(\pi/4)=1/\sqrt2$ : $R_z(\pi/2)=\tfrac{1}{\sqrt2}(I-iZ)$ and $R_x(\pi/2)=\tfrac{1}{\sqrt2}(I-iX)$ .

Step 2. Net angle from the scalar law (7) with $\hat m=\hat x,\hat n=\hat z,\alpha=\beta=\pi/2$ . The axes are perpendicular, $\hat m\cdot\hat n=0$ , so $\cos(\gamma/2)=\cos^2(\pi/4)=\tfrac12$ , giving $\gamma/2=\pi/3$ and the net angle $\gamma=2\pi/3=120°$ .

Step 3. Net axis. Multiplying, $R_x(\pi/2)R_z(\pi/2)=\tfrac12(I-iX)(I-iZ)=\tfrac12(I-iX-iZ-XZ)$ , and $XZ=-iY$ gives $\tfrac12\big(I-i(X-Y+Z)\big)$ . Matching $\cos(\gamma/2)I-i\sin(\gamma/2)\hat k\cdot\vec\sigma$ with $\sin(\pi/3)=\sqrt3/2$ gives $\hat k=(\hat x-\hat y+\hat z)/\sqrt3$ .

Step 4. Interpret. A $120°$ turn about a body diagonal cyclically permutes the coordinate axes; here it acts by conjugation as $X\to Z\to Y\to X$ (with signs), the kind of Pauli-permuting operation the Clifford group is built from. Net result: $R_x(\pi/2)R_z(\pi/2)=R_{\hat k}(2\pi/3)$ with $\hat k=(\hat x-\hat y+\hat z)/\sqrt3$ — two perpendicular quarter-turns make one diagonal third-turn, so rotations neither commute nor simply add.

2 · Extracting the axis and angle of the $\sqrt{X}$ gate (fully worked)

The gate $\sqrt X=\tfrac12\big(\begin{smallmatrix}1+i&1-i\\1-i&1+i\end{smallmatrix}\big)$ is hardware-native on many superconducting processors (the "SX" gate). Find its global phase, rotation angle, and axis.

Step 1. Strip the phase. $\det\sqrt X=\tfrac14\big[(1+i)^2-(1-i)^2\big]=\tfrac14[2i-(-2i)]=i=e^{i\pi/2}$ , so $\alpha=\tfrac12\arg\det=\pi/4$ and $V=e^{-i\pi/4}\sqrt X\in SU(2)$ .

Step 2. Angle from the trace. $\mathrm{tr}\,\sqrt X=1+i=\sqrt2\,e^{i\pi/4}$ , so $\mathrm{tr}\,V=e^{-i\pi/4}\sqrt2\,e^{i\pi/4}=\sqrt2=2\cos(\theta/2)$ . Thus $\cos(\theta/2)=1/\sqrt2$ and the rotation angle is $\theta=\pi/2$ — a quarter-turn, as the name $\sqrt X$ demands.

Step 3. Axis from a Pauli projection. $\mathrm{tr}(\sqrt X\,X)=1-i=\sqrt2\,e^{-i\pi/4}$ , so $\mathrm{tr}(V\sigma_x)=\sqrt2\,e^{-i\pi/2}=-i\sqrt2$ . The recipe $\mathrm{tr}(V\sigma_j)=-2i\sin(\theta/2)n_j$ with $\sin(\pi/4)=1/\sqrt2$ gives $-i\sqrt2=-i\sqrt2\,n_x$ , so $n_x=1$ and the other projections vanish. The axis is $\hat x$ .

Step 4. Assemble. $\sqrt X=e^{i\pi/4}R_{\hat x}(\pi/2)$ , a quarter-turn about $\hat x$ up to a $\pi/4$ phase. Applied twice, $(\sqrt X)^2=e^{i\pi/2}R_{\hat x}(\pi)=e^{i\pi/2}(-iX)=X$ — genuinely a square root of the bit-flip, two $\pi/2$ turns making the $\pi$ turn.

3 · Why Hadamard swaps the two bases (axis argument)

Show geometrically that $H$ interchanges $\{|0\rangle,|1\rangle\}$ with $\{|+\rangle,|-\rangle\}$ . $H$ is a $\pi$ -rotation about the bisector $\hat n=(\hat x+\hat z)/\sqrt2$ of $\hat x$ and $\hat z$ . A $\pi$ -rotation about the bisector of two axes exchanges them (and flips the third): $\hat z\to\hat x$ , $\hat x\to\hat z$ , $\hat y\to-\hat y$ . The poles of $\hat z$ are $|0\rangle,|1\rangle$ and those of $\hat x$ are $|+\rangle,|-\rangle$ , so swapping the axes swaps the bases. $HZH=X$ , $HXH=Z$ is read straight off the axis geometry — no matrix products.

Physics of the portrait

The sphere carries the name of a physicist who never worked on quantum computers. It came from the magnetic-resonance laboratory, where a spin- $\tfrac12$ nucleus in a field traces exactly the same geometry a qubit does under a gate.

Origins · 1 minFelix Bloch and how this was discovered

Felix Bloch

1905–1983 · nuclear magnetic resonance and the geometry of two-level systems

Felix Bloch's first fame came from solids: his 1928 doctoral work under Werner Heisenberg produced Bloch's theorem and Bloch waves, the foundation of the quantum theory of electrons in crystals. But the sphere that bears his name comes from later. In 1946 at Stanford, Bloch developed the theory of nuclear induction — the resonant response of nuclear spins to radio-frequency fields — and introduced the magnetization vector whose tip traces a sphere as the spins precess, work for which he shared the 1952 Nobel Prize with Edward Purcell. The Bloch vector and Bloch equations are the direct ancestors of the qubit picture: a driven spin- $\tfrac12$ in a rotating frame turns about an axis set by the drive phase through an angle set by the pulse area — precisely $R_{\hat n}(\theta)$ . The mathematical backbone is older — Hamilton's quaternions (1843) are the algebra of $SU(2)$ , Cartan's spinors (1913) the objects that change sign under a $2\pi$ turn — but it was Feynman, Vernon, and Hellwarth in 1957 who showed that any two-level system maps onto a rotating vector on Bloch's sphere. That is the theorem this article rests on: the qubit is a spin in disguise, and a gate is a rotation.

History

1843

— Hamilton discovers the quaternions, the algebra underlying SU(2)

1913

— Cartan introduces spinors, objects that change sign under a 2π turn

1927

— Pauli writes the spin matrices σ_x, σ_y, σ_z for the electron

1946

— Bloch's nuclear-induction theory gives the Bloch vector its name

1957

— Feynman, Vernon and Hellwarth map any two-level system to a rotating vector

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, §1.3.1 and §4.2 (Bloch sphere; rotation operators; Theorem 4.1) · Preskill, Lecture Notes for Physics 219, Ch. 2 · Watrous, The Theory of Quantum Information, §2 · Kitaev, Shen & Vyalyi, Classical and Quantum Computation, §8 (single-qubit gates; Solovay–Kitaev) · Mermin, Quantum Computer Science, Ch. 1 · Wilde, Quantum Information Theory, Ch. 3. All equations are standard. The rotation operator $R_{\hat n}(\theta)=\cos(\theta/2)I-i\sin(\theta/2)\hat n\cdot\vec\sigma$ follows from $(\hat n\cdot\vec\sigma)^2=I$ ; the $SU(2)\to SO(3)$ homomorphism and kernel $\{\pm I\}$ are classical results; the named-gate identities $X=iR_{\hat x}(\pi)$ , $Z=iR_{\hat z}(\pi)$ , $S=e^{i\pi/4}R_{\hat z}(\pi/2)$ , $T=e^{i\pi/8}R_{\hat z}(\pi/4)$ , $H=e^{i\pi/2}R_{(\hat x+\hat z)/\sqrt2}(\pi)$ , $HXH=Z$ , and $HZH=X$ follow by direct exponentiation; and the worked-example results $R_x(\pi/2)R_z(\pi/2)=R_{(\hat x-\hat y+\hat z)/\sqrt3}(2\pi/3)$ and $\sqrt X=e^{i\pi/4}R_{\hat x}(\pi/2)$ are confirmed by explicit matrix multiplication.

Prerequisites. Pauli Gates (X, Y, Z) · Single-Qubit Rotation Gates · Quantum Circuit Model. Leads to. Z-Y Decomposition · Universal Gate Sets · Circuit Identities and Equivalences.

Ask the tutor: ask me to take any single-qubit gate you like — a raw $2\times2$ matrix, a pulse, or a product of named gates — and walk it onto the sphere as an explicit axis and angle.

Why does the angle in the exponential carry a factor of one half?

Because the state vector lives in $SU(2)$ while the observable Bloch arrow lives in $SO(3)$ , related by a two-to-one homomorphism rather than an identity. Conjugation — the way a gate acts on the arrow — effectively squares the representation, so the exponent that advances the state vector by $\theta/2$ turns the arrow by the full $\theta$ . Setting $R_{\hat n}(\theta)=\exp(-i\tfrac{\theta}{2}\hat n\cdot\vec\sigma)$ makes the induced $SO(3)$ angle come out to exactly $\theta$ . The half-angle is the fingerprint of the double cover, and forgetting it makes a Pauli $Z$ — a full $\pi$ -turn — look like a half-turn.

Does a 2π rotation really do nothing, or does it change the state?

Both, depending on what you watch. Every Bloch arrow returns exactly to where it started after a $2\pi$ turn, so anything measurable about a single isolated qubit is unchanged. But the state vector is multiplied by $-1$ : $R_{\hat n}(2\pi)=-I$ . This spinor sign is invisible for one qubit, yet becomes measurable in interference: rotate a spin by $2\pi$ in one arm of an interferometer and leave the other alone, and the relative sign shifts the fringes — seen with neutrons. The sphere returns; the ket needs a full $4\pi$ to truly come home.

Is the rotation axis of a gate the same thing as its eigenbasis?

Yes, exactly. The two eigenstates sit at the poles of the rotation axis, because a rotation fixes precisely the points on its axis (up to phase). If the eigenvalues are $e^{i\lambda_\pm}$ , the axis points along the eigenstate with the larger phase and the angle is $\theta=\lambda_+-\lambda_-$ . So "diagonalise the gate" and "find its Bloch axis and angle" are the same computation: diagonalising hands you the eigenvectors (the poles) and eigenvalues (whose phases differ by the angle). A Pauli $Z$ , diagonal in $\{|0\rangle,|1\rangle\}$ , therefore has axis $\hat z$ .

Why is conjugation UVU† the same as rotating V's axis by U?

Because the map from gates to rotations is a group homomorphism. If $V$ rotates by $R_V$ and $U$ by $R_U$ , then $UVU^{\dagger}$ rotates by $R_UR_VR_U^{-1}$ — the change-of-frame formula: the rotation $R_V$ performed about the axis $R_U$ has moved $V$ 's axis to. Conjugating keeps the angle but transports the axis by $U$ 's rotation. This is why $HZH=X$ is simply "Hadamard swings the $z$ -axis onto the $x$ -axis", and every Clifford conjugation of a Pauli is a permutation of coordinate axes, provable by moving one arrow rather than multiplying matrices.

Why does this only work for one qubit — is there a Bloch sphere for two?

The clean "state is a point, gate is a rotation" picture is special to a single qubit, where the pure-state space after dropping global phase is the two-sphere and $SU(2)$ double-covers $SO(3)$ . Two qubits have a six-dimensional pure-state manifold ( $\mathbb{CP}^3$ ) and unitaries in $SU(4)$ , for which no faithful three-dimensional sphere exists — entanglement has nowhere to sit on two separate spheres. You can still draw each qubit's reduced state as a Bloch vector, but entangled states then have short vectors (each qubit looks mixed) and the correlations vanish from view. The Bloch sphere is a single-qubit tool.

How do experimentalists actually apply a rotation R_n(θ) to a qubit?

By a resonant pulse. A drive tuned to the qubit frequency, viewed in the rotating frame, appears as a static effective field about which the qubit precesses. The angle $\theta$ is the pulse area — the time integral of the drive amplitude — so a longer or stronger pulse turns further. The axis $\hat n$ lies in the equatorial plane at an azimuth equal to the drive phase, so a $90°$ phase shift swaps an $R_x$ for an $R_y$ . Rotations about $\hat z$ are usually done virtually, by relabelling the phase of later pulses, which costs no time and is essentially error-free — one reason $R_z$ is a preferred native gate.

Bloch Sphere Representation of Gates

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Composing two quarter-turns into a body-diagonal rotation (fully worked)

2 · Extracting the axis and angle of the $\sqrt{X}$ gate (fully worked)

3 · Why Hadamard swaps the two bases (axis argument)

Physics of the portrait

Going further

Before this

Where this leads

Bloch Sphere Representation of Gates

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Composing two quarter-turns into a body-diagonal rotation (fully worked)

2 · Extracting the axis and angle of the X\sqrt{X}X​ gate (fully worked)

3 · Why Hadamard swaps the two bases (axis argument)

Physics of the portrait

Going further

Before this

Where this leads

2 · Extracting the axis and angle of the $\sqrt{X}$ gate (fully worked)