Pauli Operators

The Pauli operators are the three most important $2\times2$ matrices in quantum information, and they carry a double life almost no other operators share. Wolfgang Pauli wrote them down in 1927 to describe the spin- $\tfrac12$ degree of freedom of the electron; a lifetime later they reappeared, unchanged, as the canonical observables and the canonical gates of a single qubit. That coincidence is the whole story. Each of $X$ , $Y$ , $Z$ is Hermitian, so it is a legitimate observable with real eigenvalues $\pm1$ ; each is also unitary, so it is a legitimate quantum gate. A generic operator is one or the other; the Paulis are both, because each is a self-inverse Hermitian involution, $\sigma^2=I=\sigma^\dagger\sigma$ . Three threads run out of that fact. First, structure: with the identity $I$ , the three matrices span the real four-dimensional space of Hermitian $2\times2$ matrices, so every single-qubit observable, Hamiltonian, and density operator is a real combination of $\{I,X,Y,Z\}$ — this is why a qubit carries exactly three measurable numbers and why the Bloch sphere exists. Second, dynamics: the algebra $X^2=Y^2=Z^2=I$ with pairwise anticommutation and $XY=iZ$ is the rescaled algebra of $\mathfrak{su}(2)$ , so exponentiating the Paulis generates the rotation group $SU(2)$ and the gates $R_x,R_y,R_z$ . Third, robustness: tensor products of Paulis form the Pauli group whose abelian subgroups are the stabilizer codes, and because any single-qubit error decomposes into $I,X,Y,Z$ , correcting four discrete cases tames all continuous noise. Structure, rotation, uncertainty, and error correction all flow from the same short list of relations — re-deriving them is the task below.

Key takeaways

The three Pauli matrices $X$ , $Y$ , $Z$ are the traceless, Hermitian, unitary $2\times2$ matrices; with $I$ they form a complete basis for every single-qubit operator.
Each is simultaneously an observable (Hermitian, eigenvalues $\pm1$ ) and a gate (unitary), because each squares to the identity; distinct Paulis anticommute.
Their exponentials $e^{-i\theta\,\hat n\cdot\vec\sigma/2}$ generate all Bloch-sphere rotations, making them the Lie-algebra generators of $SU(2)$ — a $2\pi$ turn returns $-I$ , not $I$ .
Any single-qubit state is $\rho=\tfrac12(I+\vec r\cdot\vec\sigma)$ with the Bloch vector built from the Pauli expectation values $r_j=\langle\sigma_j\rangle$ .
Their anticommutation is the algebraic root of the spin uncertainty relations, $\Delta\sigma_x\,\Delta\sigma_y\ge|\langle\sigma_z\rangle|$ .
Key idea: the Pauli operators are at once a Hermitian basis for single-qubit observables and the unitary generators of single-qubit rotations, and their anticommutation algebra encodes both the geometry of the Bloch sphere and the uncertainty relations.

The core insight

The Pauli operators are simultaneously a Hermitian basis for single-qubit observables and the unitary generators of single-qubit rotations; their anticommutation algebra $\sigma_j\sigma_k=\delta_{jk}I+i\epsilon_{jkl}\sigma_l$ encodes both the geometry of the Bloch sphere and the uncertainty relations.

Where this lives today

The Pauli operators are the working coordinate system of an entire industry. On today's superconducting and trapped-ion processors, single-qubit logic is delivered as $R_x$ and $R_y$ microwave pulses while $R_z$ is often "virtual" — a bookkeeping phase shift on later pulses that costs zero time and zero error, precisely because $R_z(\theta)=\operatorname{diag}(e^{-i\theta/2},e^{+i\theta/2})$ only relabels the rotating frame. Every calibration routine speaks Pauli: randomized benchmarking interleaves random Clifford gates to extract an average gate error, and Pauli twirling conjugates a noisy channel by random Paulis to turn coherent error into a stochastic Pauli channel. Read-out is Pauli too — classical-shadow tomography reconstructs an unknown $\rho$ from random single-qubit Pauli measurements, the $\langle X\rangle,\langle Y\rangle,\langle Z\rangle$ of equation (6) scaled up. And fault tolerance is Pauli all the way down: the surface code that IBM and Google are racing to scale is a stabilizer code whose checks are tensor products of $X$ 's and $Z$ 's, whose syndromes discretize continuous noise into the four Pauli errors, and whose logical qubits are defined by which Pauli strings commute with the stabilizers. The open frontier — pushing the physical error rate below threshold and decoding syndromes fast enough in real time — is stated and solved entirely in the algebra Pauli wrote down in 1927.

The physical picture

Picture the state of a single qubit as a unit arrow pointing somewhere on a sphere. The three Pauli operators are not points on that sphere but its three coordinate axes, and each poses a yes/no question with answers $+1$ and $-1$ . $Z$ asks "are you $|0\rangle$ or $|1\rangle$ ?" and reads the poles; $X$ asks the same in the equatorial basis $|+\rangle,|-\rangle$ , and $Y$ in the circular basis $|+i\rangle,|-i\rangle$ . Because the three axes are mutually perpendicular and maximally tilted — any eigenstate of one overlaps any eigenstate of another with probability exactly $\tfrac12$ — a state that answers one question sharply is completely undecided about the other two. That single geometric fact is the uncertainty principle for spin, drawn as a picture.

Figure

Three Pauli operators are the three mutually unbiased axes of the Bloch sphere: Z reads the 0 and 1 poles with eigenvalues +1 and −1, X reads the plus/minus pair, Y reads the +i/−i pair — and because any two axes overlap with probability exactly 1/2, a sharp value on one axis leaves the other two maximally uncertain

Seen as gates rather than questions, the same three matrices act on the arrow instead of reading it. $X$ is the bit-flip — the quantum NOT that swaps the poles — $Z$ is the phase-flip, and $Y=iXZ$ does both. Their continuous versions are honest rotations: $R_z(\theta)$ spins the arrow about the vertical axis, $R_x,R_y$ about the horizontal ones, with the Pauli operator the generator inside each exponential. There is a famous twist: because that generator carries a factor $\tfrac12$ , a full $2\pi$ turn multiplies the state by $-1$ , and only a $4\pi$ turn brings it home. This spinor sign is the visible signature that $SU(2)$ double-covers $SO(3)$ , and it is worth seeing as a curve with real numbers on it.

Figure

The spinor sign: the coefficient cos(θ/2) inside R_z(θ) = cos(θ/2) I − i sin(θ/2) Z runs +1 → 0 → −1 → 0 → +1 as θ sweeps 0 → π → 2π → 3π → 4π, so a full 2π rotation gives R_z(2π) = −I and only a 4π rotation restores the state — the hallmark of SU(2) double-covering SO(3)

The relation that packages the entire algebra in one line is $\sigma_j\sigma_k=\delta_{jk}I+i\epsilon_{jkl}\sigma_l$ . Its symmetric part $\delta_{jk}I$ says each Pauli squares to the identity; its antisymmetric part $i\epsilon_{jkl}\sigma_l$ is the three-dimensional cross product in matrix disguise. Quantum spin and vector geometry share one skeleton, which is why "axes" and "rotations" are an exact isomorphism, not a loose analogy. And because $\{I,X,Y,Z\}$ is a complete operator basis, any error or Hamiltonian on one qubit is a weighted mix of these four — the reason continuous decoherence collapses, after a syndrome measurement, into a discrete choice among bit-flips, phase-flips, and their product.

The core equations

Everything begins with the explicit matrices. Fixing the ordering $|0\rangle=(1,0)^T$ , $|1\rangle=(0,1)^T$ for rows and columns turns each abstract operator into four numbers you can multiply by hand.

X=\begin{pmatrix}0&1\\1&0\end{pmatrix},\quad Y=\begin{pmatrix}0&-i\\i&0\end{pmatrix},\quad Z=\begin{pmatrix}1&0\\0&-1\end{pmatrix},\quad I=\begin{pmatrix}1&0\\0&1\end{pmatrix} \qquad (1)

where: $X,Y,Z$ — the three Pauli operators, also written $\sigma_x,\sigma_y,\sigma_z$ or $\sigma_1,\sigma_2,\sigma_3$ ; $I$ — the $2\times2$ identity, sometimes $\sigma_0$ ; $|0\rangle,|1\rangle$ — computational basis states ordering rows and columns. Read off directly: all three are Hermitian and traceless, and each squared returns $I$ , so each is unitary too. Applying the matrices to the basis kets shows their names in action and why $Y$ is "both flips at once."

X|0\rangle=|1\rangle,\ X|1\rangle=|0\rangle,\quad Z|0\rangle=|0\rangle,\ Z|1\rangle=-|1\rangle,\quad Y|0\rangle=i|1\rangle,\ Y|1\rangle=-i|0\rangle \qquad (2)

where: $X$ — bit-flip (quantum NOT), swaps $|0\rangle\leftrightarrow|1\rangle$ ; $Z$ — phase-flip, fixes $|0\rangle$ and sends $|1\rangle\to-|1\rangle$ ; $Y$ — combined bit- and phase-flip, $Y=iXZ$ . The reading: $Z$ is diagonal because the poles are its eigenstates, while $X,Y$ are off-diagonal, moving probability between the poles. The single identity below contains the entire multiplication table — squares, commutators, and anticommutators at once. It is the most useful equation on the page.

\sigma_j\sigma_k=\delta_{jk}I+i\sum_{l}\epsilon_{jkl}\,\sigma_l \qquad (3)

where: $\delta_{jk}$ — Kronecker delta, $1$ if $j=k$ else $0$ ; $\epsilon_{jkl}$ — Levi-Civita symbol, $+1$ cyclic in $(x,y,z)$ , $-1$ anticyclic, $0$ if repeated; $j,k,l$ — indices over $\{x,y,z\}$ . The reading: $j=k$ makes the right side $I$ (involution); $j\ne k$ keeps only the $\epsilon$ term, giving $XY=iZ$ , $YZ=iX$ , $ZX=iY$ cyclically. Because the four Paulis are orthogonal under the trace inner product, any $2\times2$ matrix expands on them, with coefficients pulled out by a single trace.

M=\tfrac12\big(c_0 I+c_x X+c_y Y+c_z Z\big),\qquad c_\mu=\operatorname{Tr}(\sigma_\mu M) \qquad (4)

where: $M$ — an arbitrary $2\times2$ complex matrix; $\sigma_\mu$ — the four Paulis with $\sigma_0=I,\sigma_1=X,\sigma_2=Y,\sigma_3=Z$ ; $c_\mu$ — coefficients, real exactly when $M$ is Hermitian. The reading: $\{I,X,Y,Z\}$ is a coordinate system for operators, the engine behind state tomography and the discretization of errors. Exponentiating any Pauli — or any unit combination of them — produces a Bloch-sphere rotation, the bridge from static observables to dynamics and gates.

R_{\hat n}(\theta)=e^{-i\theta\,\hat n\cdot\vec\sigma/2}=\cos\tfrac\theta2\,I-i\sin\tfrac\theta2\,(\hat n\cdot\vec\sigma) \qquad (5)

where: $\hat n=(n_x,n_y,n_z)$ — a unit vector naming the rotation axis; $\vec\sigma=(X,Y,Z)$ — the vector of Paulis; $\theta$ — the rotation angle, turning the Bloch vector by $\theta$ about $\hat n$ . The reading: the closed cosine/sine form (Derivation B) exists only because $(\hat n\cdot\vec\sigma)^2=I$ , and the $\tfrac12$ in the exponent makes a $2\pi$ turn give $-I$ . Every single-qubit state, pure or mixed, is the identity plus a Pauli expansion whose coefficients are its measurable spin components.

\rho=\tfrac12\big(I+\vec r\cdot\vec\sigma\big),\qquad r_j=\operatorname{Tr}(\rho\,\sigma_j)=\langle\sigma_j\rangle \qquad (6)

where: $\rho$ — a single-qubit density operator (Hermitian, unit trace, positive semidefinite); $\vec r$ — the Bloch vector with $|\vec r|\le1$ , equal to $1$ for pure and $<1$ for mixed states; $\langle\sigma_j\rangle$ — the expectation of the $j$ -th Pauli. The reading: measuring $\langle X\rangle,\langle Y\rangle,\langle Z\rangle$ fully reconstructs $\rho$ , and $\operatorname{Tr}\rho^2=\tfrac12(1+|\vec r|^2)$ turns them into a purity. Finally, because each Pauli squares to the identity, its variance is fixed by its mean, and anticommutation forces a Robertson bound between components.

(\Delta\sigma_j)^2=1-\langle\sigma_j\rangle^2,\qquad \Delta\sigma_x\,\Delta\sigma_y\ge|\langle\sigma_z\rangle| \qquad (7)

where: $\Delta\sigma_j$ — the standard deviation $\sqrt{\langle\sigma_j^2\rangle-\langle\sigma_j\rangle^2}$ ; $\langle\sigma_z\rangle$ — the conjugate-component mean, lower-bounding the other two uncertainties. The reading: a state sharp in $Z$ ( $\langle\sigma_z\rangle=\pm1$ ) is maximally fuzzy in $X$ and $Y$ ( $\Delta\sigma_x=\Delta\sigma_y=1$ ) — the first figure made quantitative.

Derivation AThe master product identity and the algebra of rotations

The goal is to prove that the single line $\sigma_j\sigma_k=\delta_{jk}I+i\epsilon_{jkl}\sigma_l$ really contains the full algebra of the Paulis, and that its two halves are the anticommutator (fixing that each Pauli is an involution) and the commutator (reproducing the cross-product structure of three-dimensional rotations, so that $\{iX/2,iY/2,iZ/2\}$ generate $\mathfrak{su}(2)$ ). We build the identity from the explicit matrices, split it by symmetry, and read the physics off each part.

Step 1. Multiply the matrices of equation (1) directly. Equal indices give $\sigma_j^2=I$ in every case; distinct indices give $XY=\operatorname{diag}(i,-i)=iZ$ and cyclically $YZ=iX$ , $ZX=iY$ , with reversed order flipping the sign, $YX=-iZ$ . Both cases collapse into $\sigma_j\sigma_k=\delta_{jk}I+i\epsilon_{jkl}\sigma_l$ . Why: a single closed-form product rule lets every later manipulation be done symbolically, without returning to the matrix entries.

Step 2. Symmetrize in $j\leftrightarrow k$ . The delta term is symmetric and the epsilon term is antisymmetric, so the symmetric (anticommutator) part keeps only the delta: $\tfrac12\{\sigma_j,\sigma_k\}=\delta_{jk}I$ , i.e. $\{\sigma_j,\sigma_k\}=2\delta_{jk}I$ . Distinct Paulis therefore anticommute, $\{X,Y\}=0$ , and each squares to $I$ . Why: anticommutation is the precise statement that the three axes are mutually incompatible observables, and Step 3 turns it into the uncertainty relation.

Step 3. Antisymmetrize instead; now the delta cancels: $\tfrac12[\sigma_j,\sigma_k]=i\epsilon_{jkl}\sigma_l$ , so $[\sigma_j,\sigma_k]=2i\epsilon_{jkl}\sigma_l$ and $[X,Y]=2iZ$ . With $J_l=\tfrac12\sigma_l$ this reads $[J_j,J_k]=i\epsilon_{jkl}J_l$ — the $\mathfrak{su}(2)$ angular-momentum algebra ( $\hbar=1$ ). Why: the commutator is the geometric cross product rewritten for operators, which is why exponentiating the Paulis produces genuine rotations. *As a check: $XY=\operatorname{diag}(i,-i)=iZ$ and $YX=-iZ$ , so indeed $\{X,Y\}=0$ and $[X,Y]=2iZ$ , in agreement with $\delta_{xy}=0$ and $\epsilon_{xyz}=1$ .*

Go deeper · Derivation B · 1 minExponentiating a Pauli operator into a Bloch-sphere rotation

Derivation BExponentiating a Pauli operator into a Bloch-sphere rotation

The aim is to show that the matrix exponential $e^{-i\theta\,\hat n\cdot\vec\sigma/2}$ collapses to the closed form $\cos\tfrac\theta2\,I-i\sin\tfrac\theta2\,(\hat n\cdot\vec\sigma)$ , and to trace where each ingredient of a rotation comes from — why the series closes into a cosine and a sine, why the half-angle appears, and why a $2\pi$ rotation returns $-I$ . The one fact that makes it work is that a unit Pauli vector squares to the identity, so its powers cycle with period two.

Step 1. Let $A=\hat n\cdot\vec\sigma$ with $\hat n$ a unit vector, and compute its square using equation (3): $A^2=\sum_{jk}n_jn_k\sigma_j\sigma_k=\sum_{jk}n_jn_k(\delta_{jk}I+i\epsilon_{jkl}\sigma_l)$ . The epsilon term is antisymmetric in $j,k$ while $n_jn_k$ is symmetric, so it vanishes, leaving $A^2=|\hat n|^2 I=I$ . Why: this involution property is the engine of the derivation — $A^{2k}=I$ and $A^{2k+1}=A$ , so only $I$ and $A$ ever appear however many terms you keep.

Step 2. Expand the exponential as a power series, $e^{-i\theta A/2}=\sum_m \tfrac{1}{m!}(-i\theta/2)^m A^m$ , and sort by parity of $m$ . Even powers contribute $A^{2k}=I$ and odd powers $A^{2k+1}=A$ , so the sum factors into $\big(\sum_k \tfrac{(-1)^k(\theta/2)^{2k}}{(2k)!}\big)I - i\big(\sum_k \tfrac{(-1)^k(\theta/2)^{2k+1}}{(2k+1)!}\big)A$ . Why: the parity split is exactly the split between the cosine and sine series, reducing an operator problem to two scalar series in the half-angle $\theta/2$ .

Step 3. Recognize the two bracketed sums as $\cos(\theta/2)$ and $\sin(\theta/2)$ , giving $R_{\hat n}(\theta)=\cos\tfrac\theta2\,I-i\sin\tfrac\theta2\,A$ . Setting $\hat n=\hat z$ recovers $R_z(\theta)=\operatorname{diag}(e^{-i\theta/2},e^{+i\theta/2})$ . Why: the half-angle inside the exponential becomes a full angle $\theta$ on the Bloch sphere because the sphere is built from bilinears in the state, so $SU(2)$ turns twice as slowly as its $SO(3)$ image. *As a check: at $\theta=2\pi$ the formula gives $R_{\hat n}=-I$ (the spinor sign), and at $\theta=4\pi$ it returns $+I$ — confirming that $SU(2)$ double-covers $SO(3)$ , with $R_z(2\pi)=-I$ as the explicit numerical instance.*

Go deeper · Derivation C · 1 minVariance of a Pauli observable and the spin uncertainty relation

Derivation CVariance of a Pauli observable and the spin uncertainty relation

The purpose here is to turn anticommutation into a quantitative uncertainty relation, and to prove the striking fact that a Pauli observable has no free variance at all: once you know its mean, its spread is fixed, because $\sigma_j^2=I$ . We compute the second moment, then the variance, then feed the commutator into the Robertson bound to get the trade-off between conjugate spin components, and check it saturates on a pole.

Step 1. For any normalized state the second moment is trivial: $\langle\sigma_j^2\rangle=\langle I\rangle=1$ . Hence $(\Delta\sigma_j)^2=1-\langle\sigma_j\rangle^2$ , between $0$ (an eigenstate, $\langle\sigma_j\rangle=\pm1$ ) and $1$ (equatorial, $\langle\sigma_j\rangle=0$ ). Why: a two-outcome $\pm1$ observable is a Bernoulli variable, so its variance is fixed by its mean — no independent second parameter to measure.

Step 2. Apply Robertson, $\Delta A\,\Delta B\ge\tfrac12|\langle[A,B]\rangle|$ , with $A=\sigma_x$ , $B=\sigma_y$ . From Derivation A, $[\sigma_x,\sigma_y]=2i\sigma_z$ , so $\tfrac12|\langle[A,B]\rangle|=|\langle\sigma_z\rangle|$ , giving $\Delta\sigma_x\,\Delta\sigma_y\ge|\langle\sigma_z\rangle|$ . Why: the same commutator that generated rotations now bounds simultaneous knowledge — rotation and uncertainty are two readings of one antisymmetric structure.

Step 3. Test the bound on the north pole $|0\rangle$ . There $\langle\sigma_z\rangle=1$ , so the right side is $1$ ; and $\langle\sigma_x\rangle=\langle\sigma_y\rangle=0$ gives $\Delta\sigma_x=\Delta\sigma_y=1$ from Step 1, so the left side is also $1$ . The bound is saturated: a state sharp in $Z$ is maximally uncertain in $X$ and $Y$ , and their product equals the $Z$ mean. Why: saturation on the poles is the quantitative face of the mutual unbiasedness in the first figure — certainty about one axis costs all certainty about the perpendicular ones. *As a check: the pure-state identity $\langle\sigma_x\rangle^2+\langle\sigma_y\rangle^2+\langle\sigma_z\rangle^2=1$ is consistent with $(\Delta\sigma_j)^2=1-\langle\sigma_j\rangle^2$ , and the uncertainty bound saturates at $\langle\sigma_z\rangle=1$ on $|0\rangle$ .*

Limiting cases

The structural identities below are the ones worth committing to memory; every later result about qubits leans on one of them.

Case	Result	Meaning
Square to identity	$X^2=Y^2=Z^2=I$	each Pauli is an involution, so its eigenvalues are confined to $\pm1$ and it is unitary as well as Hermitian
Anticommutation	$\{\sigma_j,\sigma_k\}=2\delta_{jk}I$	distinct Paulis anticommute; this is the algebraic source of the spin uncertainty relations
Commutator	$[\sigma_j,\sigma_k]=2i\,\epsilon_{jkl}\,\sigma_l$	the cyclic $\mathfrak{su}(2)$ structure; rescaling by $\tfrac12$ gives the angular-momentum generators
Traceless and orthogonal	$\operatorname{Tr}\sigma_j=0,\ \operatorname{Tr}(\sigma_j\sigma_k)=2\delta_{jk}$	Hilbert–Schmidt orthogonality, the reason equation (4) can extract each coefficient by a trace
Determinant and identity	$\det\sigma_j=-1,\ \operatorname{Tr}I=2$	eigenvalues $\{+1,-1\}$ multiply to $-1$ ; the identity, with trace $2$ , completes the operator basis

Common trapA full 2π rotation returns the qubit to itself, and Z is the same thing as the phase gate S

Both halves are false, and both come from dropping the factor of $\tfrac12$ in the generator. First, $R_z(\theta)=e^{-i\theta Z/2}$ , not $e^{-i\theta Z}$ : the gate carries a half-angle, so $R_z(2\pi)=\cos\pi\,I-i\sin\pi\,Z=-I$ . That $-1$ is harmless as a global phase on one qubit, but decisively observable in interference or on the controlled branch of a two-qubit gate; only $R_z(4\pi)=I$ truly restores the ket. Second, do not identify the Pauli $Z$ (eigenvalues $\pm1$ , its own inverse) with the phase gate $S=\sqrt{Z}=\operatorname{diag}(1,i)$ or with $T=\sqrt{S}$ : $Z$ is a Clifford involution, while $T$ in particular is the non-Clifford resource that lifts a stabilizer circuit to universality.

Worked examples

1 · Diagonalizing a qubit Hamiltonian by Pauli decomposition (fully worked)

A single spin feels the Hamiltonian $H=\begin{pmatrix}1&2\\2&-1\end{pmatrix}$ (energy units). Expand it on the Pauli basis, then read off its energy levels and eigenstates without solving a characteristic polynomial by hand.

Step 1. Extract coefficients with $c_\mu=\operatorname{Tr}(\sigma_\mu H)$ : $c_0=\operatorname{Tr}H=0$ , $c_x=\operatorname{Tr}(XH)=4$ , $c_y=\operatorname{Tr}(YH)=0$ , $c_z=\operatorname{Tr}(ZH)=1-(-1)=2$ . By equation (4), $H=\tfrac12(4X+2Z)=2X+Z$ — a Pauli field $H=\vec c\cdot\vec\sigma$ with $\vec c=(2,0,1)$ .

Step 2. Any $\vec c\cdot\vec\sigma$ squares to $|\vec c|^2 I$ (the involution argument of Derivation B), so its eigenvalues are $\pm|\vec c|$ . Here $|\vec c|=\sqrt{4+0+1}=\sqrt5$ , giving $E_\pm=\pm\sqrt5\approx\pm 2.236$ and a gap $2\sqrt5\approx4.47$ . The energy levels are $E_\pm=\pm\sqrt5\approx\pm2.236$ .

Step 3. The eigenstates point along $\pm\hat n$ with $\hat n=\vec c/|\vec c|=(2,0,1)/\sqrt5$ ; the ground state sits at $-\hat n$ , and its energy $-\sqrt5$ matches direct diagonalization of $H$ , confirming the decomposition. A traceless qubit Hamiltonian is just a vector $\vec c$ : spectrum $\pm|\vec c|$ , eigenstates $\pm\hat n$ on the Bloch sphere — no polynomial required.

2 · A rotation gate and its measurement statistics (fully worked)

Start the qubit in $|0\rangle$ and apply $R_y(\theta)$ with $\theta=\pi/3$ . Find the output state, the probability of reading $1$ in the computational basis, and the resulting $\langle Z\rangle$ .

Step 1. Use equation (5) with $A=Y$ : $R_y(\theta)=\cos\tfrac\theta2\,I-i\sin\tfrac\theta2\,Y$ . Acting on $|0\rangle$ with $Y|0\rangle=i|1\rangle$ , the two $i$ 's cancel: $R_y(\theta)|0\rangle=\cos\tfrac\theta2\,|0\rangle+\sin\tfrac\theta2\,|1\rangle$ — the qubit swept up from the north pole by $\theta$ in the $x$ – $z$ plane.

Step 2. At $\theta=\pi/3$ , $\theta/2=\pi/6$ : $\cos\tfrac\pi6\approx0.866$ , $\sin\tfrac\pi6=\tfrac12$ , so $|\psi\rangle=0.866\,|0\rangle+0.5\,|1\rangle$ . The probability of outcome $1$ is $\sin^2\tfrac\theta2=(0.5)^2=\mathbf{0.25}$ . A $60^\circ$ rotation tips a quarter of the probability onto $|1\rangle$ .

Step 3. The $z$ -component is $\langle Z\rangle=\cos^2\tfrac\theta2-\sin^2\tfrac\theta2=\cos\theta=\cos\tfrac\pi3=\mathbf{0.5}$ — the Bloch vector is rotated by $\theta$ off the pole, so its projection on $\hat z$ is $\cos\theta$ : rotation angle and measurement statistic are the same number. Prob $(1)=\sin^2(\theta/2)=0.25$ and $\langle Z\rangle=\cos\theta=0.5$ .

Physics of the portrait

The Pauli operators are one of those rare objects that were invented for one purpose and turned out to be the exact language of another discovered decades later. Pauli built them to force the electron's two-valuedness into quantum mechanics; the quantum information era found them waiting, already the perfect basis for the qubit.

Origins · 1 minWolfgang Pauli and how this was discovered

Wolfgang Pauli

1900–1958 · the spin matrices that became quantum information's canonical operators

Wolfgang Pauli was a physicist of ferocious standards — the author, at twenty-one, of an encyclopedia article on relativity that Einstein praised, and later the terror of sloppy seminars, whose "not even wrong" became the field's sharpest insult. In 1925 he proposed that the electron carries a two-valued quantum number with "no classical description," and stated the exclusion principle that bears his name. When Goudsmit and Uhlenbeck identified that two-valuedness as spin, Pauli in 1927 supplied the mathematics that made it rigorous: the nonrelativistic theory of spin built on three $2\times2$ matrices $\sigma_x,\sigma_y,\sigma_z$ acting on a two-component wavefunction, encoding the whole of a spin- $\tfrac12$ system in the algebra $\sigma_j\sigma_k=\delta_{jk}I+i\epsilon_{jkl}\sigma_l$ . A year later Dirac's relativistic equation absorbed them into the gamma matrices, and for half a century they were simply "the spin matrices." Their second life began when the qubit was recognized as an abstract spin- $\tfrac12$ : the same three matrices are now the elementary gates of quantum computing, the generators of single-qubit rotations, and, tensored together, the Pauli group that underlies every stabilizer code. Pauli's exclusion principle also explains the periodic table and the stability of matter, winning him the 1945 Nobel Prize; he is remembered too for the "Pauli effect," the legend that apparatus broke merely when he entered the room.

History

1922

— Stern and Gerlach split a silver-atom beam, revealing spatial quantization of angular momentum

1925

— Pauli proposes a two-valued electron degree of freedom and states the exclusion principle

1927

— Pauli introduces the spin matrices σx, σy, σz in his nonrelativistic theory of electron spin

1928

— Dirac's relativistic electron equation absorbs the Pauli matrices into the gamma matrices

1997

— Gottesman recasts quantum error correction as the stabilizer formalism built on the Pauli group

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, §2.1.3 (Pauli matrices) and §4.2 (single-qubit gates and rotations) · Preskill, Lecture Notes for Physics 219, Ch. 2–4 · Watrous, The Theory of Quantum Information, Ch. 1–2 · Wilde, Quantum Information Theory, Ch. 3 · Kitaev, Shen & Vyalyi, Classical and Quantum Computation, §8 · Mermin, Quantum Computer Science, Ch. 1–2 · Gottesman, Stabilizer Codes and Quantum Error Correction (PhD thesis, 1997) for the Pauli-group formalism. The key numerical anchors throughout this article are: each Pauli has eigenvalues $\pm1$ and determinant $-1$ ; $XY=\operatorname{diag}(i,-i)=iZ$ , giving $\{X,Y\}=0$ and $[X,Y]=2iZ$ ; $(\hat n\cdot\vec\sigma)^2=I$ drives the $\cos/\sin$ resummation with $R_z(2\pi)=-I$ ; $H=\begin{pmatrix}1&2\\2&-1\end{pmatrix}=2X+Z$ has spectrum $\pm\sqrt5\approx\pm2.236$ ; $R_y(\pi/3)|0\rangle$ yields prob $(1)=\sin^2(\pi/6)=0.25$ and $\langle Z\rangle=\cos(\pi/3)=0.5$ ; and $r_j=\operatorname{Tr}(\rho\sigma_j)$ . The uncertainty bound follows from Robertson applied to $[\sigma_x,\sigma_y]=2i\sigma_z$ .

Prerequisites. The Qubit and Its State Space · Dirac Bra–Ket Notation · Observables and Hermitian Operators · Inner Product, Norm, and Normalization. Leads to. Density Operators · Projective Measurements · Expectation Values and Variance · Global and Relative Phase.

Ask the tutor: ask me to rotate any state on the Bloch sphere, decompose a matrix into Paulis, or show how a stabilizer syndrome turns continuous noise into a discrete Pauli error.

Why exactly are X, Y, Z the "right" three operators and not some other trio?

Because they are an orthogonal basis for the traceless Hermitian $2\times2$ matrices — exactly the space of non-trivial single-qubit observables. Any Hermitian qubit operator is $\tfrac12(c_0I+\vec c\cdot\vec\sigma)$ with real coefficients, so once the identity is fixed, three more operators span the rest, and $X,Y,Z$ are the simplest orthogonal choice (orthogonal under $\operatorname{Tr}(\sigma_j\sigma_k)=2\delta_{jk}$ ). You could rotate to any other $SU(2)$ -image triple, but never use fewer than three — the same three real numbers a Bloch vector carries. The trio is canonical, not unique.

If Z and R_z(θ) are different, when do I use which?

Use the Pauli $Z$ when you mean a measurement or a discrete phase-flip: eigenvalues $\pm1$ , its own inverse, the observable "computational-basis parity." Use $R_z(\theta)=e^{-i\theta Z/2}$ when you mean a continuous gate that rotates the Bloch vector by $\theta$ about the $z$ -axis. They coincide only up to a phase at $\theta=\pi$ : $R_z(\pi)=-iZ$ , differing by the global factor $-i$ . In circuit design the distinction is everything — the generator $Z$ sets the axis, the angle $\theta$ inside $R_z$ sets how far you turn, and dropping the $\tfrac12$ doubles every rotation.

Why does a qubit need a 4π rotation to return to itself when a classical object needs only 2π?

Because the qubit lives in the double cover $SU(2)$ of the rotation group $SO(3)$ . Physical directions in space repeat every $2\pi$ , but the state is a two-component spinor whose amplitudes are bilinearly related to those directions, so it turns half as fast: $R_{\hat n}(2\pi)=-I$ , and only $R_{\hat n}(4\pi)=+I$ . That sign is not an artifact — it shows up in neutron-interferometry experiments, where sending one arm through a $2\pi$ magnetic rotation flips the interference fringes relative to the other. The same factor of $\tfrac12$ that puts the sign there is the one in the generator $e^{-i\theta\sigma/2}$ .

How do the Pauli operators connect to the Bloch sphere so precisely?

Through equation (6): every single-qubit density operator is $\rho=\tfrac12(I+\vec r\cdot\vec\sigma)$ , and the three Bloch-vector components are literally the Pauli expectation values, $r_j=\langle\sigma_j\rangle$ . So the sphere is not a metaphor bolted on afterward — it is the coefficient space of the Pauli expansion, with $I$ fixing the trace to $1$ and $X,Y,Z$ supplying the coordinates. Purity $\operatorname{Tr}\rho^2=\tfrac12(1+|\vec r|^2)$ is the radius; positivity forces $|\vec r|\le1$ , so states live inside a ball and pure states on its surface. Measuring the three Paulis on many copies is exactly how you find where the arrow points.

What does "any single-qubit error is a combination of I, X, Y, Z" really buy us?

It converts an impossible problem — protecting against a continuum of errors — into a finite one. A generic noise operator $E$ expands by equation (4) as $E=c_0I+c_xX+c_yY+c_zZ$ . When a stabilizer code measures its syndrome, that measurement projects the corrupted state onto an eigenspace of the Pauli checks, collapsing the continuous superposition of errors into a single discrete Pauli: no error, a bit-flip $X$ , a phase-flip $Z$ , or both $Y$ . You only ever correct those four cases, and doing so corrects all errors in their span, including coherent rotations and amplitude damping. This "discretization of errors" is the theorem that makes fault-tolerant quantum computing conceivable at all.

Are the Pauli operators the same thing as the Pauli group used in error correction?

They are its single-qubit seed. The $n$ -qubit Pauli group is all tensor products $P_1\otimes\cdots\otimes P_n$ with each $P_i\in\{I,X,Y,Z\}$ , times the phases $\{\pm1,\pm i\}$ needed to close it under multiplication; on one qubit it is just $\{\pm1,\pm i\}\times\{I,X,Y,Z\}$ , sixteen elements. Its importance in coding is that any two elements either commute or anticommute — never anything in between — which lets you pick a large abelian subgroup, the stabilizer, whose simultaneous $+1$ eigenspace is the protected code space. So the single-qubit Paulis here, tensored up, are the entire alphabet of stabilizer quantum error correction.