The Pauli Gates
Three tiny 2 \times 2 matrices show up everywhere in quantum computing —
in error correction, in measurement, in the definition of almost every other gate. They are named
after Wolfgang Pauli, and they are the closest thing a qubit has to a "flip." A classical bit has
exactly one non-trivial operation, NOT (turn 0 into
1). A qubit lives on a whole sphere, so it has three independent
ways to be turned inside out — one for each axis. Those three flips are the
Pauli gates X, Y, and
Z.
X = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}, \qquad Y = \begin{bmatrix} 0 & -i \\ i & 0 \end{bmatrix}, \qquad Z = \begin{bmatrix} 1 & 0 \\ 0 & -1 \end{bmatrix}.
Each is a
single-qubit gate,
so each is unitary. But the Paulis are special twice over: each one is also
Hermitian (equal to its own conjugate transpose), and each one
squares to the identity, X^2 = Y^2 = Z^2 = I. That last
fact means every Pauli is its own inverse — apply it twice and you are exactly back where you started.
X is the bit-flip (quantum NOT)
Watch X act on the two basis states. Matrix times column vector:
X|0\rangle = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}\begin{bmatrix} 1 \\ 0 \end{bmatrix} = \begin{bmatrix} 0 \\ 1 \end{bmatrix} = |1\rangle, \qquad X|1\rangle = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}\begin{bmatrix} 0 \\ 1 \end{bmatrix} = \begin{bmatrix} 1 \\ 0 \end{bmatrix} = |0\rangle.
It swaps |0\rangle and |1\rangle — this is
exactly the classical NOT, which is why X is called the
bit-flip. Geometrically it is a 180^\circ rotation of the
Bloch sphere about the x-axis: it turns the north pole into the south pole.
Z is the phase-flip
Z leaves |0\rangle completely alone but attaches
a minus sign to |1\rangle:
Z|0\rangle = \begin{bmatrix} 1 & 0 \\ 0 & -1 \end{bmatrix}\begin{bmatrix} 1 \\ 0 \end{bmatrix} = |0\rangle, \qquad Z|1\rangle = \begin{bmatrix} 1 & 0 \\ 0 & -1 \end{bmatrix}\begin{bmatrix} 0 \\ 1 \end{bmatrix} = \begin{bmatrix} 0 \\ -1 \end{bmatrix} = -|1\rangle.
Nothing about a measurement in the |0\rangle/|1\rangle basis changes —
the probabilities |{\pm}1|^2 = 1 are identical. What changed is the
relative phase between the two components, which is why Z
is the phase-flip. It is a 180^\circ rotation about the
z-axis. And Y does both at once — up to a phase
it is a bit-flip and a phase-flip, Y = iXZ, a
180^\circ rotation about the y-axis.
Reading a circuit
In a circuit diagram a qubit is a horizontal wire read left to right, and each gate
is a labelled box sitting on it. Here a state |\psi\rangle enters, meets an
X box, then a Z box, and comes out as
ZX|\psi\rangle — the gate written first (leftmost on the wire) is
applied first, so in matrix notation it sits on the right.
Worked example: X and Z acting on the plus state
The most revealing test states are |{+}\rangle = \tfrac{1}{\sqrt2}(|0\rangle + |1\rangle)
and |{-}\rangle = \tfrac{1}{\sqrt2}(|0\rangle - |1\rangle). First apply
X to |{+}\rangle:
X|{+}\rangle = \tfrac{1}{\sqrt2}\begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}\begin{bmatrix} 1 \\ 1 \end{bmatrix} = \tfrac{1}{\sqrt2}\begin{bmatrix} 1 \\ 1 \end{bmatrix} = |{+}\rangle.
X swaps the two equal entries and leaves them unchanged, so
|{+}\rangle is a fixed point of the bit-flip. Now the same state
through Z:
Z|{+}\rangle = \tfrac{1}{\sqrt2}\begin{bmatrix} 1 & 0 \\ 0 & -1 \end{bmatrix}\begin{bmatrix} 1 \\ 1 \end{bmatrix} = \tfrac{1}{\sqrt2}\begin{bmatrix} 1 \\ -1 \end{bmatrix} = |{-}\rangle.
The phase-flip turns |{+}\rangle into |{-}\rangle
(and, by the same arithmetic, |{-}\rangle back into
|{+}\rangle). So Z is the bit-flip of the
{+}/{-} world — the mirror image of how X behaves
on |0\rangle/|1\rangle. That symmetry is the whole point of the three axes.
Worked example: every Pauli squares to the identity
Because a Pauli is its own inverse, applying it twice must do nothing. Check
X^2 directly by multiplying the matrix by itself:
X^2 = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix}\begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix} = \begin{bmatrix} 1 & 0 \\ 0 & 1 \end{bmatrix} = I.
The same happens for Y and Z:
Z^2 = I because (-1)^2 = 1, and
Y^2 = I because the two {\pm}i multiply to
-i^2 = 1. You can also see it on the sphere: two half-turns about the same
axis make a full turn, which is the identity. It also matches
X|0\rangle = |1\rangle followed by
X|1\rangle = |0\rangle — round trip, back to start.
- X = \left[\begin{smallmatrix} 0 & 1 \\ 1 & 0 \end{smallmatrix}\right] is the
bit-flip: X|0\rangle = |1\rangle,
X|1\rangle = |0\rangle — the quantum NOT;
- Z = \left[\begin{smallmatrix} 1 & 0 \\ 0 & -1 \end{smallmatrix}\right] is the
phase-flip: Z|0\rangle = |0\rangle,
Z|1\rangle = -|1\rangle, so Z|{+}\rangle = |{-}\rangle;
- Y = \left[\begin{smallmatrix} 0 & -i \\ i & 0 \end{smallmatrix}\right] = iXZ
does both, a 180^\circ turn about the
y-axis;
- each is Hermitian and unitary, and each
squares to I (self-inverse); geometrically each is a
180^\circ rotation of the Bloch sphere about its axis.
Suppose a qubit is |0\rangle and you sneak a Z
onto it. Then you measure in the standard basis. Nothing happened —
Z|0\rangle = |0\rangle, so you still read 0 with
certainty. Do the same to |1\rangle and you read 1:
the minus sign in -|1\rangle is a global phase there, invisible to
any measurement. But feed Z the plus state and it flips it clean over,
|{+}\rangle \to |{-}\rangle — two states that a {+}/{-}
measurement tells apart perfectly. Same gate, same qubit: utterly undetectable in one basis
and a total flip in another. That is why phase errors are the sneaky ones quantum error correction has
to hunt for separately from bit errors.
It is tempting to think of X and Z as "the same
kind of flip." They are not: X is a
180^\circ turn about the x-axis and
Z a turn about the z-axis — perpendicular
motions, and they do not commute (XZ = -ZX). Second trap:
the minus in Z|1\rangle = -|1\rangle is not an ignorable
global phase.
A global phase multiplies the whole state and never matters; but here only the
|1\rangle part is negated, so in a superposition like
|{+}\rangle it becomes a relative phase — and relative
phases are physically real and measurable. Bit-flip versus phase-flip is a genuine difference, never a
book-keeping one.