The Pauli gates X, Y, Z flip a qubit by a fixed half-turn — they are switches with only an "on". But on the Bloch sphere a qubit is just a point, and a point can be rotated by any angle you like. The rotation gates R_x(\theta), R_y(\theta), R_z(\theta) are exactly that: the continuous, dial-a-value single-qubit gates. Each one spins the Bloch vector through an angle \theta about one of the three axes, and the Paulis fall out as the special case \theta = \pi — a full half-turn.

One formula, three axes

Every rotation gate is generated by exponentiating the corresponding Pauli. For a unit axis \hat{n} with Pauli vector \vec{\sigma} = (X, Y, Z),

and this rotates the Bloch vector by angle \theta about \hat n. Note the half-angle \theta/2 sitting in the exponent — the same spin-½ bookkeeping that put a \theta/2 in the Bloch-sphere form of a state. Writing out the three axis gates as 2\times2 matrices:

R_y is the friendly one — its entries are real, so it just tilts the state in the plane. R_x carries the imaginary -i off the diagonal, and R_z is diagonal: it only ever changes phases, never the |0\rangle vs. |1\rangle probabilities.

Watch a rotation happen

Here is the x–z great circle of the Bloch sphere. The state starts at |0\rangle (top) and R_y(\theta) swings it through the full angle \theta toward |+\rangle and on to |1\rangle. Drag the dial and watch the amplitudes: at 90^\circ you land on the equator at |+\rangle, and it takes a full 180^\circ on the sphere to reach |1\rangle — even though the amplitude \sin\tfrac{\theta}{2} only ran up to a quarter turn.

Because the Bloch angle is the full \theta while the amplitudes move by \theta/2, one lap of the sphere (\theta = 2\pi) sends the state to -|0\rangle, not back to |0\rangle. More on that surprise below.

Worked example 1: R_x(\pi) is X

Set \theta = \pi. Then \cos\tfrac{\pi}{2} = 0 and \sin\tfrac{\pi}{2} = 1, so

The rotation gate equals the Pauli X up to the overall factor -i. That factor is a global phase — unobservable — so R_x(\pi) and X are the same physical operation. The same holds for every Pauli: R_y(\pi) \propto Y and R_z(\pi) \propto Z. The Paulis are simply the \theta = \pi rotations.

Worked example 2: R_z(\theta) adds a relative phase

Apply R_z(\theta) to a general state:

Pulling out the global phase e^{-i\theta/2} leaves a clean statement: R_z(\theta) inserts the relative phase e^{i\theta} between |0\rangle and |1\rangle — a rotation of the longitude \varphi on the Bloch sphere, leaving the probabilities untouched. Choosing \theta = \tfrac{\pi}{2} gives R_z(\tfrac{\pi}{2}) = e^{-i\pi/4}\,\mathrm{diag}(1, i) = e^{-i\pi/4} S, so the phase gate S is R_z(\tfrac{\pi}{2}) up to global phase.

Worked example 3: R_y(\tfrac{\pi}{2}) makes |+\rangle

With \theta = \tfrac{\pi}{2} the half-angle is 45^\circ, so \cos 45^\circ = \sin 45^\circ = \tfrac{1}{\sqrt2} and

A quarter turn about y drives |0\rangle from the north pole down to |+\rangle on the equator — a real-valued, phase-free way to build an equal superposition (like a "half-Hadamard" with no extra phases).

Rotations are a complete toolkit

Since a single-qubit gate is a unitary and every single-qubit unitary is a rotation of the Bloch sphere, three rotations are all you ever need. The Euler decomposition writes any single-qubit unitary U as

for some angles \alpha, \beta, \gamma and a global phase e^{i\delta}. So R_y and R_z alone generate every one-qubit operation, up to a phase you cannot see.