A qubit is usually introduced with two symbols and a warning. The symbols are |0> and |1>, and the warning is that a qubit is not simply in one of them. That warning is correct, but it does not give you anything to picture. The Bloch sphere does. It turns every possible pure state of a single qubit into a single point on the surface of a sphere, and every single-qubit gate into a rotation of that point. Once you can read the sphere, a large amount of quantum notation becomes geometry you can reason about.
The representation is named after physicist Felix Bloch. It works only for one qubit, and that limitation is important, but within that scope it is exact: it is not a cartoon approximation of the state. It is the state, drawn differently.
From two amplitudes to one point
A pure single-qubit state is written as |ψ> = α|0> + β|1>, where α and β are complex numbers called amplitudes. Measurement in the computational basis returns 0 with probability |α|2 and 1 with probability |β|2. Because those two probabilities must add to one, the amplitudes obey |α|2 + |β|2 = 1. That condition is called normalisation.
Two complex numbers contain four real numbers, which is too many for a point on a sphere. Two reductions fix that. Normalisation removes one degree of freedom. The second reduction is subtler: multiplying the whole state by the same phase factor, a number of the form eiγ, changes no measurement outcome at all. That overall factor is called the global phase, and because it is unobservable we are free to choose a convenient value for it.
After those two reductions, exactly two real numbers remain. They are the two angles of the sphere:
|ψ> = cos(θ/2) |0> + eiφ sin(θ/2) |1>
Here θ, the polar angle, runs from 0 to π, and φ, the azimuthal angle, runs from 0 to 2π. IBM Quantum Learning uses the same form when it derives the Bloch representation from the density matrix, with |0> at Cartesian coordinates (0, 0, 1) and |1> at (0, 0, -1).
Reading the sphere: poles, equator, and phase
The poles are the basis states
When θ = 0, the sine term vanishes and the state is exactly |0>. That is the north pole. When θ = π, the cosine term vanishes and the state is |1>, up to the phase factor. That is the south pole. At either pole φ is meaningless, for the same reason longitude is meaningless at the Earth's poles: there is no second component for a phase to be relative to.
Everywhere else, θ controls the balance between the two outcomes. The probability of measuring 1 is sin2(θ/2). Move halfway down, to θ = π/2, and that probability is exactly one half. Move closer to the south pole and 1 becomes more likely, without the state ever needing to be described as uncertain in a classical sense. The point is definite; only the measurement outcome is probabilistic.
The equator is equal superposition, distinguished by phase
Every point on the equator has θ = π/2, so every equatorial state gives 0 and 1 with equal probability. What distinguishes one equatorial point from another is φ, the relative phase between the |0> and |1> components. Four equatorial points are worth memorising:
- +X, φ = 0: |+> = (|0> + |1>) / √2
- -X, φ = π: |-> = (|0> – |1>) / √2
- +Y, φ = π/2: (|0> + i|1>) / √2, often written |+i>
- -Y, φ = 3π/2: (|0> – i|1>) / √2, often written |-i>
These four states are indistinguishable if you only measure in the computational basis: all of them return 0 or 1 half the time. They are nevertheless different states, and later gates can turn their phase differences into different measurement outcomes. That is the practical meaning of relative phase. It is invisible to one measurement and decisive for what happens next.
This is also the cleanest correction to a common slogan. An equatorial qubit is not usefully described as simply being 0 and 1 at once. Two states can both be equal superpositions and still be opposites of each other, as |+> and |-> are. Antipodal points on the sphere, points exactly opposite each other, are orthogonal states, and the sphere makes that relationship visible at a glance.
Why the formula uses θ/2
New readers often stumble on the half-angle. Why does a state at the equator, a quarter turn from the pole in ordinary geometry, correspond to θ = π/2 in the sphere's own coordinates but appear with θ/2 inside the cosine and sine? The answer is that the sphere's polar angle is θ itself. The half appears because probabilities come from squared amplitudes. Setting the |1> probability to sin2(θ/2) makes that probability vary smoothly from 0 at the north pole to 1 at the south pole as θ goes from 0 to π, which is exactly one trip from pole to pole.
The half-angle has a second consequence worth knowing. Orthogonal states sit at opposite ends of a diameter rather than at right angles, as you might expect from ordinary vectors. |0> and |1> are orthogonal, and they are antipodal on the sphere. That stretching by a factor of two is the price, and the benefit, of fitting a two-dimensional complex state space onto an ordinary sphere.
Gates are rotations
The Bloch picture pays for itself when you add gates. Every operation on a single qubit that preserves normalisation corresponds to a rotation of the Bloch vector, possibly combined with a reflection in some descriptions, and the standard gates have a direct geometric reading:
- X gate: a rotation by π about the X axis. It carries the north pole to the south pole, so it swaps |0> and |1>, the quantum analogue of a classical NOT.
- Z gate: a rotation by π about the Z axis. The poles do not move, but the equator turns halfway around: |+> becomes |->. It changes phase, not probabilities, in the computational basis.
- Y gate: a rotation by π about the Y axis, which also swaps the poles, with phase effects that differ from X.
- Hadamard gate: carries the north pole to the +X point on the equator. Applied to |0>, it creates |+>, the standard way to prepare an equal superposition.
- S and T gates: rotations about the Z axis by π/2 and π/4 respectively. They leave measurement probabilities in the computational basis unchanged while advancing the relative phase.
- RX, RY, and RZ gates: rotations by an angle you choose about the named axis. RZ changes only φ; RY changes θ and therefore changes the measurement probabilities.
In Qiskit you can watch this directly. Building a one-qubit circuit, applying a Hadamard, and plotting the resulting statevector with plot_bloch_multivector from qiskit.visualization shows the vector moved from the north pole to the equator. Adding an S gate after the Hadamard leaves it on the equator but moves it around toward +Y. That pair of experiments, done once, teaches more about phase than several pages of algebra.
from qiskit import QuantumCircuit
from qiskit.quantum_info import Statevector
from qiskit.visualization import plot_bloch_multivector
qc = QuantumCircuit(1); qc.h(0); plot_bloch_multivector(Statevector(qc))
Inside the sphere: mixed states and noise
So far every state has been pure, and every pure state lies exactly on the surface. Real qubits also have mixed states, which are probabilistic mixtures of pure states rather than a single definite superposition. In the density-matrix form used by IBM Quantum Learning, a state can be written using a Bloch vector whose length is at most one. Length one is the surface and means pure. Shorter vectors lie inside the sphere and mean mixed, with the centre representing the maximally mixed state, an equal classical mixture with no phase information left.
That interior gives you a visual language for noise. If a diagram shows a vector inside the sphere, the qubit is no longer in one pure state, and something — ignorance or noise — has been averaged over.
What the Bloch sphere cannot do
The sphere's clarity comes from its limits, and stating them prevents the most common misunderstandings.
- It describes one qubit only. Two qubits live in a four-dimensional complex space. There is no single ordinary sphere that shows an arbitrary two-qubit state, and in particular entanglement cannot be read off two separate Bloch spheres. Plotting one sphere per qubit shows only each qubit's reduced state, which for a maximally entangled pair is the centre of the sphere, a picture that hides the correlations that matter.
- It shows the state before measurement, not the outcome. Measurement in the computational basis projects the state to a pole, with probabilities set by θ. The sphere does not predict which pole a single shot will give.
- Global phase is absent by design. Two states differing only by a global phase are the same point, correctly, because they are physically indistinguishable.
Used within those bounds, the Bloch sphere is the fastest reliable mental model a programmer can build for single-qubit behaviour. Read θ as the balance between 0 and 1, read φ as the relative phase that future gates can expose, and read every gate as a rotation. Most single-qubit circuits become predictable from those three rules alone.
