The Qubit and Its State Space

The qubit is the atom of quantum information, and its entire theory is a story about one small vector space. A classical bit is a label that reads either $0$ or $1$ ; a qubit replaces that label with a unit vector in a two-dimensional complex Hilbert space $\mathcal{H}=\mathbb{C}^2$ , spanned by two orthonormal reference states $|0\rangle$ and $|1\rangle$ called the computational basis. Any pure state is a normalized superposition $|\psi\rangle=\alpha|0\rangle+\beta|1\rangle$ with complex amplitudes obeying $|\alpha|^2+|\beta|^2=1$ . That single complex-linear combination is where all of the departures from classical bits are hidden — superposition, phase, interference, and the measurement rules that turn a continuum of states into one classical bit of readout. Two facts organize everything that follows. First, the amplitudes are complex, so they carry a magnitude and a phase; the magnitudes fix the measurement probabilities through the Born rule, while the relative phase between $\alpha$ and $\beta$ is invisible in the computational basis but fully physical in a rotated basis, where it drives interference. Second, an overall global phase multiplying the whole state changes no prediction whatever, so physical states are rays, not vectors. Stripping the one normalization constraint and the one global-phase redundancy from the four real numbers in $(\alpha,\beta)$ leaves exactly two real parameters — a polar angle $\theta$ and an azimuthal angle $\phi$ — which map bijectively onto the surface of a sphere, the Bloch sphere, the picture that makes the whole subject visual. The same abstract $\mathbb{C}^2$ describes a spin- $\tfrac12$ nucleus, a two-level atom or trapped ion, a photon's polarization, and a superconducting transmon; the hardware is a labeling convention, which is exactly why a hardware-independent theory of quantum information exists. This article treats pure single-qubit states only — mixtures and open systems need the density operator.

Key takeaways

A qubit is a unit vector $|\psi\rangle=\alpha|0\rangle+\beta|1\rangle$ in the two-dimensional complex Hilbert space $\mathbb{C}^2$ , with $|\alpha|^2+|\beta|^2=1$ .
The amplitudes $\alpha,\beta$ are complex, so they store magnitude and phase — far more structure than a classical bit's single yes/no.
After fixing normalization and discarding the unobservable global phase, exactly two real parameters $(\theta,\phi)$ remain, and they are the coordinates of a point on the Bloch sphere.
Measurement in the computational basis returns $0$ with probability $|\alpha|^2$ and $1$ with probability $|\beta|^2$ (Born rule), then collapses the state to the matching basis vector.
Relative phase is physical and shows up as interference in a rotated basis; global phase is not — states are rays, not vectors.
Key idea: a single qubit is a unit vector in $\mathbb{C}^2$ whose two complex amplitudes carry continuous magnitude and phase, yet any one measurement collapses it and yields at most one classical bit.

The core insight

A single qubit is a unit vector in $\mathbb{C}^2$ ; its two complex amplitudes carry continuous magnitude and phase information, yet measurement collapses it and yields at most one classical bit.

Where this lives today

The single qubit is now an engineered object with published error bars. Superconducting transmons at Google and IBM, trapped ytterbium and barium ions at Quantinuum and IonQ, and neutral rubidium atoms in optical tweezers at QuEra all realize the same $\mathbb{C}^2$ ; on the best of them single-qubit gate fidelities exceed 99.9%, meaning an intended rotation of the Bloch vector lands within a thousandth of its target. The way experimentalists verify they hold a given state is a direct application of this article: single-qubit state tomography measures $\langle X\rangle,\langle Y\rangle,\langle Z\rangle$ on many identically prepared copies and reconstructs the Bloch vector, because no single measurement can reveal a continuous amplitude — the one-bit readout limit is a law, not a technological gap. That same limit is a resource in quantum key distribution: BB84 encodes bits alternately in the $\{|0\rangle,|1\rangle\}$ and $\{|+\rangle,|-\rangle\}$ bases, so an eavesdropper who measures in the wrong basis both randomizes her result and disturbs the state, betraying herself. The open frontier is not the single qubit but keeping it coherent: every real device is an open system whose Bloch vector shrinks toward the center as it decoheres — exactly why the pure-state picture here is the launch point for density operators, channels, and error correction.

The physical picture

Think of the classical bit as a switch with two detents. The qubit throws the switch away and hands you an arrow that may point anywhere on the surface of a globe. The north pole is $|0\rangle$ , the south pole is $|1\rangle$ , and every other direction is a genuine superposition with its own probability split and its own phase. How far the arrow leans toward a pole fixes the odds of that outcome; the compass bearing around the equator is the relative phase. Measurement is brutal: it snaps the arrow onto the axis you measure along, north or south with probabilities set by the lean, and whatever bearing it had is gone. The first figure pins a concrete state onto that globe.

Figure

A qubit as a point on the Bloch sphere at polar angle θ = 120° and azimuth φ = 60°: the north pole is outcome 0, the south pole outcome 1, and the measurement probabilities are cos²(θ/2) = 0.25 and sin²(θ/2) = 0.75

The two amplitudes hold four real numbers, but nature refuses to let you read two of them. Total probability must equal one, tying the magnitudes together and spending one number; an overall phase $e^{i\gamma}$ multiplying the whole state leaves every prediction untouched, spending another. What survives is the polar angle $\theta$ , controlling the $|0\rangle$ -versus- $|1\rangle$ split, and the azimuthal angle $\phi$ — the relative phase — dark in a computational-basis measurement but bursting into view the instant you measure along a different axis. That last point is the whole game, so the second figure isolates it. Amplitudes are not probabilities; they are (loosely) square roots of probabilities, each dressed with a phase. Two states can carry identical $|\alpha|^2$ and $|\beta|^2$ — identical statistics in the computational basis — and still be perfectly distinguishable, because their relative phase decides whether amplitudes add or cancel once you rotate the qubit before measuring. That steering of constructive against destructive addition is interference, the resource that separates quantum information from a classical coin.

The core equations

A pure qubit state is any complex-linear combination of the two computational basis states. This is the superposition principle applied to the smallest quantum system, and the definition on which everything else is built.

|\psi\rangle = \alpha\,|0\rangle + \beta\,|1\rangle, \qquad \alpha,\beta\in\mathbb{C} \qquad (1)

where: $|\psi\rangle$ — the qubit's state vector in $\mathbb{C}^2$ ; $|0\rangle,|1\rangle$ — the computational (or $Z$ ) basis, a fixed orthonormal reference; $\alpha,\beta$ — complex probability amplitudes for outcomes $0$ and $1$ . Reading it: a qubit is a complex-coefficient blend of "definitely 0" and "definitely 1." Choosing coordinates makes the algebra concrete. The two basis kets are the standard column vectors of $\mathbb{C}^2$ , orthonormal — unit length and mutually perpendicular under the inner product.

|0\rangle=\begin{pmatrix}1\\0\end{pmatrix},\quad |1\rangle=\begin{pmatrix}0\\1\end{pmatrix},\qquad \langle i|j\rangle=\delta_{ij} \qquad (2)

where: the columns represent $|0\rangle,|1\rangle$ ; $\langle i|j\rangle$ — the inner product of two basis kets; $\delta_{ij}$ — the Kronecker delta, $1$ if $i=j$ and $0$ otherwise. Reading it: $\langle0|0\rangle=\langle1|1\rangle=1$ and $\langle0|1\rangle=0$ , a perpendicular unit frame. Not every vector in $\mathbb{C}^2$ is a legal state — only those of unit length. Normalization is not a cosmetic convention; it is the demand that the total probability of all measurement outcomes equals one.

\langle\psi|\psi\rangle=|\alpha|^2+|\beta|^2=1 \qquad (3)

where: $\langle\psi|=\alpha^*\langle0|+\beta^*\langle1|$ — the bra (conjugate-transpose) dual to $|\psi\rangle$ ; $|\alpha|^2,|\beta|^2$ — the measurement probabilities for $0,1$ . Reading it: the squared amplitudes form a genuine probability distribution, so states live on the unit sphere of $\mathbb{C}^2$ . Once normalization and the unobservable global phase are removed, a pure state is fixed by just two real angles — the Bloch-sphere parametrization, the coordinate form of the picture above.

|\psi\rangle=\cos\!\tfrac{\theta}{2}\,|0\rangle+e^{i\phi}\sin\!\tfrac{\theta}{2}\,|1\rangle,\qquad 0\le\theta\le\pi,\ 0\le\phi<2\pi \qquad (4)

where: $\theta$ — polar angle from the north pole, setting the split via $p(0)=\cos^2(\theta/2)$ ; $\phi$ — azimuthal angle, the relative phase between the amplitudes. Reading it: $\theta$ tilts the arrow toward a pole, $\phi$ spins it around the equator. Measurement connects the amplitudes to observed frequencies. The Born rule says the probability of an outcome is the squared modulus of the amplitude along that basis vector, and the act of measuring collapses the state onto the outcome seen.

p(k)=|\langle k|\psi\rangle|^2,\qquad |\psi\rangle\ \xrightarrow{\ \text{outcome }k\ }\ |k\rangle \qquad (5)

where: $p(k)$ — probability of outcome $k\in\{0,1\}$ ; $\langle k|\psi\rangle$ — the amplitude found by projecting $|\psi\rangle$ onto $|k\rangle$ ; the arrow denotes irreversible collapse to the matching basis state. Reading it: in the computational basis $p(0)=|\alpha|^2$ , $p(1)=|\beta|^2$ , and after readout the pre-measurement amplitudes are gone. The computational basis is only one choice of measurement axis. Two other standard bases — the eigenstates of the Pauli $X$ and $Y$ operators — are the tool for exposing relative phase.

|\pm\rangle=\tfrac{1}{\sqrt2}\big(|0\rangle\pm|1\rangle\big),\qquad |\pm i\rangle=\tfrac{1}{\sqrt2}\big(|0\rangle\pm i|1\rangle\big) \qquad (6)

where: $|+\rangle,|-\rangle$ — the $X$ basis (along $\pm x$ ); $|+i\rangle,|-i\rangle$ — the $Y$ basis (along $\pm y$ ). Reading it: these are the poles of the other two Bloch axes, and measuring along them reads out phase the $Z$ basis cannot see.

Derivation ANormalization from the inner product

The goal is to show that the constraint $|\alpha|^2+|\beta|^2=1$ is forced, not assumed: it is exactly the statement that the state vector has unit length, which is in turn exactly the statement that the two measurement probabilities sum to one. We compute the inner product of a general state with itself and use orthonormality to kill the cross terms, reading off what "length one" demands of the amplitudes.

Step 3. Apply orthonormality, $\langle0|0\rangle=\langle1|1\rangle=1$ and $\langle0|1\rangle=\langle1|0\rangle=0$ . The two mixed terms vanish and only $\langle\psi|\psi\rangle=|\alpha|^2+|\beta|^2$ survives. Why: a perpendicular basis is what makes "length" decompose into an independent sum over components — Pythagoras in $\mathbb{C}^2$ .

Go deeper · Derivation B · 1 minFrom four real numbers to two angles on a sphere

Derivation BFrom four real numbers to two angles on a sphere

The goal is to explain why a pure qubit, which starts life as two complex numbers — four real parameters — is physically described by only two, and why those two are naturally the angles of a sphere. We count parameters, subtract the two that carry no physical information (one fixed by normalization, one by the irrelevance of global phase), and show the survivors assemble into the standard Bloch form, so the "arrow on a globe" picture is derived, not asserted.

Step 1. Write both amplitudes in polar form, $\alpha=r_0e^{i\gamma_0}$ and $\beta=r_1e^{i\gamma_1}$ . This is four real numbers, $r_0,r_1,\gamma_0,\gamma_1$ , the raw parameter budget of an unconstrained vector in $\mathbb{C}^2$ . Why: separating each amplitude into magnitude and phase lets us attack the two redundancies one at a time.

Step 2. Impose normalization $r_0^2+r_1^2=1$ . One real equation removes one magnitude, and the pair can be written with a single angle: $r_0=\cos(\theta/2)$ , $r_1=\sin(\theta/2)$ with $0\le\theta\le\pi$ . Why: any point on a unit circle in the $(r_0,r_1)$ quadrant is captured by one angle, so the two magnitudes cost only one parameter.

Step 3. Factor out the overall phase: $|\psi\rangle=e^{i\gamma_0}\big(\cos(\theta/2)|0\rangle+e^{i(\gamma_1-\gamma_0)}\sin(\theta/2)|1\rangle\big)$ . Only the difference $\phi=\gamma_1-\gamma_0$ sits between the two amplitudes. Why: a common phase multiplies both terms identically, so it cannot change any relative structure — it factors cleanly outside the bracket.

Step 4. Discard the global phase, $e^{i\gamma_0}\sim1$ , because it changes no probability in any basis (shown next). What remains is $|\psi\rangle=\cos(\theta/2)|0\rangle+e^{i\phi}\sin(\theta/2)|1\rangle$ , two real parameters $(\theta,\phi)$ — a point on the unit two-sphere, the Bloch sphere; the count is $4-1(\text{norm})-1(\text{phase})=2$ . Why: two angles on a sphere is the honest dimension of the physical state space, and it is why single-qubit dynamics is literally rotation of an arrow.

Go deeper · Derivation C · 1 minRelative phase is observable, global phase is not

Derivation CRelative phase is observable, global phase is not

The goal is to prove the single fact that most cleanly separates a qubit from a classical coin: the relative phase between amplitudes is a real, measurable degree of freedom, while a global phase multiplying the whole state is pure gauge and carries no information. We compare $|+\rangle$ and $|-\rangle$ — identical in the computational basis, differing only by a sign on $|1\rangle$ — in the $Z$ basis and then the $X$ basis, and contrast that with an overall phase factor.

Step 2. Measure in the $X$ basis instead. Compute the overlaps directly: $\langle+|+\rangle=1$ , while $\langle+|-\rangle=\tfrac12(\langle0|+\langle1|)(|0\rangle-|1\rangle)=\tfrac12(1-1)=0$ . So $|+\rangle$ gives outcome $+$ with certainty and $|-\rangle$ gives outcome $-$ with certainty — perfect discrimination. Why: rotating the measurement axis converts the hidden relative phase into a magnitude, where the Born rule can finally see it.

Step 3. Conclude that the sign — a relative phase of $\pi$ — is a genuine physical parameter: $|0\rangle+|1\rangle$ and $|0\rangle-|1\rangle$ are different states because they yield different, distinguishable statistics in a real experiment. Why: any parameter that changes a measurable distribution is by definition physical, and this one does.

Step 4. Contrast the global phase. For any $\gamma$ and any basis state $|k\rangle$ , $p(k)=|\langle k|(e^{i\gamma}|\psi\rangle)|^2=|e^{i\gamma}|^2|\langle k|\psi\rangle|^2=|\langle k|\psi\rangle|^2$ , since $|e^{i\gamma}|=1$ . No probability in any basis moves, so $e^{i\gamma}|\psi\rangle\sim|\psi\rangle$ and physical states are rays. Why: an overall phase multiplies every amplitude equally and cancels in every squared modulus, so unlike a relative phase it can never be detected.

Limiting cases

Special states and parameter limits fix the geometry:

Case	Result	Meaning
Classical-bit limit	$\alpha=1,\ \beta=0\ \Rightarrow\ \|\psi\rangle=\|0\rangle$	A basis state is a definite classical bit; measurement gives $0$ with certainty. The poles of the Bloch sphere are the two classical configurations.
Equal superposition	$\|\psi\rangle=\tfrac{1}{\sqrt2}(\|0\rangle+\|1\rangle)=\|+\rangle$	Sits on the equator; the $Z$ -basis outcome is a fair coin ( $\tfrac12,\tfrac12$ ), the point of maximal computational-basis uncertainty, yet it is $+$ with certainty in the $X$ basis.
Relative-phase sweep	$\|\psi\rangle=\tfrac{1}{\sqrt2}(\|0\rangle+e^{i\phi}\|1\rangle)$	Walks around the equator. All such states share the same $Z$ statistics but are physically distinct; $\phi$ is read out in the $X$ / $Y$ bases.
Global-phase change	$e^{i\gamma}\|\psi\rangle\sim\|\psi\rangle$	Multiplying by a unit-modulus scalar changes no probability in any basis; the global phase is gauge, so states are rays, not vectors.
Antipodal pair	$\langle\psi\|\psi_\perp\rangle=0$	Orthogonal states sit at opposite poles and are the only states one measurement can tell apart with certainty.

Common trapA superposition is not a secret coin flip

A qubit in $\tfrac{1}{\sqrt2}(|0\rangle+|1\rangle)$ is not "really 0 or 1, we just don't know which." That description is a classical probabilistic mixture — a coin already flipped, outcome hidden — a different physical object, described by a density operator, not a state vector. The tell is interference: the true superposition is $+$ with probability $1$ in the $X$ basis, whereas the hidden-coin mixture gives $+$ with probability only $\tfrac12$ . The amplitudes $\alpha,\beta$ are complex probability amplitudes whose relative phase is a physical degree of freedom that a mixture has thrown away. Collapsing "superposition" into "ignorance" erases exactly the phase that makes quantum computing possible.

Worked examples

1 · A real-amplitude qubit, read three ways (confidence builder)

Take $|\psi\rangle=0.6\,|0\rangle+0.8\,|1\rangle$ . Confirm it is a legal state, find its computational-basis probabilities, locate it on the Bloch sphere, and find its $X$ -basis statistics.

Step 1. Check normalization: $|\alpha|^2+|\beta|^2=0.6^2+0.8^2=0.36+0.64=\mathbf{1}$ . It is a unit vector, so a valid pure state (the 3–4–5 triangle in disguise).

Step 2. Born rule in the $Z$ basis: $p(0)=|0.6|^2=\mathbf{0.36}$ and $p(1)=|0.8|^2=\mathbf{0.64}$ — outcome $1$ almost two times out of three.

Step 3. Bloch coordinates. With real positive amplitudes $\phi=0$ . From $\cos(\theta/2)=0.6$ , $\theta=2\arccos(0.6)=\mathbf{106.3^\circ}$ — just below the equator, consistent with $p(1)>p(0)$ . The Bloch vector is $(\sin\theta\cos\phi,\sin\theta\sin\phi,\cos\theta)=(2\cdot0.6\cdot0.8,\ 0,\ 0.36-0.64)=\mathbf{(0.96,\ 0,\ -0.28)}$ , a unit vector as required.

Step 4. Rotate the measurement axis to the $X$ basis: $\langle+|\psi\rangle=\tfrac{1}{\sqrt2}(0.6+0.8)$ , so $p(+)=\tfrac12|0.6+0.8|^2=\tfrac12(1.96)=\mathbf{0.98}$ and $p(-)=\tfrac12(0.04)=\mathbf{0.02}$ . Although the $Z$ measurement is a near coin flip, the same state is almost deterministic in the $X$ basis — the amplitudes are nearly in-phase, so they add. The one state gives $(0.36, 0.64)$ in $Z$ but $(0.98, 0.02)$ in $X$ ; the basis you choose is part of the measurement.

2 · A complex phase changes the answer

Take $|\psi\rangle=\tfrac12|0\rangle+\tfrac{\sqrt3}{2}e^{i\pi/3}|1\rangle$ . Find the $Z$ -basis probabilities, the Bloch angles, and the $X$ -basis probabilities, and compare with Example 1's lesson about phase.

Step 1. Normalization: $|\tfrac12|^2+|\tfrac{\sqrt3}{2}|^2=\tfrac14+\tfrac34=\mathbf{1}$ ; the phase factor $e^{i\pi/3}$ has modulus one and does not affect the norm.

Step 2. $Z$ -basis Born rule: $p(0)=|\tfrac12|^2=\mathbf{0.25}$ , $p(1)=|\tfrac{\sqrt3}{2}|^2=\mathbf{0.75}$ . Reading magnitudes off equation (4), $\cos(\theta/2)=\tfrac12$ gives $\theta=\mathbf{120^\circ}$ with phase $\phi=\pi/3=\mathbf{60^\circ}$ — precisely the state drawn in the first figure.

Step 3. $X$ -basis amplitude: $\langle+|\psi\rangle=\tfrac{1}{\sqrt2}\big(\tfrac12+\tfrac{\sqrt3}{2}e^{i\pi/3}\big)$ . Write $e^{i\pi/3}=\tfrac12+i\tfrac{\sqrt3}{2}$ , so $\tfrac{\sqrt3}{2}e^{i\pi/3}=\tfrac{\sqrt3}{4}+i\tfrac34$ and the bracket is $\tfrac{2+\sqrt3}{4}+i\tfrac34$ . Why the phase now matters: unlike Example 1 the amplitudes are not collinear in the complex plane, so they add as vectors, not scalars.

Step 4. Square the modulus: $p(+)=\tfrac12\Big[\big(\tfrac{2+\sqrt3}{4}\big)^2+\big(\tfrac34\big)^2\Big]=\tfrac{16+4\sqrt3}{32}=\dfrac{4+\sqrt3}{8}\approx\mathbf{0.716}$ , hence $p(-)\approx\mathbf{0.284}$ . Had the phase been $\phi=0$ the same magnitudes would give $p(+)=\tfrac12|\tfrac12+\tfrac{\sqrt3}{2}|^2\approx0.933$ ; the $60^\circ$ relative phase has swung the $X$ -basis odds from $0.93$ to $0.72$ without touching the $Z$ -basis odds at all. Identical $Z$ statistics $(0.25, 0.75)$ , but the relative phase alone moves $p(+)$ — phase is physics.

Physics of the portrait

The qubit as a named object is astonishingly recent. The physics — spin- $\tfrac12$ , two-level atoms, the Bloch vector — was in place by the 1940s, but the recognition that such a system is a unit of information on the same footing as Shannon's bit is a product of the 1990s.

Origins · 1 minBenjamin Schumacher and how this was discovered

Benjamin Schumacher

b. 1957 · quantum information theory, quantum data compression

Benjamin Schumacher is the physicist who coined the word "qubit" and, in the same breath, gave it its information-theoretic meaning. In his 1995 paper "Quantum coding" (Physical Review A 51, 2738) he proved the quantum analogue of Shannon's noiseless coding theorem: a source of quantum states can be compressed to a number of qubits set by the von Neumann entropy, no fewer, establishing the qubit as the irreducible currency of quantum information exactly as the bit is for classical information. Schumacher has told the story that the term arose half-jokingly in 1992 conversations with William Wootters, and that he expected it to be forgotten; instead it became the defining noun of an entire field. The deeper content of his work is that a qubit is not merely a small quantum system but a resource that can be counted: it can encode more than a classical bit's worth of correlation yet still yields only one bit on measurement, and the accounting of that gap is the theory his coding theorem opened. That accounting rests entirely on the structure this article develops — the $\mathbb{C}^2$ state space, the amplitudes, the phase, and the collapse that limits readout to one classical bit — unifying the Hilbert-space viewpoint of Dirac and von Neumann with the spin picture of Pauli and Bloch under one idea: that information, not any apparatus, is the thing a qubit carries.

History

1922

— Stern and Gerlach split an atomic beam in two, exhibiting a physical two-level degree of freedom

1927

— Pauli writes the spin matrices, the operator algebra of a single two-level system

1946

— Bloch describes nuclear-spin dynamics as a vector precessing on a sphere

1982

— Feynman argues that simulating quantum systems demands quantum machines

1995

— Schumacher publishes "Quantum coding," coining the word qubit and proving quantum data compression

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, §1.2, §2.2 (the qubit and its state space) · Preskill, Quantum Information Lecture Notes, Ch. 2 · Mermin, Quantum Computer Science, Ch. 1 · Watrous, The Theory of Quantum Information, Ch. 2 · Wilde, Quantum Information Theory, Ch. 3 · Kitaev, Shen & Vyalyi, Classical and Quantum Computation, §8. Key results to keep in mind: the normalization identity $\langle\psi|\psi\rangle=|\alpha|^2+|\beta|^2=1$ and the degree-of-freedom count $4-1-1=2$ yielding Bloch form $\cos(\theta/2)|0\rangle+e^{i\phi}\sin(\theta/2)|1\rangle$ ; the $X$ -basis distinguishability $\langle+|-\rangle=0$ ; Example 1 checks $0.6^2+0.8^2=1$ , Bloch vector $(0.96,0,-0.28)$ , $p(+)=0.98$ ; Example 2 gives $p(0)=0.25$ , $p(1)=0.75$ , $p(+)=(4+\sqrt3)/8\approx0.716$ . Schumacher's coining of "qubit" and the quantum coding theorem appear in Phys. Rev. A 51, 2738 (1995).

Prerequisites. Dirac Bra-Ket Notation · Inner Product, Norm, and Normalization · Superposition Principle. Leads to. Global and Relative Phase · Born Rule · Pauli Operators · Density Operators.

Ask the tutor: Ask me to walk any state from its amplitudes to its Bloch vector and its probabilities in the $Z$ , $X$ , and $Y$ bases, and I will show you exactly where the phase hides and where it shows up.

Why is the qubit a two-dimensional complex space and not just two real numbers?

The two levels $|0\rangle$ and $|1\rangle$ give the dimension two, but the amplitudes must be complex, and the imaginary part is not decoration. A complex amplitude carries a phase, and relative phase between the amplitudes is measurable — it decides the outcome of interference experiments in rotated bases. If amplitudes were real you would lose the $Y$ basis and could not describe a general rotation of the Bloch vector. The complex structure is what upgrades a probability (one real number per outcome) to an amplitude (a magnitude and a phase).

What is the difference between global phase and relative phase, concretely?

Relative phase is the phase between the two amplitudes — the $e^{i\phi}$ in $\cos(\theta/2)|0\rangle+e^{i\phi}\sin(\theta/2)|1\rangle$ — and it is physical: sweeping $\phi$ walks the state around the equator of the Bloch sphere and changes measurable statistics in the $X$ and $Y$ bases. Global phase is a single factor $e^{i\gamma}$ multiplying the entire state at once; it cancels in every squared modulus $|\langle k|\psi\rangle|^2$ , so no experiment can detect it. The clean way to remember it: relative phase moves you to a different point on the sphere, global phase leaves you on the same point — which is why physical states are rays, not raw vectors.

If a qubit holds a continuum of states, why can I only get one bit out?

Because measurement is not readout of a stored number — it is a projection. The Born rule turns the continuous amplitudes into probabilities for a small set of discrete outcomes, and a computational-basis measurement has just two, so you receive a single bit, then collapse to that outcome and lose the amplitudes you might have read next time. To learn the continuous parameters $(\theta,\phi)$ you need many identically prepared copies and statistics over them (tomography); you cannot extract them from one system. This gap — a continuum in, one bit out — is not an engineering limitation but a structural feature, and it underlies quantum cryptography and the no-cloning theorem.

Does the Bloch sphere generalize to two or more qubits?

Not as a single sphere. One qubit's pure states are the surface of the ordinary two-sphere, which is why the picture is so useful. Two qubits live in $\mathbb{C}^2\otimes\mathbb{C}^2=\mathbb{C}^4$ , whose pure states form a much larger space that cannot be drawn as one sphere — and crucially, most two-qubit states are entangled, so neither qubit has a definite Bloch vector of its own. You can still assign each qubit a point inside its own Bloch ball via the reduced density operator, but for entangled states those points sit strictly inside (they are mixed), and the real correlations live in a joint structure the two separate balls cannot capture. The single-qubit sphere is the exception worth learning first.

Is the choice of |0⟩ and |1⟩ physically meaningful, or just a convention?

Both, in a precise sense. Choosing which two orthogonal states to call $|0\rangle$ and $|1\rangle$ is a labeling convention — you could rotate to any orthonormal pair and rename them, and the abstract theory would not notice. But once you fix your measurement apparatus it singles out a basis physically: a Stern–Gerlach magnet along $z$ measures in one basis, along $x$ in another, and these are genuinely different experiments. So the basis is convention until you commit to how you measure; then it is the axis your detector reads out. This is why one state has different "probabilities" in different bases — the state is fixed, but the question you ask it is not.

How do experimenters actually confirm a qubit is in the state they intended?

By quantum state tomography. Because one measurement yields only a bit and destroys the state, they prepare the same state many times and measure different Pauli observables across the ensemble: $Z$ estimates $\langle Z\rangle=p(0)-p(1)$ , and $X$ , $Y$ estimate $\langle X\rangle,\langle Y\rangle$ . Those three numbers are exactly the components of the Bloch vector, so together they reconstruct the state (a pure state has unit length; a shorter vector reveals decoherence), and devices report the result as a fidelity against the target. The whole procedure is the Born rule and Bloch parametrization of this article, run in reverse.