Qubits Explained: How Quantum Bits Work

From Bits to Qubits

A classical bit is the simplest possible unit of information: a single yes-or-no distinction, represented as 0 or 1. Every piece of digital information, from a text message to a 4K video, is encoded as a sequence of bits. A classical computer processes information by applying logical operations (AND, OR, NOT) to bits, transforming one sequence into another. The power of classical computing comes from doing this extremely fast (billions of operations per second) and at enormous scale (modern processors contain billions of transistors, each implementing a bit).

A qubit generalizes the classical bit by allowing states between 0 and 1. The state of a qubit is written as alpha|0> + beta|1>, where alpha and beta are complex numbers called probability amplitudes. The squared magnitude of alpha gives the probability of measuring the qubit as 0, and the squared magnitude of beta gives the probability of measuring it as 1. These probabilities must sum to 1 (|alpha|^2 + |beta|^2 = 1), but within that constraint, alpha and beta can take any complex values. This means a qubit has a continuum of possible states, not just two.

The critical difference becomes apparent with multiple qubits. Two classical bits have exactly 4 possible states (00, 01, 10, 11), and at any moment the system is in exactly one of them. Two qubits can be in a superposition of all 4 states simultaneously, with a separate amplitude for each: alpha_00|00> + alpha_01|01> + alpha_10|10> + alpha_11|11>. For N qubits, the state is described by 2^N amplitudes. Ten qubits require 1,024 amplitudes. Fifty qubits require over one quadrillion amplitudes. This exponential scaling is what gives quantum computers their potential power, because a quantum operation applied to N qubits effectively acts on all 2^N amplitudes simultaneously.

The Bloch Sphere: Visualizing a Qubit

Because the state of a single qubit is determined by two complex numbers with one constraint (normalization), it has two real degrees of freedom. These can be parameterized as two angles, theta and phi, on a sphere called the Bloch sphere. The north pole of the sphere represents |0>, the south pole represents |1>, and every other point on the surface represents a specific superposition. The equator contains equal superpositions of |0> and |1> differing only in relative phase. The state |+> = (|0> + |1>)/sqrt(2) sits on the equator at phi = 0, while |-> = (|0> - |1>)/sqrt(2) sits at phi = pi.

Quantum gates correspond to rotations of the Bloch sphere. The Hadamard gate rotates the state by 180 degrees around an axis halfway between the X and Z axes, mapping |0> to |+> (from the north pole to the equator). The Pauli-X gate rotates 180 degrees around the X axis, swapping |0> and |1> (flipping north to south pole). The T gate rotates 45 degrees around the Z axis, adjusting the phase without changing the measurement probabilities. Any single-qubit gate can be decomposed into rotations around two non-parallel axes, which is why the Hadamard and T gates together form a universal single-qubit gate set.

The Bloch sphere only works for single qubits. For two or more entangled qubits, the state cannot be decomposed into individual qubit states, and no simple geometric visualization captures the full complexity. The state of N entangled qubits lives in a 2^N-dimensional complex vector space called Hilbert space, where each dimension corresponds to one of the possible measurement outcomes. This mathematical structure is what makes quantum computing both powerful and difficult to simulate classically.

Physical Implementations of Qubits

Any quantum system with two distinguishable states can serve as a qubit, but practical implementations must satisfy stringent requirements. The qubit must maintain its quantum state long enough to complete useful computations (long coherence time). Gate operations must be performed with very low error rates (high fidelity). Qubits must be individually addressable for single-qubit gates and selectively coupleable for two-qubit gates. And the technology must be scalable to thousands or millions of qubits while maintaining quality.

Superconducting qubits are the most widely deployed technology, used by IBM, Google, Rigetti, and others. They are tiny electrical circuits made from superconducting materials (typically aluminum on silicon or sapphire substrates) that contain a Josephson junction, a thin insulating barrier that allows quantum tunneling of Cooper pairs. The circuit behaves like a quantum harmonic oscillator with discrete energy levels, and the two lowest levels serve as |0> and |1>. Gate operations are performed by applying precisely shaped microwave pulses at frequencies of 4 to 6 GHz. Coherence times have improved from nanoseconds in 2000 to hundreds of microseconds in 2026, and two-qubit gate fidelities exceed 99.5%. The main advantage is fabrication compatibility with existing semiconductor manufacturing, enabling scaling to hundreds and eventually thousands of qubits on a single chip.

Trapped ion qubits use individual atoms (typically ytterbium-171 or barium-137) confined in electromagnetic traps and manipulated with laser beams. The qubit states correspond to two energy levels of the ion's electron configuration. Trapped ions offer the longest coherence times of any qubit technology (minutes, compared to microseconds for superconducting qubits) and the highest gate fidelities (above 99.9% for both single and two-qubit gates). They also benefit from perfect uniformity, since every ytterbium-171 ion is identical by the laws of physics, eliminating the fabrication variability that plagues solid-state qubits. The challenge is speed and scalability: gate operations take microseconds rather than nanoseconds, and scaling beyond a few dozen ions in a single trap requires complex architectures with multiple trapping zones connected by ion shuttling.

Neutral atom qubits have surged in capability since 2022. Individual atoms (typically rubidium or cesium) are held in arrays of optical tweezers, focused laser beams that grip atoms at their focal points. The atoms can be arranged in arbitrary 1D, 2D, or 3D geometries, and two-qubit gates are performed by exciting atoms to high-energy Rydberg states where they interact strongly with their neighbors. QuEra and Atom Computing have demonstrated arrays of 200+ qubits with the ability to rearrange atoms dynamically during computation. The reconfigurability of atom arrays enables non-local connectivity patterns that are difficult to achieve in superconducting chips with fixed wiring.

Photonic qubits encode information in properties of individual photons, such as polarization (horizontal vs vertical) or time bin (early vs late arrival). Photons travel at the speed of light and do not interact with their environment, giving them essentially infinite coherence times. Two-qubit gates are the challenge, because photons do not naturally interact with each other. Linear optical quantum computing schemes use interference and measurement to implement effective photon-photon interactions, but they require many ancillary photons and are probabilistic. PsiQuantum and Xanadu are pursuing photonic architectures, with PsiQuantum focusing on building a million-qubit fault-tolerant photonic processor using semiconductor manufacturing.

Qubit Quality Metrics

The usability of a qubit is defined by several measurable quantities. Coherence time (T1 and T2) measures how long the qubit maintains its quantum state before noise destroys the information. T1 (energy relaxation time) measures how long the qubit stays in the excited state before decaying to the ground state. T2 (dephasing time) measures how long the qubit maintains phase coherence, which determines how well it can participate in interference-based algorithms. All quantum operations must complete within the coherence time, so the ratio of coherence time to gate time determines the maximum circuit depth.

Gate fidelity measures the accuracy of quantum operations, expressed as the probability that the gate produces the correct output state. A fidelity of 99.9% means the gate introduces an error roughly once every 1,000 operations. For single-qubit gates, fidelities above 99.99% are routinely achieved across multiple technologies. Two-qubit gate fidelities are lower, typically 99% to 99.9%, because two-qubit operations are more complex and involve interactions between qubits that are harder to control precisely. Since error rates compound multiplicatively through a circuit, the two-qubit gate fidelity is usually the limiting factor for circuit depth.

Measurement fidelity indicates how accurately the qubit state can be read out at the end of a computation. If measurement has 99% fidelity, roughly 1 in 100 measurements returns the wrong result. For superconducting qubits, readout is performed by probing a coupled resonator and has typical fidelities of 99% to 99.9%. For trapped ions, state-dependent fluorescence gives readout fidelities above 99.9%. Since quantum algorithms require many repeated measurements (shots) to determine the probability distribution of outcomes, measurement errors can be partially mitigated through statistical analysis, but high native fidelity is always preferred.

Connectivity describes which pairs of qubits can interact directly through two-qubit gates. Superconducting processors typically have nearest-neighbor connectivity on a 2D lattice, meaning each qubit can directly interact with only 2 to 4 neighbors. Operations on non-adjacent qubits require SWAP gates to move quantum information across the chip, adding overhead and errors. Trapped ion processors have all-to-all connectivity within a single trap zone, meaning any pair of ions can interact directly. Neutral atom arrays offer reconfigurable connectivity by physically moving atoms between operations.