In short: Quantum entanglement is a non-classical correlation in which two or more qubits share a single joint state, so measuring one instantly fixes the statistics of the other — and it is the resource that gives quantum computing, teleportation, and provably secure communication their power.
What "correlated qubits" really means
Classical bits are independent objects: a byte is just eight switches, and knowing one tells you nothing extra about the rest. Qubits can be entangled, meaning the whole system is described by one joint wavefunction that cannot be factored into independent single-qubit descriptions. For a two-qubit system the general state is a four-amplitude vector over the basis |00⟩, |01⟩, |10⟩, |11⟩. If that vector can be written as a tensor product (|ψ⟩A ⊗ |ψ⟩B), the qubits are separable. If it cannot, they are entangled.
The canonical example is the Bell state |Φ⁺⟩ = (|00⟩ + |11⟩)/√2. Measure qubit A in the computational basis and you get 0 or 1 with 50% probability each — pure noise locally. But the instant A reads 0, B is guaranteed to read 0, and if A reads 1, B reads 1, with 100% correlation, every single shot, no matter how far apart the qubits are. That perfect correlation on top of locally random outcomes is the signature of entanglement.
A single two-qubit state needs 4 complex amplitudes; n entangled qubits need 2n. At 50 qubits that is 250 ≈ 1.1×1015 amplitudes — why classical simulation of entangled systems explodes and why quantum hardware is interesting.
The four Bell states
There are exactly four maximally entangled two-qubit states, the Bell basis. Each is produced by a Hadamard gate followed by a CNOT applied to a different input, and they differ only in a relative sign or a bit flip:
| Bell state | Expression | Prepared from | A⊕B correlation |
|---|---|---|---|
| |Φ⁺⟩ | (|00⟩+|11⟩)/√2 | H·CNOT on |00⟩ | same (even) |
| |Φ⁻⟩ | (|00⟩−|11⟩)/√2 | H·CNOT on |10⟩ | same (even) |
| |Ψ⁺⟩ | (|01⟩+|10⟩)/√2 | H·CNOT on |01⟩ | opposite (odd) |
| |Ψ⁻⟩ | (|01⟩−|10⟩)/√2 | H·CNOT on |11⟩ | opposite (odd) |
Why it is not "faster than light"
A persistent myth is that entanglement lets you signal instantly. It does not. The local outcome at B is pure 50/50 randomness; nothing an experimenter does at A changes the marginal statistics at B. The correlation only becomes visible when the two measurement records are brought together and compared over a classical channel. This is the no-signalling theorem, and it is why quantum teleportation always requires two classical bits to be sent alongside the shared Bell pair.
Bell's inequality (CHSH form) caps any local-hidden-variable theory at S ≤ 2. Quantum mechanics predicts S = 2√2 ≈ 2.83. Loophole-free experiments (Delft, 2015) measured violations well above 2, closing the "spooky action" debate experimentally.
Entanglement in quantum computing
Entanglement is the engine behind quantum speedups. Algorithms such as Shor's factoring and Grover's search build a superposition across an entangled register, let interference amplify the right answer, then measure. Without entanglement a quantum computer is efficiently simulable classically (the Gottesman–Knill theorem shows stabilizer-only circuits are tractable), so entanglement is precisely what separates genuine quantum advantage from clever classical tricks.
In practice, two-qubit entangling gates — CNOT on superconducting transmons, Mølmer–Sørensen gates on trapped ions — are the hardest, slowest, and noisiest operations. On today's hardware a CNOT error rate near 0.5–1% dwarfs single-qubit errors around 0.05%, which is why surface-code error correction spends most of its overhead protecting entangling operations. If you want to understand how these gates are composed, see our companion piece on quantum gates, and the broader hardware picture in quantum decoherence.
Entanglement in quantum communication
Shared entanglement underpins the most practical quantum technologies shipping today. Quantum key distribution protocols such as E91 use Bell pairs so that any eavesdropper necessarily disturbs the correlations — security rests on physics, not on assumed computational hardness. Quantum teleportation transfers an unknown qubit state using one Bell pair plus two classical bits, and entanglement swapping chains pairs together to build quantum repeaters, the key to long-distance quantum networks because photons are lost exponentially in fiber (roughly 0.2 dB/km).
Designing the control electronics, cryogenic readout, and photonic links for these systems is squarely a mixed-signal VLSI problem — the same discipline our partners cover in hands-on VLSI and semiconductor training. The full CourseTron quantum track is on the courses page.
Ready to go from Bell states to running Shor's algorithm on real hardware? Enrol in CourseTron's Quantum Computing Foundations course and build entangling circuits in a browser-based simulator.
Frequently asked questions
Is entanglement the same as superposition?
No. Superposition is a property of a single qubit being in a blend of |0⟩ and |1⟩. Entanglement is a property of the joint state of two or more qubits that cannot be described independently. You need superposition to create entanglement, but a single qubit can be in superposition without being entangled with anything.
Can entanglement send information faster than light?
No. Each party sees purely random local outcomes, and the correlation only appears when results are compared over a classical (light-speed-limited) channel. The no-signalling theorem forbids superluminal communication.
How many Bell states are there?
Exactly four for two qubits: |Φ⁺⟩, |Φ⁻⟩, |Ψ⁺⟩, |Ψ⁻⟩. They form an orthonormal basis of the two-qubit space and differ only by a bit flip and/or a relative phase.
Why does entanglement matter for a quantum speedup?
Stabilizer and product-state circuits are efficiently simulable classically, so entanglement is the resource that lets a quantum computer explore a 2n-dimensional state space and use interference to extract answers no classical machine can reach efficiently.