Entanglement
Correlations stronger than anything classical, and a resource you can spend.
Two qubits can share a state that neither one has on its own. Measure them far apart and their results are correlated in a way no classical plan can reproduce. Einstein called it spooky; today we make it with two gates and spend it like currency.
What you’ll learn
- Tell a product state from an entangled state of two qubits.
- Build a Bell pair with a Hadamard and a CNOT, and predict what measuring it gives.
- Predict how often Alice's and Bob's results agree when they measure a Bell pair at different angles.
- Explain why entanglement beats every classical strategy in the CHSH game, yet cannot send a message.
- Name what entanglement is used for: teleportation, superdense coding and quantum key distribution.
Two qubits, four amplitudes
One qubit has two amplitudes. Two qubits have four classical configurations, , , , , and the quantum state gives each of them its own amplitude:
The left bit is qubit A, the right bit is qubit B, and as always the squared magnitudes add up to one.
If you prepare A and B separately, A as and B as , the joint amplitudes simply multiply: . This rule is the tensor product, written , and states built this way are called product states.
Qubit A, prepared on its own, has . Qubit B, prepared separately, has . What is the probability, in percent, that measuring both gives ?
Show the answer
10 %
For a product state the probabilities multiply too: . The qubits are independent: learning A's result tells you nothing new about B.
A state that will not split
Now consider the most famous state in quantum information:
Measure both qubits and you get or , half the time each, and never or . The two qubits always agree. Can you find a state for A and a state for B whose product is ?
Think first Try to write as . What goes wrong?
You would need , so all four numbers are nonzero. But there is no term, so you also need , which forces or to be zero. Contradiction: no product of single-qubit states gives . It is entangled.
A state that cannot be factored into single-qubit states is called entangled. Gates acting on one qubit at a time can never produce one; you need an interaction between the qubits. In the widget, the CNOT toggle applies a controlled-NOT, which flips B whenever A is .
Build a Bell pair
On a quantum computer you make from with two gates:
- A Hadamard on qubit 0 puts it in superposition: .
- A CNOT with qubit 0 as control flips qubit 1 only in the second branch: .
The Hadamard creates two branches; the CNOT copies each branch's bit onto the second qubit and ties the two qubits together. is one of four Bell states.
You prepare and measure qubit 0. The result is . What will a measurement of qubit 1 give?
- with certainty
- or , 50/50
- with certainty
- It depends on how far apart the qubits are
Show the answer
C. with certainty
Only and have nonzero amplitude. Seeing on qubit 0 leaves , so qubit 1 gives every time, at any distance.
Alice and Bob compare notes
Now send one qubit of each pair to Alice and the other to Bob. Each picks a measurement direction, an angle in the – plane of the Bloch sphere ( is the usual measurement, is ), and records or .
Their correlation is the average of the product of their two results: if they always agree, if they always disagree, if there is no relation at all. For , quantum mechanics predicts
Equal angles give : perfect agreement, whether both use , both use , or any common angle in between. Angles apart give .
Alice measures at and Bob at . What percentage of pairs give the same result?
Show the answer
75 %
. Since and the two probabilities add up to 1, .
Is this just two sealed envelopes, one red and one blue? Not quite. A source that secretly sends or at random also agrees in the basis, but in the basis its pairs agree only half the time. The Bell pair agrees at every common angle. A cleverer classical source could imitate even that; the decisive test, next, compares different angles.
Beating every classical strategy
In 1964 John Bell showed that any theory in which each particle carries pre-set answers ("local hidden variables") must obey a limit that quantum mechanics breaks. A clean modern version is the CHSH game.
A referee sends Alice a random bit and Bob a random bit . Without communicating, they answer with bits and . They win if
that is, their answers must agree, unless both received a , in which case they must differ.
Think first What is the best classical strategy, and how often does it win?
Always answer . Then , which wins whenever : three rounds out of four. No classical strategy, not even one using shared random numbers agreed in advance, wins more than 75% of the time.
With a shared Bell pair, Alice and Bob each let their question choose a measurement angle. Their outcomes are random, but correlated just right: they win of rounds. Since 75% is the classical ceiling, those extra ten points are a measurable proof that nature has no local hidden variables, a phenomenon called nonlocality.
A resource you can spend
Entanglement is more than a curiosity: it is a fuel. One shared Bell pair, called an ebit, lets you:
- Teleport an unknown qubit, using the pair plus two classical bits.
- Superdense-code: send two classical bits by transmitting a single qubit.
- Grow a secret key (QKD): if the pairs still violate a Bell inequality, no eavesdropper can know much about the results.
Each use consumes the pair. And no amount of local operations and classical communication can create entanglement from scratch: the two parties must share some to begin with, or send each other quantum systems.
The road ahead
In the Excited State level, each idea from this lesson gets its full mathematics:
| Lesson | What you will do |
|---|---|
| Composite Systems & Tensor Products | Compute Kronecker products, write two-qubit gate matrices, prove a determinant test for entanglement, and measure one qubit out of several. |
| Bell States & Creating Entanglement | Build all four Bell states and the three-qubit GHZ state, find their symmetries, and learn which gates can entangle. |
| Bell Inequalities & Nonlocality | Derive the classical CHSH bound and reach with a Bell pair. |
| Entanglement as a Resource (LOCC) | Work out what local operations and classical communication can and cannot do with entanglement, including purifying noisy pairs. |
Think first In one sentence: what makes the correlations of a Bell pair "quantum"?
No list of answers fixed in advance can reproduce them (they violate a Bell inequality), and yet they still cannot carry a message from Alice to Bob.
Key takeaways
- Two qubits have four amplitudes. Prepared independently, they form a product state with .
- An entangled state such as cannot be factored: the pair is in a definite state while each qubit alone is completely random.
- A Hadamard followed by a CNOT turns into a Bell pair.
- Bell-pair results agree at every common angle, with correlation ; no local hidden-variable model reproduces this, as the CHSH game shows (75% classical versus about 85% quantum).
- Entanglement cannot send messages, but it is a consumable resource for teleportation, superdense coding and quantum key distribution.
Check yourself
In the interactive lesson these are scored with your confidence, and what you miss comes back for review.
Which of these two-qubit states is entangled?
Show the answer
C.
always gives opposite results and cannot be factored: it is the Bell state , the "singlet". The others factor as , and .
Starting from , which circuit produces ?
- CNOT (control qubit 0), then on qubit 0
- on both qubits
- on qubit 0, then CNOT (control qubit 0)
- on qubit 0, then CNOT (control qubit 0)
Show the answer
D. on qubit 0, then CNOT (control qubit 0)
puts qubit 0 in superposition, , and the CNOT then flips qubit 1 in the branch, giving .
Alice and Bob share many Bell pairs. Can Alice send Bob a message by choosing which angle she measures at?
- No: Bob's results are 50/50 whatever Alice does; the correlation appears only when they compare notes
- Yes, instantly, because her measurement collapses his qubit
- Yes, but only at the speed of light
- Only if they agreed on the angles in advance
Show the answer
A. No: Bob's results are 50/50 whatever Alice does; the correlation appears only when they compare notes
Whatever Alice does, Bob's qubit on its own still behaves like a fair coin. The correlation is visible only after they compare records over an ordinary channel. This is the no-signalling principle.
In the CHSH game the best classical strategy wins 75% of rounds. What is the best possible win rate with shared entanglement?
- 75%: entanglement cannot help
- 100%
- About 85%, that is
- About 95%
Show the answer
C. About 85%, that is
The best quantum strategy wins , and no quantum strategy does better (Tsirelson's bound). The gap above 75% is exactly what Bell experiments measure.
Alice holds one qubit of . Examined on its own, with no information from Bob, her qubit gives:
- A definite result that matches Bob's
- 50/50 results in the basis, but always in the basis
- Results whose odds change when Bob measures
- 50/50 results in every basis
Show the answer
D. 50/50 results in every basis
Each half of a Bell pair is maximally mixed: no measurement on it alone has a predictable result, and its Bloch arrow has length zero, as you saw in the widget. The information lives entirely in the correlations.
Go deeper
- Composite Systems & Tensor Products · Two qubits, four amplitudes: how state spaces multiply.
- Bell States & Creating Entanglement · The four Bell states, and the circuits that make them.
- Bell Inequalities & Nonlocality · The CHSH game: a measurable proof that nature is not classical.
- Entanglement as a Resource (LOCC) · What you can and cannot do with local operations and classical communication.
Further reading
- M. A. Nielsen & I. L. Chuang, Quantum Computation and Quantum Information, §1.3.6 (Bell states) and §2.6 (EPR and the Bell inequality) (Cambridge, 10th anniv. ed., 2010).
- J. S. Bell, "On the Einstein Podolsky Rosen paradox", Physics 1, 195 (1964).