No-cloning and no-broadcasting theorems
Prove deterministic no-cloning from preserved inner products and distinguish independent copies from correlated broadcasting of commuting states.
On this page
The question for today
What you’ll understand
- Identify the fixed-device and exact deterministic assumptions of no-cloning.
- Demonstrate why a basis-copying gate fails on a superposition.
- Distinguish cloning from broadcasting and state the commuting-family exception.
A classical copying device can duplicate either value of an unknown bit. A quantum device can likewise copy the labels of a chosen orthogonal basis, but it cannot exactly duplicate every unknown superposition. No-cloning is a statement about a single physical operation working across a specified family of possible inputs. No-broadcasting extends the obstruction to mixed states, even when the outputs are allowed to be correlated. Both the task and the allowed exceptions must be stated carefully.
A universal copier would change inner products
Suppose a fixed device receives one unknown pure input and an input-independent blank and environment. Any deterministic quantum channel has an isometric description when its discarded environment is retained. If the two desired output copies are exactly pure, the final global state factors into those copies and a possibly input-dependent environment state. Preservation of the inner product between two alternative inputs then relates their initial overlap to its square times the final environment overlap.
The absolute environment overlap cannot exceed one. For distinct nonorthogonal inputs, their overlap magnitude lies strictly between zero and one, so its square is smaller than itself. The required equality is impossible. Orthogonal states avoid the contradiction because their overlap is zero; identical rays avoid it because the magnitude is one. A fixed basis copier is therefore consistent with quantum mechanics, whereas a deterministic exact copier for a family containing two distinct nonorthogonal pure states is not.
Copying basis labels creates correlations
Controlled-NOT copies the zero–one label into a blank zero: zero-zero stays zero-zero and one-zero becomes one-one. By linearity, a superposition input produces a superposition of these two correlated terms. It does not produce the tensor square of the original superposition, which would also contain cross-label terms with products of amplitudes. Each local output can consequently be mixed even though the total output is pure.
The theorem does not prevent repeatedly preparing a state from a known classical recipe. It also does not forbid approximate copies with a defined fidelity, or certain heralded probabilistic copying tasks on restricted families. A heralded success branch changes the deterministic assumptions and must include its success probability. Quantum error correction encodes information into correlations and reverses a specified class of errors; it does not leave several independent perfect copies of an arbitrary logical state available for separate inspection.
Broadcasting weakens the output requirement
Cloning a mixed state requires the product output rho tensor rho. Broadcasting asks only that both output marginals equal rho, allowing correlations between them. A single fixed channel can broadcast a finite-dimensional family of density operators exactly if and only if the operators commute pairwise. This is a theorem, not a conclusion obtained merely by applying the pure-state inner-product proof to arbitrary mixed matrices.
For the commuting direction, choose a common eigenbasis and copy that classical basis label. The output is a mixture of matched labels, with both marginals equal to the input. Its correlations generally prevent it from being a tensor product of independent copies. Noncommuting families cannot be broadcast by one input-independent deterministic channel, even under this weaker requirement. A state-specific preparation device is outside that unknown-family task, so saying that a known fixed mixed state can be prepared twice does not challenge the theorem.
Work it through
A copier for basis labels fails on plus
Apply controlled-NOT to . Compare the joint output with the desired two-copy state, and calculate either output's squared fidelity with plus.
Expand the input as . Controlled-NOT changes only the second term, giving , a Bell vector.
The desired product has all four computational amplitudes equal to one half. The actual vector has no zero-one or one-zero amplitude, so it is a different joint state.
Tracing either Bell factor removes the off-diagonal reference terms and leaves I/2. Squared fidelity with the pure plus target is then , whereas either ideal independent copy would have fidelity one.
Controlled-NOT copies the chosen basis labels but produces an entangled Bell pair from plus and a blank; neither marginal is a perfect plus copy.
The living notebook
Explore the model
Interactive lab · quantum
An unknown basis distribution omits phase
For each fresh comparison, Reset and set both controls. Reset and set angle 90 degrees and phase 0 degrees, then in another fresh run set both controls to angle 90 and phase 180 degrees. Predict equal computational populations but Bloch x=+1 and −1. The comparison shows why reproducing one basis distribution does not specify a quantum-state copy. The lab has one pure qubit only; it implements no copier, output pair, mixed marginal, broadcasting process, or measurement collapse.
- Probability |0⟩
- 71.1%
- Probability |1⟩
- 28.9%
- Bloch (x, y, z)
- 0.78, 0.45, 0.42
A normalized two-level pure state, not an interacting field simulation. Playback adds one radian per second to the relative phase; basis probabilities stay fixed. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forA copying claim must preserve the full unknown state, including phase information beyond a selected probability distribution.
Pause & reason
Check your understanding
Your turn
Try it yourself
Use controlled-NOT with a blank zero to broadcast . Write the output, check both marginals, and compare its probability of 11 with the independent-copy target.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Propagate each classical basis projector separately, then use the product rule only for the independent-copy target.
Compare with the solution
The output is . Tracing either factor gives diag(3/4,1/4), so broadcasting succeeds. The joint probability of 11 is 1/4; in it would be . The outputs are correlated rather than independently cloned. The same fixed construction broadcasts every state diagonal in this basis, illustrating the commuting-family exception.
Keep the idea
Take it with you
- Exact deterministic copying fails for distinct nonorthogonal pure inputs under one fixed device.
- A basis copier can preserve a classical label while destroying the local coherence of a superposition.
- Broadcasting permits correlations but is possible for an unknown finite-dimensional family exactly when its states commute.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.