The No-Cloning Theorem: Why CNOT Copies Bits but Not Qubits

The No-Cloning Theorem: Why CNOT Copies Bits but Not Qubits

CNOT copies computational-basis bits, but it cannot copy an arbitrary unknown qubit. A tiny Qiskit circuit shows why |+> becomes entangled instead of cloned.

A CNOT gate can turn |0⟩|0⟩ into |0⟩|0⟩ and |1⟩|0⟩ into |1⟩|1⟩. That looks like copying. But place an arbitrary qubit in the control and the same gate does not leave two independent copies. It entangles the qubits instead. This small distinction is the heart of the no-cloning theorem.

What the theorem forbids

The no-cloning theorem says that no single physical process can perfectly copy every arbitrary unknown quantum state. It does not say that nothing quantum can ever be copied. A known computational-basis state can be copied into a blank target with a CNOT, and a state whose preparation recipe is known can be prepared again. The impossible task is taking one unknown state and making a perfect independent copy for every possible input.

Here is the key idea in a short proof. Suppose a machine U clones any state, with the second register always initialized to the same blank state |0⟩. Take two possible inputs, |ψ⟩ and |φ⟩, whose inner product is c = ⟨ψ|φ⟩. Before copying, the overlap between the complete inputs is c. After a perfect copy, the overlap between the outputs is c × c = c². But a unitary operation preserves inner products, so it would require c = c². That is true only when c is 0 or 1. A universal copier therefore cannot work for arbitrary non-orthogonal states.

A qubit that CNOT does not clone

Try the state |+⟩ = (|0⟩ + |1⟩) / √2 as the control, with the target in |0⟩. By linearity, the CNOT produces:

(|00⟩ + |11⟩) / √2

This is an entangled Bell state, not two copies of |+⟩. Two independent copies would instead be |+⟩|+⟩ = (|00⟩ + |01⟩ + |10⟩ + |11⟩) / 2. The output amplitudes are different, so the gate has not cloned the input.

from qiskit import QuantumCircuit
from qiskit.quantum_info import Statevector

qc = QuantumCircuit(2)
qc.h(0)       # put the control in |+>
qc.cx(0, 1)   # target began in |0>

state = Statevector.from_instruction(qc)
print(state.draw(output="text"))

The statevector shows amplitude only on |00⟩ and |11⟩, each with magnitude 1 / √2. Qiskit documents Statevector.from_instruction for evolving a circuit from the all-zero input and the text drawing option in its Statevector API reference.

Why programmers should care

The theorem explains why measuring or probing an unknown qubit is not like checking a file and keeping an untouched backup. A measurement can change what can be learned from later measurements, and a perfect extra copy is not available as a workaround. This is one reason eavesdropping attempts in protocols such as BB84 can leave detectable disturbance, although the no-cloning theorem alone is not a complete security proof.

Quantum error correction does not evade the theorem by making spare copies. It encodes quantum information across a larger entangled state and uses syndrome measurements to identify errors without reading out the encoded logical state. The useful rule of thumb is simple: copying known classical labels is possible; a universal copy button for unknown quantum states is not.