Simple Algorithm
Quantum Computing · Excited State · 24 min

Teleportation & Superdense Coding

Trading entanglement and classical bits for quantum states, and back.

One shared Bell pair plus two classical bits can move an unknown qubit across any distance; one Bell pair plus one qubit can carry two classical bits. In this lesson you derive both protocols line by line, build them as circuits, show why neither beats light, and quantify what noisy entanglement costs.

What you’ll learn

  • Use the identities ∣βxy⟩=(I⊗XyZx)∣Φ+⟩\ket{\beta_{xy}}=(\id\otimes X^yZ^x)\ket{\Phi^+} to encode and decode two bits in a Bell pair.
  • Derive the teleportation identity and Bob's corrections Zm1Xm2Z^{m_1}X^{m_2}, and implement them with deferred measurement.
  • Show that Bob's qubit is maximally mixed until the classical bits arrive, and that no copy is left behind.
  • Compute teleportation fidelity with noisy entanglement, Fˉ=(2f+1)/3\bar F=(2f+1)/3, and explain the classical 2/32/3 benchmark.
  • State the resource inequalities and argue that both protocols are optimal.

The Bell basis as a Pauli toolkit

The four Bell states form an orthonormal basis of two qubits. Label them by two bits, ∣βxy⟩=12(∣0y⟩+(−1)x∣1yˉ⟩)\ket{\beta_{xy}}=\tfrac{1}{\sqrt2}\big(\ket{0y}+(-1)^x\ket{1\bar y}\big), so that β00=Φ+\beta_{00}=\Phi^+, β01=Ψ+\beta_{01}=\Psi^+, β10=Φ−\beta_{10}=\Phi^- and β11=Ψ−\beta_{11}=\Psi^-. The key identity: every Bell state is Φ+\Phi^+ with a Pauli applied to just one of the qubits, either one.

∣βxy⟩=(I⊗XyZx)∣Φ+⟩=(ZxXy⊗I)∣Φ+⟩\ket{\beta_{xy}}=(\id\otimes X^{y}Z^{x})\ket{\Phi^+}=(Z^{x}X^{y}\otimes\id)\ket{\Phi^+}
xyxyBell statePauli on the first qubit(H⊗I) CNOT(H\otimes\id)\,\mathrm{CNOT} maps it to
00Φ+=12(∣00⟩+∣11⟩)\Phi^+=\tfrac{1}{\sqrt2}(\ket{00}+\ket{11})I\id∣00⟩\ket{00}
01Ψ+=12(∣01⟩+∣10⟩)\Psi^+=\tfrac{1}{\sqrt2}(\ket{01}+\ket{10})XX∣01⟩\ket{01}
10Φ−=12(∣00⟩−∣11⟩)\Phi^-=\tfrac{1}{\sqrt2}(\ket{00}-\ket{11})ZZ∣10⟩\ket{10}
11Ψ−=12(∣01⟩−∣10⟩)\Psi^-=\tfrac{1}{\sqrt2}(\ket{01}-\ket{10})ZXZX (XX first, then ZZ)∣11⟩\ket{11}
Check your understanding

Alice and Bob share Φ+\Phi^+. Alice applies ZZ to her qubit and sends it to Bob, who applies CNOT and HH and measures both qubits. What does he read?

  1. 1010, because the pair became Φ−\Phi^-
  2. 0101, because the pair became Ψ+\Psi^+
  3. 0000: a local ZZ cannot change a shared entangled state
  4. A random result, since ZZ only changes a phase
Show the answer

A. 1010, because the pair became Φ−\Phi^-

(Z⊗I)Φ+=Φ−=β10(Z\otimes\id)\Phi^+=\Phi^-=\beta_{10}, and the decoder maps β10↦∣10⟩\beta_{10}\mapsto\ket{10}.

Superdense coding

Protocol (Bennett & Wiesner 1992). Alice and Bob share Φ+\Phi^+ in advance.

  1. To send the bits (x,y)(x,y), Alice applies ZxXyZ^xX^y to her qubit, turning the pair into ∣βxy⟩\ket{\beta_{xy}}.
  2. She sends her qubit to Bob.
  3. Bob, now holding both qubits, performs the Bell measurement and reads (x,y)(x,y).

Build it below. The first two columns make the shared Φ+\Phi^+; qubit 0 is Alice's and qubit 1 is Bob's.

Think first Could Alice squeeze 3 bits into one transmitted qubit by sharing a bigger entangled state with Bob, say with Bob holding many qubits?

No. With one transmitted qubit and any amount of prior entanglement, at most 2 bits can be conveyed: sending a dd-dimensional system carries at most 2log⁡2d2\log_2 d bits, even with unlimited shared entanglement. Superdense coding is already optimal.

Teleportation, line by line

Now the protocol of Bennett, Brassard, Crépeau, Jozsa, Peres & Wootters (1993). Alice holds the unknown ∣ψ⟩=α∣0⟩+β∣1⟩\ket\psi=\alpha\ket0+\beta\ket1 on qubit 0 and shares Φ+\Phi^+ on qubits 1 (hers) and 2 (Bob's).

DerivationThe teleportation identity
  1. ∣ψ⟩0∣Φ+⟩12=12(α∣000⟩+α∣011⟩+β∣100⟩+β∣111⟩)\ket\psi_0\ket{\Phi^+}_{12}=\tfrac{1}{\sqrt2}\big(\alpha\ket{000}+\alpha\ket{011}+\beta\ket{100}+\beta\ket{111}\big)

    Expand the product state.

  2. ∣00⟩=∣Φ+⟩+∣Φ−⟩2,∣11⟩=∣Φ+⟩−∣Φ−⟩2,∣01⟩=∣Ψ+⟩+∣Ψ−⟩2,∣10⟩=∣Ψ+⟩−∣Ψ−⟩2\ket{00}=\tfrac{\ket{\Phi^+}+\ket{\Phi^-}}{\sqrt2},\quad\ket{11}=\tfrac{\ket{\Phi^+}-\ket{\Phi^-}}{\sqrt2},\quad\ket{01}=\tfrac{\ket{\Psi^+}+\ket{\Psi^-}}{\sqrt2},\quad\ket{10}=\tfrac{\ket{\Psi^+}-\ket{\Psi^-}}{\sqrt2}

    Rewrite Alice's two qubits (0 and 1) in the Bell basis.

  3. =12[∣Φ+⟩(α∣0⟩+β∣1⟩)+∣Φ−⟩(α∣0⟩−β∣1⟩)+∣Ψ+⟩(β∣0⟩+α∣1⟩)+∣Ψ−⟩(α∣1⟩−β∣0⟩)]=\tfrac12\Big[\ket{\Phi^+}\big(\alpha\ket0+\beta\ket1\big)+\ket{\Phi^-}\big(\alpha\ket0-\beta\ket1\big)+\ket{\Psi^+}\big(\beta\ket0+\alpha\ket1\big)+\ket{\Psi^-}\big(\alpha\ket1-\beta\ket0\big)\Big]

    Collect terms. Nothing physical has happened yet: this is the same state, re-expanded.

  4. =12∑x,y∈{0,1}∣βxy⟩01⊗XyZx∣ψ⟩2=\tfrac12\sum_{x,y\in\{0,1\}}\ket{\beta_{xy}}_{01}\otimes X^{y}Z^{x}\ket\psi_2

    Recognize Bob's four states as ∣ψ⟩\ket\psi, Z∣ψ⟩Z\ket\psi, X∣ψ⟩X\ket\psi and XZ∣ψ⟩XZ\ket\psi.

  5. outcome (m1,m2) ⇒ Bob holds Xm2Zm1∣ψ⟩,Zm1Xm2 Xm2Zm1∣ψ⟩=∣ψ⟩\text{outcome }(m_1,m_2)\ \Rightarrow\ \text{Bob holds }X^{m_2}Z^{m_1}\ket\psi,\qquad Z^{m_1}X^{m_2}\,X^{m_2}Z^{m_1}\ket\psi=\ket\psi

    Alice's Bell measurement returns (m1,m2)=(x,y)(m_1,m_2)=(x,y), each with probability ¼ whatever ∣ψ⟩\ket\psi is. Bob applies Xm2X^{m_2} and then Zm1Z^{m_1}.

No signalling, no cloning, and noisy entanglement

DerivationBob's qubit before the call
  1. ρB=∑m1,m214 Xm2Zm1 ψ Zm1Xm2\rho_B=\sum_{m_1,m_2}\tfrac14\,X^{m_2}Z^{m_1}\,\psi\,Z^{m_1}X^{m_2}

    Bob does not know the outcome, so he averages over the four equally likely branches.

  2. =14(ψ+ZψZ+XψX+YψY)=\tfrac14\big(\psi+Z\psi Z+X\psi X+Y\psi Y\big)

    XZ=−iYXZ=-iY, and the phase cancels in the conjugation.

  3. =I2for every ψ=\tfrac{\id}{2}\qquad\text{for every }\psi

    Write ψ=12(I+r⃗⋅σ⃗)\psi=\tfrac12(\id+\vec r\cdot\vec\sigma). Conjugation by each Pauli keeps its own Bloch component and flips the other two, so the four terms cancel r⃗\vec r completely. This Pauli twirl is the no-signalling theorem at work.

Real entanglement is noisy. Suppose Alice and Bob share a mixed two-qubit state ρ\rho instead of Φ+\Phi^+. Its fully entangled fraction f=max⁡Φ⟨Φ∣ρ∣Φ⟩f=\max_\Phi\bra\Phi\rho\ket\Phi is the largest overlap with any maximally entangled state. Running the standard protocol (with Bob's corrections adjusted to the best Φ\Phi) gives average teleportation fidelity (Horodecki, Horodecki & Horodecki 1999):

Fˉ=2f+13.\bar F=\frac{2f+1}{3}.
Predict before you look

Alice and Bob teleport through a Werner state p Φ++(1−p)I4p\,\Phi^++(1-p)\tfrac{\id}{4} with p=0.8p=0.8. What average fidelity do they get?

Show the answer

0.9

f=1+3(0.8)4=0.85f=\tfrac{1+3(0.8)}{4}=0.85, so Fˉ=2(0.85)+13=0.9\bar F=\tfrac{2(0.85)+1}{3}=0.9, which equals 1+p2\tfrac{1+p}{2}. Teleporting through a Werner state acts on ∣ψ⟩\ket\psi as a depolarizing channel that shrinks its Bloch vector by pp.

Resource accounting

ConversionResource inequalityProtocol
Entanglement + classical bits → qubit1 ebit+2 cbits ≥ 1 qubit1\ \text{ebit}+2\ \text{cbits}\ \ge\ 1\ \text{qubit}Teleportation
Entanglement + qubit → classical bits1 ebit+1 qubit ≥ 2 cbits1\ \text{ebit}+1\ \text{qubit}\ \ge\ 2\ \text{cbits}Superdense coding
Qubit → entanglement1 qubit ≥ 1 ebit1\ \text{qubit}\ \ge\ 1\ \text{ebit}Send half of a Bell pair
Classical communication → entanglementImpossibleLOCC cannot create entanglement

Both protocols are optimal. Superdense coding cannot beat 2 bits per qubit (the entanglement-assisted Holevo bound). And teleportation cannot use fewer than 2 classical bits per qubit, by an argument that combines the two protocols:

Check your understanding

Why must every teleportation protocol use at least 2 classical bits per qubit?

  1. Because a qubit stores two real numbers, a magnitude and a phase
  2. Because Bob's corrections need one bit per Pauli, and there are two Paulis
  3. Otherwise, followed by superdense coding, it would let shared entanglement boost a classical channel beyond its capacity
  4. Because with fewer bits Bob's qubit would remain entangled with Alice's
Show the answer

C. Otherwise, followed by superdense coding, it would let shared entanglement boost a classical channel beyond its capacity

If 1 bit plus entanglement could teleport a qubit, using that qubit for superdense coding would deliver 2 bits for 1 bit sent. But before the message arrives, Bob's systems are independent of Alice's data (no-signalling), so everything he learns must flow through the transmitted bits, and kk bits carry at most kk bits of information. Hence 2 bits are necessary.

Key takeaways

  • ∣ψ⟩∣Φ+⟩=12∑xy∣βxy⟩XyZx∣ψ⟩\ket\psi\ket{\Phi^+}=\tfrac12\sum_{xy}\ket{\beta_{xy}}X^yZ^x\ket\psi: a Bell measurement leaves Bob with a known Pauli of ∣ψ⟩\ket\psi, undone by Zm1Xm2Z^{m_1}X^{m_2}.
  • Superdense coding: ZxXyZ^xX^y on Alice's half makes ∣βxy⟩\ket{\beta_{xy}}, and Bob's CNOT plus HH reads (x,y)(x,y). Two bits per qubit is optimal.
  • Before the bits arrive Bob holds I/2\id/2 (a Pauli twirl), so there is no signalling; Alice keeps nothing, so there is no cloning.
  • With noisy entanglement Fˉ=(2f+1)/3\bar F=(2f+1)/3; beating 2/32/3 requires, and certifies, entanglement.

Check yourself

In the interactive lesson these are scored with your confidence, and what you miss comes back for review.

Question 1

Alice's Bell measurement returns (m1,m2)=(0,1)(m_1,m_2)=(0,1). Which operation must Bob apply?

  1. XX
  2. ZZ
  3. ZXZX
  4. Nothing
Show the answer

A. XX

For outcome (0,1)(0,1) Bob holds X1Z0∣ψ⟩=X∣ψ⟩X^{1}Z^{0}\ket\psi=X\ket\psi, and applying XX restores ∣ψ⟩\ket\psi.

Question 2

Alice and Bob teleport through a Werner state with p=1/3p=1/3, so f=1/2f=1/2. What is the average teleportation fidelity?

  1. 1/21/2
  2. 5/65/6
  3. 1/31/3
  4. 2/32/3
Show the answer

D. 2/32/3

Fˉ=2f+13=23\bar F=\tfrac{2f+1}{3}=\tfrac23: exactly the classical measure-and-prepare benchmark. Werner states with p≤1/3p\le1/3 are separable, so this is no surprise.

Question 3

In superdense coding, Eve intercepts the qubit that Alice sends, but never touches Bob's. What can she learn about the 2-bit message?

  1. Both bits, by measuring in the Bell basis
  2. One of the two bits
  3. Nothing at all
  4. The parity x⊕yx\oplus y
Show the answer

C. Nothing at all

The intercepted qubit's reduced state is I/2\id/2 for all four messages, so no measurement on it alone can reveal anything.

Question 4

In the deferred-measurement teleportation circuit, what state are Alice's two qubits in after Bob's corrections?

  1. ∣00⟩\ket{00}
  2. A state entangled with Bob's qubit
  3. A copy of ∣ψ⟩\ket\psi on qubit 0
  4. ∣+⟩∣+⟩\ket+\ket+, independent of ∣ψ⟩\ket\psi
Show the answer

D. ∣+⟩∣+⟩\ket+\ket+, independent of ∣ψ⟩\ket\psi

In every branch ∣m1m2⟩\ket{m_1m_2} Bob ends with ∣ψ⟩\ket\psi, so the total state is (12∑∣m1m2⟩)⊗∣ψ⟩=∣+⟩∣+⟩∣ψ⟩\big(\tfrac12\sum\ket{m_1m_2}\big)\otimes\ket\psi=\ket+\ket+\ket\psi: a product state, independent of ∣ψ⟩\ket\psi. No copy remains with Alice.

Question 5

Which resource conversion is impossible?

  1. 1 ebit + 2 cbits → 1 qubit
  2. Unlimited classical communication and local operations → 1 ebit
  3. 1 ebit + 1 qubit → 2 cbits
  4. 1 qubit → 1 ebit
Show the answer

B. Unlimited classical communication and local operations → 1 ebit

LOCC cannot create entanglement from scratch. The others are teleportation, superdense coding, and simply sending half of a Bell pair.

Go deeper

Further reading

  • C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres & W. K. Wootters, "Teleporting an unknown quantum state via dual classical and Einstein–Podolsky–Rosen channels", Physical Review Letters 70, 1895 (1993).
  • C. H. Bennett & S. J. Wiesner, "Communication via one- and two-particle operators on Einstein–Podolsky–Rosen states", Physical Review Letters 69, 2881 (1992).
  • M. Horodecki, P. Horodecki & R. Horodecki, "General teleportation channel, singlet fraction, and quasidistillation", Physical Review A 60, 1888 (1999).
  • M. A. Nielsen & I. L. Chuang, Quantum Computation and Quantum Information, §1.3.7 and §2.3 (Cambridge, 2010).