Simple Algorithm
Quantum Computing · Excited State · 22 min

Entanglement as a Resource (LOCC)

What you can and cannot do with local operations and classical communication.

Put Alice and Bob in distant labs joined only by a telephone line. Each can do anything quantum locally, and they may talk as much as they like, but they cannot send qubits. Under these rules entanglement cannot be created, only spent and reshaped, which turns it into a currency with exchange rates.

What you’ll learn

  • Define LOCC and prove that it cannot create entanglement from separable states.
  • Account for teleportation and superdense coding as exchanges between ebits, classical bits and qubits.
  • Quantify pure-state entanglement by the entropy of entanglement, and decide simple conversions with Schmidt-rank monotonicity and majorization.
  • Explain the monogamy of entanglement and of nonlocality, and why it underpins quantum cryptography.
  • Recognize why mixed-state entanglement needs new tools: distillation, the PPT criterion and witnesses.

Distant labs and a phone line

LOCC stands for local operations and classical communication. Alice and Bob may apply any unitaries, add ancillas and measure in their own labs, and broadcast their measurement results, in as many rounds as they like; later operations may depend on earlier results. What they may not do is send quantum systems to each other, or apply a joint gate such as a CNOT across the two labs.

This is the natural setting of quantum networks: classical communication is cheap and reliable, while sending qubits is expensive and noisy. Entanglement distributed in advance bridges the gap.

DerivationLOCC cannot create entanglement
  1. σ=∑ipi αi⊗βi\sigma=\sum_ip_i\,\alpha_i\otimes\beta_i

    A separable state is a mixture of product states. LOCC can prepare any of them: Alice samples ii, announces it, and each party prepares their factor.

  2. (Ak⊗Bk)(αi⊗βi)(Ak⊗Bk)†=(AkαiAk†)⊗(BkβiBk†)(A_k\otimes B_k)(\alpha_i\otimes\beta_i)(A_k\otimes B_k)^\dagger=(A_k\alpha_iA_k^\dagger)\otimes(B_k\beta_iB_k^\dagger)

    Product Kraus operators map product operators to product operators.

  3. Λ(σ)=∑i,kpi (AkαiAk†)⊗(BkβiBk†)\Lambda(\sigma)=\sum_{i,k}p_i\,(A_k\alpha_iA_k^\dagger)\otimes(B_k\beta_iB_k^\dagger)

    Summing, and normalizing each term, gives again a mixture of product states: separable in, separable out. The same holds branch by branch when measurement outcomes are kept.

  4. So entanglement behaves like a resource: LOCC can consume or reshape it, never produce it. A Bell pair delivered in advance, one ebit, has value.

Check your understanding

Alice and Bob start in ∣00⟩\ket{00} and may use LOCC only. Which state can they prepare?

  1. ∣Φ+⟩=12(∣00⟩+∣11⟩)\ket{\Phi^+}=\tfrac{1}{\sqrt2}(\ket{00}+\ket{11})
  2. 12(∣00⟩ ⁣⟨00∣+∣11⟩ ⁣⟨11∣)\tfrac12\big(\dyad{00}+\dyad{11}\big)
  3. 0.9 ∣00⟩+0.1 ∣11⟩\sqrt{0.9}\,\ket{00}+\sqrt{0.1}\,\ket{11}
  4. ∣Ψ−⟩=12(∣01⟩−∣10⟩)\ket{\Psi^-}=\tfrac{1}{\sqrt2}(\ket{01}-\ket{10})
Show the answer

B. 12(∣00⟩ ⁣⟨00∣+∣11⟩ ⁣⟨11∣)\tfrac12\big(\dyad{00}+\dyad{11}\big)

Alice flips a coin, tells Bob the result, and both prepare ∣00⟩\ket{00} or ∣11⟩\ket{11} accordingly. The result is perfectly correlated in the ZZ basis, yet separable: classical correlation, not entanglement. The other three states are entangled, so LOCC cannot make them.

Ebits as currency

Once entanglement is shared, LOCC becomes powerful. Write [qq][qq] for one ebit, [c→c][c\to c] for one classical bit sent and [q→q][q\to q] for one qubit sent. The basic protocols are exchange rates between these resources:

ProtocolConsumesDelivers
Teleportation11 ebit + 2+\ 2 classical bits11 qubit transmitted
Superdense coding11 ebit + 1+\ 1 qubit sent22 classical bits
Entanglement swappingebits A–R and R–B, plus 22 classical bits from R11 ebit between A and B, who never interacted
Resource inequalities: [qq]+2[c→c]≥[q→q][qq]+2[c\to c]\ge[q\to q] and [qq]+[q→q]≥2[c→c][qq]+[q\to q]\ge2[c\to c].
Think first Could a cleverer teleportation protocol get by with only one classical bit per qubit?

No. If 11 ebit plus 11 classical bit could teleport a qubit, then superdense coding through that teleported qubit would deliver 22 classical bits using 22 ebits and only 11 transmitted bit. But pre-shared entanglement cannot increase what a classical channel carries: Bob's half of the entanglement is independent of Alice's message (no-signalling), so everything he learns about the message must pass through the bits he receives. Hence teleportation needs at least 22 classical bits per qubit.

How much entanglement does a pure state hold?

For a pure state, write the Schmidt decomposition ∣ψ⟩=∑iλi ∣ai⟩∣bi⟩\ket\psi=\sum_i\sqrt{\lambda_i}\,\ket{a_i}\ket{b_i}. Local unitaries change the bases, never the Schmidt coefficients λi\lambda_i, so every entanglement property of ∣ψ⟩\ket\psi is a function of λ=(λ1,λ2,… )\lambda=(\lambda_1,\lambda_2,\dots). The entropy of entanglement is the von Neumann entropy of either reduced state,

E(ψ)=S(ρA)=−∑iλilog⁡2λi,E(\psi)=S(\rho_A)=-\sum_i\lambda_i\log_2\lambda_i,

measured in ebits: E=1E=1 for a Bell pair and E=0E=0 for a product state.

How many Bell pairs is a partially entangled state worth? With many copies, Alice and Bob can turn nn copies of ∣ψ⟩\ket\psi into about nE(ψ)nE(\psi) Bell pairs (entanglement concentration), and conversely make nn copies of ∣ψ⟩\ket\psi from about nE(ψ)nE(\psi) Bell pairs plus a sublinear amount of communication (entanglement dilution). Both work with vanishing error as n→∞n\to\infty (Bennett, Bernstein, Popescu & Schumacher, 1996), so for pure states the exchange rate is unique and reversible: the entropy of entanglement is the measure.

Predict before you look

Alice and Bob share n=1000n=1000 copies of 0.9 ∣00⟩+0.1 ∣11⟩\sqrt{0.9}\,\ket{00}+\sqrt{0.1}\,\ket{11}. In the large-nn limit, about how many Bell pairs is this batch worth?

Show the answer

470 Bell pairs

E=−0.9log⁡20.9−0.1log⁡20.1≈0.137+0.332=0.469E=-0.9\log_2 0.9-0.1\log_2 0.1\approx0.137+0.332=0.469 ebits per copy, so about 469469 Bell pairs. At finite nn the yield is somewhat lower, approaching nEnE with corrections of order n\sqrt n.

Single copies: rank and majorization

With a single copy and no error allowed, the rules are stricter than "entropy cannot increase". The first constraint is the Schmidt rank, the number of nonzero Schmidt coefficients.

DerivationSchmidt rank never increases under LOCC
  1. ∣ψ⟩=∑ijCij ∣i⟩∣j⟩,Schmidt rank(ψ)=rank⁡C\ket\psi=\sum_{ij}C_{ij}\,\ket i\ket j,\qquad\text{Schmidt rank}(\psi)=\rank C

    The Schmidt coefficients are the squared singular values of the coefficient matrix CC.

  2. (A⊗B)∣ψ⟩=∑ij(A C BT)ij ∣i⟩∣j⟩(A\otimes B)\ket\psi=\sum_{ij}\big(A\,C\,B^{T}\big)_{ij}\,\ket i\ket j

    A product Kraus operator acts on CC by multiplication from both sides.

  3. rank⁡(A C BT)≤rank⁡C\rank\big(A\,C\,B^{T}\big)\le\rank C

    Rank never grows under multiplication. Every branch of an LOCC protocol has this form, so even with small probability, LOCC cannot raise the Schmidt rank: it cannot turn a product state into an entangled one, or a rank-2 state into a rank-3 state.

Monogamy: entanglement cannot be shared freely

If Alice's qubit is maximally entangled with Bob's, it cannot be entangled, or even correlated, with anything else. More generally, if ρAB\rho_{AB} is pure then every extension to a third system is a product, ρABC=ρAB⊗ρC\rho_{ABC}=\rho_{AB}\otimes\rho_C. Entanglement is monogamous, unlike classical correlation: a classical bit can be perfectly correlated with any number of copies.

DerivationNonlocality is monogamous too
  1. Alice measures A0A_0 or A1A_1, Bob B0B_0 or B1B_1, Charlie C0C_0 or C1C_1. Alice uses the same two measurements in her CHSH test with Bob and in her test with Charlie.

  2. SAB+SAC=⟨A0B0⟩+⟨A0C1⟩+⟨A1B0⟩−⟨A1C1⟩⏟Bob measures B0, Charlie C1+⟨A0C0⟩+⟨A0B1⟩+⟨A1C0⟩−⟨A1B1⟩⏟Bob measures B1, Charlie C0S_{AB}+S_{AC}=\underbrace{\expval{A_0B_0}+\expval{A_0C_1}+\expval{A_1B_0}-\expval{A_1C_1}}_{\text{Bob measures } B_0,\ \text{Charlie } C_1}+\underbrace{\expval{A_0C_0}+\expval{A_0B_1}+\expval{A_1C_0}-\expval{A_1B_1}}_{\text{Bob measures } B_1,\ \text{Charlie } C_0}

    Regroup the eight correlators into two CHSH-like sums.

  3. In the first bracket Bob and Charlie can measure simultaneously (B0B_0 and C1C_1): together they act like one party with a single setting and two output bits. A party with a single setting can always be simulated locally (its outputs can be generated in advance), so each bracket is at most 2.

  4. SAB+SAC≤4S_{AB}+S_{AC}\le4

    If Alice and Bob violate CHSH, Alice and Charlie cannot, with the same measurements on Alice's side. Only no-signalling was used, to ensure each correlator does not depend on what the third party measures, so this holds in any no-signalling theory.

Monogamy is why entanglement-based QKD works: if Alice and Bob observe a near-maximal CHSH violation, an eavesdropper's system is nearly uncorrelated with Alice's outcomes. This is the heart of Device-Independent QKD.

Check your understanding

Alice and Bob share exactly ∣Φ+⟩\ket{\Phi^+}. Eve holds a third system CC. What can the joint state be?

  1. Eve's system may be entangled with Alice's qubit, as long as it is not entangled with Bob's
  2. Eve may hold a classical copy of Alice's ZZ-measurement outcome
  3. Anything, as long as Eve's system is not entangled with ABAB
  4. Only ρABC=∣Φ+⟩ ⁣⟨Φ+∣⊗ρC\rho_{ABC}=\dyad{\Phi^+}\otimes\rho_C: Eve is completely uncorrelated
Show the answer

D. Only ρABC=∣Φ+⟩ ⁣⟨Φ+∣⊗ρC\rho_{ABC}=\dyad{\Phi^+}\otimes\rho_C: Eve is completely uncorrelated

A pure state of ABAB admits only product extensions, ρABC=∣Φ+⟩ ⁣⟨Φ+∣⊗ρC\rho_{ABC}=\dyad{\Phi^+}\otimes\rho_C. Eve can know nothing about Alice's outcomes, which is the ideal situation for key distribution.

Mixed states: the road ahead

Real pairs are noisy. A mixed state is separable if ρ=∑ipi αi⊗βi\rho=\sum_ip_i\,\alpha_i\otimes\beta_i, and entangled otherwise. For mixed states, "entangled", "useful for distillation" and "violates a Bell inequality" come apart. The two-qubit Werner state ρp=p ∣Ψ−⟩ ⁣⟨Ψ−∣+(1−p) I4\rho_p=p\,\dyad{\Psi^-}+(1-p)\,\tfrac{\id}{4} shows all three regimes:

Range of ppWhat ρp\rho_p can do
p≤13p\le\tfrac13Separable: preparable by LOCC, worthless as entanglement
13<p≤12\tfrac13<p\le\tfrac12Entangled and distillable, yet it admits a local hidden-variable model for all projective measurements (Werner, 1989)
12<p≤12\tfrac12<p\le\tfrac{1}{\sqrt2}Entangled and distillable, but no CHSH violation
p>12≈0.71p>\tfrac{1}{\sqrt2}\approx0.71Violates CHSH
Entanglement is necessary for nonlocality, but not sufficient.

Three questions open up, each a Rydberg-level lesson:

  • Can noisy pairs be purified? Yes: by LOCC on many copies, Alice and Bob can distill fewer, better pairs. Every entangled two-qubit state is distillable.
  • How do you even tell whether a mixed state is entangled? Separable states have a positive partial transpose (the PPT criterion), and for two qubits that test is exact.
  • Is all entanglement distillable? No. In larger dimensions there are entangled PPT states that cannot be distilled at all, yet cost ebits to create: bound entanglement.

Key takeaways

  • LOCC maps separable states to separable states: entanglement cannot be created, only consumed and reshaped, so it is a resource.
  • Exchange rates: teleportation spends 11 ebit + 2+\ 2 classical bits per qubit; superdense coding spends 11 ebit + 1+\ 1 qubit per 22 classical bits.
  • Pure states: E=S(ρA)E=S(\rho_A) governs asymptotic, reversible conversion; for single copies the Schmidt rank cannot increase and Nielsen's majorization criterion decides deterministic conversion.
  • Entanglement and nonlocality are monogamous (CAB2+CAC2≤CA(BC)2C_{AB}^2+C_{AC}^2\le C_{A(BC)}^2; SAB+SAC≤4S_{AB}+S_{AC}\le4), the root of quantum key security.
  • For mixed states, entanglement, distillability and nonlocality come apart, which motivates distillation, PPT and witnesses.

Check yourself

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

Question 1

Which task is impossible by LOCC?

  1. Turning ∣Φ+⟩\ket{\Phi^+} into 0.9 ∣00⟩+0.1 ∣11⟩\sqrt{0.9}\,\ket{00}+\sqrt{0.1}\,\ket{11} with certainty
  2. Turning 0.9 ∣00⟩+0.1 ∣11⟩\sqrt{0.9}\,\ket{00}+\sqrt{0.1}\,\ket{11} into ∣Φ+⟩\ket{\Phi^+} with certainty
  3. Preparing 12(∣00⟩ ⁣⟨00∣+∣11⟩ ⁣⟨11∣)\tfrac12\big(\dyad{00}+\dyad{11}\big) from ∣00⟩\ket{00}
  4. Teleporting a qubit using one shared Bell pair and two classical bits
Show the answer

B. Turning 0.9 ∣00⟩+0.1 ∣11⟩\sqrt{0.9}\,\ket{00}+\sqrt{0.1}\,\ket{11} into ∣Φ+⟩\ket{\Phi^+} with certainty

By Nielsen's theorem a deterministic conversion needs λmax⁡(ψ)≤λmax⁡(ϕ)\lambda_{\max}(\psi)\le\lambda_{\max}(\phi), but 0.9>0.50.9>0.5. The best success probability is 0.20.2.

Question 2

A pure state has Schmidt coefficients (12,14,14)(\tfrac12,\tfrac14,\tfrac14). What is its entropy of entanglement?

  1. log⁡23≈1.58\log_2 3\approx1.58 ebits
  2. 11 ebit
  3. 1.51.5 ebits
  4. 0.50.5 ebits
Show the answer

C. 1.51.5 ebits

E=−12log⁡212−2⋅14log⁡214=12+1=1.5E=-\tfrac12\log_2\tfrac12-2\cdot\tfrac14\log_2\tfrac14=\tfrac12+1=1.5 ebits.

Question 3

To teleport 5 qubits, Alice and Bob need at least…

  1. 5 ebits and 10 classical bits
  2. 5 ebits and 5 classical bits
  3. 10 ebits and 5 classical bits
  4. 5 ebits and no communication
Show the answer

A. 5 ebits and 10 classical bits

Each teleported qubit consumes one ebit and two classical bits, and neither cost can be reduced: [qq]+2[c→c]≥[q→q][qq]+2[c\to c]\ge[q\to q] is tight.

Question 4

Alice uses the same two measurements in CHSH tests with Bob and with Charlie, and she and Bob observe SAB=2.7S_{AB}=2.7. What upper bound does no-signalling place on SACS_{AC}?

  1. 22
  2. 222\sqrt2
  3. 1.31.3
  4. 00
Show the answer

C. 1.31.3

Monogamy of nonlocality: SAB+SAC≤4S_{AB}+S_{AC}\le4, so SAC≤1.3S_{AC}\le1.3. In particular Alice and Charlie cannot violate CHSH.

Question 5

For the Werner state ρp=p ∣Ψ−⟩ ⁣⟨Ψ−∣+(1−p) I4\rho_p=p\,\dyad{\Psi^-}+(1-p)\,\tfrac{\id}{4} with p=0.6p=0.6, which statement is true?

  1. It is separable
  2. It violates the CHSH inequality
  3. It is bound entangled (entangled but undistillable)
  4. It is entangled and distillable, but violates no CHSH inequality
Show the answer

D. It is entangled and distillable, but violates no CHSH inequality

p=0.6>13p=0.6>\tfrac13, so it is entangled, and every entangled two-qubit state is distillable. CHSH needs p>1/2≈0.71p>1/\sqrt2\approx0.71.

Go deeper

Further reading

  • M. A. Nielsen & I. L. Chuang, Quantum Computation and Quantum Information, §12.5 (Cambridge, 10th anniv. ed., 2010).
  • R. Horodecki, P. Horodecki, M. Horodecki & K. Horodecki, "Quantum entanglement", Rev. Mod. Phys. 81, 865 (2009).
  • C. H. Bennett, H. J. Bernstein, S. Popescu & B. Schumacher, "Concentrating partial entanglement by local operations", Phys. Rev. A 53, 2046 (1996).
  • V. Coffman, J. Kundu & W. K. Wootters, "Distributed entanglement", Phys. Rev. A 61, 052306 (2000).