Concurrence and Wootters' Formula

Entanglement of a pure state is easy to quantify: trace out one half, and the entropy of what remains tells you everything. Entanglement of a mixed state is a different animal. A density matrix can be torn into pure pieces in uncountably many ways, and the honest amount of entanglement is the cheapest such decomposition — the one that, on average, uses the least entanglement to build the state. This convex-roof minimisation is a search over an infinite, curved landscape of ensembles, and for a generic system it is intractable. Then, for the smallest interesting case — two qubits — William Wootters found that the whole minimisation collapses to a single algebraic formula, built from an object called the concurrence. The concurrence lives on a symmetry unique to qubits: an antiunitary "spin flip" that time-reverses each spin, $|\psi\rangle\mapsto(\sigma_y\otimes\sigma_y)|\psi^*\rangle$ . The overlap of a state with its own flipped partner is invariant under local unitaries and vanishes exactly on product states, so its magnitude is already a raw entanglement gauge; for a pure state it equals twice the modulus of the amplitude-matrix determinant, $C=2|ad-bc|$ . Wootters' 1998 achievement was to extend this to any mixed two-qubit $\rho$ by spin-flipping the density matrix, forming the non-Hermitian product $\rho\tilde\rho$ , and combining its sorted eigenvalue roots into $C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\}$ . Because the entanglement of formation is a fixed monotone function of $C$ , this single number solves — exactly, and for the only time in the whole theory — the convex roof. Concurrence is the workhorse of two-qubit entanglement: it powers decoherence studies, condensed-matter phase diagnostics, and the monogamy inequalities that limit how entanglement is shared.

Key takeaways

The concurrence $C(\rho)$ measures two-qubit entanglement on a scale from $0$ (separable) to $1$ (maximally entangled, e.g. a Bell state).
Wootters' formula is closed-form: $C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\}$ , where the $\lambda_i$ are the decreasing square roots of the eigenvalues of $\rho\tilde\rho$ .
The spin-flipped state $\tilde\rho=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y)$ carries the antiunitary time-reversal symmetry that makes qubits special.
For a pure state $C=2|ad-bc|$ — twice the modulus of the amplitude-matrix determinant, a single local-unitary invariant.
Concurrence hands you the entanglement of formation exactly: $E_F=h\!\big(\tfrac{1+\sqrt{1-C^2}}{2}\big)$ , the only case where this convex-roof measure is solved.
Its square, the tangle $\tau=C^2$ , obeys the Coffman–Kundu–Wootters monogamy inequality, the first rigorous statement that entanglement cannot be freely shared.
Key idea: the antiunitary spin flip $|\psi\rangle\mapsto(\sigma_y\otimes\sigma_y)|\psi^*\rangle$ defines a local-unitary-invariant bilinear form that vanishes on product states; its convex-roof extension is analytically computable, giving concurrence and with it a closed-form entanglement of formation for two qubits.

The core insight

The antiunitary spin flip $|\psi\rangle\mapsto(\sigma_y\otimes\sigma_y)|\psi^*\rangle$ defines a bilinear form whose magnitude $|\langle\psi|\tilde\psi\rangle|$ is invariant under local unitaries and vanishes exactly on product states; extending that form to mixed states by convex roof yields the concurrence, an entanglement monotone that Wootters computed in closed form and that fixes the entanglement of formation for any two qubits.

Where this lives today

Because concurrence is the one entanglement measure that is both exactly computable and operationally meaningful for two qubits, it has become the default diagnostic wherever pairwise entanglement is tracked. In condensed-matter physics the two-site reduced density matrix of a spin chain — extracted from measured correlation functions in Heisenberg or transverse-field Ising models — is fed straight into Wootters' formula, and the resulting concurrence peaks or its derivative diverges at quantum critical points, making entanglement a probe of phase transitions. In open-system dynamics concurrence is the standard witness for entanglement sudden death: under local decoherence a two-qubit state's concurrence can hit exactly zero in finite time, long before the individual coherences have decayed. Experimental groups reconstruct $\rho$ by state tomography on photon pairs, trapped ions, and superconducting qubits and quote the concurrence as the headline entanglement figure. And the tangle's monogamy inequality is a research frontier of its own — the 3-tangle distinguishes GHZ-type from W-type tripartite entanglement, and generalised monogamy relations constrain entanglement in quantum networks and even sharpen the black-hole firewall paradox.

The physical picture

Write a pure two-qubit state as a $2\times2$ matrix of amplitudes, $M=\left(\begin{smallmatrix}a&b\\c&d\end{smallmatrix}\right)$ with rows labelled by the first qubit and columns by the second. A product state is one in which the matrix factorises into a column times a row — its rows are proportional, so its determinant vanishes. Entanglement is precisely the failure of that factorisation, and the determinant $ad-bc$ is exactly how much it fails. The concurrence is twice its modulus, and because a determinant is unchanged (up to phase) by independent basis changes on each side, this number does not move when Alice and Bob each spin their own qubit — it is a genuine local invariant. The first figure makes the determinant into a picture: a Bell state has the biggest possible determinant, a product state has none.

Figure

Concurrence equals twice the modulus of the amplitude-matrix determinant. For the Bell state the determinant is 1/2 so C = 1 (maximally entangled); for the product state one row is a multiple of the other, the determinant is 0, and C = 0 (separable)

The determinant picture is the pure-state face of concurrence; the mixed-state formula is what made Wootters' result famous. It looks like magic — spin-flip the density matrix, multiply, take eigenvalue roots, subtract — but each step is the residue of a hard optimisation that was solved and then packed away. What the number ultimately computes is the entanglement of formation: dial a Bell state against white noise and $C$ tells you, at each noise level, how many ebits it would cost to synthesise the state. The second figure is the payoff — the closed-form curve $E_F(C)$ that no other entanglement measure possesses.

Figure

The closed-form entanglement of formation as a function of concurrence, E_F = h((1 + √(1 − C²))/2) with h the binary entropy. It rises monotonically and convexly from 0 at C = 0 to 1 bit at C = 1. Marked: the Werner state at p = 1/2 (C = 0.25, E_F = 0.118 bit), a partially entangled state at C = 0.6 (E_F = 0.469 bit), and the Bell state at C = 1 (E_F = 1 bit)

The core equations

The construction starts from the spin flip, an antiunitary operation that time-reverses both qubits. Applied to a density matrix it conjugates the entries and sandwiches them between two $\sigma_y$ factors, producing a companion state whose overlap with the original diagnoses entanglement.

\tilde\rho=(\sigma_y\otimes\sigma_y)\,\rho^{*}\,(\sigma_y\otimes\sigma_y) \qquad (1)

where: $\rho$ — the two-qubit density matrix in the computational basis; $\rho^{*}$ — its entrywise complex conjugate in that basis; $\sigma_y=\left(\begin{smallmatrix}0&-i\\i&0\end{smallmatrix}\right)$ — the Pauli- $y$ matrix; $\tilde\rho$ — the spin-flipped (time-reversed) density matrix, itself a legitimate density matrix. Wootters' formula is the centrepiece. Form the non-Hermitian product $\rho\tilde\rho$ , take the square roots of its (real, non-negative) eigenvalues, sort them downward, and subtract the smaller three from the largest.

C(\rho)=\max\{0,\;\lambda_1-\lambda_2-\lambda_3-\lambda_4\},\qquad \lambda_1\ge\lambda_2\ge\lambda_3\ge\lambda_4\ge0 \qquad (2)

where: $\lambda_i$ — the square roots of the eigenvalues of $\rho\tilde\rho$ , equivalently the eigenvalues of the Hermitian $R=\sqrt{\sqrt\rho\,\tilde\rho\,\sqrt\rho}$ ; $C(\rho)$ — the concurrence, $0\le C\le1$ . Reading: the $\max$ with zero is essential — when noise dominates, $\lambda_1$ fails to beat the other three and the state is separable. For a pure state the machinery reduces to a plain overlap: the state against its own spin-flipped partner. The magnitude of that inner product is the concurrence, needing no eigenvalues at all.

C(\psi)=|\langle\psi|\tilde\psi\rangle|,\qquad |\tilde\psi\rangle=(\sigma_y\otimes\sigma_y)|\psi^{*}\rangle \qquad (3)

where: $|\psi\rangle$ — a normalised pure two-qubit state; $|\tilde\psi\rangle$ — its spin-flipped partner. A Bell state is (up to phase) its own flip, giving overlap $1$ ; a product state is maximally unlike its flip, giving overlap $0$ . Expanding that overlap in the computational basis yields the single most useful pure-state expression: twice the modulus of the amplitude-matrix determinant.

C(\psi)=2\,|ad-bc|,\qquad |\psi\rangle=a|00\rangle+b|01\rangle+c|10\rangle+d|11\rangle \qquad (4)

where: $a,b,c,d$ — amplitudes with $|a|^2+|b|^2+|c|^2+|d|^2=1$ ; $ad-bc$ — the determinant of the amplitude matrix, an $\mathrm{SL}(2)\times\mathrm{SL}(2)$ invariant. Reading: a product state has proportional rows and hence zero determinant, so equation (4) reproduces separability as $C=0$ . The prize is the entanglement of formation, which for two qubits is a fixed increasing function of $C$ — the only closed-form solution of a convex roof known anywhere in the theory.

E_F(\rho)=h\!\left(\frac{1+\sqrt{1-C(\rho)^2}}{2}\right),\qquad h(x)=-x\log_2 x-(1-x)\log_2(1-x) \qquad (5)

where: $E_F$ — entanglement of formation in ebits (bits); $h$ — the binary Shannon entropy; $C$ — the concurrence of $\rho$ . Reading: $C$ and $E_F$ carry the same information, but $C$ is the algebraically convenient handle and $E_F$ is the operational one — the asymptotic cost in Bell pairs to make the state. Finally, the square of the concurrence — the tangle — is the quantity that is shared among qubits, and it obeys a strict conservation law: entanglement is monogamous.

\tau=C^2,\qquad C_{AB}^2+C_{AC}^2\le C_{A(BC)}^2 \qquad (6)

where: $\tau$ — the two-qubit tangle; $C_{AB}$ — concurrence between $A$ and $B$ after tracing out the third qubit; $C_{A(BC)}$ — concurrence between $A$ and the joint system $BC$ . The total tangle qubit $A$ can distribute is capped, so strong entanglement with $B$ forces weak entanglement with $C$ — the Coffman–Kundu–Wootters inequality.

Derivation APure-state concurrence is twice the determinant

The goal is to start from the abstract definition $C(\psi)=|\langle\psi|\tilde\psi\rangle|$ and turn it into the concrete, computable invariant $C=2|ad-bc|$ , showing along the way that the spin flip really does act as the antiunitary time reversal claimed above and that the resulting number is manifestly independent of any global phase on the state. We evaluate the double flip on each basis ket, assemble the flipped partner, take the overlap, and read off the magnitude.

Step 1. Work out the action of $\sigma_y\otimes\sigma_y$ on the computational basis using $\sigma_y|0\rangle=i|1\rangle$ and $\sigma_y|1\rangle=-i|0\rangle$ : one finds $(\sigma_y\otimes\sigma_y)|00\rangle=-|11\rangle$ , $(\sigma_y\otimes\sigma_y)|11\rangle=-|00\rangle$ , while $(\sigma_y\otimes\sigma_y)|01\rangle=|10\rangle$ and $(\sigma_y\otimes\sigma_y)|10\rangle=|01\rangle$ . Why: each single-qubit $\sigma_y$ swaps $0$ and $1$ with an $\pm i$ factor, so the tensor product either double-swaps with a real sign or produces a cross term — this is the qubit time reversal.

Step 2. Conjugate the state, $|\psi^{*}\rangle=a^{*}|00\rangle+b^{*}|01\rangle+c^{*}|10\rangle+d^{*}|11\rangle$ , then apply the flip to get $|\tilde\psi\rangle=-a^{*}|11\rangle+b^{*}|10\rangle+c^{*}|01\rangle-d^{*}|00\rangle$ . Why: complex conjugation is the antilinear half of the antiunitary; without it the operation would be an ordinary unitary and would not detect entanglement.

Step 3. Form the overlap $\langle\psi|\tilde\psi\rangle$ by pairing matching basis labels: $a^{*}(-d^{*})+b^{*}(c^{*})+c^{*}(b^{*})+d^{*}(-a^{*})=-2(a^{*}d^{*}-b^{*}c^{*})=-2(ad-bc)^{*}$ . Why: only the four surviving terms contribute, and they collect into precisely twice the conjugated determinant — every cross term either matches or vanishes by orthonormality.

Step 4. Take the modulus: $C(\psi)=|\langle\psi|\tilde\psi\rangle|=2|ad-bc|$ . The conjugate and the overall sign disappear under the absolute value, and any global phase $e^{i\theta}$ on $|\psi\rangle$ cancels between bra and ket. Why: an entanglement measure must not depend on unphysical phase, and the modulus enforces exactly that, leaving a genuine $\mathrm{SL}(2)\times\mathrm{SL}(2)$ invariant. *For the Bell state $a=d=1/\sqrt2$ , $b=c=0$ gives $C=2|1/2|=1$ ; for a product state the rows of the amplitude matrix are proportional so $ad-bc=0$ and $C=0$ .*

Go deeper · Derivation B · 1 minWootters' formula as the solved convex roof

Derivation BWootters' formula as the solved convex roof

The goal is to explain why the four sorted eigenvalue roots of $\rho\tilde\rho$ are the right objects and how they solve the convex-roof minimisation that defines mixed-state entanglement, converting an optimisation over infinitely many ensembles into one algebraic expression. We define the convex roof, exhibit the spectral data that bounds every decomposition, and quote the existence of the optimal ensemble that saturates the bound — the content of Wootters' theorem.

Step 1. The mixed-state concurrence is the convex roof $C(\rho)=\min\sum_k p_k\,C(\psi_k)$ , minimised over all pure-state ensembles $\{p_k,|\psi_k\rangle\}$ that reproduce $\rho=\sum_k p_k|\psi_k\rangle\langle\psi_k|$ . Why: entanglement should count the cheapest way to build the state, so the honest measure averages pure-state concurrence over the least-entangled decomposition, not an arbitrary one.

Step 2. Symmetrise the flip: define the Hermitian, positive-semidefinite $R=\sqrt{\sqrt\rho\,\tilde\rho\,\sqrt\rho}$ , whose eigenvalues $\lambda_1\ge\lambda_2\ge\lambda_3\ge\lambda_4$ are the square roots of the eigenvalues of $\rho\tilde\rho$ . Why: the raw product $\rho\tilde\rho$ is non-Hermitian, but it is similar to a positive operator, so its spectrum is real and non-negative and $R$ gives an honest, orderable set of singular-value-like numbers.

Step 3. For any decomposition the average concurrence is bounded below by $\lambda_1-\lambda_2-\lambda_3-\lambda_4$ , and Wootters proves an optimal ensemble exists — built in the "magic" Bell basis where the flip acts as complex conjugation — that attains it whenever the bound is positive. Why: the magic basis diagonalises the symmetric bilinear form behind the flip, letting one rotate the ensemble to equalise pure-state concurrences and hit the spectral lower bound exactly.

Step 4. Clamp at zero: when $\lambda_1\le\lambda_2+\lambda_3+\lambda_4$ the optimal average concurrence is $0$ and the state is separable, so $C(\rho)=\max\{0,\lambda_1-\lambda_2-\lambda_3-\lambda_4\}$ . Why: a concurrence can never be negative, and the crossing point $\lambda_1=\lambda_2+\lambda_3+\lambda_4$ is precisely where a separable decomposition first becomes available. *For pure $\rho=|\psi\rangle\langle\psi|$ only $\lambda_1$ is nonzero and the formula collapses to $C=\lambda_1=|\langle\psi|\tilde\psi\rangle|$ , recovering equation (3); the general four-eigenvalue formula reproduces the Werner result on direct computation.*

Go deeper · Derivation C · 1 minConcurrence of the Werner state

Derivation CConcurrence of the Werner state

The goal is to evaluate Wootters' formula on the canonical noisy Bell state $\rho_W=p\,|\Phi^+\rangle\langle\Phi^+|+\tfrac{1-p}{4}\mathbb{I}$ and extract the sharp entanglement threshold in $p$ , demonstrating the whole algorithm end to end and confirming that the concurrence boundary coincides with the partial-transpose boundary for this family. We diagonalise in the Bell basis, where the spin flip is trivial, read off the sorted roots, and combine them.

Step 1. Rewrite $\rho_W$ in the Bell basis: the white-noise term spreads weight $\tfrac{1-p}{4}$ over all four Bell states, and the coherent term adds $p$ to $|\Phi^+\rangle$ , so the Bell-basis probabilities are $\tfrac{1+3p}{4}$ for $|\Phi^+\rangle$ and $\tfrac{1-p}{4}$ for each of the other three. Why: $\rho_W$ is diagonal in the Bell basis, which is the natural frame because the Bell states are eigenstates of the spin flip.

Step 2. Because each Bell state maps to itself (up to sign) under the flip, $\rho\tilde\rho$ is diagonal in the Bell basis with eigenvalues equal to the squares of these probabilities, so the sorted roots are the probabilities themselves: $\lambda_1=\tfrac{1+3p}{4}$ and $\lambda_2=\lambda_3=\lambda_4=\tfrac{1-p}{4}$ . Why: for a flip-diagonal state $\tilde\rho=\rho$ , so $\rho\tilde\rho=\rho^2$ and the roots of its eigenvalues are just the eigenvalues of $\rho$ .

Step 3. Combine: $\lambda_1-\lambda_2-\lambda_3-\lambda_4=\tfrac{1+3p}{4}-3\cdot\tfrac{1-p}{4}=\tfrac{6p-2}{4}=\tfrac{3p-1}{2}$ , so $C(\rho_W)=\max\{0,\tfrac{3p-1}{2}\}$ . Why: three equal subtracted roots make the bookkeeping trivial, and the result is positive exactly when the coherent weight beats the combined noise.

Step 4. Locate the threshold: $\tfrac{3p-1}{2}>0\iff p>\tfrac13$ , and below $p=\tfrac13$ the $\max$ clamps $C$ to zero. Why: this coincides with the point where the partial transpose of $\rho_W$ first develops a negative eigenvalue, so for two qubits the concurrence and PPT boundaries agree exactly. *Spot checks: at $p=1$ , $C=1$ (pure Bell); at $p=\tfrac12$ , $C=\tfrac{3/2-1}{2}=0.25$ ; at $p=\tfrac13$ , $C=0$ .*

Limiting cases

Concurrence takes especially transparent values on the states that anchor the theory, from the vacuum-like maximally mixed state up to the maximally entangled Bell states.

Case	Result	Meaning
Pure product state	$\|\psi\rangle=\|0\rangle\otimes\|+\rangle,\quad C=0$	Separable: rows of the amplitude matrix are proportional, $ad-bc=0$ , and $E_F=0$ .
Bell state	$\|\Phi^+\rangle=\tfrac{1}{\sqrt2}(\|00\rangle+\|11\rangle),\quad C=1$	Maximally entangled: $a=d=\tfrac{1}{\sqrt2}$ , $b=c=0$ , so $C=2\|ad\|=1$ and $E_F=1$ bit.
Werner state	$\rho_W=p\|\Phi^+\rangle\langle\Phi^+\|+\tfrac{1-p}{4}\mathbb{I},\quad C=\max\{0,\tfrac{3p-1}{2}\}$	Entangled iff $p>\tfrac13$ , the same threshold the PPT criterion gives for this family.
Maximally mixed	$\rho=\tfrac14\mathbb{I},\quad C=0$	All four $\lambda_i=\tfrac14$ , so $\lambda_1-\lambda_2-\lambda_3-\lambda_4=-\tfrac12<0$ and the $\max$ returns $0$ .
Rank-2 W marginal	$\rho_{AB}=\mathrm{Tr}_C\|W\rangle\langle W\|,\quad C_{AB}=\tfrac23$	A mixed state with $C=\tfrac23$ ; the two pairwise tangles saturate the CKW bound, so the 3-tangle vanishes.

Common trapRoots, not eigenvalues — and sort before you subtract

The single most common error in applying Wootters' formula is to use the eigenvalues of $\rho\tilde\rho$ directly. They are wrong twice over. First, they scale like probability squared, so subtracting them does not even have the right dimension for a concurrence; the correct $\lambda_i$ are their square roots, equal to the eigenvalues of the Hermitian $R=\sqrt{\sqrt\rho\,\tilde\rho\,\sqrt\rho}$ . Second, the roots must be sorted in decreasing order before you subtract the smaller three from the largest — feeding them in unsorted can turn an entangled state into an apparent separable one or vice versa. A useful sanity check: the four $\lambda_i$ are non-negative and, for a normalised $\rho$ , their squares sum to $\mathrm{Tr}(\rho\tilde\rho)\le1$ , with equality only for pure states.

Worked examples

1 · The Werner state at p = 1/2 (fully worked)

Take the two-qubit Werner state $\rho_W=\tfrac12|\Phi^+\rangle\langle\Phi^+|+\tfrac14\cdot\tfrac12\,\mathbb{I}$ , i.e. $p=\tfrac12$ . Compute its concurrence, its tangle, and its entanglement of formation, and say how much entanglement it actually carries.

Step 1. Read off the Bell-basis spectrum from the derivation. With $p=\tfrac12$ the sorted roots of $\rho\tilde\rho$ are $\lambda_1=\tfrac{1+3p}{4}=\tfrac{1+1.5}{4}=\tfrac{2.5}{4}=0.625$ and $\lambda_2=\lambda_3=\lambda_4=\tfrac{1-p}{4}=\tfrac{0.5}{4}=0.125$ .

Step 2. Apply Wootters' formula: $C=\max\{0,\;0.625-0.125-0.125-0.125\}=\max\{0,\;0.25\}=\mathbf{0.25}$ . This agrees with the closed form $C=\tfrac{3p-1}{2}=\tfrac{0.5}{2}=0.25$ , and since $p=\tfrac12>\tfrac13$ the state is on the entangled side of the threshold.

Step 3. The tangle is $\tau=C^2=0.25^2=\mathbf{0.0625}$ — the shareable amount of entanglement, small because the state is heavily diluted with noise.

Step 4. Convert to entanglement of formation. First $\sqrt{1-C^2}=\sqrt{1-0.0625}=\sqrt{0.9375}=0.9682$ , so the effective Schmidt weight is $\mu_+=\tfrac{1+0.9682}{2}=0.9841$ . Then $E_F=h(0.9841)=-0.9841\log_2 0.9841-0.0159\log_2 0.0159=\mathbf{0.118}$ bits. So $\rho_W$ at $p=\tfrac12$ has $C=0.25$ , tangle $0.0625$ , and $E_F\approx0.118$ bit — genuinely entangled, but carrying barely a tenth of a Bell pair's worth of entanglement.

2 · A partially entangled pure state, two ways

Consider the pure state $|\psi\rangle=0.8\,|00\rangle+0.6\,|11\rangle$ . Find its concurrence from the determinant, then confirm it and compute $E_F$ from the Schmidt weights.

Step 1. Identify the amplitudes: $a=0.8$ , $d=0.6$ , $b=c=0$ , and check normalisation $0.8^2+0.6^2=0.64+0.36=1$ . The determinant is $ad-bc=(0.8)(0.6)-0=0.48$ .

Step 2. Apply the determinant formula: $C=2|ad-bc|=2(0.48)=\mathbf{0.96}$ — close to maximal, as expected for amplitudes near equal.

Step 3. Cross-check via the Schmidt picture. The state is already in Schmidt form with squared coefficients $\mu_+=0.64$ and $\mu_-=0.36$ , so $C=2\sqrt{\mu_+\mu_-}=2\sqrt{0.64\times0.36}=2\sqrt{0.2304}=2(0.48)=0.96$ . The two routes agree.

Step 4. Compute the entanglement of formation. For a pure state $E_F$ is just the entanglement entropy $h(\mu_+)=h(0.64)=-0.64\log_2 0.64-0.36\log_2 0.36=\mathbf{0.943}$ bits, consistent with equation (5) since $\tfrac{1+\sqrt{1-0.96^2}}{2}=\tfrac{1+0.28}{2}=0.64$ . The state has $C=0.96$ and carries $0.943$ bit of entanglement — almost, but not quite, a full Bell pair.

Physics of the portrait

Concurrence is one of a cluster of results — no-cloning, teleportation, entanglement of formation, monogamy — that all trace back to the same person, and to a period in the 1980s and 1990s when the abstract theory of entanglement was being turned into a quantitative science.

Origins · 1 minWilliam Wootters and how this was discovered

William Wootters

born 1951 · concurrence and the entanglement of formation

William K. Wootters, an American physicist trained under John Wheeler at the University of Texas, is one of the founders of quantitative entanglement theory, and a striking number of the field's cornerstones carry his name. In 1982, with Wojciech Zurek, he proved the no-cloning theorem — that an unknown quantum state cannot be copied — in a single short paper. In 1993 he was one of the six authors of the quantum teleportation protocol. But his most technically remarkable result is this one: in 1997, with Scott Hill, he introduced the concurrence and solved the entanglement of formation for rank-2 states; then in 1998, in "Entanglement of formation of an arbitrary state of two qubits" (Phys. Rev. Lett. 80, 2245), he closed the general two-qubit case, showing that the forbidding convex-roof minimisation collapses to the eigenvalues of $\rho\tilde\rho$ — the only fully solved instance to this day. In 2000, with Valerie Coffman and Joydip Kundu, he proved the monogamy inequality that made the tangle the currency of shared entanglement. Working for decades at Williams College rather than a large research university, Wootters built much of the vocabulary — cloning, teleportation, concurrence, monogamy — with which quantum information now speaks.

History

1935

— Schrödinger names entanglement and calls it the characteristic trait of quantum mechanics

1982

— Wootters and Zurek prove the no-cloning theorem

1997

— Hill and Wootters introduce the concurrence and solve the entanglement of formation for rank-2 states

1998

— Wootters closes the general two-qubit entanglement of formation with the concurrence formula

2000

— Coffman, Kundu, and Wootters prove the CKW monogamy inequality for the tangle

Going further

Texts. Nielsen & Chuang, Quantum Computation and Quantum Information, §12.5 (entanglement of formation and concurrence) · Preskill, Lecture Notes on Quantum Computation, ch. 4–5 · Watrous, The Theory of Quantum Information, ch. 6 (entanglement measures) · Wilde, Quantum Information Theory, ch. 11 · Horodecki, Horodecki, Horodecki & Horodecki, "Quantum entanglement," Rev. Mod. Phys. 81, 865 (2009).

Primary sources. Wootters, "Entanglement of formation of an arbitrary state of two qubits," Phys. Rev. Lett. 80, 2245 (1998) · Hill & Wootters, "Entanglement of a pair of quantum bits," Phys. Rev. Lett. 78, 5022 (1997) · Coffman, Kundu & Wootters, "Distributed entanglement," Phys. Rev. A 61, 052306 (2000) · Wootters & Zurek, "A single quantum cannot be cloned," Nature 299, 802 (1982). Key numerical landmarks worth keeping in mind: (1) the double-flip gives $(\sigma_y\otimes\sigma_y)|00\rangle=-|11\rangle$ from $\sigma_y|0\rangle=i|1\rangle$ , and the overlap evaluates to $\langle\psi|\tilde\psi\rangle=-2(ad-bc)^{*}$ , giving $C=2|ad-bc|$ ; (2) the Werner spectrum is $\lambda_1=\tfrac{1+3p}{4}$ , $\lambda_{2,3,4}=\tfrac{1-p}{4}$ , hence $C=\max\{0,\tfrac{3p-1}{2}\}$ with threshold $p=\tfrac13$ matching the PPT boundary; (3) the pure-state concurrence satisfies $C=2\sqrt{\mu_+\mu_-}=\sqrt{1-(\mu_+-\mu_-)^2}$ in terms of Schmidt weights; (4) $E_F=h\big(\tfrac{1+\sqrt{1-C^2}}{2}\big)$ gives $1$ bit at $C=1$ and $0$ at $C=0$ , with $E_F(0.25)=0.118$ bit and $E_F(0.96)=0.943$ bit; (5) the W-state marginal concurrence is $C_{AB}=\tfrac23$ and the monogamy bound is saturated: $2(\tfrac23)^2=\tfrac89=C_{A(BC)}^2$ .

Prerequisites. Separable versus Entangled States · Schmidt Decomposition · Entanglement Entropy of Pure States · Bell States and EPR Pairs. Leads to. Entanglement of Formation · Entanglement Witnesses · PPT Criterion and the Peres–Horodecki Test.

Ask the tutor: Give me any two-qubit state — a Bell pair, a Werner state at your favourite noise level, or a raw density matrix — and I will spin-flip it, run Wootters' formula step by step, and hand you back the concurrence, tangle, and entanglement of formation.

Why is two qubits the only case where the entanglement of formation is solved?

The convex-roof minimisation is intractable in general because it searches over all pure-state decompositions of $\rho$ , an infinite curved set with no algebraic shortcut. Two qubits are rescued by a lucky coincidence of low dimension: the spin flip $\sigma_y\otimes\sigma_y$ is an antiunitary symmetry whose associated bilinear form can be simultaneously diagonalised in the "magic" Bell basis, and this lets Wootters construct the optimal ensemble explicitly. From $2\times3$ upward no comparable symmetry exists, the optimal decomposition is not known in closed form, and $E_F$ can only be bounded or computed numerically. Concurrence generalises (as the I-concurrence or via convex-roof extensions), but the clean formula does not — it is genuinely special to a pair of qubits.

What is the spin flip actually doing physically?

On a single qubit, $\sigma_y$ combined with complex conjugation is the time-reversal operator for a spin-1/2: it reverses the sense of the spin, sending a state on the Bloch sphere to its antipode. Applying it to both qubits and conjugating produces a "partner" state $|\tilde\psi\rangle$ . The key fact is how overlap with this partner behaves: a product state points maximally away from its own flipped partner, giving zero overlap, while a Bell state is (up to phase) invariant under the flip, giving unit overlap. So $|\langle\psi|\tilde\psi\rangle|$ literally measures how much a state resembles its own time reverse, and that resemblance turns out to be exactly its entanglement. The antiunitary character — the conjugation — is essential; a mere unitary could not distinguish entangled from product states this way.

How is concurrence related to negativity and to the PPT test?

They agree on the separable/entangled boundary for two qubits but not on the amount of entanglement. The PPT test asks only whether the partial transpose is positive, and negativity quantifies how negative it gets; concurrence instead quantifies the entanglement of formation, the cost to build the state. For the Werner family both vanish at the same threshold $p=\tfrac13$ , but away from it they differ numerically (at $p=\tfrac12$ the concurrence is $0.25$ while the negativity is $0.125$ ). Use concurrence when you want the entanglement of formation or the tangle; use negativity when you want a measure computable in any dimension, including where concurrence has no closed form.

Does C(ρ) = 0 really guarantee the state is separable?

For two qubits, yes — this is one of the clean facts that makes the case special. Concurrence vanishing is equivalent to the existence of a separable decomposition, which by the Peres–Horodecki theorem is in turn equivalent to the partial transpose being positive. All three conditions — zero concurrence, separability, positive partial transpose — coincide for $2\times2$ (and $2\times3$ ) systems. This equivalence breaks in higher dimensions, where bound-entangled states have positive partial transpose yet are entangled, and where a vanishing generalised concurrence no longer certifies separability. So "C = 0 means separable" is a theorem you may rely on for a pair of qubits, and only there.

What does the tangle add beyond the concurrence?

The tangle $\tau=C^2$ is the quantity that behaves additively when entanglement is shared, and that is what makes it, rather than the concurrence itself, the natural currency of monogamy. The Coffman–Kundu–Wootters inequality $C_{AB}^2+C_{AC}^2\le C_{A(BC)}^2$ holds for the squared concurrences — it is false for the concurrences themselves. Physically, a qubit has a fixed budget of tangle it can distribute among its partners, so entangling it strongly with one partner necessarily leaves less for others. The leftover, $C_{A(BC)}^2-C_{AB}^2-C_{AC}^2$ , defines the 3-tangle, a genuine measure of tripartite entanglement that is nonzero for GHZ states and zero for W states — the first hint of how entanglement structure grows richer beyond two parties.

Why must I take square roots of the eigenvalues of ρρ̃ rather than the eigenvalues themselves?

Because $\rho\tilde\rho$ is a product of two density-matrix-like operators, its eigenvalues scale like probability squared — they are dimensionally one power too high to be combined into a linear entanglement measure. The physically meaningful quantities are the singular-value-like $\lambda_i=\sqrt{\mathrm{eig}_i(\rho\tilde\rho)}$ , which are equal to the eigenvalues of the Hermitian, positive operator $R=\sqrt{\sqrt\rho\,\tilde\rho\,\sqrt\rho}$ and which scale linearly in probability. Only these combine correctly as $\lambda_1-\lambda_2-\lambda_3-\lambda_4$ . Skipping the square root is the single most frequent mistake in applying the formula, and it is usually caught by the sanity check that the correct $\lambda_i$ satisfy $\sum_i\lambda_i^2=\mathrm{Tr}(\rho\tilde\rho)\le1$ .