---
title: Distributed Monogamy of Entanglement
url: https://www.emergentmind.com/topics/distributed-monogamy-of-entanglement
type: topic
---

# Distributed Monogamy of Entanglement

Distributed monogamy of entanglement denotes the family of constraints that govern how bipartite entanglement can be allocated across multipartite quantum systems. In its standard form, monogamy compares the entanglement between a distinguished subsystem and the remainder with the sum of its pairwise entanglements to individual parties; in its distributed form, the notion is sharpened by recursive $n$-party inequalities, partition-dependent trade-offs, and, in recent work, operational limits on extracting EPR pairs from subsets of parties in a network [2201.00366]. A complementary review perspective emphasizes that distributed monogamy also includes partition-sensitive relations such as “entanglement polygon” or “triangle” inequalities, which track entanglement across many bipartitions rather than only node-to-node marginals [2512.21992].

## 1. Foundational formulations

The prototypical monogamy relation is the Coffman–Kundu–Wootters inequality for three qubits,
\[
C^2(A\mid BC)\ge C^2(A\mid B)+C^2(A\mid C),
\]
where $C$ is Wootters’ concurrence and the square $C^2$ is the tangle. In generalized CKW form, for a bipartite entanglement measure $E$ and an $n$-partite state $\rho_{A_1A_2\cdots A_n}$, one writes
\[
E(\rho_{A_1\mid A_2\cdots A_n})\ge \sum_{i=2}^n E(\rho_{A_1A_i}),
\]
or, more commonly in the qubit literature, one studies powered variants $E^\alpha$ for a suitable exponent [2201.00366].

Distributed monogamy extends this perspective in two directions. First, it includes recursive network inequalities such as
\[
E^\alpha(\rho_{A\mid B_1\cdots B_{n-1}})\ge \sum_{i=1}^{n-1}E^\alpha(\rho_{AB_i}),
\]
which are obtained by iterating a tripartite monogamy relation. Second, it includes partition-aware constraints: in a four-partite pure state $\lvert\psi\rangle_{ABCD}$ one may compare $E(A\mid BCD)$ not only with $E(A\mid B)+E(A\mid C)+E(A\mid D)$ but also with quantities such as $E(AB\mid CD)+E(AC\mid BD)+E(AD\mid BC)$, thereby tracking the full sharing structure across bipartitions [2512.21992].

A related unifying concept is the monogamy power. Guo defined the monogamy power $\alpha(E)$ of an entanglement measure as the infimum of exponents for which $E^\alpha$ satisfies the exact $n$-party monogamy inequality for all multipartite states, and similarly introduced the polygamy power for assisted entanglement. In that framework, once a given power makes the tripartite inequality hold, larger powers inherit monogamy for arbitrary $n$ [1703.01935].

## 2. Equality-based monogamy and the disentangling condition

A major conceptual shift came from the equality-based definition of monogamy. Instead of beginning with a fixed additive inequality, one says that a bipartite entanglement measure $E$ is monogamous if, for every tripartite state $\rho_{ABC}$,
\[
E(\rho_{A\mid BC})=E(\rho_{AB}) \quad\Longrightarrow\quad E(\rho_{AC})=0.
\]
This “disentangling condition” captures the idea that if all of $A$’s entanglement with $BC$ is already exhausted by $AB$, then no entanglement remains across $AC$ [1809.08532].

For continuous measures, this condition is equivalent to the existence of a dimension-dependent exponent $\alpha>0$ such that
\[
E^\alpha(\rho_{A\mid BC})\ge E^\alpha(\rho_{AB})+E^\alpha(\rho_{AC})
\]
for all tripartite states in bounded dimension. The exponent is generally non-universal and dimension-dependent, but the equivalence shows that equality-based monogamy and powered CKW-type inequalities are two formulations of the same underlying constraint [1809.08532].

The structural content of the disentangling condition is strongest for measures that, on pure bipartite states, are given by a strictly concave function of the reduced density matrix. If
\[
E(\lvert\psi\rangle_{AB})=f(\rho_A),\qquad \rho_A=\operatorname{Tr}_B\lvert\psi\rangle\langle\psi\rvert,
\]
with $f$ unitary-invariant and strictly concave, then for a pure tripartite state $\lvert\psi\rangle_{ABC}$ satisfying $E(\lvert\psi\rangle_{A\mid BC})=E(\rho_{AB})$, subsystem $B$ factorizes as $B_1\otimes B_2$ up to a local unitary, and one can write
\[
\lvert\psi\rangle_{ABC}=\lvert\phi\rangle_{AB_1}\otimes \lvert\chi\rangle_{B_2C}.
\]
Consequently, $\rho_{AC}$ is product and $E(\rho_{AC})=0$. The same strict-concavity hypothesis also yields monogamy for convex-roof extensions on mixed tripartite states: if $E_F(\rho_{A\mid BC})=E_F(\rho_{AB})$, then $\rho_{AC}$ is separable and $E_F(\rho_{AC})=0$ [1809.08532].

This immediately produces a distributed version. By iterating the tripartite inequality over $B_1,\dots,B_{n-1}$, one arrives at
\[
E^\alpha(\rho_{A\mid B_1\cdots B_{n-1}})\ge \sum_{i=1}^{n-1}E^\alpha(\rho_{AB_i}),
\]
which is precisely a distributed monogamy inequality over an $n$-party network [1809.08532].

## 3. Entanglement of formation and the rehabilitation of monogamy

The entanglement of formation (EOF) has been central to debates on monogamy because the unsquared quantity does not generally satisfy the naive three-party inequality
\[
E_F(A\!:\!BC)\ge E_F(A\!:\!B)+E_F(A\!:\!C).
\]
For pure tripartite states, Fanchini and collaborators expressed the failure of this linear inequality through the EOF-tangle
\[
\tau_A=S(\rho_A)-\delta^\leftarrow_{AB}-\delta^\leftarrow_{AC}
      =\tfrac12\bigl[\Delta^\leftarrow_{AB}+\Delta^\leftarrow_{AC}\bigr],
\]
where $\delta^\leftarrow$ is quantum discord and $\Delta^\leftarrow=J^\leftarrow-\delta^\leftarrow$ is the discrepancy between classical and quantum correlations. In this formulation, $\tau_A<0$ exactly when the total quantum-correlation share exceeds the total classical share. For the GHZ family, $\tau_A=S(\rho_A)>0$, whereas for the W family one has $\tau_A<0$ for all nontrivial parameters; the symmetric W state gives $\tau_A\approx-0.546$ [1110.1054].

The later resolution was that the apparent non-monogamy of EOF is specific to the unsquared quantity. Oliveira, de Oliveira, and collaborators showed that the square of EOF obeys exactly the same CKW-type inequality as squared concurrence:
\[
E_F^2(A:BC)\ge E_F^2(A:B)+E_F^2(A:C).
\]
They further extended this to arbitrary $N$-qubit states,
\[
E_F^2(A_1:A_2\ldots A_N)\ge \sum_{i=2}^N E_F^2(A_1:A_i),
\]
by combining generalized CKW monogamy for $C^2$ with the fact that $f(x)=E_F^2(\sqrt{x})$ is a strictly increasing convex function of $x=C^2$ on two-qubit states [1312.7287].

This result changed the status of EOF in the monogamy literature. The unsquared measure fails the naive additive inequality, but the squared measure satisfies the full CKW-style distributed constraint on multiqubit systems. At the same time, even linear EOF cannot be freely shared: for pure three-qubit states one numerically finds the strict upper bound
\[
E_F(1:2)+E_F(1:3)\le 1.18819,
\]
with the maximal value nearly attained by a W-type state that also maximizes the single-qubit entropy $S(\rho_1)=1$ [1312.7287].

EOF is also tightly linked to classical correlations through the Koashi–Winter identity,
\[
E_F(1:2)+J^\leftarrow(1:3)=S(\rho_1),
\]
which implies, for pure three-qubit states,
\[
E_F(1:2)+E_F(1:3)=D(1:2)+D(1:3)
=2S(\rho_1)-\bigl[J^\leftarrow(1:2)+J^\leftarrow(1:3)\bigr].
\]
States approaching the EOF-sum bound have nearly maximal $S(\rho_1)=1$ and reduced classical correlation [1312.7287].

A broader multipartite picture emerges from monogamy-like relations between EOF and discord. For pure tripartite states one has
\[
E_{A\mid B}+E_{A\mid C}=D_{A\mid B}+D_{A\mid C},
\]
and for arbitrary pure $N$-partite states there are cyclic equalities in which sums of EOF terms across selected bipartitions exactly match sums of discord terms after the conditional-entropy contributions cancel telescopically [1802.00471].

## 4. Strict concavity, non-monogamous monotones, and common misconceptions

The strict-concavity criterion also clarifies why some LOCC-monotones fail monogamy. Guo’s partial-norm family comprises the partial-norm of entanglement $E_2$, minimal partial-norm $E_{\min}$, reinforced minimal partial-norm $E'_{\min}$, and partial negativity $\hat N$. On pure states, their reduced-state functions are concave but not strictly concave; after convex-roof extension they remain bona fide LOCC-monotones, yet they are not monogamous [2212.06521].

The mechanism of failure is explicit. For $E_2$ one can construct a tripartite state with
\[
E_2(A\mid BC)=E_2(AB)
\]
while $\rho_{AC}$ has nonzero partial-transpose negativity, so $E_2(A\mid C)>0$. Analogous constructions show that $E_{\min}$ and $E'_{\min}$ are likewise non-monogamous for $\dim A>2$, and that $\hat N(A\mid BC)=\hat N(A\mid B)$ can hold simultaneously with $\hat N(A\mid C)>0$ [2212.06521].

This establishes a sharp distinction between concavity and strict concavity. With strict concavity, equality in the Jensen-type step forces all reduced states in an optimal decomposition to coincide, which in turn yields factorization and disentangling. With mere concavity, equality can be saturated by distinct spectra, and residual entanglement on the unused cut can survive. From the perspective of quantum networks, the paper states that any entanglement measure intended to control or bound the sharing of bipartite links must be informationally complete, i.e. must have a strictly concave reduced-state function, if one wants guaranteed monogamy [2212.06521].

A related misconception concerns quantum correlations beyond entanglement. Streltsov, Adesso, Piani, and Bruss proved that any measure of correlations that is monogamous on all states and satisfies positivity, local-unitary invariance, and no increase under attaching an uncorrelated ancilla must vanish on all separable states. In that sense, only entanglement measures can be strictly monogamous on all states. Their restricted positive example is the geometric discord, which satisfies the monogamy inequality on all pure states of three qubits but not on general mixed states [1112.3967].

## 5. Dimension dependence and weighted hierarchies

Distributed monogamy is not dimension-universal for all entanglement measures. Lancien, Di Plinio, Price, and Winter formalized general quantitative monogamy as the existence of a continuous trade-off function
\[
E_{A:BC}(\rho_{ABC})\ge f(E_{A:B}(\rho_{AB}),E_{A:C}(\rho_{AC}))
\]
with $f(x,y)\ge \max(x,y)$ and strict inequality on some region. They showed that every additive and suitably normalized entanglement measure that faithfully captures the fully antisymmetric state fails any nontrivial dimension-independent monogamy relation; this includes the regularized entanglement of formation $E_F^\infty$ and the regularized relative entropy of entanglement $E_R^\infty$. They also proved that non-regularized $E_F$ and $E_R$ generically violate any dimension-independent monogamy on random induced states in large dimension [1604.02189].

What survives is dimension-dependent monogamy. For any fixed finite dimensions $d_A,d_B,d_C$, explicit inequalities restore strict trade-offs. For EOF one such bound is
\[
E_F(\rho_{A:BC})\ge
\max\Bigl\{
E_F(\rho_{A:B})+\frac{c}{d_A d_C[\log d_{A,C}]^8[E_F(\rho_{A:C})]^8},
\,
E_F(\rho_{A:C})+\frac{c}{d_A d_B[\log d_{A,B}]^8[E_F(\rho_{A:B})]^8}
\Bigr\},
\]
for universal constant $c>0$ and $d_{A,B}=\min(d_A,d_B)$, $d_{A,C}=\min(d_A,d_C)$ [1604.02189].

Within multiqubit systems, a separate development produced hierarchies of tighter distributed inequalities. A unified framework identifies the smallest monogamy exponent $\alpha_c$ for several measures on qubit systems: $\alpha_c=2$ for concurrence, $\alpha_c=\sqrt2$ for entanglement of formation, and $\alpha_c=1$ for Tsallis–$q$, Rényi–$q$, and unified-$(q,s)$ entanglement in the parameter ranges stated in the original theorem [2007.04520]. Under ordering assumptions on pairwise entanglements, later works replaced the unit coefficients in CKW-type sums by larger weights, first $(\alpha/2)^k$ and $(\alpha/\sqrt2)^k$ for concurrence and EOF, and then the strictly larger factors $(2^{\alpha/2-1})^k$ and $(2^{\alpha/\sqrt2}-1)^k$ [1702.03405]. Further refinements introduced parameter-dependent prefactors, Hamming-weight constructions, and additional cross-terms for concurrence, negativity, CREN, and general powered entanglement measures, always with the stated goal of obtaining tighter lower bounds than earlier CKW-type estimates in multiqubit systems [1803.11355].

These hierarchies do not replace the equality-based picture; rather, they quantify how much slack remains once monogamy is already known to hold for a powered measure. They are therefore a quantitative refinement of distributed monogamy, not an alternative definition.

## 6. Multipartite geometry, operational limits, and recent extensions

Distributed monogamy also appears in geometric constraints involving genuine multipartite entanglement. For an $n$-qubit pure state and a nodal qubit $A_1$, the monogamy score
\[
\delta_1^{\mathcal Q}(\lvert\Psi\rangle)
=
\mathcal Q(\rho_{A_1:\mathrm{rest}})
-\sum_{k=2}^n \mathcal Q(\rho_{A_1A_k})
\]
is bounded above by a function of the generalized geometric measure $\mathcal E(\lvert\Psi\rangle)$ whenever the largest Schmidt coefficient arises from a $1\mid \mathrm{rest}$ bipartition. For entropic measures such as EOF, discord, squashed entanglement, work-deficit, and relative-entropy entanglement, the bound is
\[
\delta^{\mathcal Q}\le h(\mathcal E),
\]
while for squared concurrence it is
\[
\delta_{C^2}\le 4\,\mathcal E(1-\mathcal E).
\]
For pure three-qubit states this upper bound is universal [1505.01748].

A related three-qubit program uses genuine multipartite source entanglement $E_S$ as an upper bound for the sum of pairwise EOF squares. For GHZ-class states, the square of source entanglement serves as an upper bound for $E_{AB}^2+E_{AC}^2+E_{BC}^2$, with some exceptions for specific non-generic GHZ states. Numerical evidence supports the same inequality for W-class states. By contrast, accessible entanglement $E_a$ is generally too weak: the analogous bound with $E_a^2$ is almost always violated once any reduced pair carries entanglement [2409.00865].

The most explicitly operational formulation to date is based on extendibility. For integer $k\ge2$, $k$-extendibility quantifies how many copies of one subsystem can be simulated by the environment. A finer notion, fractional or $(p,q)$-extendibility, is invariant under tensor products and monotonic under local processing. Using it, a recent theorem establishes that for any state on $AB_1B_2\dots B_n$, the maximum average probability of extracting an EPR pair from a random subset of $k\le n/2$ systems among the $B_i$’s is exactly the fraction $k/n$ in the ideal case, and at most $k/n+\Delta(\epsilon)$ under $\epsilon$-approximate extraction [2607.08591]. In formula form,
\[
P_{\mathrm{ave}}
=
\frac{1}{\binom{n}{k}}\sum_{|S|=k}p_S
\le
\frac{k}{n}
\]
when the decoders recover a perfect maximally entangled pair from each chosen subset [2607.08591].

This operational distributed monogamy has a direct channel-theoretic consequence: any quantum erasure channel with erasure probability more than a half cannot simulate a less noisy erasure channel, even with asymptotically many uses of the noisier channel [2607.08591]. A plausible implication is that distributed monogamy is no longer only a property of static multipartite states; it also functions as an obstruction in asymptotic channel simulation.

A complementary qualitative dual is converse monogamy. For a tripartite pure state, if one reduced pair is entangled and non-distillable, then the other two reduced pairs are distillable; in the terminology of Chen and Hayashi, weak entanglement on one cut forces strong distillable entanglement on the remaining cuts [1109.1349]. Together with the equality-based, powered, geometric, and operational formulations, this suggests that distributed monogamy is best understood not as a single inequality but as a layered principle governing which multipartite entanglement configurations are kinematically, geometrically, and operationally compatible.

Source: https://www.emergentmind.com/topics/distributed-monogamy-of-entanglement