Bit Threads, Multiflow, and Optimization Duality

On a static Riemannian bulk slice in Einstein gravity, the RT minimum is dual to a maximum flux problem. A bit-thread flow is a divergenceless vector field whose pointwise norm is bounded by $1/(4G_N)$; its maximal flux through $A$ equals the area of the least homologous cut divided by $4G_N$. The equivalence is a continuum max-flow/min-cut theorem, not an independent entropy prescription. Multiflow results require several antisymmetric flows sharing a common capacity constraint. Time dependence and higher-derivative functionals lie outside the original theorem unless a new convex formulation is proved.

Required background. RT supplies the static min-cut problem, and convex duality supplies strong duality and constraint qualifications.

Helpful background. Field variations and boundary terms controls flux constraints, and wedge nesting supplies the geometric applications.

The static max-flow/min-cut theorem

Let $(\Sigma,g_{ij})$ be a regulated static bulk slice and $A\subset\partial\Sigma$. A feasible flow $v^i$ obeys

$$\nabla_i v^i=0, \qquad |v|\equiv\sqrt{g_{ij}v^iv^j}\leq\frac{1}{4G_N}.$$

Its flux through $A$ is

$$\Phi_A(v)=\int_A\sqrt{h_A}\,n_i v^i.$$

For any cut $m$ homologous to $A$, there is a region $R$ with $\partial R=A\cup(-m)$. The divergence theorem gives

$$\Phi_A(v)=\int_m\sqrt{h_m}\,n_i v^i \leq\frac{\operatorname{Area}(m)}{4G_N}.$$

Taking the supremum over flows and then the infimum over cuts proves weak duality. Under the regularity and compactness assumptions of the Riemannian max-flow/min-cut theorem, a calibrating flow exists and saturates the minimal cut:

$$\max_{v:\,\nabla\cdot v=0,\ |v|\leq1/(4G_N)} \Phi_A(v) =\min_{m\sim A}\frac{\operatorname{Area}(m)}{4G_N}.$$

Freedman and Headrick established this equivalence and its holographic interpretation (Freedman and Headrick 2017, §§2–3). The flow is generally nonunique even when the minimal surface is unique; only the maximal flux is fixed.

An explicit interval flow in AdS3

Take the hyperbolic plane $H^2$, the $t=0$ slice of Poincaré $\mathrm{AdS}_3$, and an interval $A$. Introduce Fermi coordinates $(\rho,\sigma)$ about its RT geodesic, for which

$$ds^2=L^2\left(d\rho^2+\cosh^2\rho\,d\sigma^2\right).$$

The geodesic is $\rho=0$; regulated $\sigma$-length is the same dimensionless length that gives $\mathcal L_\gamma=L\int d\sigma$. Define

$$v=\frac{1}{4G_NL\cosh\rho}\,\partial_\rho.$$

Because $\sqrt g=L^2\cosh\rho$,

$$\nabla\cdot v =\frac{1}{L^2\cosh\rho} \partial_\rho\!\left(L^2\cosh\rho\,v^\rho\right)=0.$$

Its norm is

$$|v|=\frac{1}{4G_N\cosh\rho} \leq\frac{1}{4G_N},$$

with equality exactly at the geodesic. The flux through that cut is therefore

$$\Phi_A(v)=\int_{\rho=0}ds\,|v| =\frac{\mathcal L_\gamma}{4G_N} =\frac{c}{3}\log\!\left(\frac{\ell}{\epsilon}\right).$$

Weak duality already bounded every flow by this number, so the explicit flow proves maximality and reproduces the RT interval entropy. Thread integral curves fan out away from the bottleneck; their density is not a literal count of microscopic Bell pairs.

Multiflow capacity is extra structure

Partition the boundary into regions $A_1,\ldots,A_N$. A multiflow consists of

$$v_{ij}=-v_{ji}, \qquad \nabla\cdot v_{ij}=0,$$

where $v_{ij}$ has boundary support only on $A_i\cup A_j$. A common pointwise capacity condition is

$$\sum_{i<j}|v_{ij}|\leq\frac{1}{4G_N}.$$

The total flow leaving $A_i$ is $v_i=\sum_jv_{ij}$. A multiflow theorem can choose the $v_{ij}$ so that several region fluxes simultaneously saturate their cuts. This simultaneous feasibility is stronger than constructing a separate max flow for each region; independently optimal flows may overfill the same bottleneck when superposed. Multiflow existence underwrites thread proofs of monogamy-type inequalities (Cui et al. 2019, §§2–3).

Where the original proof stops

Time dependence. HRT is a spacetime extremization, while the flow theorem lives on one Riemannian slice. Choosing an arbitrary slice reproduces the slicing failure exhibited on the HRT page. Covariant thread proposals introduce causal or form-valued constraints, but each requires its own duality theorem and hypotheses.

Higher derivatives. If entropy is $\int\sqrt h\,F(R,K,\ldots)$, a simple pointwise norm bound on a vector does not generally produce that cut functional. Extrinsic-curvature dependence can destroy the convex, local capacity structure. Reusing $|v|\leq1/(4G_N)$ would return area and therefore fail the known higher-curvature correction.

Quantum corrections. Bulk entropy is nonlocal in the cut and is not represented by the original local capacity bound. “Quantum bit threads” are proposed extensions, not consequences of the classical theorem.

These are direct adversarial controls: if the problem is time dependent, higher derivative, or quantum, derive the new primal and dual optimization and prove strong duality before using thread language quantitatively. Otherwise the strongest licensed claim remains the static Einstein max-flow/min-cut equivalence.

Handoff

Tensor-network models provide a discrete comparison but not a derivation of continuum gravity. Entropy cones uses multiflows and graph cuts within their classical static domain.

The chapter overview contains the structure diagram and validity and failure diagram. They are embedded there once so that their shared chapter-level context is not repeated on every article.

For the chapter-wide comparison of assumptions, counterevidence, falsifiers, and claim ceilings, see the claim-domain table.

References

• Cui, S. X., Hayden, P., Headrick, M., Stoica, B., and Walter, M. (2019). “Bit threads and holographic monogamy.” Communications in Mathematical Physics 376, 609–648. DOI.
• Freedman, M., and Headrick, M. (2017). “Bit threads and holographic entanglement.” Communications in Mathematical Physics 352, 407–438. DOI.