---
title: Bridge Criterion in Knot Theory
url: https://www.emergentmind.com/topics/bridge-criterion
type: topic
---

# Bridge Criterion in Knot Theory

The Bridge Criterion is a local characterization of when a reduction along a bridge disk actually lowers bridge number. For a knot \(K\) in \(n\)-bridge position with respect to a Heegaard surface \(S\) in a closed orientable \(3\)-manifold \(M\), it states that reducing \(K\) along a bridge disk \(D\) yields an \((n-1)\)-bridge position if and only if there is a bridge disk \(E\) on the opposite side of \(S\) such that \((D,E)\) is a cancelling pair. In the formulation of Hayashi and Shimokawa, this gives a precise bridge-theoretic analogue of the stabilization criterion for Heegaard splittings, and in the unknot case it also controls the existence of a spanning disk compatible with the bridge structure [1606.07224].

## 1. Formal setup: Heegaard splittings, bridge splittings, and bridge disks

Let \(M\) be a closed, orientable \(3\)-manifold and \(S \subset M\) a closed, orientable surface that splits \(M\) into two compression bodies; in the setting under discussion, \(S\) is a Heegaard surface and the closures of the components of \(M \setminus S\) are handlebodies. A Heegaard splitting is a decomposition
\[
M \;=\; V \cup_S W,
\]
where \(V\) and \(W\) are handlebodies with common boundary \(\partial V=\partial W=S\) [1606.07224].

A bridge splitting of \((M,K)\) is such a Heegaard splitting together with a position of a knot \(K \subset M\) so that \(K\cap V\) and \(K\cap W\) are collections of \(n\) disjoint, boundary-parallel arcs in \(V\) and \(W\), respectively. One then says that \(K\) is in \(n\)-bridge position with respect to \(S\), and writes
\[
(M,K)\;=\;(V,V\cap K)\cup_S(W,W\cap K).
\]
Equivalently, if \(H_1\) and \(H_2\) denote the closures of the components of \(M\setminus S\), then
\[
K\cap H_i=\text{a union of }n\text{ boundary-parallel arcs in }H_i,\qquad i=1,2.
\]

In \(S^3\), when \(S\) is a \(2\)-sphere giving a decomposition \(S^3=B^3_1\cup_S B^3_2\), it is called a bridge sphere. A bridge is one of the arcs of \(K\cap V\) or \(K\cap W\). For a bridge arc \(a\subset K\cap V\), a bridge disk is a properly embedded disk
\[
D\subset V
\]
with boundary
\[
\partial D=a\cup b,
\]
where \(b\subset S\) is an arc with \(\partial a=\partial b\subset S\), and \(D\cap K=a\). Proper embedding means \(\partial D\subset S\cup K\) and \(\operatorname{int}(D)\) is disjoint from \(S\cup K\).

A pair \((D,E)\) of bridge disks with \(D\subset V\) and \(E\subset W\) is a cancelling pair if the disks lie on opposite sides of \(S\), \(D\cap E\) consists of a single point \(p\) on \(K\), and the intersection is otherwise minimal:
\[
\operatorname{int}(D)\cap \operatorname{int}(E)=\varnothing,\qquad \partial D\cap \partial E=\{p\}.
\]
When such a pair exists, the bridge splitting is said to be perturbed.

## 2. Reduction along a bridge disk and the criterion itself

Let \(D\subset V\) be a bridge disk with \(\partial D=a\cup b\), where \(a=D\cap K\) is a bridge arc in \(V\) and \(b=D\cap S\) is an arc in \(S\). A reduction of \(K\) along \(D\) is the isotopy supported in a neighborhood of \(D\) that slides \(a\) along \(D\) to \(b\) and then slightly past \(S\) into \(W\), producing a new arc \(a'\subset W\) with the same endpoints on \(S\). Formally, there is an ambient isotopy \(\{\phi_t\}_{t\in[0,1]}\) of \(M\), supported near \(D\), such that \(\phi_0=\mathrm{id}\), \(\phi_1(K)\) is identical to \(K\) outside a neighborhood of \(D\), and \(\phi_1(a)=a'\subset W\) with \(\partial a'=\partial a\subset S\). The swept region is a rectangle \(R\subset W\) with opposite sides \(a'\) and \(b\) and the other sides \(\alpha,\beta\subset S\) meeting \(b\) at the endpoints \(\partial a'\) [1606.07224].

The Bridge Criterion is the following theorem:
\[
\textbf{Theorem.}
\]
Let \(K\subset M\) be a knot in \(n\)-bridge position with respect to a Heegaard surface \(S\) in a \(3\)-manifold \(M\),
\[
(M,K)\;=\;(V,V\cap K)\cup_S(W,W\cap K),
\]
with \(V\) and \(W\) handlebodies and \(K\cap V\), \(K\cap W\) equal to \(n\) boundary-parallel arcs. Let \(D\subset V\) be a bridge disk. Then a reduction of \(K\) along \(D\) yields an \((n-1)\)-bridge position with respect to \(S\) if and only if there exists a bridge disk \(E\subset W\) such that \((D,E)\) is a cancelling pair.

This criterion identifies precisely when a local isotopy across \(S\) is a genuine bridge-lowering move. The existence of the dual disk \(E\) is neither auxiliary nor optional: it is exactly the condition that the reduction cancels a perturbation rather than merely changing the embedding. Conversely, if no such \(E\) exists, then reduction along \(D\) does not lower the bridge number, and it may fail to produce a bridge position at all.

## 3. Structure of the proof

The forward implication from a cancelling pair to a bridge-lowering reduction is the standard direction. If \((D,E)\) is a cancelling pair, the bridge splitting is perturbed, and sliding the bridge arc \(a=D\cap K\) along \(D\) across \(S\) cancels the local maximum–minimum pair introduced by the perturbation. The resulting position has exactly one fewer bridge on each side, hence an \((n-1)\)-bridge position [1606.07224].

The reverse implication is the substantive part. Assume that reduction along \(D\) yields an \((n-1)\)-bridge position. After the reduction, the arc \(a\) becomes \(a'\subset W\). Let \(c\subset W\) be the new bridge containing \(a'\), let \(C\subset W\) be a bridge disk for \(c\), and let \(R\subset W\) be the rectangle swept out during the reduction, with opposite sides \(a'\) and \(b\).

The proof first simplifies intersections between \(C\) and \(R\). Circle components and arc components whose endpoints lie on \(\alpha\), \(\beta\), or \(b\) are removed by local isotopies, leaving only essential arcs of the type connecting \(\alpha\) to \(\beta\). After minimizing intersections, one sets
\[
k = |R\cap C|-1,
\]
interpreted as the number of parallel “lower cap” arcs that remain. Reversing the reduction pushes \(a'\) back along \(R\cup D\) to \(\alpha\cup a\cup\beta\), carrying \(C\) along. This produces \(k\) parallel lower caps in \(V\) beneath \(D\), and yields
\[
C' = C\cap W,
\]
a \(k\)-punctured disk.

The next stage chooses a bridge disk \(E\) for a bridge adjacent to \(a\) so that \(|C'\cap E|\) is minimal. The punctures of \(C'\) are labeled \(1,2,\dots,k\) in nested order around \(b\) on \(S\). After minor adjustments, a combinatorial analysis shows that \(C'\cap E\) consists only of arcs, that any boundary-parallel arc in \(C'\) must have one endpoint on \(b\) and the other on \(d=C\cap S\), and that the label sequence along \(E\cap S\) has a constrained form beginning with \(b\), then \(1,2,\dots,k\), followed by a string of \(d\)'s, then \(k,\dots,1\), then \(b\), and so on.

An outermost-arc argument then forces every outermost arc to have label pair \((k,k)\). A second lemma proves that \(k=0\): otherwise an “innermost loop with a fat vertex” argument in the capped disk \(\overline{C}\) contradicts the non-boundary-parallel condition. Thus \(C'\) is an actual disk rather than a punctured disk.

At that point, a final minimization shows that all arcs of \(C'\cap E\) can be removed except the bridge \(e\) itself. If extra intersections remained, an outermost arc and cut-and-paste argument would produce a new bridge disk with fewer intersections, contradicting minimality. Hence
\[
D\cap E
\]
is exactly one point on \(K\), and \((D,E)\) is a cancelling pair.

The proof therefore converts the apparently global assumption that the reduction yields an \((n-1)\)-bridge position into the existence of an explicitly dual bridge disk on the opposite side. The criterion is thus genuinely biconditional.

## 4. The unknot case and spanning disks

For the unknot in \(S^3\), the criterion has a geometric application to spanning disks. If \(K\subset S^3\) is an unknot in \(n\)-bridge position with respect to a bridge sphere \(S\), and a reduction along a bridge disk \(D\) yields an \((n-1)\)-bridge position, then \(K\) bounds a disk
\[
F\subset S^3
\]
such that \(D\) is a subdisk of \(F\) and
\[
F\cap S
\]
is a union of \(n\) arcs [1606.07224].

The argument begins by applying the Bridge Criterion to obtain a bridge disk \(E\) on the opposite side such that \((D,E)\) is a cancelling pair. Let
\[
a=D\cap K,\qquad b=E\cap K,\qquad p=D\cap E\in K.
\]
One simultaneously isotopes \(a\) and \(b\) along their disks, fixing \(p\), to arcs \(a'\subset W\) and \(b'\subset V\), sweeping out regions \(\Delta_1\) and \(\Delta_2\) that contain \(D\) and \(E\), respectively.

In the reduced \((n-1)\)-bridge position, the unknot bounds a disk \(F_0\) with
\[
F_0\cap S
\]
equal to \(n-1\) arcs. Horizontal isotopies are then used so that \((\Delta_1\cup\Delta_2)\cap F_0\) consists only of \(a'\cup b'\) and no other arcs with endpoints in \(\operatorname{int}(a')\) or \(\operatorname{int}(b')\). One then reconstructs a spanning disk for the original knot by setting
\[
F = F_0\cup \Delta_1\cup \Delta_2.
\]
A small isotopy of the portion of \(F\) in \(V\) across \(S\), near \(\Delta_1\cup\Delta_2\), restores \(F\cap S\) to a collection of \(n\) arcs while keeping \(D\subset F\) as a subdisk.

This theorem sharpens earlier unknot statements recorded in the paper: rather than only asserting the existence of a spanning disk adapted to the bridge sphere, it identifies the exact circumstance in which a chosen bridge disk \(D\) sits inside such a spanning disk, namely when reduction along \(D\) cancels a perturbation.

## 5. Relation to perturbation, bridge number, and other bridge-theoretic criteria

The Bridge Criterion is directly analogous to the classical stabilization criterion for Heegaard splittings: a Heegaard splitting
\[
M=V\cup_S W
\]
is stabilized if there are compressing disks \(D\subset V\) and \(E\subset W\) with \(|D\cap E|=1\), and genus-lowering compression occurs precisely in that situation. In the bridge setting, the corresponding statement is that reduction along a bridge disk lowers \(n\) by one exactly when there is a dual bridge disk on the opposite side meeting it in a single point on \(K\) [1606.07224].

For a fixed surface \(S\), the integer \(n\) is the bridge number of \(K\) with respect to \(S\). The theorem therefore gives a recognition principle for genuine bridge-lowering moves: one searches for cancelling pairs \((D,E)\), and reductions along such disks cancel perturbations. The absence of any such \(E\) implies that no reduction along \(D\) can lower the bridge number. This is a local criterion, formulated in terms of a specific bridge disk and its opposite-side dual.

In related literature, the phrase “Bridge Criterion” also appears in broader senses. One use is for high-distance bridge surfaces: large distance in the curve complex is presented as a criterion forcing strong geometric and topological restrictions, including exclusion of low-complexity essential or meridional surfaces and restrictions on competing bridge surfaces [1203.4294]. Another use is the rectangle condition on a bridge sphere: if the pants decompositions arising from maximal collections of essential cut disks satisfy the rectangle condition, then the Hempel distance satisfies
\[
d(\Sigma)\ge 2,
\]
so the bridge sphere is strongly irreducible and not perturbed [1404.7165]. A further diagrammatic criterion is the \(2\)-connected condition of Jang, Kobayashi, Ozawa, and Takao: if every graph \(\mathcal{G}_{i,j,\varepsilon}\) in a bridge diagram is \(2\)-connected, then the bridge sphere is strongly irreducible and hence unperturbed, a criterion used to construct links with locally minimal but not globally minimal bridge positions [2412.04621].

These usages are related but not identical. The criterion of Hayashi and Shimokawa decides when a specified reduction lowers bridge number; the distance, rectangle, and \(2\)-connected criteria instead obstruct perturbation or weak reducibility of the bridge sphere as a whole. Together they show that bridge theory contains both local cancellation criteria and global complexity criteria.

## 6. Edge cases, examples, and significance

Several constraints and edge cases are built into the theorem. The statements hold in any closed orientable \(3\)-manifold equipped with a Heegaard splitting; in \(S^3\), the surface is often a bridge sphere, but the criterion is not restricted to that case [1606.07224]. When \(n=1\), a cancelling pair exists if and only if reduction across \(D\) yields a \(0\)-bridge position, meaning that \(K\) lies entirely on \(S\); the paper notes that this is a degenerate case in conventions where bridge number is usually taken to be at least \(1\).

The existence of a cancelling pair is not unique. Different bridge disks \(E\subset W\) may serve as the dual disk to a given \(D\subset V\). Conversely, absence of any such \(E\) means that reduction along \(D\) cannot lower bridge number. The paper also records a stronger warning: a reduction might fail to produce a bridge position at all.

A standard example where the criterion applies starts with a knot in \((n-1)\)-bridge position relative to a bridge sphere \(S\), then perturbs it locally near \(S\) by pushing a small subarc across the sphere, thereby creating an additional local maximum and local minimum. In the resulting \(n\)-bridge position there are bridge disks \(D\) and \(E\) on opposite sides of \(S\) whose boundaries meet at the unique point where the pushed arc crosses \(S\), so \((D,E)\) is a cancelling pair. Reducing along either disk removes the perturbation and returns the knot to \((n-1)\)-bridge position.

By contrast, if a knot is already in a genuinely minimal \(n\)-bridge position with no perturbations, and one chooses a bridge disk \(D\) whose boundary arc on \(S\) is “deep” among nested arcs and not adjacent to any opposite-side bridge disk meeting it in a single point on \(K\), the swept rectangle \(R\) forces any bridge disk for the new arc to intersect \(R\) in non-removable type-\((6)\) arcs. Reversing the move produces \(k>0\) lower caps beneath \(D\), no cancelling pair exists, and the reduction does not yield an \((n-1)\)-bridge position. In such a case the resulting embedding may even fail to be a bridge position.

The significance of the theorem is therefore precise rather than heuristic. It does not merely say that perturbations can sometimes be removed by reductions; it characterizes exactly which reductions remove them. For the unknot, it also translates that local bridge-theoretic condition into the existence of a spanning disk containing the chosen bridge disk and meeting the bridge sphere in exactly \(n\) arcs. In this sense, the Bridge Criterion connects local isotopy, perturbation theory, and the geometry of spanning surfaces in a single equivalence.

Source: https://www.emergentmind.com/topics/bridge-criterion