---
title: Hurwitz Words in Sₙ Factorizations
url: https://www.emergentmind.com/topics/hurwitz-words
type: topic
---

# Hurwitz Words in Sₙ Factorizations

Searching arXiv for the specified paper to ground the article in the cited source.
Hurwitz words are reduced transposition-factorizations of the long cycle \(c=(1\,2\,\cdots\,n)\) in the symmetric group \(S_n\). For \(n\ge 2\), they form a combinatorial model in which algebraic factorizations, non-crossing structures, and graph-theoretic metrics interact tightly. In Khachatryan’s study of the Hurwitz graph, the vertex set consists of all reduced words for \(c\) in terms of transpositions, edges encode local Hurwitz shifts, and the resulting graph is analyzed through its radius, diameter bounds, center, and a geometric-tree correspondence; an extended version of the graph, based on longer pull moves, exhibits a different center and improved metric bounds [1508.02620].

## 1. Reduced transposition-factorizations of the long cycle

Let \(S_n\) denote the symmetric group on \(\{1,2,\dots,n\}\), and let
\[
c=(1\,2\,\cdots\,n)\in S_n.
\]
Write
\[
T=\{(i\,j)\mid 1\le i<j\le n\}
\]
for the set of transpositions in \(S_n\). The relevant length function is
\[
\ell(\pi)=\min\{\,k : \pi=t_1t_2\cdots t_k,\;\;t_i\in T\},
\]
and \(\ell(c)=n-1\).

A Hurwitz word of length \(n-1\) is a tuple
\[
w=(t_1,t_2,\dots,t_{n-1})\in T^{\,n-1}
\]
such that
\[
t_1t_2\cdots t_{n-1}=c
\]
and this factorization is reduced. The set of all such words is
\[
\mathcal F_n=\bigl\{\,w=(t_1,\dots,t_{n-1})\in T^{n-1}\mid t_1\cdots t_{n-1}=c\text{ and is reduced}\bigr\}.
\]

This formulation places Hurwitz words at the intersection of Coxeter-type length, factorization theory, and non-crossing combinatorics. The non-crossing-partition lattice \(\mathrm{NC}(n)\), via Kreweras, enters because its maximal chains correspond bijectively to reduced \(T\)-factorizations of \(c\). The summary given for the paper also records that there are \((n-1)!\) Hurwitz words in \(\mathcal F_n\) [1508.02620].

## 2. Local shifts and the Hurwitz graph

The Hurwitz graph \(H(S_n)\) is built from two elementary Hurwitz moves, or local shifts, indexed by \(1\le i\le n-2\). For a word \((t_1,\dots,t_{n-1})\), the right-shift \(R_i\) is
\[
R_i\bigl(t_1,\dots,t_i,t_{i+1},\dots,t_{n-1}\bigr)
=
\bigl(t_1,\dots,t_{i-1},\,t_{i+1},\,t_i^{\,t_{i+1}},\,t_{i+2},\dots,t_{n-1}\bigr),
\]
where \(t_i^{\,t_{i+1}}=t_{i+1}^{-1}t_it_{i+1}\). The left-shift \(L_i\) is
\[
L_i\bigl(t_1,\dots,t_i,t_{i+1},\dots,t_{n-1}\bigr)
=
\bigl(t_1,\dots,t_{i-1},\,t_{i+1}^{\,t_i},\,t_i,\,t_{i+2},\dots,t_{n-1}\bigr).
\]
These satisfy \(L_i=R_i^{-1}\).

The Hurwitz graph \(H(S_n)\) is the undirected graph with vertex set \(\mathcal F_n\), where two words are adjacent if they differ by exactly one application of some \(R_i\) or \(L_i\). The graph is known to be connected. In this model, adjacency is not mere permutation of letters: the conjugation built into each move preserves the total product while changing the reduced factorization in a controlled local way [1508.02620].

The significance of this definition is structural. It converts the space of reduced transposition-factorizations of a fixed permutation into a metric object, so questions about factorization equivalence become questions about graph distance, eccentricity, radius, diameter, and centrality.

## 3. Radius and diameter bounds of \(H(S_n)\)

A principal metric invariant of the Hurwitz graph is its radius. The theorem cited from Adin–Roichman states that
\[
\mathrm{rad}\bigl(H(S_n)\bigr)=\binom{n-1}{2}=\tfrac12(n-1)(n-2).
\]
Since any graph of radius \(R\) has diameter at most \(2R\), this gives
\[
\mathrm{diam}\bigl(H(S_n)\bigr)\le 2\binom{n-1}{2}=(n-1)(n-2).
\]

These formulas place the ordinary Hurwitz graph in a quadratic metric regime. The radius already determines that central vertices are at distance at most \(\binom{n-1}{2}\) from every other vertex, and the diameter bound shows that the full graph cannot be larger than twice that scale in graph distance. The paper’s focus is not merely on these global bounds, but on identifying concrete central vertices and understanding how their combinatorics is reflected in a geometric-tree model [1508.02620].

## 4. The center through the geometric-tree correspondence

A key device is the map
\[
I:\mathcal F_n\longrightarrow G_n,
\]
where \(G_n\) is the graph whose vertices are the non-crossing spanning trees on \(n\) points in convex position, with adjacency defined by differing by exactly one edge. For a Hurwitz word \(w=(t_1,\dots,t_{n-1})\), the image \(I(w)\) is the non-crossing tree whose edges are the transpositions \(t_i\). Goulden–Yong’s theorem identifies this with a bijection between \(\mathcal F_n\) and the set of linear extensions of a certain tree-order on edges.

A non-crossing geometric tree \(T\) is called a boundary caterpillar if its spine is a path of consecutive boundary vertices, that is, a cyclic interval \(\{i,i+1,\dots,i+k\}\). The paper proves that every word \(w\in\mathcal F_n\) such that \(I(w)\) is a boundary caterpillar satisfies
\[
\max_{u\in\mathcal F_n} d_{H(S_n)}(w,u)\le \binom{n-1}{2},
\]
and hence \(w\) lies in the center of \(H(S_n)\).

The paper further proves a broader existence statement: for any non-crossing tree \(T\in G_n\), there exists at least one \(w\in\mathcal F_n\) with \(I(w)=T\) such that \(w\) is central in \(H(S_n)\). Consequently, every tree in \(G_n\) can be realized by some central Hurwitz word, while boundary caterpillars provide an explicit large family of central vertices. The center can therefore be characterized as
\[
\mathrm{Center}\bigl(H(S_n)\bigr)
=
\bigl\{\,w\in\mathcal F_n : d_{H(S_n)}(w,u)\le\binom{n-1}{2}\;\forall\,u\in\mathcal F_n\bigr\}.
\]

This geometric-tree viewpoint clarifies that centrality is not confined to a single tree shape. A plausible implication is that the geometry of \(G_n\) organizes, but does not by itself uniquely determine, metric centrality in \(H(S_n)\), since distinct linear extensions of the same non-crossing tree can behave differently unless a specific central realization is selected [1508.02620].

## 5. The extended Hurwitz graph

Khachatryan introduces an extended graph by allowing a letter to be pulled across several positions in one move. For \(1\le i<j\le n-1\), define the extended left-pull and extended right-pull by
\[
L_{i,j}=L_{j-1}L_{j-2}\cdots L_i,
\qquad
R_{i,j}=R_iR_{i+1}\cdots R_{j-1}.
\]
Then \(L_{i,j}\) carries the \(i\)th letter to the \(j\)th slot, while \(R_{i,j}\) carries the \(j\)th letter to the \(i\)th slot, with conjugation along the way.

The extended Hurwitz graph \(\EuScript H(S_n)\) has the same vertex set \(\mathcal F_n\), but two words are adjacent if one can be obtained from the other by exactly one extended move \(L_{i,j}\) or \(R_{i,j}\). In this enlarged adjacency structure, every star-word—meaning a word whose tree \(I(w)\) is a star centered at some vertex \(i\)—has eccentricity exactly \(n-2\). Hence
\[
\mathrm{rad}\bigl(\EuScript H(S_n)\bigr)=n-2,
\]
and the set of all stars is the center of \(\EuScript H(S_n)\).

The same analysis yields a diameter bound. Because pulling changes the tree by at most one edge,
\[
\mathrm{diam}\bigl(\EuScript H(S_n)\bigr)\le 2(n-2).
\]
Using the result of Hernando–Hurtado–Marquez–Mora–Noy that
\[
\mathrm{diam}(G_n)\ge n-5,
\]
together with the fact that \(I:\EuScript H(S_n)\to G_n\) is surjective and sends adjacent words either to identical trees or to adjacent trees, one obtains
\[
n-5\le \mathrm{diam}\bigl(\EuScript H(S_n)\bigr)\le 2(n-2).
\]

The extended graph therefore compresses distances dramatically: the radius drops from \(\binom{n-1}{2}\) in \(H(S_n)\) to \(n-2\) in \(\EuScript H(S_n)\). At the same time, the center becomes sharply characterized by star-words, in contrast with the ordinary Hurwitz graph, where every non-crossing tree has at least one central realization [1508.02620].

## 6. Structural summary and proof pattern

The principal invariants discussed in the paper can be organized as follows.

| Object | Radius | Diameter information |
|---|---:|---:|
| \(H(S_n)\) | \(\binom{n-1}{2}\) | \(\le (n-1)(n-2)\) |
| \(\EuScript H(S_n)\) | \(n-2\) | \(n-5\le \mathrm{diam}\le 2(n-2)\) |

For centers, the ordinary and extended graphs differ qualitatively.

| Graph | Central vertices |
|---|---|
| \(H(S_n)\) | Contains every linear extension of any boundary caterpillar; for every \(T\in G_n\), there exists at least one central \(w\) with \(I(w)=T\) |
| \(\EuScript H(S_n)\) | Precisely the set of star-words |

The proofs proceed by induction on \(n\). In the ordinary Hurwitz graph, when \(I(w)\) has a leaf \(i\) next to \(i+1\), the transposition corresponding to that leaf can be pushed to the front or back of another word \(u\) in at most \(n-2\) moves, after which the argument descends to \(n-1\) points. In the extended graph, pull moves interact directly with the tree model, and the observation that a pull changes the tree by at most one edge links distances in \(\EuScript H(S_n)\) to distances in \(G_n\) [1508.02620].

From a broader perspective, Hurwitz words provide a concrete instance in which reduced factorizations, non-crossing spanning trees, and graph centers can be studied within a common framework. The paper’s central contribution is to show that the center of the ordinary Hurwitz graph is extensive and tree-sensitive, whereas the center of the extended graph is rigidly concentrated on stars.

Source: https://www.emergentmind.com/topics/hurwitz-words