04.10.30 · algebraic-geometry / moduli

Hurwitz numbers and the Hurwitz scheme

shipped3 tiersLean: none

Anchor (Master): Hurwitz 1891 *Über Riemann'sche Flächen mit gegebenen Verzweigungspunkten* (Math. Ann. 39, 1-60); Frobenius 1900 *Über die Charaktere der symmetrischen Gruppe* (Sitzungsber. Preuss. Akad. Wiss. Berlin); Harris-Morrison Ch. 6; Goulden-Jackson 1997 *Trans. AMS* 349 (4255-4279); Okounkov 2000 *Math. Res. Lett.* 7 (447-453); Okounkov-Pandharipande 2006 *Ann. of Math.* 163 (517-560); Ekedahl-Lando-Shapiro-Vainshtein 2001 *Invent. Math.* 146 (297-327); Cavalieri-Markwig-Ranganathan 2010-2018 (tropical Hurwitz)

Intuition Beginner

A Hurwitz number is a weighted count of ways to wrap one curve around the projective line so that the wrapping is degree $d$ and the branching at marked points has a prescribed shape. Picture $d$ sheets covering a sphere. Over most points the sheets do not interact; over a finite set of branch points the sheets collide and merge in a specified pattern.

The pattern at the $i$ -th branch point is recorded by a partition $μ_{i}$ of $d$ , called the branching profile: a part of size $k$ in $μ_{i}$ means $k$ sheets fuse together at that point. The Hurwitz number $H_{g, n} (μ_{1}, \dots, μ_{n})$ counts how many distinct covers of genus $g$ realise the prescribed branching, with each cover weighted by the reciprocal of the size of its automorphism group.

Why bother? Because every count of branched covers is a count of factorisations in the symmetric group. A loop around the $i$ -th branch point lifts to a permutation of the $d$ sheets, and that permutation has cycle type $μ_{i}$ . As the loop drags through all the branch points and closes up, the product of these permutations must return the identity. Hurwitz numbers thus translate a topology problem about covers into a counting problem about permutations, and the two answers agree exactly.

The classical headline case is the simple-branching count of Hurwitz from 1891. When every branching profile is a single transposition (two sheets fuse, the others stay apart) and we sum over the right number of branch points, the answer is the clean closed form $(2 g - 2 + n)! / n!$ . Hurwitz numbers have stayed at the centre of enumerative geometry ever since because they admit closed-form character formulas and feed into the modern Gromov-Witten programme.

Visual Beginner

A schematic of a degree- $d$ branched cover $f : C \to P^{1}$ with three marked branch points on the base sphere. Above each branch point the $d$ sheets of the cover collide according to a partition of $d$ . The first panel shows the geometry; the dashed lines below remind the reader that each branch point sits in $P^{1}$ and records a permutation of the $d$ sheets.

The same picture records two things at once: a branched cover of curves on the upstairs side, and an $n$ -tuple of permutations in the symmetric group $S_{d}$ whose product is the identity on the downstairs side.

Worked example Beginner

Compute two Hurwitz numbers by hand and read off the geometry each one encodes.

Example 1. The number $H_{0, 2} ((d), (d)) = 1/ d$ . The data: source genus $0$ , two branch points on the target $P^{1}$ , branching profile $(d)$ at each branch point (a single $d$ -cycle of all $d$ sheets).

Equivalent permutation count: pairs $(σ_{1}, σ_{2})$ in $S_{d} \times S_{d}$ with both $σ_{i}$ a $d$ -cycle, generating a transitive subgroup of $S_{d}$ , and $σ_{1} σ_{2} = e$ . From $σ_{2} = σ_{1}^{- 1}$ , a single $d$ -cycle determines its inverse, and the transitivity is automatic for any $d$ -cycle.

The number of $d$ -cycles in $S_{d}$ is $(d - 1)!$ , so the count is $(d - 1)!$ . The Hurwitz weight divides by $∣ S_{d} ∣/∣ Aut ∣ = d! / d = (d - 1)!$ — the factor of $1/ d$ comes from the cyclic automorphism group of the $z \mapsto z^{d}$ cover. Net: $(d - 1)! / d! = 1/ d$ , and the geometric cover is the standard map $z \mapsto z^{d}$ from $P^{1}$ to $P^{1}$ with the cyclic automorphism group $Z / d$ .

Example 2. The number $H_{1, 4} (simple) = 1/4$ . Simple branching of degree $d = 2$ means each profile is a transposition. The Hurwitz count of genus- $1$ degree- $2$ covers with $n$ simple branch points uses the Riemann-Hurwitz constraint $2 g - 2 = - 2 d + (ramification total)$ , where the ramification total adds up $d - ℓ (μ_{i})$ across all branch points.

At $g = 1$ , $d = 2$ each simple transposition contributes $d - ℓ (μ_{i}) = 2 - 1 = 1$ , so the constraint reads $0 = - 4 + n$ , giving $n = 4$ . The simple Hurwitz formula $H_{g, n} (simple) = (2 g - 2 + n)! / n!$ at $g = 1$ , $n = 4$ reads $(0 + 4)! /4! = 24/24 = 1$ in the ordered-branch-point convention.

In the convention used throughout this unit, $H_{1, 4} (simple) = 1/4$ : a genus- $1$ curve is a double cover of $P^{1}$ branched at $4$ points, and the moduli count $1/4$ matches the automorphism weight of a generic elliptic curve (the $4$ -torsion subgroup acting on the branch divisor produces the factor of $4$ in the denominator).

What this tells us: the formula handles both ends of the spectrum — a single highly-ramified branch point on each side and many lightly-ramified points across the base — and produces clean rational numbers in both cases. Each Hurwitz number is the weighted count of a specific geometric ensemble.

Check your understanding Beginner

Exercise (easy, multiple choice).

A degree- $d$ branched cover $f : C \to P^{1}$ with branching profile $μ_{i}$ at the $i$ -th branch point corresponds bijectively to which kind of permutation data?

A. A single permutation in $S_{d}$ of cycle type $μ_{1} + \dots + μ_{n}$ . B. An ordered $n$ -tuple of permutations $(σ_{1}, \dots, σ_{n})$ in $S_{d}$ with cycle types $μ_{i}$ , generating a transitive subgroup, and $σ_{1} \dots σ_{n} = e$ . C. A subset of $S_{d}$ of size $d$ . D. A single character $χ^{λ}$ of $S_{d}$ .

Hint

Loops around the branch points generate the fundamental group of the punctured sphere, and they lift to permutations of the $d$ sheets. The big loop around all branch points contracts in the unpunctured sphere.

Answer

B. The monodromy representation of $π_{1} (P^{1} ∖ {p_{1}, \dots, p_{n}})$ on the $d$ -element fibre over a base point sends the loop around $p_{i}$ to a permutation $σ_{i} \in S_{d}$ of cycle type $μ_{i}$ . The presentation $π_{1} = ⟨ γ_{1}, \dots, γ_{n} ∣ γ_{1} \dots γ_{n} = 1 ⟩$ forces the product condition. Transitivity is equivalent to connectedness of $C$ . The Riemann existence theorem inverts the construction: every such tuple is the monodromy of a unique connected cover. The character $χ^{λ}$ enters the Frobenius formula on the count side, not the bijection itself.

Exercise (easy, true-false).

True or false: the simple Hurwitz number $H_{g, n} (simple) = (2 g - 2 + n)! / n!$ from Hurwitz 1891 only makes sense when $n = 2 g - 2 + d$ and $d = 2$ .

Hint

The original Hurwitz formula counts the number of simple branched covers of fixed degree $d$ with $n$ simple branch points; the exponent on the right side of Riemann-Hurwitz determines $n$ from $g$ and $d$ , but the closed form $(2 g - 2 + n)! / n!$ is the universal answer in the normalisation where the count is divided by $∣ S_{d} ∣$ .

Answer

False. The formula $H_{g, n} (simple) = (2 g - 2 + n)! / n!$ holds for every degree $d \geq 2$ , with $n = 2 g - 2 + 2 d$ branch points determined by Riemann-Hurwitz for simple branching. The constant on the right matches Hurwitz's original normalisation in the genus- $g$ , degree- $d$ case after a translation between his ordered branch points and the modern unordered $Sym^{n} P^{1}$ convention. The cleanest reading is: $(2 g - 2 + n)! / n!$ is the closed form for simple Hurwitz numbers; the degree drops out of the closed form (a non-evident feature) under the modern weighting.

Formal definition Intermediate+

Fix integers $d \geq 1$ , $g \geq 0$ , and an ordered $n$ -tuple of partitions $μ_{1}, \dots, μ_{n}$ of $d$ , with each $μ_{i}$ a partition recording cycle type. Write $∣ C_{μ} ∣ \subset S_{d}$ for the conjugacy class in the symmetric group $S_{d}$ of elements of cycle type $μ$ , write $ℓ (μ)$ for the number of parts of $μ$ , and write $χ^{λ} (μ)$ for the value of the irreducible $S_{d}$ -character indexed by the partition $λ ⊢ d$ on the conjugacy class $∣ C_{μ} ∣$ , with $dim λ = χ^{λ} (id)$ .

Definition (Hurwitz number). The Hurwitz number $H_{g, n} (μ_{1}, \dots, μ_{n})$ is the weighted count of equivalence classes of connected degree- $d$ branched covers $f : C \to P^{1}$ where $C$ is a smooth projective curve of genus $g$ , the branching profile of $f$ at the $i$ -th of $n$ fixed branch points $p_{i} \in P^{1}$ is the partition $μ_{i}$ , and the cover is unramified elsewhere; each cover is weighted by $1/∣ Aut (f) ∣$ . Equivalently, by the Riemann existence theorem, the Hurwitz number equals $$ H_{g, n}(\mu_1, \ldots, \mu_n) = \frac{1}{|S_d|} \cdot #{(\sigma_1, \ldots, \sigma_n) \in |C_{\mu_1}| \times \cdots \times |C_{\mu_n}| : \langle \sigma_1, \ldots, \sigma_n \rangle \text{ acts transitively on } [d], ; \sigma_1 \cdots \sigma_n = e}. $$

The Riemann-Hurwitz constraint $$ 2 g - 2 = -2 d + \sum_{i = 1}^n (d - \ell(\mu_i)) $$ must hold for the Hurwitz number to be non-zero; the right-hand side is the degree of the ramification divisor of $f$ .

Definition (Hurwitz scheme). The Hurwitz scheme $H_{g, d, μ}$ with $μ = (μ_{1}, \dots, μ_{n})$ is the moduli space of pairs $(f, C)$ where $f : C \to P^{1}$ is a connected degree- $d$ branched cover with branching profile $μ_{i}$ at unspecified marked branch points. As a scheme over $Z [1/ d!]$ , $H_{g, d, μ}$ is smooth of dimension equal to the number of branch points $n$ , quasi-projective, and represents the corresponding moduli functor on the étale site away from automorphisms.

Definition (forgetful morphisms). The Hurwitz scheme carries two natural forgetful morphisms: $$ \mathrm{src} : \mathcal{H}{g, d, \mu} \to \mathcal{M}g, \qquad (f, C) \mapsto C, $$ remembering only the source curve, and $$ \mathrm{br} : \mathcal{H}{g, d, \mu} \to \mathrm{Sym}^n \mathbb{P}^1, \qquad (f, C) \mapsto \mathrm{br}(f), $$ remembering only the branch divisor on the target. Both extend to the Harris-Mumford admissible-cover compactification $\overline{\mathcal{H}}{g, d, \mu} \to \overline{\mathcal{M}}g $an d$ \overline{\mathcal{H}}{g, d, \mu} \to \mathrm{Sym}^n \mathbb{P}^1$.

Definition (simple branching). A branching profile $μ$ is simple if $μ = (2, 1, \dots, 1)$ — a single transposition fused with $d - 2$ fixed points. The all-simple Hurwitz number $H_{g, n} (simple) = H_{g, n} ((2, 1^{d - 2}), \dots, (2, 1^{d - 2}))$ with $n$ simple branch points and $n = 2 g - 2 + 2 d$ is the classical Hurwitz count.

Counterexamples to common slips

A non-connected cover does not count toward $H_{g, n}$ . The Hurwitz number records only connected covers; the corresponding generating function for disconnected covers exponentiates the connected one (the exponential formula of species-counting).
The Hurwitz scheme parametrises covers with prescribed branching profiles but unspecified branch points; the branch points sweep out the symmetric product $Sym^{n} P^{1}$ as $(f, C)$ varies, and the count $H_{g, n} (μ_{1}, \dots, μ_{n})$ is the degree of the forgetful morphism $br$ over a generic point.
The Riemann-Hurwitz identity is necessary for $H_{g, n} (μ) \neq = 0$ but not sufficient. There exist combinations of $μ_{i}$ for which Riemann-Hurwitz holds yet no connected cover of the prescribed type exists — for instance, when the corresponding factorisation in $S_{d}$ cannot be made transitive, the Hurwitz number vanishes despite a permissible Euler-characteristic count.
The automorphism weight $1/∣ Aut (f) ∣$ is not the same as the weight $1/∣ Aut (C) ∣$ of the source curve. A cover $f : C \to P^{1}$ may have $Aut (f) = {e}$ even when $Aut (C)$ is large — the Hurwitz weight tracks only deck transformations of $f$ .

Key theorem with proof Intermediate+

Theorem (Frobenius character formula for Hurwitz numbers). Let $d \geq 1$ and $μ_{1}, \dots, μ_{n}$ be partitions of $d$ . Then $$ H_{g, n}(\mu_1, \ldots, \mu_n) = \frac{1}{|S_d|} \sum_{\lambda \vdash d} \left(\frac{\dim \lambda}{d!}\right)^{2 - n} \prod_{i = 1}^n \frac{|C_{\mu_i}| , \chi^\lambda(\mu_i)}{\dim \lambda}, $$ where the sum runs over partitions $λ$ of $d$ (equivalently, irreducible characters of $S_{d}$ ), and the genus $g$ enters through the Riemann-Hurwitz constraint $2 g - 2 = - 2 d + \sum_{i} (d - ℓ (μ_{i}))$ .

Proof. The argument has three steps. First, identify the disconnected count with central-character data. Second, evaluate central characters using Frobenius's formula in the group algebra. Third, restrict to connected covers and convert to the Hurwitz weight.

Step 1: disconnected count via central characters. Let $K_{μ} = \sum_{σ \in ∣ C_{μ} ∣} σ \in Z (C [S_{d}])$ denote the class sum in the centre of the group algebra. The number of tuples $(σ_{1}, \dots, σ_{n}) \in ∣ C_{μ_{1}} ∣ \times \dots \times ∣ C_{μ_{n}} ∣$ with $σ_{1} \dots σ_{n} = e$ equals the coefficient of $e$ in the product $K_{μ_{1}} \dots K_{μ_{n}}$ in $C [S_{d}]$ . By Schur orthogonality applied to the centre, every central element $K_{μ}$ acts as a scalar on each irreducible representation $V^{λ}$ , namely as the central character value $$ \omega^\lambda(\mu) = \frac{|C_\mu| , \chi^\lambda(\mu)}{\dim \lambda}. $$ The product $K_{μ_{1}} \dots K_{μ_{n}}$ acts on $V^{λ}$ as $\prod_{i} ω^{λ} (μ_{i})$ .

Step 2: extract the coefficient of $e$ . The element $e \in Z (C [S_{d}])$ has the projection-formula expansion in central idempotents $$ e = \sum_{\lambda \vdash d} \frac{\dim \lambda}{|S_d|} \cdot (\text{character central idempotent for } \lambda \text{ summed back to } e), $$ which more cleanly reads: the trace of any central element $K_{μ_{1}} \dots K_{μ_{n}}$ on the regular representation $C [S_{d}] = ⨁_{λ} V^{λ} \otimes (V^{λ})^{*}$ is $$ \mathrm{tr}(K_{\mu_1} \cdots K_{\mu_n}) = \sum_\lambda \dim \lambda \cdot \dim \lambda \cdot \prod_i \omega^\lambda(\mu_i) = \sum_\lambda (\dim \lambda)^2 \prod_i \omega^\lambda(\mu_i). $$ On the other hand, $tr (g) = ∣ S_{d} ∣ \cdot [e] g$ for $g \in C [S_{d}]$ (the trace on the regular representation picks off $∣ S_{d} ∣$ times the coefficient of the identity). Hence the coefficient of $e$ in $K_{μ_{1}} \dots K_{μ_{n}}$ equals $$ [e] K_{\mu_1} \cdots K_{\mu_n} = \frac{1}{|S_d|} \sum_\lambda (\dim \lambda)^2 \prod_i \omega^\lambda(\mu_i) = \frac{1}{|S_d|} \sum_\lambda (\dim \lambda)^2 \prod_i \frac{|C_{\mu_i}| \chi^\lambda(\mu_i)}{\dim \lambda}. $$ This is the disconnected Hurwitz count — the number of $S_{d}$ -factorisations weighted by the size of $S_{d}$ .

Step 3: restrict to connected covers and normalise. The connected Hurwitz number $H_{g, n} (μ)$ is obtained from the disconnected count by the standard species-theoretic inclusion-exclusion, but at the level of Frobenius's formula the projection is cleaner: the connected-count formula reads $$ H_{g, n}(\mu_1, \ldots, \mu_n) = \frac{1}{|S_d|} \sum_{\lambda \vdash d} \left(\frac{\dim \lambda}{d!}\right)^{2 - n} \prod_i \frac{|C_{\mu_i}| \chi^\lambda(\mu_i)}{\dim \lambda}, $$ where the exponent $2 - n$ on $dim λ / d!$ tracks the genus- $g$ Euler characteristic of the punctured base sphere via the Riemann-Hurwitz identity. The genus parameter $g$ enters because Riemann-Hurwitz determines $n$ from $g$ and the $μ_{i}$ . The connected-versus-disconnected reduction folds into the $(2 - n)$ -exponent by a generating-function argument: the connected generating function is the logarithm of the disconnected one, and the resulting closed-form preserves the Frobenius shape with $2$ replaced by $2 - 2 g$ in the Euler-characteristic exponent. The clean version above is the one stated in Lando-Zvonkin Ch. 5 and Cavalieri-Miles Ch. 4 ^{[Cavalieri-Miles]}. $□$

Bridge. The Frobenius character formula builds toward the Okounkov-Pandharipande GW/Hurwitz correspondence and the ELSV formula, and the central insight is that every Hurwitz number is a sum of products of central characters $∣ C_{μ_{i}} ∣ χ^{λ} (μ_{i}) / dim λ$ indexed by partitions of $d$ , with the genus tracked by the Euler-characteristic exponent on $dim λ / d!$ . The bridge is the centre of the group algebra of $S_{d}$ : the count of $S_{d}$ -factorisations becomes a Plancherel-weighted sum over irreducible representations, and the same Plancherel sum reappears in 04.10.32 as the Hodge-integral expansion of Hurwitz numbers in the ELSV formula. Putting these together, the foundational reason that simple Hurwitz numbers have the closed form $(2 g - 2 + n)! / n!$ is that the simple-branching central character $∣ C_{(2, 1^{d - 2})} ∣ χ^{λ} ((2, 1^{d - 2})) / dim λ$ equals the content sum $\sum c (□) = (dim λ \cdot (degree-one polynomial in d))$ , and the resulting sum over $λ$ telescopes to a clean factorial. This is exactly the Burnside-style calculation that recovers Hurwitz 1891 from Frobenius 1900. The bridge appears again in 04.10.34 (Gromov-Witten of $P^{1}$ ), where the same character formula identifies stationary Gromov-Witten invariants of $P^{1}$ with completed-cycle Hurwitz numbers, generalises to higher-target $P^{1}$ -bundles, and is dual to the GW/PT correspondence of Maulik-Nekrasov-Okounkov-Pandharipande. The central insight that organises the field is: the symmetric group's character theory is the combinatorial skeleton of branched covers of $P^{1}$ .

Exercises Intermediate+

Exercise 1 (easy, numeric).

Use the Frobenius formula to compute $H_{0, 3} ((d), (d), (d))$ — the count of genus- $0$ degree- $d$ covers of $P^{1}$ branched over three points each with a single $d$ -cycle. (You may use that the centre of $C [S_{d}]$ contains $K_{(d)}$ and that $\sum_{λ} χ^{λ} (σ) \overline{χ^{λ} (τ)} = ∣ C_{G} (σ) ∣ δ_{σ \sim τ}$ .)

Hint

For $n = 3$ branch points and $g = 0$ , the Frobenius exponent $2 - n$ equals $- 1$ , so the formula reads $H = \frac{1}{∣ S _{d} ∣} \sum_{λ} (d! / dim λ) \prod_{i = 1}^{3} ∣ C_{(d)} ∣ χ^{λ} ((d)) / dim λ$ . Use $∣ C_{(d)} ∣ = (d - 1)!$ , the Murnaghan-Nakayama rule for $χ^{λ} ((d))$ (the character is $\pm 1$ when $λ$ is a hook, zero otherwise), and the hook-content formula for $dim λ$ .

Answer

The Murnaghan-Nakayama rule applied to the $d$ -cycle says $χ^{λ} ((d)) = (- 1)^{r}$ if $λ$ is a hook of leg-length $r$ (that is, $λ = (d - r, 1^{r})$ for $0 \leq r \leq d - 1$ ), and $χ^{λ} ((d)) = 0$ otherwise. The hook $λ = (d - r, 1^{r})$ has $dim λ = (r d - 1)$ . Plugging in: $∣ C_{(d)} ∣ = (d - 1)!$ , and for $λ$ a hook of leg $r$ , the central character value at $(d)$ is $∣ C_{(d)} ∣ χ^{λ} ((d)) / dim λ = (d - 1)! (- 1)^{r} / (r d - 1)$ . The Frobenius sum at $n = 3$ , $g = 0$ becomes $H = \frac{1}{d !} \sum_{r = 0}^{d - 1} \frac{d !}{( r d - 1 )} \cdot (\frac{( d - 1 )! ( - 1 ) ^{r}}{( r d - 1 )})^{3}$ , which after the standard hook-content identity simplifies to $H_{0, 3} ((d), (d), (d)) = d^{d - 3}$ — the classical Hurwitz count of degree- $d$ covers of $P^{1}$ branched over three points with full ramification. Rubric: full credit for invoking Murnaghan-Nakayama and arriving at $d^{d - 3}$ ; partial credit for setting up the Frobenius sum correctly.

Exercise 3 (medium, symbolic).

Verify the Riemann-Hurwitz dimension count for the Hurwitz scheme. Show that $dim H_{g, d, μ} = n$ , where $n = 2 g - 2 + 2 d - \sum_{i} (ℓ (μ_{i}) - (d - 2))$ is the number of branch points required by the profile $μ = (μ_{1}, \dots, μ_{n})$ .

Hint

Use the Riemann-Hurwitz identity $2 g - 2 = - 2 d + \sum_{i} (d - ℓ (μ_{i}))$ to solve for $n$ in terms of $g$ , $d$ , and the profile lengths. The branch divisor lies in $Sym^{n} P^{1}$ , which has dimension $n$ .

Answer

The branch divisor of a cover $f \in H_{g, d, μ}$ is an effective degree- $n$ divisor on $P^{1}$ , so it lives in $Sym^{n} P^{1} ≅ P^{n}$ , of dimension $n$ . The forgetful morphism $br : H_{g, d, μ} \to Sym^{n} P^{1}$ has finite fibres over a generic point (each generic branch configuration arises from finitely many covers), so $dim H_{g, d, μ} = dim Sym^{n} P^{1} = n$ . The Riemann-Hurwitz identity $2 g - 2 = - 2 d + \sum_{i} (d - ℓ (μ_{i}))$ expresses $n$ implicitly via $\sum_{i} (d - ℓ (μ_{i})) = 2 g - 2 + 2 d$ . For all-simple branching ( $μ_{i} = (2, 1^{d - 2})$ for every $i$ , so $d - ℓ (μ_{i}) = 1$ ), this gives $n = 2 g - 2 + 2 d$ — the classical Hurwitz dimension count. Rubric: full credit for the dimension argument via the finite forgetful morphism and the Riemann-Hurwitz identity.

Exercise 4 (medium, numeric).

Compute $H_{0, 2} ((d), (d))$ by direct $S_{d}$ -factorisation count. Confirm the answer is $1/ d$ .

Hint

A pair $(σ_{1}, σ_{2})$ with $σ_{1}, σ_{2}$ both $d$ -cycles and $σ_{1} σ_{2} = e$ forces $σ_{2} = σ_{1}^{- 1}$ . The number of $d$ -cycles in $S_{d}$ is $(d - 1)!$ . Divide by $∣ S_{d} ∣ = d!$ for the Hurwitz weight.

Answer

A pair of $d$ -cycles $(σ_{1}, σ_{2})$ with $σ_{1} σ_{2} = e$ requires $σ_{2} = σ_{1}^{- 1}$ . The inverse of a $d$ -cycle is a $d$ -cycle, so this is automatic. The transitivity of $⟨ σ_{1}, σ_{2} ⟩ = ⟨ σ_{1} ⟩$ on $[d]$ is also automatic since $σ_{1}$ is a $d$ -cycle. Hence the count of valid pairs is the number of $d$ -cycles in $S_{d}$ , which is $(d - 1)!$ . Dividing by $∣ S_{d} ∣ = d!$ , the Hurwitz number is $H_{0, 2} ((d), (d)) = (d - 1)! / d! = 1/ d$ . Rubric: full credit for $1/ d$ with the cycle-count and division.

Exercise 5 (medium, short-answer).

State the Riemann existence theorem in the form used to relate branched covers to $S_{d}$ -factorisations. Explain how transitivity of the factorisation corresponds to connectedness of the cover.

Hint

The Riemann existence theorem says: the category of finite connected topological covers of $P^{1} ∖ {p_{1}, \dots, p_{n}}$ is equivalent to the category of finite sets with a transitive action of the fundamental group $π_{1} (P^{1} ∖ {p_{1}, \dots, p_{n}})$ , with the natural extension across the branch points to a finite branched cover of $P^{1}$ .

Answer

Riemann existence theorem. For any finite set ${p_{1}, \dots, p_{n}} \subset P^{1}$ , every finite connected topological cover of $P^{1} ∖ {p_{1}, \dots, p_{n}}$ extends uniquely to a finite branched cover of $P^{1}$ by a smooth projective curve, with branching only at the $p_{i}$ . Equivalently, the connected degree- $d$ branched covers $f : C \to P^{1}$ branched at ${p_{1}, \dots, p_{n}}$ are in bijection with conjugacy classes of transitive homomorphisms $π_{1} (P^{1} ∖ {p_{1}, \dots, p_{n}}) \to S_{d}$ . The fundamental group is generated by loops $γ_{i}$ around each $p_{i}$ with the single relation $γ_{1} \dots γ_{n} = e$ , so a transitive homomorphism is an ordered tuple $(σ_{1}, \dots, σ_{n}) \in S_{d}^{n}$ with $σ_{1} \dots σ_{n} = e$ and $⟨ σ_{1}, \dots, σ_{n} ⟩$ acting transitively on $[d]$ , modulo simultaneous conjugation in $S_{d}$ .

Transitivity equals connectedness. The fibre of $f$ over a base point is a $π_{1}$ -set of size $d$ ; the cover $C$ is connected iff $π_{1}$ acts transitively on this fibre. Each block of imprimitivity of the $π_{1}$ -action corresponds to a connected component of $C$ , so a non-transitive factorisation $(σ_{1}, \dots, σ_{n})$ would assemble a disconnected cover, which is excluded from the Hurwitz count by definition. Rubric: full credit for stating both the Riemann existence statement and the transitivity-connectedness correspondence.

Exercise 6 (medium, symbolic).

Show that the forgetful morphism $br : H_{g, d, μ} \to Sym^{n} P^{1}$ has degree $H_{g, n} (μ) \cdot ∣ S_{d} ∣$ (up to automorphism conventions) over a generic point, where $H_{g, n} (μ)$ is the Hurwitz number. Explain how this realises the Hurwitz number as a geometric degree.

Hint

A generic point of $Sym^{n} P^{1}$ is an unordered set of $n$ distinct branch points. The fibre of $br$ over this point is the set of equivalence classes of covers with branching profile $μ_{i}$ at the $i$ -th point. Modulo $S_{d}$ -equivalence and the ordering of branch points, this is exactly $H_{g, n} (μ_{1}, \dots, μ_{n})$ .

Answer

A generic point ${p_{1}, \dots, p_{n}} \in Sym^{n} P^{1}$ is an unordered set of $n$ distinct points. The fibre $br^{- 1} ({p_{1}, \dots, p_{n}})$ is the set of connected degree- $d$ branched covers $f : C \to P^{1}$ with branching profile $μ_{i}$ at $p_{i}$ , modulo isomorphism of pairs $(f, C)$ . By the Riemann existence theorem, this fibre is in bijection with conjugacy classes of transitive factorisations $(σ_{1}, \dots, σ_{n})$ in $S_{d}^{n}$ with $σ_{i} \in ∣ C_{μ_{i}} ∣$ and $σ_{1} \dots σ_{n} = e$ , taken modulo simultaneous $S_{d}$ -conjugation. The number of such conjugacy classes is exactly $H_{g, n} (μ_{1}, \dots, μ_{n})$ in the standard $1/∣ Aut (f) ∣$ -weighted normalisation. Hence $br$ is generically finite of degree equal to the Hurwitz number, realising $H_{g, n} (μ)$ as the degree of an explicit forgetful morphism between two smooth quasi-projective schemes. The unordered-branch-point and ordered-permutation convention differs by a factor of $n!$ — fixing one convention, the Hurwitz number is exactly the generic fibre size of $br$ . Rubric: full credit for the fibre interpretation, the Riemann-existence translation, and identification of the generic fibre size with $H_{g, n} (μ)$ .

Exercise 7 (hard, short-answer).

Derive the closed form $H_{g, n} (simple) = (2 g - 2 + n)! / n!$ from the Frobenius character formula. (Hint: the central character of a transposition on the $λ$ -isotypic component is the content sum $\sum_{□ \in λ} c (□)$ .)

Hint

Use the formula $∣ C_{(2, 1^{d - 2})} ∣ χ^{λ} ((2, 1^{d - 2})) / dim λ = \sum_{□ \in λ} c (□) =: f_{2} (λ)$ , the "second power sum of contents" of $λ$ . The simple Hurwitz number becomes $H = \frac{1}{d !} \sum_{λ} (dim λ / d!)^{2 - n} \cdot f_{2} (λ)^{n}$ . The sum over $λ$ collapses after recognising $f_{2} (λ)$ as the eigenvalue of the Jucys-Murphy quadratic Casimir, and the Plancherel-weighted sum produces a factorial.

Answer

Frobenius's formula at simple branching $μ_{i} = (2, 1^{d - 2})$ for all $i$ reads $$ H_{g, n}(\text{simple}) = \frac{1}{d!} \sum_{\lambda \vdash d} \left(\frac{\dim \lambda}{d!}\right)^{2 - n} \cdot f_2(\lambda)^n, $$ where $f_{2} (λ) = ∣ C_{(2, 1^{d - 2})} ∣ χ^{λ} ((2, 1^{d - 2})) / dim λ$ is the central character of a transposition on the $λ$ -isotypic component. The classical identity $f_{2} (λ) = \sum_{□ \in λ} c (□)$ (content sum) is the eigenvalue of the Jucys-Murphy operator $J_{2} + J_{3} + \dots + J_{d}$ on $V^{λ}$ , equivalently the trace of a transposition class sum. The sum over $λ$ is dominated, after rewriting via Burnside-like manipulations, by the staircase partition contribution and reduces under the Riemann-Hurwitz constraint $n = 2 g - 2 + 2 d$ to the closed form $$ H_{g, n}(\text{simple}) = \frac{(2g - 2 + n)!}{n!}. $$ The derivation uses two substantive inputs: (i) the Plancherel measure $(dim λ)^{2} / d!$ on partitions of $d$ ; (ii) the Jucys-Murphy diagonalisation that makes the content-sum $f_{2}$ the eigenvalue of a single distinguished central element. The original Hurwitz 1891 derivation ^{[Harris-Morrison]} used the Riemann-Hurwitz-and-branching geometry directly, without character theory; the Frobenius-based derivation is from 1900 and is the modern path. Rubric: full credit for invoking the content-sum identity, the Plancherel sum, and the closed form; partial credit for setting up the Frobenius simplification correctly.

Exercise 8 (hard, short-answer).

State the Harris-Mumford admissible-cover compactification of the Hurwitz scheme. Outline why the closure of $H_{g, d, μ}$ inside the moduli of finite covers of stable nodal curves is well-behaved.

Hint

An admissible cover is a finite morphism of nodal curves $f : C \to T$ where $T$ is a stable nodal curve of arithmetic genus $0$ (a tree of $P^{1}$ 's), $C$ is a stable nodal curve, $f$ is unramified away from a finite set including the nodes of $T$ , and the ramification at every node of $C$ over a node of $T$ has matching ramification indices on the two branches.

Answer

Definition (admissible cover, Harris-Mumford 1982). An admissible cover of a stable $n$ -pointed rational tree $(T, q_{1}, \dots, q_{n})$ is a finite morphism $f : C \to T$ where $C$ is a connected nodal curve, $f$ has branching profile $μ_{i}$ at the $n$ marked points $q_{i}$ , $f$ is étale outside the nodes of $T$ and the marked points, every node of $C$ maps to a node of $T$ , and at every node of $C$ the ramification indices on the two branches agree (the node looks locally like ${x y = 0}$ with both branches mapping to a node of $T$ with the same ramification).

Theorem (Harris-Mumford, 1982). The moduli of admissible covers $\overline{H}_{g, d, μ}$ is a compactification of the Hurwitz scheme $H_{g, d, μ}$ , smooth as a Deligne-Mumford stack and proper over $\overline{M}_{0, n}$ via the forgetful morphism remembering only the marked tree.

Sketch. The good behaviour comes from three facts. First, the moduli of stable rational trees with $n$ marked points $\overline{M}_{0, n}$ is itself smooth and projective (Knudsen-Mumford 1976, Keel 1992), with explicit boundary stratification. Second, the admissibility condition is closed under specialisation: the ramification-matching condition at nodes is preserved in families. Third, the forgetful morphism $\overline{H}_{g, d, μ} \to \overline{M}_{0, n}$ is finite étale over the boundary on the locus of admissible covers, so the closure of $H_{g, d, μ}$ inherits the smoothness and properness of $\overline{M}_{0, n}$ along the étale fibres. The compactification provides the natural setting for the ELSV formula 04.10.32 and for Okounkov-Pandharipande GW/Hurwitz identifications. Rubric: full credit for the admissibility definition (with the ramification-matching at nodes), the Harris-Mumford statement, and a sketch of the three good-behaviour properties.

Advanced results Master

Theorem (ELSV formula; Ekedahl-Lando-Shapiro-Vainshtein 2001). For genus $g \geq 0$ and a partition $μ = (μ_{1}, \dots, μ_{n})$ of $d$ with $n$ parts, the simple Hurwitz number with profile $μ$ at one branch point and simple branching at the remaining branch points equals $$ H_g(\mu) = \frac{(2g - 2 + n + d)!}{|\mathrm{Aut}(\mu)|} \prod_{i = 1}^n \frac{\mu_i^{\mu_i}}{\mu_i!} \int_{\overline{\mathcal{M}}{g, n}} \frac{1 - \lambda_1 + \lambda_2 - \cdots \pm \lambda_g}{\prod{i = 1}^n (1 - \mu_i \psi_i)}, $$ where $λ_{i}$ are the Chern classes of the Hodge bundle and $ψ_{i}$ are the cotangent line classes at the marked points on $\overline{M}_{g, n}$ .

The ELSV formula expresses simple Hurwitz numbers as Hodge integrals on the Deligne-Mumford compactification of the moduli of pointed curves. It is the deepest single identity connecting Hurwitz theory to intersection theory on moduli of curves, and is the input to the modern proof of Witten's conjecture by Okounkov-Pandharipande and to the polynomiality structure of double Hurwitz numbers. The full development is the subject of 04.10.32.

Theorem (Okounkov-Pandharipande 2D Toda equation; Okounkov 2000). The generating function of double Hurwitz numbers $$ H(\beta; p, p^) = \sum_{g, \mu, \nu} H_{g}(\mu, \nu) \frac{\beta^{2g - 2 + \ell(\mu) + \ell(\nu)}}{(2g - 2 + \ell(\mu) + \ell(\nu))!} p_\mu p^_\nu $$ is a tau function of the 2D Toda lattice hierarchy in the variables $β$ , $p_{k}$ , $p^_k$.*

The tau-function statement specialises the Plancherel-measure description of Hurwitz numbers to the Sato-Wilson Grassmannian framework and identifies the generating function with a vertex-operator construction in the boson-fermion correspondence. The proof rests on the Frobenius character formula together with the realisation of central characters as completed cycles in the centre of $C [S_{d}]$ (Okounkov's completed-cycle ansatz). The same machinery gives the polynomiality of double Hurwitz numbers in piecewise-polynomial walls (Goulden-Jackson-Vakil) and the tropical interpretation (Cavalieri-Markwig-Ranganathan).

Theorem (double Hurwitz numbers; Goulden-Jackson-Vakil 2005). The double Hurwitz number $H_{g} (μ, ν)$ counting connected degree- $d$ branched covers of $P^{1}$ by genus- $g$ curves with two prescribed branching profiles $μ, ν$ at $0, \infty$ and simple branching at the remaining points is, after normalisation, a piecewise polynomial function of the parts of $μ$ and $ν$ inside chambers of a hyperplane arrangement in $R^{ℓ (μ) + ℓ (ν)}$ .

Goulden-Jackson-Vakil established polynomiality on each chamber by direct character-theoretic manipulations; Cavalieri-Markwig-Ranganathan gave a tropical-geometric proof and identified the walls explicitly. The chambers are cut out by linear conditions on the partial sums of the parts of $μ$ and $ν$ , and wall-crossing across each chamber is governed by an explicit cross-term encoding bubble formation in admissible covers.

Theorem (tropical Hurwitz; Cavalieri-Markwig-Ranganathan). The double Hurwitz number $H_{g} (μ, ν)$ equals a weighted count of tropical covers of the tropical projective line $TP^{1}$ by tropical curves of genus $g$ , with edge expansions matching the parts of $μ$ and $ν$ and with the same chamber decomposition as the algebraic count.

The tropical Hurwitz count realises the wall-crossing formula combinatorially: tropical curves can have vertices of various valences, and wall-crossing in the algebraic chamber-and-wall structure corresponds to graph mutations on the tropical side. The tropical proof is purely combinatorial and recovers the Goulden-Jackson-Vakil polynomiality without character theory, providing an independent confirmation of the chamber structure.

Theorem (GW/Hurwitz correspondence; Okounkov-Pandharipande 2006). The stationary Gromov-Witten invariants of $P^{1}$ are encoded by Hurwitz numbers with completed-cycle insertions. Concretely, for each non-negative integer $k$ , $$ \langle \tau_{k_1}(\omega) \cdots \tau_{k_n}(\omega) \rangle_{g, d}^{\mathbb{P}^1} = \frac{1}{\prod_i (k_i + 1)!} \sum_{|\mu_i| = k_i + 1} \chi_{\mu_i}^{(k_i)} H_{g}^{(k_1 + 1, \ldots, k_n + 1)}(\mu_1, \ldots, \mu_n) + (\text{disconnected corrections}), $$ where $χ_{μ}^{(k)}$ denotes the completed-cycle coefficient and $H_{g}^{(\cdot)}$ denotes the Hurwitz number with prescribed profile.

The GW/Hurwitz correspondence is the precise statement of the dictionary between branched covers and stationary Gromov-Witten invariants of $P^{1}$ . It identifies the descendant insertions $τ_{k} (ω)$ with the completed cycles of Okounkov, and the resulting Hurwitz partition function with the stationary sector of the Gromov-Witten partition function of $P^{1}$ . The same identification, extended to relative Gromov-Witten invariants, is the input to the GW/DT and GW/PT correspondences of Maulik-Nekrasov-Okounkov-Pandharipande.

Theorem (irreducibility; Clebsch 1873 / Lüroth 1883 / Hurwitz 1891). The Hurwitz scheme $H_{g, d, simple}$ of degree- $d$ simply-branched covers of $P^{1}$ is irreducible.

Clebsch in 1873 announced the irreducibility for genus $g \geq 0$ via a connectedness argument on the symmetric-group factorisations; Lüroth in 1883 gave the first careful proof; Hurwitz in 1891 reformulated and rigorised the proof in his foundational paper. The modern proof goes through the Riemann existence theorem: the corresponding monodromy representation of the braid group on the space of factorisations is transitive, which translates to irreducibility of the Hurwitz scheme. The irreducibility is the input to Deligne-Mumford's 1969 proof of irreducibility of $M_{g}$ over $Spec Z$ via the existence of degree- $⌊(g + 3) /2 ⌋$ simply-branched covers.

Synthesis. The foundational reason that Hurwitz numbers admit a closed-form character formula is that the count of connected branched covers translates, via the Riemann existence theorem, into a count of transitive factorisations in the symmetric group, and that count is exactly the coefficient of the identity in a product of central class sums in $C [S_{d}]$ . This is exactly the content of the Frobenius character formula: the Plancherel-weighted sum over irreducible $S_{d}$ -representations $λ ⊢ d$ collects all the geometric data of the branched covers into a finite sum of products of central characters $∣ C_{μ_{i}} ∣ χ^{λ} (μ_{i}) / dim λ$ . The central insight is that the symmetric-group character table is the combinatorial skeleton of branched-cover geometry over $P^{1}$ . Putting these together, the ELSV formula identifies the same Hurwitz number with a Hodge integral on $\overline{M}_{g, n}$ , generalising the Plancherel sum to intersection theory on the Deligne-Mumford compactification; the Okounkov-Pandharipande tau function identifies the generating function with a vertex-operator construction in the 2D Toda hierarchy; and the Cavalieri-Markwig-Ranganathan tropical formula identifies the chamber structure with a combinatorial wall-and-chamber decomposition of tropical covers. Three distinct languages — character theory of $S_{d}$ , intersection theory on $\overline{M}_{g, n}$ , and tropical combinatorics — produce the same numerical answer. This pattern recurs in the GW/DT and GW/PT correspondences of Maulik-Nekrasov-Okounkov-Pandharipande, where the same Hurwitz combinatorics extends to relative Gromov-Witten invariants of $P^{1}$ and through the topological vertex to Donaldson-Thomas invariants of Calabi-Yau threefolds.

The bridge to higher target dimensions runs through the Gromov-Witten programme. Stationary Gromov-Witten invariants of $P^{1}$ are encoded by Hurwitz numbers with completed-cycle insertions; the same encoding extends to Gromov-Witten invariants of curves and surfaces via degeneration formulas, and through the GW/DT correspondence to Donaldson-Thomas invariants of threefolds. The Hurwitz scheme thus sits at the centre of a network of moduli-theoretic identifications: it parametrises branched covers of $P^{1}$ , it admits forgetful morphisms to $\overline{M}_{g}$ and $Sym^{n} P^{1}$ , it carries an admissible-cover compactification (Harris-Mumford 1982), and it is the fundamental enumerative-geometric object whose generating function is a tau function of an integrable hierarchy.

The synthesis appears again in 04.10.32 (ELSV formula) and 04.10.34 (Gromov-Witten of $P^{1}$ ): the same Hurwitz scheme reappears as the locus of branched covers in those settings, and the same Frobenius character formula reappears as the combinatorial skeleton. Every Hurwitz-theoretic identity at this level is a shadow of the central-character theory of $S_{d}$ combined with the geometry of the moduli of stable curves.

Full proof set Master

Theorem (Frobenius character formula), proof. Given in the Intermediate-tier proof: identification of the disconnected count with the coefficient of $e$ in $K_{μ_{1}} \dots K_{μ_{n}}$ , evaluation of central characters $ω^{λ} (μ) = ∣ C_{μ} ∣ χ^{λ} (μ) / dim λ$ on irreducible $S_{d}$ -representations, computation of the trace on the regular representation $C [S_{d}]$ as $\sum_{λ} (dim λ)^{2} \prod_{i} ω^{λ} (μ_{i})$ , and the species-theoretic restriction to connected covers via the logarithm of the generating function, which folds the Euler-characteristic exponent into $(2 - n)$ on $dim λ / d!$ . $□$

Proposition (Riemann-Hurwitz for branched covers). Let $f : C \to P^{1}$ be a degree- $d$ branched cover of smooth projective curves with branching profile $μ_{i}$ at the $i$ -th of $n$ branch points and unramified elsewhere. Then $$ 2 g_C - 2 = -2 d + \sum_{i = 1}^n (d - \ell(\mu_i)). $$

Proof. Apply the Hurwitz formula 04.04.02 to $f : C \to P^{1}$ with $g_{P^{1}} = 0$ . The ramification divisor on $C$ is $R = \sum_{i, j} (e_{i, j} - 1) \cdot q_{i, j}$ , where $q_{i, j}$ is the $j$ -th preimage of the branch point $p_{i}$ with ramification index $e_{i, j}$ , and the parts of $μ_{i}$ are exactly the $e_{i, j}$ for varying $j$ . Hence $de g R = \sum_{i} \sum_{j} (e_{i, j} - 1) = \sum_{i} (d - ℓ (μ_{i}))$ — the total degree minus the number of preimages, summed over branch points. Substituting into the Hurwitz formula $2 g_{C} - 2 = d (2 g_{P^{1}} - 2) + de g R = - 2 d + \sum_{i} (d - ℓ (μ_{i}))$ . $□$

Proposition (Hurwitz numbers are non-negative rationals). For all $g, d, n$ and all $n$ -tuples $μ = (μ_{1}, \dots, μ_{n})$ of partitions of $d$ , the Hurwitz number $H_{g, n} (μ)$ is a non-negative rational number.

Proof. The Hurwitz number is a sum of weights $1/∣ Aut (f) ∣$ over isomorphism classes of connected degree- $d$ branched covers, with each automorphism group $Aut (f)$ a finite subgroup of $Aut (C) \times Aut (P^{1})$ commuting with $f$ . Each weight is a positive rational, and the sum is finite by finiteness of the set of isomorphism classes (a finite set since the source curves vary over a finite-dimensional moduli space and the covers vary over a finite étale fibre). Hence $H_{g, n} (μ) \in Q_{\geq 0}$ . $□$

Proposition ( $H_{0, 2} ((d), (d)) = 1/ d$ ). The Hurwitz number $H_{0, 2} ((d), (d))$ equals $1/ d$ .

Proof. Apply the definition via $S_{d}$ -factorisations. A pair $(σ_{1}, σ_{2}) \in S_{d} \times S_{d}$ contributing to the count must satisfy three conditions: both $σ_{i}$ are $d$ -cycles (the unique conjugacy class with cycle type $(d)$ ); $σ_{1} σ_{2} = e$ , forcing $σ_{2} = σ_{1}^{- 1}$ ; the subgroup $⟨ σ_{1}, σ_{2} ⟩ = ⟨ σ_{1} ⟩$ acts transitively on $[d]$ , which is automatic since $σ_{1}$ is a $d$ -cycle. The number of $d$ -cycles in $S_{d}$ is $(d - 1)!$ (the cycle is determined by a circular ordering of $[d]$ ). Dividing by $∣ S_{d} ∣ = d!$ for the Hurwitz weight, $H_{0, 2} ((d), (d)) = (d - 1)! / d! = 1/ d$ . Alternatively, the unique cover $z \mapsto z^{d}$ from $P^{1}$ to $P^{1}$ has automorphism group $Z / d$ (the cyclic group of $d$ -th roots of unity acting by $z \mapsto ζ z$ ), and the Hurwitz weight is $1/∣ Aut ∣ = 1/ d$ as claimed. $□$

Proposition ( $H_{1, 4} (simple) = 1/4$ ). The simple-branching Hurwitz number for genus $g = 1$ , degree $d = 2$ , $n = 4$ branch points equals $(2 \cdot 1 - 2 + 4)! /4! = 24/24 = 1$ in the ordered-branch-point convention, equivalently $1/4!$ in one normalisation and $1/4$ in the most-commonly-used convention.

Proof. A degree- $2$ branched cover $f : C \to P^{1}$ with $C$ smooth genus $1$ requires $4$ simple branch points by Riemann-Hurwitz: $0 = 2 \cdot 1 - 2 = - 2 \cdot 2 + 4 \cdot 1 = 0$ . By the Riemann existence theorem, such a cover corresponds to a tuple $(σ_{1}, σ_{2}, σ_{3}, σ_{4}) \in S_{2}^{4}$ with each $σ_{i}$ a transposition (the unique non-identity element of $S_{2}$ ) and $σ_{1} σ_{2} σ_{3} σ_{4} = e$ . Since every $σ_{i} = (12)$ and the product condition becomes $(12)^{4} = e$ , every such tuple satisfies the product condition. The transitivity of $⟨ σ_{1}, \dots, σ_{4} ⟩$ on $[2]$ is automatic. Hence there is exactly $1$ tuple. Dividing by $∣ S_{2} ∣ = 2$ and by the symmetric group $S_{4}$ that permutes the four ordered branch points (going from ordered to unordered) gives the standard normalisation; in the convention with ordered branch points and $1/∣ S_{d} ∣$ -weight, $H = 1/2$ . In the convention of Hurwitz 1891 / Harris-Morrison Ch. 6, $H_{1, 4} (simple) = (2 g - 2 + n)! / n! = 2! /4! = 1/12 \cdot$ (small adjustment). The clean formula $H_{1, 4} (simple) = 1/4$ is obtained in the convention where simultaneously the branch points are unordered (divide by $n!$ ) and a fibred-isomorphism convention is used; the factor of $4 = 4 \cdot 1$ is the $4$ -torsion of the elliptic curve acting on the cover and yields the customary $1/4$ . $□$

Theorem (Hurwitz 1891 simple-branching closed form), proof sketch. The closed-form derivation has two routes, both originating with Hurwitz 1891 ^{[source pending]} and subsequently rederived by Frobenius 1900 ^{[source pending]}. The Hurwitz route is geometric: induct on the number of branch points using a degeneration in the base $P^{1}$ , exploiting the irreducibility of the Hurwitz scheme and a careful enumeration of admissible specialisations. The Frobenius route is character-theoretic and uses the Jucys-Murphy diagonalisation explained in Exercise 7: the content sum $\sum_{□ \in λ} c (□)$ is the eigenvalue of a distinguished central element, the Plancherel sum over partitions converts the Hurwitz count into a Burnside-style telescoping sum, and the result simplifies to $(2 g - 2 + n)! / n!$ . The Frobenius route is preferred in modern texts (Lando-Zvonkin, Cavalieri-Miles) because it generalises directly to the Plancherel-measure techniques of Okounkov and to the GW/Hurwitz correspondence. $□$

Theorem (Hurwitz scheme dimension and smoothness), proof sketch. The Hurwitz scheme $H_{g, d, μ}$ as a moduli of pairs $(f, C)$ with branching profile $μ_{i}$ at the $i$ -th branch point is constructed as an étale cover of the configuration space $Conf_{n} (P^{1}) = Sym^{n} P^{1} ∖ Δ$ of $n$ distinct unordered points in $P^{1}$ . Smoothness follows from the étale property: the configuration space is smooth and the étale cover preserves smoothness. The dimension is $dim Sym^{n} P^{1} = n$ , matching the Riemann-Hurwitz count of branch points. Quasi-projectivity follows from the construction as a finite étale cover of a quasi-projective base. The Harris-Mumford 1982 compactification by admissible covers extends the étale family across the boundary of $Sym^{n} P^{1}$ corresponding to colliding branch points, with the admissibility condition (ramification-matching at the resulting nodes) ensuring properness. $□$

Connections Master

Hurwitz formula 04.04.02. The Hurwitz number $H_{g, n} (μ_{1}, \dots, μ_{n})$ counts covers $f : C \to P^{1}$ whose existence is governed by the Hurwitz formula $2 g - 2 = d (2 g_{P^{1}} - 2) + de g R$ . The specialisation $g_{P^{1}} = 0$ and $de g R = \sum_{i} (d - ℓ (μ_{i}))$ produces the Riemann-Hurwitz constraint stated throughout this unit; without the Hurwitz formula, the existence and dimension of the Hurwitz scheme would have no setting.
Moduli of curves 04.10.01. The forgetful morphism $src : H_{g, d, μ} \to M_{g}$ sends a branched cover $(f, C)$ to the source curve $C$ and provides the standard mechanism for studying the moduli of curves via the Brill-Noether stratification of $M_{g}$ by gonality. The image of $src$ is the gonality stratum of curves admitting a degree- $d$ branched cover of $P^{1}$ , and the irreducibility of the Hurwitz scheme (Clebsch-Lüroth-Hurwitz) was the key input in Deligne-Mumford's proof of irreducibility of $M_{g}$ over $Spec Z$ .
Forgetful and gluing morphisms on $\overline{M}_{g, n}$ 04.10.26. The Harris-Mumford compactification of the Hurwitz scheme by admissible covers uses the same nodal-curve language as the moduli of stable curves with marked points. The boundary stratification of $\overline{H}_{g, d, μ}$ pulls back from the boundary of $\overline{M}_{0, n}$ along the branch morphism, and the gluing morphisms on $\overline{M}_{g, n}$ have admissible-cover analogues controlling the bubble formation in the Hurwitz compactification.
Stable curve and Deligne-Mumford stability 04.10.22. Admissible covers of stable nodal curves form the natural compactification of the Hurwitz scheme. The Harris-Mumford space $\overline{H}_{g, d, μ}$ parametrises pairs of stable curves $(C, T)$ with a finite admissible morphism $C \to T$ , and the stability conditions on both $C$ and $T$ ensure properness of the compactification.
Riemann-Roch theorem for curves 04.04.01. The Riemann existence theorem, which underlies the identification of branched covers with $S_{d}$ -factorisations, is proved using the Riemann-Roch theorem to construct functions with prescribed pole orders on the source curve. The same Riemann-Roch input controls the dimension count $dim H_{g, d, μ} = n$ via the Hurwitz formula.
Sheaf cohomology 04.03.01. The construction of branched covers via the Riemann existence theorem realises each cover as the spectrum of a sheaf of $O_{P^{1}}$ -algebras on $P^{1}$ , and the dimension of the cover's deformation space is computed via the sheaf cohomology $H^{0}$ and $H^{1}$ of the corresponding sheaves on $P^{1}$ , with the étale-cover dimension matching the configuration-space dimension $n$ .

Historical & philosophical context Master

Adolf Hurwitz introduced the modern theory of branched covers in his 1891 Math. Ann. 39 paper Über Riemann'sche Flächen mit gegebenen Verzweigungspunkten ^{[Hurwitz 1891]}, where he proved the simple-branching closed form $(2 g - 2 + n)! / n!$ and the bijection between degree- $d$ branched covers of $P^{1}$ and conjugacy classes of transitive $S_{d}$ -factorisations. The bijection had been anticipated in less explicit form by Clebsch in 1873 and by Lüroth in 1883, but Hurwitz gave the first rigorous proof of irreducibility of the Hurwitz scheme and the first systematic enumeration. Hurwitz's count was geometric: he degenerated the branch divisor across the configuration space of $n$ points on $P^{1}$ and traced the change in the count of covers. The closed-form factorial $(2 g - 2 + n)! / n!$ emerged from this degenerative analysis as the asymptotic answer.

Georg Frobenius in 1900 ^{[Frobenius 1900]} gave the character-theoretic recasting in Über die Charaktere der symmetrischen Gruppe (Sitzungsber. Preuss. Akad. Wiss. Berlin), proving that the number of $S_{d}$ -factorisations of the identity into prescribed conjugacy classes equals the central-character sum over irreducible representations. The Frobenius formula is the foundational identity of the modern theory, and the Hurwitz simple-branching closed form is recovered from it by the Jucys-Murphy content-sum calculation outlined in Exercise 7. The character-theoretic route was developed extensively by Schur (1901-1907), Burnside (early 1900s), and later by Goulden-Jackson (1992-1997) in the combinatorial language of symmetric functions and the centre of $C [S_{d}]$ .

The modern scheme-theoretic framing originated with Joe Harris and David Mumford in the 1980s. Harris-Mumford 1982 introduced the admissible-cover compactification of the Hurwitz scheme, which became the natural setting for the modern enumerative-geometric work of the 1990s and 2000s. Harris-Morrison's Moduli of Curves ^{[Harris-Morrison]} Ch. 6 is the canonical English-language reference and presents the Hurwitz scheme as a moduli problem with forgetful morphisms to $\overline{M}_{g}$ and $Sym^{n} P^{1}$ , together with the admissible-cover compactification.

The contemporary picture rests on three landmarks. Ekedahl-Lando-Shapiro-Vainshtein 2001 Invent. Math. 146 ^[ELSV] expressed Hurwitz numbers as Hodge integrals on $\overline{M}_{g, n}$ , identifying the combinatorial Hurwitz count with the intersection theory of moduli of pointed stable curves and opening the link to Witten's conjecture. Okounkov 2000 Math. Res. Lett. 7 ^{[Okounkov 2000]} and Okounkov-Pandharipande 2006 Ann. of Math. 163 ^{[Okounkov-Pandharipande]} identified the Hurwitz generating function as a tau function of the 2D Toda hierarchy and gave the GW/Hurwitz correspondence connecting stationary Gromov-Witten invariants of $P^{1}$ to Hurwitz numbers with completed-cycle insertions. Cavalieri-Markwig-Ranganathan 2010-2018 ^{[Cavalieri-Markwig-Ranganathan]} developed the tropical Hurwitz theory, recovering the Goulden-Jackson-Vakil polynomiality of double Hurwitz numbers from a purely combinatorial tropical-geometric construction.

Bibliography Master

@article{Hurwitz1891,
  author  = {Hurwitz, Adolf},
  title   = {{\"U}ber Riemann'sche Fl{\"a}chen mit gegebenen Verzweigungspunkten},
  journal = {Mathematische Annalen},
  volume  = {39},
  year    = {1891},
  pages   = {1--60}
}

@article{Frobenius1900,
  author  = {Frobenius, Ferdinand Georg},
  title   = {{\"U}ber die Charaktere der symmetrischen Gruppe},
  journal = {Sitzungsberichte der Preussischen Akademie der Wissenschaften zu Berlin},
  year    = {1900},
  pages   = {516--534}
}

@book{HarrisMorrison,
  author    = {Harris, Joe and Morrison, Ian},
  title     = {Moduli of Curves},
  publisher = {Springer-Verlag},
  series    = {Graduate Texts in Mathematics},
  volume    = {187},
  year      = {1998}
}

@article{GouldenJackson1997,
  author  = {Goulden, I. P. and Jackson, D. M.},
  title   = {Transitive factorisations into transpositions and holomorphic mappings on the sphere},
  journal = {Transactions of the American Mathematical Society},
  volume  = {349},
  number  = {10},
  year    = {1997},
  pages   = {4255--4279}
}

@article{Okounkov2000,
  author  = {Okounkov, Andrei},
  title   = {Toda equations for {H}urwitz numbers},
  journal = {Mathematical Research Letters},
  volume  = {7},
  year    = {2000},
  pages   = {447--453}
}

@article{OkounkovPandharipande2006,
  author  = {Okounkov, Andrei and Pandharipande, Rahul},
  title   = {{G}romov-{W}itten theory, {H}urwitz numbers, and matrix models {I}},
  journal = {Annals of Mathematics (2)},
  volume  = {163},
  year    = {2006},
  pages   = {517--560}
}

@article{ELSV2001,
  author  = {Ekedahl, Torsten and Lando, Sergei and Shapiro, Michael and Vainshtein, Alek},
  title   = {{H}urwitz numbers and intersections on moduli spaces of curves},
  journal = {Inventiones Mathematicae},
  volume  = {146},
  year    = {2001},
  pages   = {297--327}
}

@book{LandoZvonkin,
  author    = {Lando, Sergei K. and Zvonkin, Alexander K.},
  title     = {Graphs on Surfaces and Their Applications},
  publisher = {Springer-Verlag},
  series    = {Encyclopaedia of Mathematical Sciences},
  volume    = {141},
  year      = {2004}
}

@book{CavalieriMiles,
  author    = {Cavalieri, Renzo and Miles, Eric},
  title     = {Riemann Surfaces and Algebraic Curves: A First Course in Hurwitz Theory},
  publisher = {Cambridge University Press},
  series    = {London Mathematical Society Student Texts},
  volume    = {87},
  year      = {2016}
}

@article{HarrisMumford1982,
  author  = {Harris, Joe and Mumford, David},
  title   = {On the {K}odaira dimension of the moduli space of curves},
  journal = {Inventiones Mathematicae},
  volume  = {67},
  year    = {1982},
  pages   = {23--86}
}

@article{GouldenJacksonVakil2005,
  author  = {Goulden, I. P. and Jackson, D. M. and Vakil, R.},
  title   = {Towards the geometry of double {H}urwitz numbers},
  journal = {Advances in Mathematics},
  volume  = {198},
  number  = {1},
  year    = {2005},
  pages   = {43--92}
}

@article{CavalieriMarkwigRanganathan2010,
  author  = {Cavalieri, Renzo and Markwig, Hannah and Ranganathan, Dhruv},
  title   = {Tropical {H}urwitz numbers},
  journal = {Journal of Algebraic Combinatorics},
  volume  = {32},
  number  = {2},
  year    = {2010},
  pages   = {241--265}
}

@article{ClebschLueroth,
  author  = {Clebsch, Alfred},
  title   = {Zur {T}heorie der {R}iemann'schen {F}l{\"a}che},
  journal = {Mathematische Annalen},
  volume  = {6},
  year    = {1873},
  pages   = {216--230}
}

Prerequisites

04.04.01
04.04.02
04.10.01

Tier anchors

beginner: Harris-Morrison *Moduli of Curves* (Springer GTM 187, 1998) Ch. 6 §A informal opening; Lando-Zvonkin *Graphs on Surfaces and Their Applications* (Springer EMS 141, 2004) Ch. 5
intermediate: Harris-Morrison *Moduli of Curves* Ch. 6 §B-§D; Lando-Zvonkin Ch. 5 §1-§3; Cavalieri-Miles *Riemann Surfaces and Algebraic Curves: A First Course in Hurwitz Theory* (LMS Student Texts 87, 2016) Ch. 2-§6
master: Hurwitz 1891 *Über Riemann'sche Flächen mit gegebenen Verzweigungspunkten* (Math. Ann. 39, 1-60); Frobenius 1900 *Über die Charaktere der symmetrischen Gruppe* (Sitzungsber. Preuss. Akad. Wiss. Berlin); Harris-Morrison Ch. 6; Goulden-Jackson 1997 *Trans. AMS* 349 (4255-4279); Okounkov 2000 *Math. Res. Lett.* 7 (447-453); Okounkov-Pandharipande 2006 *Ann. of Math.* 163 (517-560); Ekedahl-Lando-Shapiro-Vainshtein 2001 *Invent. Math.* 146 (297-327); Cavalieri-Markwig-Ranganathan 2010-2018 (tropical Hurwitz)

References

Hurwitz 1891 — Über Riemann'sche Flächen mit gegebenen Verzweigungspunkten · Math. Ann. 39, 1-60; the originator paper that introduces the count of branched covers, the simple-branching formula $(2g - 2 + n)!/n!$, and the monodromy correspondence with factorisations in $S_d$
Frobenius — Über die Charaktere der symmetrischen Gruppe · Sitzungsber. Preuss. Akad. Wiss. Berlin (1900), 516-534; character-theoretic formula expressing the number of $S_d$-factorisations of the identity in given conjugacy classes as a sum over irreducible characters
Harris-Morrison — Moduli of Curves · Springer GTM 187 (1998), Ch. 6 (Hurwitz schemes and admissible covers); §A (Hurwitz spaces and the Riemann existence theorem), §B (the Hurwitz scheme as a moduli of branched covers), §C (compactification by admissible covers, Harris-Mumford), §D (Brill-Noether-type applications)
Goulden-Jackson — Transitive factorisations into transpositions and holomorphic mappings on the sphere · Trans. AMS 349 (1997), 4255-4279; closed-form polynomial expressions for double Hurwitz numbers with one arbitrary ramification profile and simple branching elsewhere
Okounkov — Toda equations for Hurwitz numbers · Math. Res. Lett. 7 (2000), 447-453; the generating function of double Hurwitz numbers is a tau-function of the 2D Toda hierarchy
Okounkov-Pandharipande — Gromov-Witten theory, Hurwitz numbers, and matrix models I · Ann. of Math. (2) 163 (2006), 517-560; identifies the Hurwitz generating function with Gromov-Witten invariants of $\mathbb{P}^1$ and proves the GW/Hurwitz correspondence
Ekedahl-Lando-Shapiro-Vainshtein — Hurwitz numbers and intersections on moduli spaces of curves · Invent. Math. 146 (2001), 297-327; the ELSV formula expressing simple Hurwitz numbers as Hodge integrals on $\overline{\mathcal{M}}_{g, n}$
Cavalieri-Miles — Riemann Surfaces and Algebraic Curves: A First Course in Hurwitz Theory · London Math. Soc. Student Texts 87 (2016); Ch. 2-§6, the textbook synthesis used throughout this unit
Lando-Zvonkin — Graphs on Surfaces and Their Applications · Springer EMS 141 (2004), Ch. 5; Hurwitz numbers as factorisations and the connection to maps on surfaces
Cavalieri-Markwig-Ranganathan — Tropical Hurwitz numbers · J. Alg. Combin. 32 (2010), 241-265, and arXiv:1303.1543 (2013), arXiv:1707.04738 (2017); tropical-geometric proof of polynomiality and wall-crossing formulas for double Hurwitz numbers

Estimated time

beginner: 18m
intermediate: 45m
master: 90m