Table of (almost)-Belyi maps with Five Exceptional Points

Mark van Hoeij and Vijay Kunwar.

Definition 1: Let P1 = C ⋃ {∞} denote the Riemann sphere. Let k,l,m be positive integers (or , see footnote 1) and let f: P1 → P1 be a rational function. The (k,l,m)-exceptional points of f are:

Footnote 1: If k = ∞ then all roots of f are (k,l,m)-exceptional points (and likewise for roots of 1-f resp. 1/f if l resp. m is ).

Footnote 2: Definition 1 differs slightly from Hoeij+Vidunas which defined exceptional points as
{roots of f of order ≠ k} {roots of 1-f of order ≠ l} {roots of 1/f of order ≠ m}.
In the files on this website, count=5 refers to definition 1, while Count=5 refers to the definition from Hoeij+Vidunas. The Count=5 table differs from the count=5 table below in two ways: The Belyi(i) ∈ Belyi(i+1) cases disappear (explanation) and there are 5 and 1 additional Belyi maps in the first two entries of the first column.

Goal: (see section "Motivation") Up to Mobius-equivalence, list every rational function f with 5 (k,l,m)-exceptional points, with (k,l,m) = (∞,2,m) and m ∈ {3,4,6}.

(k,l,m) Belyi ∉ Belyi(1) Belyi ∈ Belyi(1)
indirectly + directly
Belyi(1) ∉ Belyi(2) Belyi(1) ∈ Belyi(2) Belyi(2)
(∞,2,3) 411 maps 9 + 266 maps 65 families 3 families 2 families
(∞,2,4) 121 maps 3 + 23 maps 20 families no cases no cases
(∞,2,6) 54 maps 2 + 5 maps 12 families no cases no cases

Completeness: Proving that the above table is complete is the goal of the paper and the accompanying website.

Belyi maps: Functions whose branched set is ⊆ {0, 1, ∞}.

Belyi(1) resp. Belyi(2): These columns list functions with 1 resp. 2 branchpoints outside of {0, 1, ∞}. Since these branchpoint(s) are allowed to vary, these functions appear in 1 resp. 2 dimensional families.

Belyi(1) maps: These are the highlight of the table because

  1. Combined with the Heun table they contain many of the Belyi maps (see column Belyi ∈ Belyi(1)).
  2. It is not a priori obvious that all of the Belyi(1) families should allow "perfect parametrizations" meaning parametrizations that: Details in items 3-5 below.
  3. Our Belyi(1) families can be rationally parametrized. We use the variable s to parametrize these families, so each of our 1-dimensional families of Belyi(1) maps in C(x) is represented by a single f ∈ Q(s)(x). Values of s where f gives a Belyi map, with the same x-degree as f, are listed in the column Belyi ∈ Belyi(1).
  4. Every Belyi(1) function with count=5 is Mobius-equivalent to precisely one f in our Belyi(1) table, for precisely one value of s.
  5. For a Belyi(1) function f, that one branchpoint outside of {0, 1, ∞}, lets call it t, can be moved through P1 - {0, 1, ∞}. That gives an action of the braid group on the set of all Belyi(1) maps that ramify above {0, 1, t, ∞}. For each f ∈ Q(s)(x) in our tables the following holds: Writing t in terms of s gives a Belyi map t ∈ Q(s), f covers precisely one braid orbit O which (up to Mobius-equivalence) has precisely |O| = degs(t) elements (key to prove uniqueness for the value of s).

count > 5: Computing all Belyi(1) maps is problematic for count=6. For Belyi maps, here are all dessins with Count ≤ 6. You can obtain the dessins for count ≤ 7 or Count ≤ 7 by downloading our algorithm in section 4.4.