arXiv reader

Klein’s ten planar dessins of degree 11, and beyond

Gareth A. Jones and Alexander K. Zvonkin

arXiv
2104.12015v2 · PDF · 17 March 2022
Categories
math.GR, math.AG, math.NT
Licence
CC BY 4.0 The site shows the text.

Abstract

We reinterpret ideas in Klein’s paper on transformations of degree 11 from the modern point of view of dessins d’enfants, and extend his results by considering dessins of type (3,2,p) and degree p or p+1, where p is prime. In many cases we determine the passports and monodromy groups of these dessins, and in a few small cases we give drawings which are topologically (or, in certain examples, even geometrically) correct. We use the Bateman-Horn Conjecture and extensive computer searches to support a conjecture that there are infinitely many primes of the form p=(q^n-1)/(q-1) for some prime power q, in which case infinitely many groups {\rm PSL}_n(q) arise as permutation groups and monodromy groups of degree p (an open problem in group theory).

Items 42

Notation

Notation 3.1 (Where to put labels)

09WY

We put a label of an edge on its left side while moving from its black end to its white one. In this way the labels corresponding to a cycle of z or, equivalently, to a face, will be situated inside this face and will be rotated by z around the center of the face in the direction corresponding to the orientation of the surface. This convention is illustrated in Figure ?.

(When a dessin is drawn on the plane, its outer face gives an impression that its labels are rotated in the direction opposite to the orientation of the plane. But this is an illusion. In fact, the dessin should be considered not on the plane but on a sphere, and the center of the outer face is situated on the “opposite side” of this sphere. Looking from this center, the rotation of the labels of the outer face does correspond to the orientation of the sphere.)

When y^2=1 and the white vertices are omitted, so that \mathcal D is a map, one can regard G as permuting its half-edges, or else directed edges. Thus, an edge of a map bears two labels while a free edge bears only one. A free edge may be considered as an outgoing directed edge.

Example

Example 3.2

U6EV

The permutations corresponding to the map in Figure ? are as follows:

\begin{aligned} x & = & (1,15,16)(2,3,4)(5,9)(6,7,8,10,11,12), \\ y & = & (1,6)(2,9)(3,13,8,14)(4,7)(5,10)(11,12), \\ z & = & (1,12,10,9,4,6,16,15)(2,5,8,13)(3,14,7). \end{aligned}

The monodromy group of this map is the alternating group {\rm A}_{16}. See also Figures ? and ? for other examples of maps with labelled edges.

Example

Example 4.1

0ZIJ

We will show that the group L={\rm PSL}_2(11) has, up to automorphisms (equivalently, up to conjugacy in {\rm PGL}_2(11)\cong {\rm Aut}\,L), a single generating triple x, y, z satisfying (?). The character table of L, produced by GAP [GAP] and given here, has columns corresponding to the conjugacy classes, with the headings na, nb, etc. (nA, nB, etc in ATLAS notation [ATLAS]) indicating successive classes of elements of order n; the rows, labelled X.j, correspond to irreducible characters \chi_j for j=1, 2,\ldots. The table includes entries A, ^*A :=(-1\mp\sqrt 5)/2 and {\rm B}, /{\rm B}:=(-1\pm i\sqrt{11})/2, while a full stop denotes an entry 0.

The paper has a listing here, which this page does not show.

In this example we take \mathcal X and \mathcal Y to be the classes 3a and 2a, and \mathcal Z to be either of the classes 11a or 11b. For each choice of \mathcal Z the character sum in formula (?) has the value

\Sigma_L=1+\frac{(-1)\cdot 1\cdot {\rm B}}{5}+\frac{(-1)\cdot 1\cdot /{\rm B}}{5}+\frac{1\cdot (-2)\cdot (-1)}{10}+\frac{1\cdot 2\cdot (-1)}{10} =1-\frac{{\rm B}+/{\rm B}}{5}=\frac{6}{5}.

Now the centralisers C_L(x), C_L(y) and C_L(z) of typical elements x, y and z of \mathcal X, \mathcal Y and \mathcal Z have orders 6, 12 and 11, so that

\frac{|L|^2}{|C(x)|\cdot|C(y)|\cdot|C(z)|}=\frac{2^4\cdot 3^2\cdot 5^2\cdot 11^2}{6\cdot 12 \cdot 11}=2\cdot 5^2\cdot11.

Multiplying this by \Sigma_L gives 2^2\cdot 3\cdot 5\cdot 11 triples (x, y, z)\in{\mathcal X}\times{\mathcal Y}\times{\mathcal Z} with xyz=1 for each choice of \mathcal Z, and hence 2^3\cdot 3\cdot 5\cdot 11 in total. Each triple must generate L since this group has no proper subgroups of order divisible by 66 (in fact, none of index less that 11). Since {\rm Aut}\,L acts semi-regularly on generating triples, dividing by its order |{\rm PGL}_2(11)|=2^3\cdot 3\cdot 5\cdot 11 shows that {\rm Aut}\,L has a single orbit on such triples, as claimed.

Since the normal subgroups of any group \Delta with quotient isomorphic to a group G are in bijective correspondence with the orbits of {\rm Aut}\,G, acting by composition, on epimorphisms \Delta\to G, it follows that the triangle group

\Delta:=\Delta(3,2,11)=\langle X, Y, Z\mid X^3=Y^2=Z^{11}=XYZ=1\rangle

has a single normal subgroup with quotient isomorphic to L. Now L has two conjugacy classes of subgroups of index 11, all isomorphic to {\rm A}_5, and these classes are transposed by its automorphism group {\rm PGL}_2(11) (see [ATLAS] or [Dic, §259]); they lift to two conjugacy classes of subgroups of index 11 in \Delta, transposed by conjugation in the extended triangle group

\Delta[3,2,11]=\langle R_0, R_1, R_2\mid R_i^2=(R_1R_2)^3=(R_2R_0)^2=(R_0R_1)^{11}=1\rangle,

which contains \Delta with index 2, and they correspond to a chiral pair of dessins \mathcal D of type (3,2,11) degree 11 with monodromy group isomorphic to L. This calculation shows that they are the only such dessins, so they are the dessins {\mathcal M}_1 and \overline{\mathcal M}_1 discussed earlier.

Used by Example 4.3
Example

Example 4.2

NOJL

Next we count triples satisfying (?) in the simple group M={\rm M}_{11} of order 7920=2^4\cdot 3^2\cdot 5\cdot 11. The character table of M, given by GAP, is shown here:

The paper has a listing here, which this page does not show.

We must take \mathcal X and \mathcal Y to be the unique conjugacy classes 3a and 2a of elements of order 3 and 2, and \mathcal Z to be either of the classes 11a or 11b of elements of order 11. We see that the character sum for M is

\Sigma_M=1+\frac{1\cdot 2\cdot (-1)}{10}+\frac{1\cdot (-2)\cdot (-1)}{10}+\frac{1\cdot (-2)\cdot (-1)}{10}=\frac{6}{5}.

By [ATLAS] the centralisers in M of x, y and z have orders 18, 48 and 11, so

\frac{|M|^2}{|C(x)|\cdot|C(y)|\cdot|C(z)|}=\frac{2^8\cdot3^4\cdot 5^2\cdot 11^2}{18\cdot 48\cdot 11}=2^3\cdot 3\cdot 5^2\cdot 11.

Multiplying this by 2\Sigma_M gives 2^5\cdot 3^2\cdot 5\cdot 11 triples in M. Now M has a single conjugacy class of 12 subgroups L\cong {\rm PSL}_2(11), and we saw in Example 4.1 that each of these is generated by 2^3\cdot 3\cdot 5\cdot 11 triples, so 2^5\cdot 3^2\cdot 5\cdot 11 triples in M generate proper subgroups. Thus none of these triples can generate M. (See [Con91, Wol90] for confirmation that {\rm M}_{11} is not a quotient of \Delta.)

The groups {\rm M}_{11} and {\rm M}_{12} were introduced by Mathieu [Mat61] in 1861, so they would have been known to Klein; unlike {\rm M}_{12} (see [JW, §2.1.6], for example), {\rm M}_{11} is not the monodromy group of any orientable cubic map, since an extension of the argument used here shows that it is not a quotient of \Gamma.

Example

Example 4.3

QOX2

Now we count triples satisfying (?) in A={\rm A}_{11}. In this case there are three conjugacy classes of elements of order 3, and two classes of elements of order 2. A simple argument shows that if elements x, y\in A of orders 3 and 2 generate a transitive group than they must have cycle structures 3^31^2 and 2^41^3, so they belong to the conjugacy classes \mathcal X={\rm 3c} and \mathcal Y={\rm 2b} of A, while z must be in one of the two classes 11a and 11b of 11-cycles in A.

The character table of A is much too large to give here. However, the only irreducible characters \chi with \chi(x)\chi(y)\chi(z)\ne 0 are the principal character \chi_1 and the characters \chi_2, \chi_6 and \chi_{11} of degrees 10, 120 and 210, so we can give here the relevant portion of the character table, concerning these classes and characters:

The paper has a listing here, which this page does not show.

The character sum

\Sigma_A:=\sum_{\chi}\frac{\chi(x)\chi(y)\chi(z)}{\chi(1)}

has the value

1+\frac{1\cdot 2\cdot(-1)}{10}+\frac{3\cdot(-8)\cdot (-1)}{120}+\frac{3\cdot 2\cdot 1}{210}=1-\frac{1}{5}+\frac{1}{5}+\frac{1}{35}=\frac{36}{35}=\frac{2^2\cdot 3^2}{5\cdot 7}.

Now |A|=11!/2=2^7\cdot 3^4\cdot 5^2\cdot 7\cdot11, and by [ATLAS] the centralisers of x, y and z in A have orders 162=2\cdot 3^4, 1152=2^7\cdot 3^2 and 11, so

\frac{|A|^2}{|C(x)|\cdot|C(y)|\cdot|C(z)|}=2^6\cdot 3^2\cdot 5^4 \cdot 7^2\cdot 11.

Multiplying this by 2\Sigma_A shows that the number of triples in A satisfying (?) is

t:=2^9\cdot 3^4\cdot 5^3\cdot 7\cdot 11=10\times 11!.

From this total we must subtract the number of triples generating proper subgroups of A. As we have seen, the only such subgroups are those isomorphic to {\rm PSL}_2(11). Each subgroup L\cong {\rm PSL}_2(11) in A is contained in a unique subgroup M\cong {\rm M}_{11}; there are two conjugacy classes of 2520 such subgroups M in A, each containing one conjugacy class of 12 subgroups L, so the number of such subgroups L in A is 2\cdot 2520\cdot 12=60480=2^6\cdot 3^3\cdot 5\cdot 7. As shown in Example 4.1, each subgroup L is generated by 2^3\cdot 3\cdot 5\cdot 11 triples satisfying (?), so the number of triples in A generating subgroups L\cong {\rm PSL}_2(11) is 2^9\cdot 3^4\cdot 5^2\cdot 7\cdot 11=t/5=2\times 11!.

Since there are 10\times 11! triples in A, of which 2\times 11! generate proper subgroups, there are 8\times11! triples generating A, forming eight orbits under {\rm Aut}\,A={\rm S}_{11}; these correspond to eight normal subgroups of \Delta with quotient A, and hence to eight regular dessins with (orientation-preserving) automorphism group A. Now A has a single conjugacy class of subgroups of index 11, namely the natural point-stabilisers isomorphic to {\rm A}_{10}, so these lift to eight more conjugacy classes of subgroups of index 11 in \Delta, in addition to the two resulting from Example 4.1. These give eight more dessins \mathcal D of type (3,2,11) and degree 11, all with monodromy group G\cong {\rm A}_{11}. Thus we have a total of ten dessins of the required degree and type. This confirms the earlier enumeration, and also confirms that for {\mathcal M}_1 and \overline{\mathcal M}_1 the monodromy group is {\rm PSL}_2(11), whereas for the other eight dessins it is {\rm A}_{11}. Generating triples for these dessins are given in Subsection ?.

Theorem

Theorem 8.1

366J

The transitive permutation groups G of prime degree are the following:

  • (a) subgroups G of {\rm AGL}_1(p) containing the translation subgroup;

  • (b) G={\rm A}_p or {\rm S}_p for a prime p\ge 5;

  • (c) G={\rm PSL}_2(11), {\rm M}_{11} or {\rm M}_{23} with p=11, 11 or 23;

  • (d) groups G such that {\rm PSL}_n(q)\le G\le {\rm P\Gamma L}_n(q) where the degree (q^n-1)/(q-1) of its natural representation(s) is prime.

Corollary

Corollary 8.2

PXJM

The only perfect groups of prime degree are the simple groups {\rm A}_p, {\rm PSL}_2(11), {\rm M}_{11}, {\rm M}_{23} and {\rm PSL}_n(q) appearing in parts (b), (c) and (d) of Theorem 8.1.

Used by Proof
Theorem

Theorem 8.3

0G41

For each prime p\ne 2,3,7 there is a dessin \mathcal D of degree p and type (3,2,p) with monodromy group {\rm A}_p.

Proof of Theorem 8.3

First suppose that p\equiv 2 mod (3), say p=3k+2. There is an obvious map for p=5, so we may assume that p\ge 11, and hence k\ge 3. Let \mathcal D be the map on the left in Figure ?, where there are k vertices of degree 3, so that \mathcal D has passport (3^k1^2, 2^{k+1}1^k, p^1). We aim to show that its monodromy group G is {\rm A}_p. By Corollary 8.2 and the discussion of sporadic examples in case (c) it is sufficient to show that G is not {\rm PSL}_2(11) acting with degree 11 or a group {\rm PSL}_n(q) acting naturally for some n\ge 2 and prime power q. In fact, when p=11, so that k=3, \mathcal D is isomorphic to \overline{\mathcal M}_3 in Figure ?, and we have seen that this has monodromy group {\rm A}_{11}. We may therefore suppose, for a contradiction, that G={\rm PSL}_n(q) acting naturally, and that p\ge 17, so that k\ge 5.

If \pi denotes the permutation character of {\rm S}_p, counting fixed points of elements, we see that the involution y corresponding to \mathcal D has fixed point ratio

\frac{\pi(y)}{\pi(1)}=\frac{k}{p}=\frac{k}{3k+2}\ge\frac{5}{17}.

We will compare this with the corresponding values for involutions in {\rm PSL}_n(q). Any involution in {\rm PSL}_2(q) fixes at most two points, so n\ne 2 since \pi(y)=k\ge 5. For n\ge 3 the involutions g\in{\rm PSL}_n(q) have fixed point ratios \pi(g)/\pi(1) bounded above by

b_e=\frac{q^{n-2}+q^{n-3}+\cdots+q+1}{q^{n-1}+q^{n-2}+\cdots+q+1} \quad\hbox{or}\quad b_o=\frac{(q^{n-3}+\cdots+q+1)+(q+1)}{q^{n-1}+q^{n-2}+\cdots+q+1}

as q is even or odd: these bounds are attained when g fixes a hyperplane or two disjoint subspaces of dimensions n-3 and 1, respectively, corresponding to matrices in {\rm SL}_n(q) with Jordan normal form given by block matrices

\Big(\,\begin{array}{cc} J & O \\ O & I \end{array}\,\Big) \quad\hbox{where}\quad J=\Big(\,\begin{array}{cc} 1 & 1 \\ 0 & 1 \end{array}\,\Big) \quad\hbox{or}\quad \Big(\,\begin{array}{cc} -1 & 0 \\ 0 &-1 \end{array}\,\Big) \quad\hbox{and}\quad I=I_{n-2}.

It is straightforward to check that if q is odd then b_o<5/17 and hence G\ne{\rm PSL}_n(q), unless n=q=3, when b_o=5/13; however, this exceptional case corresponds to p=13\not\equiv 2 mod (3). Thus q is even, so q=2 since b_e<1/q. This gives 3k+2=p=2^n-1 and hence 2^n\equiv 0 mod (3), which is impossible.

We can apply a similar argument when p\equiv 1 mod (3), say p=3k+1. We may assume that p\ge 13, that is, k\ge 4. By deleting a vertex of valency 1 from \mathcal D we obtain the map {\mathcal D}' of degree p, type (3,2,p) and passport (3^k1^1, 2^k1^{k+1},p^1) shown on the right in Figure ?. The corresponding involution y now has fixed point ratio

\frac{\pi(y)}{\pi(1)}=\frac{k+1}{p}=\frac{k+1}{3k+1}\ge\frac{5}{13}.

As before, if G={\rm PSL}_n(q) then \pi(y)>2 implies that n\ge 3. We have already seen that if q is odd then b_o<5/17, so that b_o<5/13 and hence G\ne{\rm PSL}_n(q), unless n=q=3; however, in this case p=13, and GAP shows that {\mathcal D}' has monodromy group {\rm A}_{13} rather than {\rm PSL}_3(3), so q must be even. Again, q=2 since b_e<1/q, so if any element of G={\rm PSL}_n(2) has fixed points they must form a projective subspace. Thus k=\pi(y) is odd and hence p=3k+1 is even, a contradiction. \square

Warning

Warning

ZEDQ

here, in spite of an almost universal convention on notation, the unnamed prime of which q is a power is not the prime p; instead, p will denote the degree (q^n-1)/(q-1) of G.

Lemma

Lemma 8.4

QQY6

Given integers n\ge 2 and e\ge 1, the polynomial

f(t)=\frac{t^{ne}-1}{t^e-1}=1+t^e+t^{2e}+\cdots+t^{(n-1)e}

is irreducible in {\mathbb Z}[t] if and only if n is prime and e is a power n^i\;(i\ge 0) of n.

Used by Proof
Proof of Lemma 8.4

If k\in\mathbb{N} the cyclotomic polynomial \Phi_k(x) is, by definition, the polynomial with integer coefficients whose roots are the primitive kth roots of unity. It is irreducible and has degree \varphi(k), where \varphi is the Euler totient function. For any n\in\mathbb{N} we have x^n-1=\prod_{d|n}\Phi_d(x) (see [BS, § 5.2.1] or [NZM, § 4.3, Problem 26]). Putting x=t^e gives

f(t) = \frac{t^{ne}-1}{t^e-1} = \prod_d \Phi_d(t),

with the product over all d which divide ne but not e. Thus f is irreducible if and only if there is just one such divisor d (which is ne itself, of course). By considering the prime power decompositions of e and ne one can see that this happens if and only if n is prime and e is a power of n. \square

Lemma

Lemma 8.5

XPWD

Let q=r^e where r is prime, and let n\ge 2. If p=(q^n-1)/(q-1) is prime then

  • (a) n is prime and e is a power n^i\;(i\ge 0) of n;

  • (b) q\not\equiv 1 mod (n);

  • (c) {\rm PSL}_n(q)={\rm SL}_n(q).

Proof of Lemma 8.5

(a) This follows from Lemma 8.4, using the fact that p=f(r)=\prod_d\Phi_d(r) as in (Proof), with each \Phi_d(r)\ne\pm1. (To see the latter, note that \Phi_d(r)=\prod_{\zeta}(r-\zeta) where \zeta ranges over the primitive d-th roots of unity, so that |r-\zeta|>1 for each \zeta.)

(b) If q\equiv 1 mod (n) then p=1+q+\cdots+q^{n-1}\equiv 0 mod (n) with p>n.

(c) For any q and n we have {\rm PSL}_n(q)={\rm SL}_n(q)/Z, where Z=\{\lambda I\in{\rm GL}_n(q)\mid \lambda^n=1\}\cong{\rm C}_d is the centre of {\rm SL}_n(q) and d=\gcd(q-1,n). If p is prime then d=1 by (a) and (b). \square

Remark

Remark 9.1

V9UI

Li [W.Li] has suggested a modified version of the Bateman–Horn Conjecture in which \prod_i\deg f_i in the denominator is removed and each factor 1/\ln(t) within the integral is replaced with 1/\ln(f_i(t)); the lower limit of integration may also be adjusted to avoid a logarithmic singularity at f_i(t)=1. The result is asymptotically equivalent to the original conjecture, but in cases such as Sophie Germain primes, involving a non-monic polynomial f_i, the approximation is significantly better.

Conjecture

Conjecture 9.2

K35D

There are infinitely many projective primes of the form p=1+q+q^2 where q is prime.

Conjecture

Conjecture 9.3

0UWG

For each fixed prime n\ge 3 and each fixed power e of n, there are infinitely many projective primes of the form p=1+q+\cdots+q^{n-1} where q is the e-th power of a prime.

Theorem

Theorem 10.1

EUVX

For each Fermat prime p=2^{2^k}+1\ge 5 there are 2^{2^k-k-1} dessins \mathcal D of type (3,2,p) and degree p with monodromy group G={\rm PSL}_2(2^{2^k}), acting naturally. They have genus (p-5)/12, while their regular covers \mathcal R, which are mutually non-isomorphic, have genus p(p-4)(p-5)/12. Dessins \mathcal D and \mathcal R are both mirror symmetric.

Used by Example 10.2
Example

Example 10.2

HWEF

If k=1 and p=5 we obtain one planar dessin \mathcal D, shown on the left in Figure ?, ?, with monodromy group G={\rm PSL}_2(4)\cong {\rm A}_5; its regular cover \mathcal R is the dodecahedron, and \mathcal D is the quotient of \mathcal R by a point-stabiliser in G, isomorphic to {\rm A}_4. If k=2 and p=17 we obtain two dessins of genus 1; they are shown in Section ?, Figure ?. Their regular covers are of genus 221. If k=3 and p=257 we obtain 16 dessins of genus 21, with regular covers of genus 1\,365\,441. If k=4 and p=65\,537 we obtain 2048 dessins of genus 5461, with regular covers of genus 23\,454\,100\,602\,881. No Fermat primes are known for k\ge 5, and it is conjectured that none exist, but if any do, they are covered by Theorem 10.1.

The paper has a figure here, which this page does not show.
Theorem

Theorem 10.3

760X

If q is a prime power such that p:=1+q+q^2 is prime, there are (p-1)/3e dessins of type (3,2,p) and degree p with monodromy group {\rm PSL}_3(q) acting naturally, where q is the e-th power of a prime. They are all chiral. If q is odd they have passport (3^{q(q+1)/3}1^1, 2^{(q^2-1)/2}1^{q+2}, p^1) and genus (q-3)(q+1)/12, whereas if q=2^e their passport is (3^{q(q+1)/3}1^1, 2^{q^2/2}1^{q+1}, p^1) and their genus is q(q-2)/12.

Example

Example 10.4

DTRF

The smallest example for the above theorem is the chiral pair of dessins of degree p=7 with monodromy group {\rm PSL}_3(2)\cong {\rm PSL}_2(7); one of this pair is shown on the right in Figure ?, ?.

Example

Example 10.5

QQS1

Let p=13 and G={\rm PSL}_3(3). The required conjugacy classes are (in ATLAS [ATLAS] notation) {\mathcal X}=3B, {\mathcal Y}=2A and {\mathcal Z}=13A, 13B, 13C or 13D, giving four dessins \mathcal D of type (3,2,13) and degree 13; they form two chiral pairs, each pair having the same regular cover. Each dessin \mathcal D (see Figure ?, or [APZ, Orbit 13.1, Fig. 8.22(A)]) has passport (3^41^1, 2^41^5, 13^1) and genus 0, while the two regular covers have genus 253. (These are the duals of the regular maps R253.1 and R253.2 of type \{3,13\}_8 and \{3,13\}_{12} in [Con].)

The paper has a figure here, which this page does not show.
Example

Example 10.6

13DC

Let p=31, with G={\rm PSL}_3(5). In this case {\mathcal X}=3A, {\mathcal Y}=2A and there are ten possible classes {\mathcal Z}=31A, 31B, \ldots. These give ten dessins \mathcal D of type (3,2,31) and degree 31, forming five chiral pairs with five regular covers. Each dessin \mathcal D has passport (3^{10}1^1, 2^{12}1^7, 31^1) and genus 1, while the regular covers have genus 25\,001. Generators of the monodromy groups of these ten dessins \mathcal D, as well as the diagram of one of them, are given in Subsection ?.

Example

Example 10.7

DRO3

Let p=73 and G={\rm PSL}_3(8). Here {\mathcal X}=3A and {\mathcal Y}=2A, and there are 24 classes {\mathcal Z}=73A,\ldots of elements of order 73. Since e=3 we obtain 24/3=8 dessins \mathcal D. They have passport (3^{24}1^1, 2^{32}1^9, 73^1) and genus 4; their permutation representations are given in Subsection ?. They form four chiral pairs, each pair having a regular cover of genus 1\,260\,673.

Example

Example 10.8

IQFH

Let p=2^3-1=7 and G={\rm PSL}_3(2)\;(={\rm GL}_3(2)\cong {\rm PSL}_2(7)). The only conjugacy classes of elements of order 3 or 2 in G are 3A and 2A, with cycle structures 3^21^1 and 2^21^3 in both of the natural representations of G, on points and lines of the Fano plane {\mathbb P}^2({\mathbb F}_2); there are two classes 7A, 7B of elements of order p=7. The only character which does not vanish in the character sum is \chi_1, so \Sigma=1. Since |C(x)|=3, |C(y)|=8=2^3, |C(z)|=7 and |G|=168=2^3\cdot 3\cdot 7, the number of triples of type (3,2,7) in G is

2\cdot \frac{2^6\cdot 3^2\cdot 7^2}{3\cdot 2^3\cdot 7}=2^4\cdot 3\cdot 7=2|G|=|{\rm Aut}\,G|.

(Note that {\rm Out}\,G\cong {\rm C}_2, induced by the duality of the Fano plane, or equivalently by conjugation of {\rm PSL}_2(7) in {\rm PGL}_2(7).) These triples all generate G, so \Delta=\Delta(3,2,7) has a unique normal subgroup with quotient G, corresponding to a regular dessin \mathcal R of type (3,2,7) with automorphism group G. There are two conjugacy classes of subgroups of index 7 in G, the stabilisers of points and lines in the Fano plane; these lift to two conjugacy classes of subgroups of index 7 in \Delta, corresponding to two dessins \mathcal D of type (3,2,7) and degree 7, each having regular cover \mathcal R. Each dessin \mathcal D has passport (3^21^1, 2^21^3, 7^1), so it has genus 0, while \mathcal R has genus 3. One of the two dessins \mathcal D is shown on the right in Figure ?, ?, and the other is its mirror image \overline{\mathcal D}. These two maps, corresponding to the two choices for the conjugacy class \mathcal Z or to the two choices for the class of subgroups of index 7, are not isomorphic since the outer automorphism of G transposes both pairs. The dessin \mathcal R is Klein’s heptagonal map on his quartic curve, introduced in [Kle78]; it is denoted by \{7,3\}_8 in [CM] and is the dual of R3.1 in [Con]. As the common regular cover of a chiral pair of dessins (or by its uniqueness) this map is isomorphic to its mirror image.

Example

Example 10.9

38EN

Let p=31 (again), but now in its role as a Mersenne prime 2^5-1, with G={\rm PSL}_5(2)={\rm GL}_5(2). There are two classes each of elements of orders 3 and 2, namely 3A and 3B, with cycle structures 3^81^7 and 3^{10}1^1, and 2A and 2B, with cycle structures 2^81^{15} and 2^{12}1^7, together with six classes 31A, \ldots of elements of order 31. The inequality 2a+b\ge p-1 discussed earlier implies that \mathcal X and \mathcal Y must be 3B and 2B. For each of the six choices of \mathcal Z, only \chi_1 and \chi_6 appear in the character sum, so

\Sigma=1+\frac{8\cdot(-5)\cdot1}{280}=1-\frac{1}{7}=\frac{6}{7}.

We have |C(x)|=180=2^2\cdot 3^2\cdot 5, |C(y)|=1536=2^9\cdot 3, |C(z)|=31 and |G|=9\,999\,360=2^{10}\cdot 3^2\cdot 5\cdot 7\cdot 31, so the total number of triples in G is

6\cdot 2^9\cdot 3\cdot 5\cdot 7^2\cdot 31\cdot\frac{6}{7}=2^{11}\cdot 3^3\cdot 5\cdot 7\cdot 31=6|G|=3|{\rm Aut}\,G|.

Each triple generates G, so \Delta has three normal subgroups with quotient G. There are two conjugacy classes of subgroups of index 31 in G, the stabilisers of points and hyperplanes in {\mathbb P}^4({\mathbb F}_2), giving six dessins \mathcal D of degree 31, with three chiral pairs each having the same regular cover \mathcal R. Each dessin \mathcal D has passport (3^{10}1^1,2^{12}1^7,31^1), so it has genus 1, while the regular covers \mathcal R have genus 672\,001. Generators of the monodromy groups of these six dessins \mathcal D, as well as the diagram of one of them, are given in Subsection ?.

Remark

Remark 10.10 (Goormaghtigh conjecture)

ZR57

Note that 31 appears twice as a projective prime, as both 1+5+5^2 and 1+2+2^2+2^3+2^4. In 1917 Goormaghtigh [Goo] conjectured that this example and 1+2+2^2+\cdots+2^{12}=8191=1+90+90^2 are the only repetitions among numbers of the form (q^n-1)/(q-1) for integers q, n\ge 2. Although 8191 is prime (and hence a projective prime for (q,n)=(2,13)), 90 is not a prime power, so only the first example is a repetition of projective primes. A computer search of prime degrees up to 10^{18} has revealed no further examples. The conjecture is still open. For additional information see [Goo-wiki].

Theorem

Theorem 10.11

T6MP

For each Mersenne prime p=2^n-1\ge 7 there are at least (p-1)/2n chiral pairs of dessins of type (3,2,p), degree p and genus given by (?), with monodromy group G={\rm PSL}_n(2)={\rm GL}_n(2).

Used by Proof
Proof of Theorem 10.11

We will use the Frobenius formula (?) to show that for each conjugacy class \mathcal Z of Singer cycles in G there a triple (x,y,z) of type (3,2,p) where x and y satisfy r=s=1 (as above) and z\in{\mathcal Z}. The calculation is aided by the fact that most of the non-principal irreducible characters of G vanish on the Singer cycles z (see [LRS, §4.2], for example), and hence do not contribute to the character sum \Sigma. Among the other characters, that of degree p-1, the non-principal constituent of the natural permutation character, vanishes on x, while the Steinberg character of degree 2^{n(n-1)/2} vanishes on y, so they can also be ignored. The remaining characters have sufficiently large degrees, compared with their values at x, y and z (see [LRS, §4.1,4.2] for full details), that their total contribution to \Sigma has modulus less than the contribution 1 of the principal character, so that \Sigma>0 and hence the required triples exist. As before, they all generate G by [Ber]. There are (p-1)/n conjugacy classes \mathcal Z of Singer cycles z in G, forming (p-1)/2n mutually inverse pairs; each pair are transposed by the outer automorphism group of G, as are the two permutation representations of G of degree p; thus we obtain (at least) (p-1)/2n chiral pairs of dessins \mathcal D, all with the same passport and genus, as calculated above. \square

Conjecture

Conjecture 10.12

03TL

For each projective prime p=1+q+\cdots+q^{n-1}>3 there exists at least one dessin of type (3,2,p) and degree p with monodromy group {\rm PSL}_n(q).

Theorem

Theorem 10.13

DQYJ

Suppose that p is a prime such that there is a dessin of type (3,2,p) and degree p with monodromy group G. Then p\ge 5, and one of the following holds:

  • (a) G={\rm A}_p, acting naturally, with p\ne 7;

  • (b) p=11 and G={\rm PSL}_2(p), acting on the cosets of a subgroup isomorphic to {\rm A}_5;

  • (c) p is a projective prime (q^n-1)/(q-1) for some prime n and prime-power q, and G={\rm PSL}_n(q)={\rm SL}_n(q) acting naturally on points or hyperplanes of {\mathbb P}^{n-1}(q).

Theorem

Theorem 11.1

REUW

For each prime p the dessin {\mathcal D}(p) is the only regular dessin of type (3,2,p) with automorphism group G={\rm PSL}_2(p), and {\mathcal D}_0(p) is the only dessin of degree p+1 and type (3,2,p) with monodromy group G.

Theorem

Theorem 12.1

RCX5

The perfect permutation groups of degree p+1 (p prime), containing a p-cycle, are the following:

  • (a) {\rm A}_{p+1} for primes p\ge 5,

  • (b) {\rm PSL}_2(p) for primes p\ge 5,

  • (c) affine groups {\rm AGL}_n(2)  for Mersenne primes p=2^n-1\ge 5,

  • (d) Mathieu groups {\rm M}_{11}  and {\rm M}_{12}  for p=11 and {\rm M}_{24}  for p=23,

all except {\rm M}_{11} (see below) in their natural representations.

Used by Proof
Theorem

Theorem 12.2

GHXR

For each Mersenne prime p=2^n-1>7 there are dessins of type (3, 2, p) and degree p+1 with passport (3^{(p-1)/3}1^2, 2^{(p+1)/2}, p^11^1), genus (p+5)/12 and monodromy group {\rm AGL}_n(2).

Proof of Theorem 12.2

We will use the fact that G:={\rm AGL}_n(2) is a semidirect product of its translation subgroup T\cong({\rm C}_2)^n by G_0:={\rm GL}_n(2), the stabiliser of 0. Let (x, y, z) be a generating triple for {\rm GL}_n(2) of type (3, 2, p), chosen as in the proof of Theorem 10.11 by taking r=s=1, so that \alpha=1 and \beta=(n+1)/2. Now let u=xa, v=yb and w=zc, where a, b, c\in T. We will try to choose a, b and c so that (u,v,w) is a generating triple for G of type (3, 2, p).

Since x^3=1 we have

u^3=(xa)^3=x^3a^{x^2}a^xa=a^{x^2+x+1},

so u^3=1 if and only if a\in T_x:=\ker(x^2+x+1). (Here we regard x^2+x+1 as a linear transformation of the vector space T over {\mathbb F}_2.) Similarly v^2=1 if and only if b\in T_y:=\ker(y+1), and w^p=1 if and only if c\in T_z:=\ker(z^{p-1}+z^{p-2}+\cdots+z+1). Also

uvw=xa.yb.zc=xya^yzb^zc=xyza^{yz}b^zc=a^{yz}b^zc,

so uvw=1 if and only if a^{yz}b^zc=1.

Since z is a Singer cycle in G_0, T_z=T and hence the condition on c is vacuous, whereas the Jordan forms for x and y used in the proof of Theorem 10.11 show that the subspaces T_x and T_y have dimensions n-1 and (n+1)/2. Within these subspaces one can choose a and b arbitrarily, and then define c=a^{yz}b^z, so that a^{vyz}b^zc=1 and hence (u, v, w) is a triple of type (3,2,p).

The cycle structures of u, v and w can be found as follows. One can regard {\rm AGL}_n(2) as acting naturally on T, with elements of T and {\rm GL}_n(2) acting by translations and by linear transformations. The element u=xa acts on T by t\mapsto t^xa, so it fixes an element t\in T if and only if a=t(1-x), using additive notation for the vector space T. Now the Jordan form for x shows that T(1-x)=T_x, so for every choice of a\in T_x there are |\ker(1-x)|=2 fixed points t\in T for u. Since u has order 3 it has (p-1)/3 cycles of length 3. In the case of v the Jordan form for y shows that T(1-y) is a subspace of codimension 1 in T_y, so by choosing b\in T_y\setminus T(1-y) we can ensure that v has no fixed points, and therefore consists of (p+1)/2 transpositions. Since w has order p it has cycle structure p^11^1.

Now let us define H:=\langle u, v, w\rangle\le G. We will prove that if p>7 then H=G. Factoring out T maps H onto \langle x, y, z\rangle=G_0, so HT=G. If H contains a non-identity translation t\in T then since H contains w it contains all the non-identity translations t^w, so H\ge T and hence H=G, as required.

We may therefore assume, for a contradiction, that H\cap T=1, so that H is a complement for T in G. Since the first cohomology group H^1({\rm GL}_n(2),T) for {\rm GL}_n(2) on its natural module T is trivial for all n>3 (see [Bell]) there is a single conjugacy class of such complements, so H is a point-stabiliser in the natural action of G. This contradicts the fact that v has no fixed points, by our choice of the translation b.

Thus (u, v, w) is a generating triple for G, so it corresponds to a dessin of type (3,2,p) and degree p+1 with monodromy group G. Its genus follows from its passport, determined above. \square

Example

Example 12.3

U8UQ

Taking n=5 we obtain a dessin of type (3,2,31), degree 32 and genus 3, with passport (3^{10}1^2, 2^{16}, 31^11^1). This should not be confused with the Galois orbit 32.1 of six planar dessins with the same type, degree and monodromy group in [APZ]: these have passport (3^{10}1^2, 2^{12}1^8, 31^11^1), and they correspond to taking b\in T(1-y), so that v fixes |\ker(1-y)|=8 points (see Subsection ?).

Theorem

Theorem 12.4

FARU

The alternating group {\rm A}_{p+1}, p prime, is the monodromy group of a dessin of type (3,2,p) and degree p+1 if and only if p=3 or p\ge 11.

Proof of Theorem 12.4

For p=2, 5 or 7 the triangle group \Delta(3,2,p) does not map onto {\rm A}_{p+1}: this is obvious if p=2 or 5, and for p=7 it is well known that {\rm A}_8 is not a Hurwitz group: the triples of type (3,2,7) in this group all generate proper subgroups isomorphic to {\rm PSL}_2(7). On the other hand, {\rm A}_4 is the monodromy group of the dessin of type (3,2,3) and degree 4 on the left in Figure ?, ?, ?, the quotient of the tetrahedron \{3,3\} by {\rm C}_3. (Note that {\rm A}_4 is isomorphic to {\rm PSL}_2(3) in case (b).)

The paper has a figure here, which this page does not show.

For p=11 and 13 it is sufficient by Theorem 12.1 to note that the dessins of degree 12 and 14 in Figure ?, ?, ? are not isomorphic to the unique dessins {\mathcal D}_0(p) with monodromy group {\rm PSL}_2(p) shown in Figures ? and ?.

The paper has a figure here, which this page does not show.

Finally, let p\ge 17. If p\equiv 2 mod (3), say p=3k+2 for some k\ge 5, consider the dessin \mathcal D in Figure ?, where there are k vertices of degree 3, so that the degree of \mathcal D is p+1 and its passport is (3^k1^3, 2^{k+3}1^{k-3}, p^11^1). The monodromy group of \mathcal D cannot be {\rm PSL}_2(p) since a non-identity element of that group has at most two fixed points, whereas x has three; similarly it cannot be {\rm AGL}_n(2) for any n since the fixed point set of any element of that group is either empty or an affine subspace of order 2^d for some d; finally, it cannot be a Mathieu group since \mathcal D is not isomorphic to any of the dessins in Figures ? or ? or their mirror images, so it must be {\rm A}_{p+1}.

If p\equiv 1 mod (3), say p=3k+4 for some k\ge5, one can apply a similar argument to the dessin {\mathcal D}' obtained by adding two vertices of valency 1 to the rightmost free edges of \mathcal D, so that the passport is (3^k1^5, 2^{k+5}1^{k-5}, p^11^1).\square

Theorem

Theorem 12.5

YXWV

Suppose that p is a prime such that there is a dessin of type (3,2,p) and degree p+1 with monodromy group G. Then one of the following holds:

  • (a) G={\rm A}_{p+1}, with p=3 or p\ge 11,

  • (b) G={\rm PSL}_2(p),

  • (c) G={\rm AGL}_n(2) for some Mersenne prime p=2^n-1 \ge 31,

  • (d) G={\rm M}_{12} with p=11 or G={\rm M}_{24} with p=23,

each in its natural representation. In case (a) there are exponentially many dessins as p\to\infty, even if we restrict to planar dessins. In case (b) there is a single dessin {\mathcal D}_0(p) for each prime p. In case (c) there is at least one dessin for each Mersenne prime p. In case (d) there are two chiral pairs for each of the two groups G.

Remark

Remark 14.1

WPVI

Here and everywhere, if two dessins \mathcal{D} and \overline{\mathcal{D}} of degree n form a chiral pair and \mathcal{D}=(x,y,z) then \overline{\mathcal{D}} can be obtained as (x',y',z')=(x^{-1},y^{-1},yx). Notice that z'=yx does not necessarily have any canonical form. It can, however, be “standardized” by a common conjugation of x',y',z' in {\rm S}_{n}.

Remark

Remark 14.2

4IKJ

The next Fermat prime, after 17, is 257. Let us give a cursory glance at the group {\rm PSL}_2(256). In this group, there are 128 conjugacy classes of cycles of order 257, and there are eight automorphisms of the field \mathbb{F}_{256}. Therefore, there are 16 non-isomorphic maps generating this group. Their passport is (3^{85}1^2,2^{128}1^1,257^1), and their genus is g=21, in accordance with Theorem 10.1. This theorem also implies that all these maps are mirror symmetric.

Remark

Remark 14.3

KQHT

We see that there are different ways to draw maps on surfaces of genus g>0: compare Figures ?, ?, ?, ?. One may choose a representation according to his or her taste and convenience.

Also, this time, in Figure ?, we decided once again to put the labels explicitly. We recommend to the reader to compare them with the permutations given in the above triple M1 and, specifically, to “go” along the cycle (1,2,\ldots,31).

Source and licence

The text on this page is from arXiv:2104.12015v2 by Gareth A. Jones and Alexander K. Zvonkin, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2104/2104.12015.jsonl.