This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Finite Abelian Groups with Toroidal Subgroup Lattices

Richard A. Moy Lee University, 1120 N. Ocoee St., Cleveland, TN 37311
Abstract.

In this paper, we determine the genus of the subgroup lattice of several families of abelian groups. In doing so, we classify all finite abelian groups whose subgroup lattices can be embedded into the torus.

1. Introduction

In every introductory abstract algebra course, students draw subgroup lattices, and they quickly learn that most subgroup lattices cannot be drawn without crossing edges. However, the groups whose subgroup lattice can be drawn on the plane without crossing edges were determined in [6] and [3]. Their results naturally lead us to the question of which groups have subgroup lattices that can be drawn on the 11-torus without crossing edges. In [2], the REU participants determined which cyclic groups have subgroup lattice that embed into the 11-torus. In this paper, we classify the abelian groups whose subgroup lattice graphs are genus one. Following the notation in [4], we formally define the subgroup lattice graph.

Definition 1.

The subgroup lattice graph Γ(G)\Gamma(G) of a finite group GG is a graph whose vertices are the subgroups of the group and two vertices, H1H_{1} and H2H_{2} are connected by an edge if an only if H1H2H_{1}\leq H_{2} and there is no subgroup KK such that H1<K<H2H_{1}<K<H_{2}.

Definition 2.

The genus γ(Γ)\gamma(\Gamma) of a graph Γ\Gamma is the minimum gg such that there exists an embedding of Γ\Gamma into the orientable surface SgS_{g} of genus gg.

Let n\mathbb{Z}_{n} denote the cyclic group of size nn. We now state the main result of this paper.

Theorem 1.

If GG is a finite abelian group, then γ(Γ(G))=1\gamma(\Gamma(G))=1 if and only if

  • G=nG=\mathbb{Z}_{n} where n=p1e1p2e2p3n=p_{1}^{e_{1}}p_{2}^{e_{2}}p_{3} where p1,p2,p3p_{1},p_{2},p_{3} are distinct primes and (e1,e2){(2,2),(3,2),(3,3)}(e_{1},e_{2})\in\{(2,2),(3,2),(3,3)\}, or

  • G=nG=\mathbb{Z}_{n} where n=p1p2p3p4n=p_{1}p_{2}p_{3}p_{4} where p1,p2,p3,p4p_{1},p_{2},p_{3},p_{4} are distinct primes, or

  • G{4×4,8×8,2×2×q,4×2×q}G\in\{\mathbb{Z}_{4}\times\mathbb{Z}_{4},\mathbb{Z}_{8}\times\mathbb{Z}_{8},\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q},\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}\} where q2q\neq 2 is prime, or

  • G{9×9,3×3×q}G\in\{\mathbb{Z}_{9}\times\mathbb{Z}_{9},\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q}\} where q3q\neq 3 is prime, or

  • G=25×25G=\mathbb{Z}_{25}\times\mathbb{Z}_{25}.

2. Cyclic Groups

In [6], the authors determined all abelian groups whose subgroup lattice graphs are planar. For example, consider the subgroup lattice in Figure 1 of the cyclic group p13p22\mathbb{Z}_{p_{1}^{3}p_{2}^{2}} where p1,p2p_{1},p_{2} are distinct primes. The lattice is simply the lattice of divisors where two divisors are connected if and only if their ratio is prime.

p13p22\langle p_{1}^{3}p_{2}^{2}\ranglep12p22\langle p_{1}^{2}p_{2}^{2}\ranglep1p22\langle p_{1}p_{2}^{2}\ranglep22\langle p_{2}^{2}\ranglep13p2\langle p_{1}^{3}p_{2}\ranglep12p2\langle p_{1}^{2}p_{2}\ranglep1p2\langle p_{1}p_{2}\ranglep2\langle p_{2}\ranglep13\langle p_{1}^{3}\ranglep12\langle p_{1}^{2}\ranglep1\langle p_{1}\rangle1\langle 1\rangle
Figure 1. Subgroup Lattice of p13p22\mathbb{Z}_{p_{1}^{3}p_{2}^{2}}

We readily see that the subgroup lattice of p13p22\mathbb{Z}_{p_{1}^{3}p_{2}^{2}} is simply the graph Cartesian product of the path graph on 3 vertices and the path graph on 4 vertices. This leads us to the following definition in which we follow the notation from [5].

Definition 3.

A kk-dimensional grid graph Γ=Γ(e1,,ek)\Gamma=\Gamma(e_{1},\dots,e_{k}) is the graph Cartesian product of kk paths, where the ithi^{th} path has length eie_{i}.

From this definition, the following lemma follows immediately.

Lemma 1.

The subgroup lattice of the cyclic group p1e1pkek\mathbb{Z}_{p_{1}^{e_{1}}\dots p_{k}^{e_{k}}} where the pip_{i} are distinct primes and eie_{i}\in\mathbb{N} is precisely Γ(e1,,ek)\Gamma(e_{1},\dots,e_{k}).

In [5], the authors prove many results about the genus of grid graphs Γ(e1,,ek)\Gamma(e_{1},\dots,e_{k}). In particular, they prove the following result which lists all grid graphs with genus 0 or 11. The first two bullet points of the below theorem had already been discovered by Starr and Turner [6] and the last two bullet points had been discovered by Berry, McGinnis, and Sanchez [2]. However, these earlier results used the language of subgroup lattice graphs of cyclic groups as opposed to grid graphs.

Proposition 1 ([2, 5, 6]).

A grid graph Γ(e1,,ek)\Gamma(e_{1},\dots,e_{k}) embeds on the torus if and only if

  • k2k\leq 2,

  • k=3k=3 and at most one of e1,e2,e3e_{1},e_{2},e_{3} is greater than 1,

  • k=3k=3 and (e1,e2,e3){(2,2,1),(3,2,1),(3,3,1)}(e_{1},e_{2},e_{3})\in\{(2,2,1),(3,2,1),(3,3,1)\}, or

  • k=4k=4 and (e1,e2,e3,e4)=(1,1,1,1)(e_{1},e_{2},e_{3},e_{4})=(1,1,1,1).

Thus, all cyclic groups whose subgroup lattice graphs are genus zero (the first two bullet points above) or one (the last two bullet points above) have been classified.

2.1. Tools for Finding the Genus of Subgroup Lattice Graphs of Cyclic Groups

Although Proposition 1 resolves the question of which cyclic groups have subgroup lattices of genus one, there are several results in [5] that let us determine the genus of the subgroup lattice of cyclic groups in other cases. One of the most useful results is the exact genus of grid graphs Γ(e1,,ek)\Gamma(e_{1},\dots,e_{k}) for which at least 3 of the eie_{i} are odd. Let E(Γ)E(\Gamma) and V(Γ)V(\Gamma) represent the number of edges and vertices respectively in a grid graph Γ\Gamma. The following result is due to White [8]; however, it has been rephrased in the language of [5].

Proposition 2 (Theorem 4,[8]).

Let Γ=Γ(e1,,ek)\Gamma=\Gamma(e_{1},\dots,e_{k}) be a kk-dimensional grid graph where k3k\geq 3 and at least three eie_{i} are odd, for i=1,,ki=1,\dots,k. Then GG is a lower-embeddable grid graph, i.e.

γ(Γ)=1+|E(Γ)|4|V(Γ)|2=1+12i=1k(ei+1)[12i=1keiei+11].\gamma(\Gamma)=1+\frac{|E(\Gamma)|}{4}-\frac{|V(\Gamma)|}{2}=1+\frac{1}{2}\prod_{i=1}^{k}{(e_{i}+1)}\left[\frac{1}{2}\sum_{i=1}^{k}{\frac{e_{i}}{e_{i}+1}-1}\right].

This leads to the immediate corollary.

Corollary 1.

If G=p1e1pkekG=\mathbb{Z}_{p_{1}^{e_{1}}\dots p_{k}^{e_{k}}} is a cyclic group with k3k\geq 3 and eie_{i} odd for at least three ii, then:

(1) γ(Γ(G))=1+12i=1k(ei+1)[12i=1keiei+11].\gamma(\Gamma(G))=1+\frac{1}{2}\prod_{i=1}^{k}{(e_{i}+1)}\left[\frac{1}{2}\sum_{i=1}^{k}{\frac{e_{i}}{e_{i}+1}-1}\right].

Here is an easy consequence of the previous result.

Corollary 2.

Consider the subgroup lattice graph of the cyclic group G=p1np2p3p4G=\mathbb{Z}_{p_{1}^{n}p_{2}p_{3}p_{4}} where pip_{i} are distinct primes. Then

(2) γ(Γ(G))=1+12(n+1)23[12(nn+1+312)1]=n.\gamma(\Gamma(G))=1+\frac{1}{2}\cdot(n+1)\cdot 2^{3}\cdot\left[\frac{1}{2}\left(\frac{n}{n+1}+3\cdot\frac{1}{2}\right)-1\right]=n.

Another easy consequence is the genus of the kk-hypercube, which is the subgroup lattice graph of p1pk\mathbb{Z}_{p_{1}\dots p_{k}} where pip_{i} are distinct primes.

Corollary 3.

Consider the subgroup lattice graph of the cyclic group G=p1pkG=\mathbb{Z}_{p_{1}\dots p_{k}} where pip_{i} are distinct primes and k3k\geq 3. Then

(3) γ(Γ(G))=1+122k[12i=1k121]=1+2k3(k4).\gamma(\Gamma(G))=1+\frac{1}{2}\cdot 2^{k}\cdot\left[\frac{1}{2}\sum_{i=1}^{k}{\frac{1}{2}}-1\right]=1+2^{k-3}(k-4).

We now state two additional results from [5] that will be needed in the proof of Theorem 1.

Proposition 3 (Theorem 3, [5]).

For any e1,e2e_{1},e_{2}\in\mathbb{N}, we have

(4) γ(Γ(e1,e2,1))=e12e22.\gamma(\Gamma(e_{1},e_{2},1))=\left\lfloor\frac{e_{1}}{2}\right\rfloor\left\lfloor\frac{e_{2}}{2}\right\rfloor.
Proposition 4 (Theorem 4, [5]).

For any e1e_{1}\in\mathbb{N}, we have that

(5) γ(Γ(e1,2,2))=e1.\gamma(\Gamma(e_{1},2,2))=e_{1}.

2.2. Lower and Upper Bounds on the Genus of a Grid Graph

While the previous section provided exact formulas for the genus of various grid graphs, there are several inequalities for the genus that will be useful. The first result is a lower bound for the genus of any grid graph.

Proposition 5 (Corollary 1, [5]).

Let Γ(e1,,ek)\Gamma(e_{1},\dots,e_{k}) be a grid graph with k>1k>1. Then

(6) γ(Γ(e1,,ek))1+|E|4|V|2=1+12i=1k(ei+1)[12i=1keiei+11]\gamma(\Gamma(e_{1},\dots,e_{k}))\geq 1+\frac{|E|}{4}-\frac{|V|}{2}=1+\frac{1}{2}\prod_{i=1}^{k}{(e_{i}+1)}\left[\frac{1}{2}\sum_{i=1}^{k}{\frac{e_{i}}{e_{i}+1}-1}\right]

where |E||E| and |V||V| are the number of edges and vertices respectively in Γ(e1,,ek)\Gamma(e_{1},\dots,e_{k}).

Proposition 6 (Proposition 3, [5]).

Let Γ=Γ(e1,e2,e3)\Gamma=\Gamma(e_{1},e_{2},e_{3}) be a 33-dimensional grid with at least one even grid parameter.

  1. (a)

    If exactly one grid parameter is even (the third parameter), then

    (7) γ(Γ(e1,e2,e3)γ(Γ(e1,e2,e31))+(e1+1)(e2+1)41.\gamma(\Gamma(e_{1},e_{2},e_{3})\leq\gamma(\Gamma(e_{1},e_{2},e_{3}-1))+\frac{(e_{1}+1)(e_{2}+1)}{4}-1.
  2. (b)

    If exactly two grid parameters are even (the second & third parameters), then

    γ(Γ(e1,e2,e3)γ(Γ(e1,e21,e31))+(e1+1)(e2+e3)41.\gamma(\Gamma(e_{1},e_{2},e_{3})\leq\gamma(\Gamma(e_{1},e_{2}-1,e_{3}-1))+\frac{(e_{1}+1)(e_{2}+e_{3})}{4}-1.
  3. (c)

    If all three grid parameters are even, then

    γ(Γ(e1,e2,e3))γ(Γ(e11,e21,e31))+e1e2+e1e3+e2e341\gamma(\Gamma(e_{1},e_{2},e_{3}))\leq\gamma(\Gamma(e_{1}-1,e_{2}-1,e_{3}-1))+\frac{e_{1}e_{2}+e_{1}e_{3}+e_{2}e_{3}}{4}-1

While these formulas are useful, one sometimes needs to show that a graph has a lower bound for its genus by finding a graph minor with a particular genus. When one combines multiple copies of the same graph, e.g. K3,3K_{3,3} one can use the block decomposition of a graph to compute its genus. We follow the discussion of block decompositions found in [5].

Definition 4.

A block of a graph Γ\Gamma is a maximal 22-connected subgraph BB of Γ\Gamma. Given a connected graph Γ\Gamma, there exists a unique collection of blocks ={B1,,Bk}\mathcal{B}=\{B_{1},\dots,B_{k}\} such that i=1kBi=Γ\cup_{i=1}^{k}{B_{i}}=\Gamma, which is called the block decomposition of Γ\Gamma.

The following theorem connecting the genus of a graph to the genus of the blocks in its block decomposition is due to Battle, Harary, Kodama, and Youngs [1].

Theorem 2 (Theorem 1, [1]).

If Γ\Gamma is a graph with block decomposition {B1,,Bk}\{B_{1},\dots,B_{k}\}, then γ(Γ)=i=1kγ(Bi)\gamma(\Gamma)=\sum_{i=1}^{k}{\gamma(B_{i})}.

Using these results, we can prove the following theorem.

Theorem 3.

Let G=nG=\mathbb{Z}_{n} be a cyclic group with subgroup lattice graph Γ(G)\Gamma(G) where n=p1e1pke1n=p_{1}^{e_{1}}\dots p_{k}^{e_{1}} with pip_{i} distinct primes. Then γ(Γ(G))=2\gamma(\Gamma(G))=2 if and only if

  • k=3k=3 and (e1,e2,e3){(4,2,1),(4,3,1),(5,2,1),(5,3,1),(2,2,2)}(e_{1},e_{2},e_{3})\in\{(4,2,1),(4,3,1),(5,2,1),(5,3,1),(2,2,2)\} or

  • k=4k=4 and (e1,e2,e3,e4)=(2,1,1,1)(e_{1},e_{2},e_{3},e_{4})=(2,1,1,1).

We have γ(Γ(G))=3\gamma(\Gamma(G))=3 if and only if

  • k=3k=3 and (e1,e2,e3){(6,2,1),(6,3,1),(7,2,1),(7,3,1),(3,2,2)}(e_{1},e_{2},e_{3})\in\{(6,2,1),(6,3,1),(7,2,1),(7,3,1),(3,2,2)\} or

  • k=4k=4 and (e1,e2,e3,e4)=(3,1,1,1)(e_{1},e_{2},e_{3},e_{4})=(3,1,1,1).

We have γ(Γ(G))=4\gamma(\Gamma(G))=4 if

  • k=3k=3 and

    (e1,e2,e3){((8,2,1),(8,3,1),(9,2,1),(9,3,1),(4,4,1),(5,4,1),(5,5,1),(4,2,2),(3,3,2)}(e_{1},e_{2},e_{3})\in\{((8,2,1),(8,3,1),(9,2,1),(9,3,1),(4,4,1),(5,4,1),(5,5,1),(4,2,2),(3,3,2)\}

    or

  • k=4k=4 and (e1,e2,e3,e4)=(4,1,1,1)(e_{1},e_{2},e_{3},e_{4})=(4,1,1,1).

Remark 1.

The above list of cyclic groups with genus 44 subgroup lattice graphs is exhaustive with the possible exception of p12p22p3p4\mathbb{Z}_{p_{1}^{2}p_{2}^{2}p_{3}p_{4}}. Using (6), one can show it has genus at least 44, and using other results from [5], one can show it has genus at most 66. The exact genus is unknown to the author.

Proof.

Using equations (4), (5), and (2), we can easily show that all the subgroup lattice graphs above have the specified genus with the exception of G=p13p23p32G=\mathbb{Z}_{p_{1}^{3}p_{2}^{3}p_{3}^{2}} where pip_{i} are distinct primes.

Using (7), we see that γ(Γ(p13p23p32))4\gamma(\Gamma(\mathbb{Z}_{p_{1}^{3}p_{2}^{3}p_{3}^{2}}))\leq 4. Now, we must show it is at least four. Below, we have drawn a graph that is combination of four K3,3K_{3,3} graphs.

[Uncaptioned image]

Omit all the dotted edges and black vertices. Contract the blue, green, purple, brown, and orange paths to single vertices. Note that there are two separate blue paths, one with solid blue vertices and one with blue vertices outlined with a black circle. These paths are contracted to two different vertices. The same phenomenon occurs with the green, orange, and brown vertices. Doing so results in the following minor of Γ(p13p23p32)\Gamma(\mathbb{Z}_{p_{1}^{3}p_{2}^{3}p_{3}^{2}}).

[Uncaptioned image]

Using Theorem 2 and the fact that γ(K3,3)=1\gamma(K_{3,3})=1 we deduce that the genus of the above graph minor is 44. Hence, we have deduced that γ(Γ(p13p23p32))=4.\gamma(\Gamma(\mathbb{Z}_{p_{1}^{3}p_{2}^{3}p_{3}^{2}}))=4.

Now, we must show that cyclic groups not appearing in the list above or in the list from Proposition 1 have genus at least 5 with the exception of Γ(p12p22p3p4)\Gamma(\mathbb{Z}_{p_{1}^{2}p_{2}^{2}p_{3}p_{4}}). For the remainder of the proof, suppose that GG is a cyclic group not appearing in the list of groups in Proposition 1 or Theorem 3.

Suppose k=3k=3 and e1e2e3e_{1}\geq e_{2}\geq e_{3}.

If e3=1e_{3}=1 and e2=2e_{2}=2 or 33, then e110e_{1}\geq 10 and thus γ(Γ(G))5\gamma(\Gamma(G))\geq 5 by (4). If e3=1e_{3}=1 and e2=4e_{2}=4 or 55, then e16e_{1}\geq 6 and thus γ(Γ(G))6\gamma(\Gamma(G))\geq 6 by (4). If e3=1e_{3}=1 and e26e_{2}\geq 6 then e16e_{1}\geq 6 and thus γ(Γ(G))9\gamma(\Gamma(G))\geq 9 by (4).

If e3=2e_{3}=2 and e2=2e_{2}=2 then e15e_{1}\geq 5 and thus γ(Γ(G))5\gamma(\Gamma(G))\geq 5 by (5). If e3=2e_{3}=2 and e23e_{2}\geq 3 then e14e_{1}\geq 4 and thus γ(Γ(G))>4\gamma(\Gamma(G))>4 by (6).

If e33e_{3}\geq 3 then e1,e23e_{1},e_{2}\geq 3. Since γ(Γ(p13p23p33))=5\gamma(\Gamma(\mathbb{Z}_{p_{1}^{3}p_{2}^{3}p_{3}^{3}}))=5 using (1), we deduce that γ(Γ(G))5\gamma(\Gamma(G))\geq 5.

Suppose k=4k=4 and e1e2e3e4e_{1}\geq e_{2}\geq e_{3}\geq e_{4}. If e2=e3=e4=1e_{2}=e_{3}=e_{4}=1, then e15e_{1}\geq 5 and thus γ(Γ(G))5\gamma(\Gamma(G))\geq 5 by (2). If e3=e4=1e_{3}=e_{4}=1 and e2=2e_{2}=2 then e13e_{1}\geq 3 and thus γ(Γ(G))6\gamma(\Gamma(G))\geq 6 by (6). If e3=e4=1e_{3}=e_{4}=1 and e23e_{2}\geq 3 then e13e_{1}\geq 3 and γ(Γ(G))9\gamma(\Gamma(G))\geq 9 by (6). If e32e_{3}\geq 2 then e1e22e_{1}\geq e_{2}\geq 2 and thus γ(Γ(G))>7\gamma(\Gamma(G))>7 by (6).

Suppose k=5k=5 and e1e2e3e4e5e_{1}\geq e_{2}\geq e_{3}\geq e_{4}\geq e_{5}. Then we know γ(Γ(G))5\gamma(\Gamma(G))\geq 5 since γ(Γ(p1p2p3p4p5))=5\gamma(\Gamma(\mathbb{Z}_{p_{1}p_{2}p_{3}p_{4}p_{5}}))=5 by (3). ∎

Thus, with the exception of p12p22p3p4\mathbb{Z}_{p_{1}^{2}p_{2}^{2}p_{3}p_{4}}, we have completed the classification of cyclic groups with subgroup lattice graphs of genus up to 44.

3. Non-Cyclic Abelian Groups

Now that we have classified all cyclic groups whose subgroup lattice graph has genus up to 4 with the exception of p12p22p3p4\mathbb{Z}_{p_{1}^{2}p_{2}^{2}p_{3}p_{4}}, we proceed with determining the non-cyclic abelian groups whose subgroup lattice has genus one. The following theorem classifies all non-cyclic abelian groups whose subgroup lattices have genus zero. The planar cyclic groups are described in the first two bullet points of Proposition 1.

Theorem 4 (Theorem 14, [6]).

Let GG be a finite abelian group. Then Γ(G)\Gamma(G) is planar if and only if GG is isomorphic to a planar cyclic group or to pa×p\mathbb{Z}_{p^{a}}\times\mathbb{Z}_{p} where pp is a prime and aa\in\mathbb{N}.

First, we examine a few families of graphs that will be needed to prove Theorem 1.

Lemma 2.

Let GnG_{n} be the graph pictured below.

[Uncaptioned image]

The genus of GnG_{n} is γ(Gn)=n24\gamma(G_{n})=\lceil\frac{n-2}{4}\rceil.

Proof.

The graph GnG_{n} has a clear K3,nK_{3,n} minor and γ(K3,n)=n24\gamma(K_{3,n})=\lceil\frac{n-2}{4}\rceil. Hence, γ(Gn)n24\gamma(G_{n})\geq\lceil\frac{n-2}{4}\rceil. To demonstrate that GnG_{n} can be drawn on a surface of genus n24\lceil\frac{n-2}{4}\rceil, we will construct a set of cycles that traverse each edge in both directions (e.g. define the faces of the embedding of the graph on a surface). We will first consider the case when n2mod4n\equiv 2\mod 4. The cycles are as follows:

  • (b,α1,β1)(b,\alpha_{1},\beta_{1}), (b,β2,α2)(b,\beta_{2},\alpha_{2}), (b,α3,β3)(b,\alpha_{3},\beta_{3}), (b,β4,α4)(b,\beta_{4},\alpha_{4}), \dots, (b,βn,αn)(b,\beta_{n},\alpha_{n})

  • (b,β1,c,β2)(b,\beta_{1},c,\beta_{2}), (b,β3,c,β4)(b,\beta_{3},c,\beta_{4}), \dots, (b,βn1,c,βn)(b,\beta_{n-1},c,\beta_{n})

  • (b,α2,a,α3)(b,\alpha_{2},a,\alpha_{3}), (b,α4,a,α5)(b,\alpha_{4},a,\alpha_{5}), \dots, (b,αn,a,α1)(b,\alpha_{n},a,\alpha_{1})

  • (c,β1,α1,a,α1+n2,β1+n2)(c,\beta_{1},\alpha_{1},a,\alpha_{1+\frac{n}{2}},\beta_{1+\frac{n}{2}}), (c,β3,α3,a,α3+n2,β3+n2)(c,\beta_{3},\alpha_{3},a,\alpha_{3+\frac{n}{2}},\beta_{3+\frac{n}{2}}), \dots, (c,βn2,αn2,a,αn,βn)(c,\beta_{\frac{n}{2}},\alpha_{\frac{n}{2}},a,\alpha_{n},\beta_{n}),
    (c,βn2+2,αn2+2,a,α2,β2)(c,\beta_{\frac{n}{2}+2},\alpha_{\frac{n}{2}+2},a,\alpha_{2},\beta_{2}), \dots, (c,βn1,αn1,a,αn21,βn21)(c,\beta_{n-1},\alpha_{n-1},a,\alpha_{\frac{n}{2}-1},\beta_{\frac{n}{2}-1})

Using these cycles, observe that this embedding of the graph GnG_{n} into some surface has 5n2\frac{5n}{2} faces, 2n+32n+3 vertices, and 5n5n edges. Using Euler’s formula, we deduce that 22g=5n25n+2n+3=n2+3.2-2g=\frac{5n}{2}-5n+2n+3=-\frac{n}{2}+3. This equation yields g=n24=n24g=\frac{n-2}{4}=\lceil\frac{n-2}{4}\rceil since n2mod4n\equiv 2\mod 4. Observe that the fact that n2mod4n\equiv 2\mod 4 is used inherently in the definition of the 66-cycles in the fourth bullet point.

Thus we have proved our result in the case when n2mod4n\equiv 2\mod 4. The other cases follow from the fact that GnG_{n} is a subgraph of Gn+1G_{n+1}, Gn+2G_{n+2} and Gn+2G_{n+2}. If n2mod4n\not\equiv 2\mod 4, there exists a unique integer m{n+1,n+2,n+3}m\in\{n+1,n+2,n+3\} satisfying m2mod4m\equiv 2\mod 4 with n24=m24\lceil\frac{n-2}{4}\rceil=\lceil\frac{m-2}{4}\rceil. ∎

We then use this result to prove the genus of Γ(p2×p2)\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}).

Proposition 7.

For a prime pp, γ(Γ(p2×p2))=p14\gamma(\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}))=\lceil\frac{p-1}{4}\rceil.

Proof.

The subgraph lattice graph of p2×p2\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}} is pictured below in Figure 2.

Refer to caption
Figure 2. Subgroup Lattice Graph of p2×p2\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}

The red vertices represent the p+1p+1 subgroups of size pp. The blue vertices represent the p2+pp^{2}+p cyclic subgroups of size p2p^{2} and the orange vertex represents the unique non-cyclic subgroup of size p2p^{2}. The green vertices represent the p+1p+1 subgroups of size p3p^{3}.

It is easy to see that Γ(p2×p2)\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}) has a Gp+1G_{p+1} minor by merging all the blue vertices with the respective red vertex they connect to. Hence, γ(Γ(p2×p2))p14\gamma(\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}))\geq\lceil\frac{p-1}{4}\rceil. It is also relatively easy to see that since Gp+1G_{p+1} embeds onto a surface of genus p14\lceil\frac{p-1}{4}\rceil, we can also embed Γ(p2×p2)\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}) on the same surface. This is because one can cut out a small neighborhood of the edges α1β1,αp+1βp+1\alpha_{1}\beta_{1},\dots\alpha_{p+1}\beta_{p+1} and replace it by the corresponding red vertex, green vertex, and pp blue vertices along with the edges that connect them. For example, see Figure 3.

Refer to caption
Figure 3.

Hence, we have proven that γ(Γ(p2×p2))=p14\gamma(\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}))=\lceil\frac{p-1}{4}\rceil. ∎

Lemma 3.

Let HnH_{n} be the graph pictured below.

[Uncaptioned image]

If nn is odd, the genus of HnH_{n} is γ(Hn)=2n24\gamma(H_{n})=2\lceil\frac{n-2}{4}\rceil.

Proof.

Observe that HnH_{n} contains two copies of GnG_{n}. By deleting four of the red edges and using Theorem 2, one deduces that γ(Hn)2n24\gamma(H_{n})\geq 2\lceil\frac{n-2}{4}\rceil.

Now, we consider the case when n1mod4n\equiv 1\mod 4. To demonstrate that HnH_{n} can be drawn on a surface of genus 2n242\lceil\frac{n-2}{4}\rceil, we will construct a set of cycles that traverse each edge in both directions exactly once. The cycles are as follows:

  1. (a)

    (b0,α1,β1)(b_{0},\alpha_{1},\beta_{1}), (b0,β2,α2)(b_{0},\beta_{2},\alpha_{2}), …, (b0,βn1,αn1)(b_{0},\beta_{n-1},\alpha_{n-1}), (b0,αn,βn)(b_{0},\alpha_{n},\beta_{n})

  2. (b)

    (b1,δ1,γ1)(b_{1},\delta_{1},\gamma_{1}), (b1,γ2,δ2)(b_{1},\gamma_{2},\delta_{2}), …, (b1,γn1,δn1)(b_{1},\gamma_{n-1},\delta_{n-1}), (b1,δn,γn)(b_{1},\delta_{n},\gamma_{n})

  3. (c)

    (b0,β1,c0,β2)(b_{0},\beta_{1},c_{0},\beta_{2}), (b0,α2,a0,α3)(b_{0},\alpha_{2},a_{0},\alpha_{3}), …, (b0,βn2,c0,βn1)(b_{0},\beta_{n-2},c_{0},\beta_{n-1}), (b0,αn1,a0,αn)(b_{0},\alpha_{n-1},a_{0},\alpha_{n})

  4. (d)

    (b1,δ2,c1,δ1)(b_{1},\delta_{2},c_{1},\delta_{1}), (b1,γ3,a1,γ2)(b_{1},\gamma_{3},a_{1},\gamma_{2}), …, (b1,δn2,c1,δn1)(b_{1},\delta_{n-2},c_{1},\delta_{n-1}), (b1,γn,a1,γn1)(b_{1},\gamma_{n},a_{1},\gamma_{n-1})

  5. (e)

    (c0,β1,α1,a0,α(n+3)/2,β(n+3)/2)(c_{0},\beta_{1},\alpha_{1},a_{0},\alpha_{(n+3)/2},\beta_{(n+3)/2}), (a0,α2,β2,c0,β(n+5)/2,α(n+5)/2)(a_{0},\alpha_{2},\beta_{2},c_{0},\beta_{(n+5)/2},\alpha_{(n+5)/2}), …,
    (c0,β(n3)/2,α(n3)/2,a0,αn1,βn1)(c_{0},\beta_{(n-3)/2},\alpha_{(n-3)/2},a_{0},\alpha_{n-1},\beta_{n-1}), (a0,α(n1)/2,β(n1)/2,c0,βn,αn)(a_{0},\alpha_{(n-1)/2},\beta_{(n-1)/2},c_{0},\beta_{n},\alpha_{n})

  6. (f)

    (a1,γ1,δ1,c1,δ(n+3)/2,γ(n+3)/2)(a_{1},\gamma_{1},\delta_{1},c_{1},\delta_{(n+3)/2},\gamma_{(n+3)/2}), (c1,δ2,γ2,a1,γ(n+5)/2,δ(n+5)/2)(c_{1},\delta_{2},\gamma_{2},a_{1},\gamma_{(n+5)/2},\delta_{(n+5)/2}), …,
    (a1,γ(n3)/2,δ(n3)/2,c1,δn1,γn1)(a_{1},\gamma_{(n-3)/2},\delta_{(n-3)/2},c_{1},\delta_{n-1},\gamma_{n-1}), (c1,δ(n1)/2,γ(n1)/2,a1,γn,δn)(c_{1},\delta_{(n-1)/2},\gamma_{(n-1)/2},a_{1},\gamma_{n},\delta_{n})

  7. (g)

    (b0,aa,a0,α1)(b_{0},a_{a},a_{0},\alpha_{1}), (b0,βn,c0,b1)(b_{0},\beta_{n},c_{0},b_{1}), (b0,b1,γ1,a1)(b_{0},b_{1},\gamma_{1},a_{1}), (c1,δn,b1,c0)(c_{1},\delta_{n},b_{1},c_{0})

  8. (h)

    (c0,β(n+1)/2,α(n+1)/2,a0,a1,γ(n+1)/2,δ(n+1)/2,c1)(c_{0},\beta_{(n+1)/2},\alpha_{(n+1)/2},a_{0},a_{1},\gamma_{(n+1)/2},\delta_{(n+1)/2},c_{1})

Using these cycles, observe that this embedding of the graph HnH_{n} into some surface has 5n+25n+2 faces, 4n+64n+6 vertices, and 10n+510n+5 edges. Using Euler’s formula, we deduce that 22g=5n+2(10n+5)+4n+6=n+3.2-2g=5n+2-(10n+5)+4n+6=-n+3. This equation yields g=n12=2n24g=\frac{n-1}{2}=2\lceil\frac{n-2}{4}\rceil since n1mod4n\equiv 1\mod 4. Observe that we inherently used the fact that n1mod4n\equiv 1\mod 4 when we constructed the 66-cycles in bullets (e) and (f) since we need n12\frac{n-1}{2} to be even to have an equal number of cycles beginning with c0c_{0} versus a0a_{0} (or a1a_{1} versus c1c_{1}) in our list of 66-cycles.

Thus we have proved our result in the case when n1mod4n\equiv 1\mod 4. The case when n3mod4n\equiv 3\mod 4 follows from the fact that HnH_{n} is a subgraph of Hn+2H_{n+2}. ∎

Proposition 8.

For an odd prime pp, γ(Γ(p3×p2))=2p24\gamma(\Gamma(\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}}))=2\lceil\frac{p-2}{4}\rceil. When p=2p=2, γ(Γ(8×4))=1\gamma(\Gamma(\mathbb{Z}_{8}\times\mathbb{Z}_{4}))=1.

Proof.

The subgroup lattice of p3×p2\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}} is pictured below in Figure 4.

Refer to caption
Figure 4. Subgroup Lattice Graph of p3×p2\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}}

The red vertices represent the p+1p+1 subgroups of size pp, and the red vertex with a black outline is the subgroup generated by (p2,0)(p^{2},0). The black vertices in the third row from the bottom represent the p(p+1)p(p+1) cyclic subgroups of size p2p^{2}, and the white vertex represents the unique non-cyclic subgroup of size p2p^{2}. The blue vertices without a black outline represent the p2p^{2} cyclic subgroups of size p3p^{3} and the p+1p+1 blue vertices with a black outline represent the p+1p+1 non-cyclic subgroups of size p3p^{3}. Lastly, the green vertices are the p+1p+1 subgroups of size p4p^{4}, and the green vertex outlined in black unique subgroup of size p4p^{4} that is entirely p2p^{2} torsion.

It is relatively easy to see that Γ(p3×p2)\Gamma(\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}}) has a HpH_{p} minor. Hence, γ(Γ(p3×p2))2p24\gamma(\Gamma(\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}}))\geq 2\lceil\frac{p-2}{4}\rceil. It is also relatively easy to see that since HnH_{n} embeds onto a surface of genus 2p242\lceil\frac{p-2}{4}\rceil, we can embed Γ(p3×p2)\Gamma(\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}}) on the same surface. This is because one can cut out a small neighborhood of the edges γ1δ1,γp+1δp+1\gamma_{1}\delta_{1},\dots\gamma_{p+1}\delta_{p+1} and replace it by the corresponding black vertex, green vertex, and pp blue vertices along with the edges that connect them. For example, see Figure 5 below.

Refer to caption
Figure 5.

A similar operation allows one to cut out a small neighborhood of the edges α1β1,αp+1βp+1\alpha_{1}\beta_{1},\dots\alpha_{p+1}\beta_{p+1} and replace it by the corresponding red vertex, blue vertex, and pp black vertices along with the edges that connect them.

Now consider the case when p=2p=2. We know that γ(Γ(8×4))1\gamma(\Gamma(\mathbb{Z}_{8}\times\mathbb{Z}_{4}))\geq 1 by Theorem 4. To demonstrate that γ(Γ(8×4))=1\gamma(\Gamma(\mathbb{Z}_{8}\times\mathbb{Z}_{4}))=1, we simply need to show that we can draw Γ(8×4)\Gamma(\mathbb{Z}_{8}\times\mathbb{Z}_{4}) on the 1-torus which we do here in Figure 6.

Refer to caption
Figure 6. Subgroup Lattice of 8×4\mathbb{Z}_{8}\times\mathbb{Z}_{4} Drawn on the Torus

Proposition 9.

The genus of Γ(p×p×q)\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q}) for primes pqp\neq q is p12\lceil\frac{p-1}{2}\rceil.

Proof.

The subgroup lattice of p×p×q\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q} is pictured below in Figure 7.

Refer to caption
Figure 7. Subgroup Lattice Graph of p×p×q\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q}

The blue vertices represent the p+1p+1 subgroups of size pp, the aa vertex represents the trivial subgroup, the bb vertex represents the unique subgroup of size p2p^{2}, the cc vertex represents the unique subgroup of size qq, the green vertices represent the p+1p+1 subgroups of size pqpq and the vertex dd is the full group.

If one contracts the edges 0a0c,,papc0_{a}0_{c},\dots,p_{a}p_{c} and deletes edges bdbd and acac, then one obtains a K4,p+1K_{4,p+1} graph with genus (42)(p+12)4=p12\lceil\frac{(4-2)(p+1-2)}{4}\rceil=\lceil\frac{p-1}{2}\rceil. Hence, γ(Γ(p×p×q))p12\gamma(\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q}))\geq\lceil\frac{p-1}{2}\rceil.

First, we will prove the case when pp is an oddodd prime. To show that we can embed the subgroup lattice on a surface of genus p12\lceil\frac{p-1}{2}\rceil, we will exhibit a set of cycles that traverse each edge in both directions exactly once. The cycles are as follows:

  • (a,0a,b,1a)(a,0_{a},b,1_{a}), (a,2a,b,3a)(a,2_{a},b,3_{a}), \dots, (a,(p1)a,b,pa)(a,(p-1)_{a},b,p_{a})

  • (c,1c,d,0c)(c,1_{c},d,0_{c}), (c,3c,d,2c)(c,3_{c},d,2_{c}), \dots, (c,pc,d,(p1)c)(c,p_{c},d,(p-1)_{c})

  • (a,c,0c,0a)(a,c,0_{c},0_{a}), (a,pa,pc,c)(a,p_{a},p_{c},c)

  • (b,d,1c,1a)(b,d,1_{c},1_{a}), (b,2a,2c,d)(b,2_{a},2_{c},d)

  • (a,1a,1c,c,2c,2a)(a,1_{a},1_{c},c,2_{c},2_{a}), (a,3a,3c,c,4c,4a)(a,3_{a},3_{c},c,4_{c},4_{a}), \dots, (a,(p2)a,(p2)c,c,(p1)c,(p1)a)(a,(p-2)_{a},(p-2)_{c},c,(p-1)_{c},(p-1)_{a})

  • (b,0a,0c,d,pc,pa)(b,0_{a},0_{c},d,p_{c},p_{a}), (b,4a,4c,d,3c,3a)(b,4_{a},4_{c},d,3_{c},3_{a}), \dots, (b,(p1)a,(p1)c,d,(p2)c,(p2))a)(b,(p-1)_{a},(p-1)_{c},d,(p-2)_{c},(p-2))a)

Using these cycles, observe that this embedding of Γ(p×p×q)\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q}) into some surface has 2p+42p+4 faces, 2(p+3)2(p+3) vertices, and 5p+75p+7 edges. Using Euler’s formula, we deduce that 22g=2p+4(5p+7)+2p+6=p+32-2g=2p+4-(5p+7)+2p+6=-p+3. This equation yields g=p12=p12g=\frac{p-1}{2}=\lceil\frac{p-1}{2}\rceil since pp is an odd prime. Hence, we have shown that γ(Γ(p×p×q))=p12\gamma(\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q}))=\lceil\frac{p-1}{2}\rceil when pp is an odd prime.

When p=2p=2, we know that γ(Γ(2×2×q))>0\gamma(\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}))>0 by Theorem 4. Since, Γ(2×2×q)\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}) is a subgraph of Γ(3×3×q)\Gamma(\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q}), we deduce that γ(Γ(2×2×q))=1=212\gamma(\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}))=1=\lceil\frac{2-1}{2}\rceil. ∎

Now, we compile our results to prove that all the graphs in the statement of Theorem 1 have genus one.

Proposition 10.

The following groups have subgroup lattices with genus one.

  • G{4×4,8×4,2×2×p,4×2×q}G\in\{\mathbb{Z}_{4}\times\mathbb{Z}_{4},\mathbb{Z}_{8}\times\mathbb{Z}_{4},\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{p},\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}\} where q2q\neq 2 is prime, or

  • G{9×9,3×3×q}G\in\{\mathbb{Z}_{9}\times\mathbb{Z}_{9},\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q}\} where q3q\neq 3 is prime, or

  • G=25×25G=\mathbb{Z}_{25}\times\mathbb{Z}_{25}.

Proof.

All the groups listed in this proposition, with the exception of 4×2×q\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}, have already been shown to have subgroup lattice graphs of genus one via Propositions 7, 8, and 9. Thus all that remains is to show the same result for 4×2×q\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}. Since it does not appear in the list of planar groups in Theorem 4, it must have genus at least one. So all we must do is show that its subgroup lattice graph can be drawn on the 11-torus

The subgroup lattice of 4×2×q\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q} is below in Figure 8. The blue vertices represent the three cyclic subgroups of size 22 and the blue vertex outlined in black is the subgroup generated by (2,0,0)(2,0,0). The red vertex is the unique non-cyclic subgroup of size 44. The green vertices represent the three cyclic subgroups of size 2q2q and the green vertex outlined in black is the subgroup generated by (2,0,1)(2,0,1). The orange vertex is the unique non-cyclic subgroup of size 4q4q.

Refer to caption
Figure 8. Subgroup Lattice Graph of 4×2×q\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}

Figure 9 shows how Γ(4×2×q)\Gamma(\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}) can be drawn on the 11-torus.

Refer to caption
Figure 9. Subgroup Lattice Graph of 4×2×q\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q} Drawn on the Torus

3.1. Eliminating Graphs

In this subsection, we will determine why various families of graphs have genus at least two and thus cannot be embedded into the 1-torus. First, recall that the girth of a graph is the length of its shortest cycle. Now, we recall the following result.

Proposition 11 (Proposition 1, [5]).

Suppose Γ\Gamma is a connected graph whose girth is at least 44. Then γ(Γ)1+|E|4|V|2\gamma(\Gamma)\geq 1+\frac{|E|}{4}-\frac{|V|}{2}.

Corollary 4.

If GG is a finite group whose subgroup lattice is not a tree then

(8) γ(Γ(G))1+|E|4|V|2.\gamma(\Gamma(G))\geq 1+\frac{|E|}{4}-\frac{|V|}{2}.
Remark 2.

The only abelian groups whose subgroup lattice is a tree are pk\mathbb{Z}_{p^{k}}.

Proof.

We proceed by showing that a subgroup lattice graph has girth at least 44. Clearly a subgroup lattice cannot have a cycle of length two since there is at most one edge between any two vertices. Could there be three subgroups S1,S2,S3S_{1},S_{2},S_{3} that form the vertices of a 33-cycle in the subgroup lattice? If so, one deduces that one subgroup, say S1S_{1}, is a subgroup of the other two. Furthermore, one of the remaining subgroups, say S2S_{2}, is a subgroup of the last subgroup, S3S_{3}. Hence S1<S2<S3S_{1}<S_{2}<S_{3}. This is a contradiction with the fact that S1S_{1} and S3S_{3} are connected by an edge in the subgroup lattice graph. Hence, we conclude that a 33-cycle is impossible in a subgroup lattice graph. ∎

Lemma 4.

If pp is a prime, then γ(Γ(p×p×p))2\gamma(\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{p}))\geq 2.

Proof.

There are p2+p+1p^{2}+p+1 subgroups of size pp and p2+p+1p^{2}+p+1 subgroups of size p2p^{2}. Since a subgroup of size p2p^{2} contains p21p^{2}-1 elements of order pp, it thus contains p21p1=p+1\frac{p^{2}-1}{p-1}=p+1 subgroups of size pp. Thus, each subgroup of size p2p^{2} connects to p+1p+1 subgroups of size pp.

Using (8), the subgroup lattice graph has 2(p2+p+1)+2=2p2+2p+42(p^{2}+p+1)+2=2p^{2}+2p+4 vertices and 2(p2+p+1)+(p+1)(p2+p+1)=p3+4p2+4p+32(p^{2}+p+1)+(p+1)(p^{2}+p+1)=p^{3}+4p^{2}+4p+3 edges. Since the girth of the graph is 44,

γ(Γ(p×p×p))1+p3+4p2+4p+342p2+2p+42=14(p31).\gamma(\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{p}))\geq 1+\frac{p^{3}+4p^{2}+4p+3}{4}-\frac{2p^{2}+2p+4}{2}=\frac{1}{4}(p^{3}-1).

This is strictly bigger than 11 since p2p\geq 2. ∎

Lemma 5.

The subgroup lattices of the groups below have genus at least 22.

  • 2×2×3×3\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{3}\times\mathbb{Z}_{3}

  • 4×4×q\mathbb{Z}_{4}\times\mathbb{Z}_{4}\times\mathbb{Z}_{q} for a prime q2q\neq 2

  • 2×2×q1×q2\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}} for distinct primes q1,q22q_{1},q_{2}\neq 2

  • 3×3×q1×q2\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}} for distinct primes q1,q23q_{1},q_{2}\neq 3

Proof.

The graph Γ(2×2×3×3)\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{3}\times\mathbb{Z}_{3}) has 30 vertices and 76 edges. Using (8), one computes that γ(Γ(2×2×3×3))1+764302=5.\gamma(\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{3}\times\mathbb{Z}_{3}))\geq 1+\frac{76}{4}-\frac{30}{2}=5.

For a prime q2q\neq 2, the graph Γ(4×4×q)\Gamma(\mathbb{Z}_{4}\times\mathbb{Z}_{4}\times\mathbb{Z}_{q}) has 30 vertices and 63 edges. Using (8), one computes that γ(Γ(4×4×q))1+634302=1.75.\gamma(\Gamma(\mathbb{Z}_{4}\times\mathbb{Z}_{4}\times\mathbb{Z}_{q}))\geq 1+\frac{63}{4}-\frac{30}{2}=1.75.

For distinct primes q1,q22q_{1},q_{2}\neq 2, the graph Γ(2×2×q1×q2)\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}}) has 20 vertices and 44 edges. Using (8), one computes that γ(Γ(2×2×q1×q2))1+444202=2.\gamma(\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}}))\geq 1+\frac{44}{4}-\frac{20}{2}=2.

For distinct primes q1,q23q_{1},q_{2}\neq 3, the graph Γ(3×3×q1×q2)\Gamma(\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}}) has 24 vertices and 56 edges. Using (8), one computes that γ(Γ(3×3×q1×q2))1+564242=3.\gamma(\Gamma(\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}}))\geq 1+\frac{56}{4}-\frac{24}{2}=3.

Proposition 12.

The following groups have subgroup lattice graphs with genus at least 22.

  • For a prime q2q\neq 2, 8×2×q\mathbb{Z}_{8}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q}

  • For a prime q3q\neq 3, 9×3×q\mathbb{Z}_{9}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q}

  • 16×4\mathbb{Z}_{16}\times\mathbb{Z}_{4}

  • 8×8\mathbb{Z}_{8}\times\mathbb{Z}_{8}

  • For a prime q2q\neq 2, 2×2×q2\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}} and 4×2×q2\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}}

Proof.

To prove this proposition in every case except 4×2×q2\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}}, we will display minors with genus 22 of the subgroup lattice graphs of the groups listed above. The graphs with the identified minors are displayed below in Figures 10, 11, 12, 13, and 14.

Refer to caption
Figure 10. Subgroup Lattice Graph of 8×2×q\mathbb{Z}_{8}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q} for a Prime q2q\neq 2
Refer to caption
Figure 11. Subgroup Lattice Graph of 9×3×q\mathbb{Z}_{9}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q} for a Prime q3q\neq 3
Refer to caption
Figure 12. Subgroup Lattice Graph of 16×4\mathbb{Z}_{16}\times\mathbb{Z}_{4}
Refer to caption
Figure 13. Subgroup Lattice Graph of 8×8\mathbb{Z}_{8}\times\mathbb{Z}_{8}
Refer to caption
Figure 14. Subgroup Lattice Graph of 2×2×q2\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}} for a Prime q2q\neq 2

In all of the subgroup lattice graph diagrams in Figures 10, 11, 12, 13, and 14, merging all the red vertices with a black outline and deleting all the dotted edges results in the following graph minor.

[Uncaptioned image]

By Theorem 2 and the fact that γ(K3,3)=1\gamma(K_{3,3})=1, we have shown that each of the above graphs have genus at least 22 except in the case of 4×2×q2\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}}. This result follows immediately from the fact that Γ(2×2×q2)\Gamma(\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}}) is a subgraph of Γ(4×2×q2)\Gamma(\mathbb{Z}_{4}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q^{2}}). ∎

Lemma 6.

The genus of Γ(p×p×q2)\Gamma(\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q^{2}}) for primes pqp\neq q is greater than or equal to p1p-1.

Proof.

The subgroup lattice of p×p×q2\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q^{2}} is pictured below.

Refer to caption
Figure 15. Subgroup Lattice Graph of p×p×q2\mathbb{Z}_{p}\times\mathbb{Z}_{p}\times\mathbb{Z}_{q^{2}}

The vertices between each pair of red vertices represent p+1p+1 subgroups, and there are 3p+33p+3 red vertices total. If we merge the red vertices in groups of three as shown in the picture and remove the dotted edges, we obtain a K6,p+1K_{6,p+1} graph which has genus (62)(p+12)4=p1\lceil\frac{(6-2)(p+1-2)}{4}\rceil=p-1.

3.2. Proof of Theorem 1

Now, we can proceed with the proof of the main theorem.

Proof of Theorem 1.

Proposition 1 classifies all cyclic groups with subgroup lattice graphs of genus 11 which are p12p22p3\mathbb{Z}_{p_{1}^{2}p_{2}^{2}p_{3}}, p13p22p3\mathbb{Z}_{p_{1}^{3}p_{2}^{2}p_{3}}, p13p23p3\mathbb{Z}_{p_{1}^{3}p_{2}^{3}p_{3}} and p1p2p3p4\mathbb{Z}_{p_{1}p_{2}p_{3}p_{4}} where the pip_{i} are distinct primes.

We now turn to the non-cyclic abelian groups. Proposition 10 demonstrates that every non-cyclic abelian group listed in Theorem 1 has genus one.

Now, we must demonstrate that every non-cyclic abelian group not appearing in Theorem 1 or Theorem 4 has a subgroup lattice with genus at least two.

Let

(9) G=i=1npikiG=\bigoplus_{i=1}^{n}{\mathbb{Z}_{p_{i}^{k_{i}}}}

be a non-cyclic abelian group with a subgroup lattice group with genus 11 that does not appear in the list of Theorem 1 or Theorem 4. Since GG is not cyclic, pi=pjp_{i}=p_{j} for some iji\neq j in (9).

First, let’s consider the case where all the pip_{i} are the same in (9). Lemma 4 shows that n=2n=2 necessarily. Because of Theorem 4, we only need to investigate the case k1,k22k_{1},k_{2}\geq 2. Proposition 7 shows us that γ(Γ(p2×p2))2\gamma(\Gamma(\mathbb{Z}_{p^{2}}\times\mathbb{Z}_{p^{2}}))\geq 2 when p7p\geq 7. Now, we need only consider the case where p=2p=2, 33, or 55 and k13k_{1}\geq 3 and k22k_{2}\geq 2. Proposition 8 shows us that γ(Γ(p3×p2))2\gamma(\Gamma(\mathbb{Z}_{p^{3}}\times\mathbb{Z}_{p^{2}}))\geq 2 when p3.p\geq 3. So, we are reduced to considering cases where p=2p=2 and k14k_{1}\geq 4 and k22k_{2}\geq 2 or k13k_{1}\geq 3 and k23k_{2}\geq 3. Proposition 12 shows that these groups have subgroup lattices with genus at least 22. Hence, Theorem 1 contains the complete classification of non-cyclic abelian group with subgroup lattice with genus equal to one in the case where all the pip_{i} are the same in (9).

Now, we consider the case where the collection of primes {pi}\{p_{i}\} in (9) contains exactly two primes. Recall that at least one of the primes must be repeated in order for GG to be non-cyclic. Proposition 9 shows that if γ(Γ(G))=1\gamma(\Gamma(G))=1, then the prime repeated must be 22 or 33.

Consider the case where p1=p2=3p_{1}=p_{2}=3. Proposition 12 forces k1=k2=1k_{1}=k_{2}=1. Since GG cannot appear in the list of Theorem 1, if n=3n=3, then k3>1k_{3}>1 necessarily. However, Lemma 6 rules out this case. Hence, we restrict ourselves to the case n=4n=4 which implies p3=p4=2p_{3}=p_{4}=2 necessarily. However, Lemma 5 rules out this possibility.

Now consider the case where p1=p2=2p_{1}=p_{2}=2. Proposition 12 forces either k1=k2=1k_{1}=k_{2}=1 or k1=2,k2=1k_{1}=2,k_{2}=1. Since GG cannot appear in the list of Theorem 1, if n=3n=3, then k3>1k_{3}>1 necessarily. However, Proposition 12 rules out this case. Hence, we restrict ourselves to the case n=4n=4 which implies p3=p4=3p_{3}=p_{4}=3 necessarily. However, Lemma 5 rules out this possibility.

Hence, Theorem 1 contains the complete classification of non-cyclic abelian group with subgroup lattice with genus equal to one in the case that collection of primes {pi}\{p_{i}\} in (9) contains exactly two distinct primes.

Lastly, we consider the case where the collection of primes {pi}\{p_{i}\} in (9) contains at least three distinct primes. Proposition 9 shows that if γ(Γ(G))=1\gamma(\Gamma(G))=1, then the prime repeated must be 22 or 33. However, Lemma 5 shows that γ(Γ(G))2\gamma(\Gamma(G))\geq 2 in this case since our group GG would contain either 2×2×q1×q2\mathbb{Z}_{2}\times\mathbb{Z}_{2}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}} or 3×3×q1×q2\mathbb{Z}_{3}\times\mathbb{Z}_{3}\times\mathbb{Z}_{q_{1}}\times\mathbb{Z}_{q_{2}} as a subgroup.

Hence, Theorem 1 contains the complete classification of non-cyclic abelian group with subgroup lattice with genus equal to one. ∎

4. Conclusion

The prospect of writing out explicit lists of abelian groups whose subgroup lattice graphs have genus gg for every gg seems daunting. However, there are various classes of abelian groups for which this task seems possible. In particular, the authors of [5] have a conjecture for the genus of the subgroup lattice graphs of cyclic groups with three distinct primes factors. In particular, they conjecture that the inequalities in Proposition 6 are actually equalities. Further investigation is warranted.

5. Acknowledgements

All the subgroup lattice graphs were constructed using SAGE [7]. The author would like to thank Colin Starr for introducing him to this problem in 2017 while the author was a Visiting Assistant Professor at Willamette University.

References

  • [1] Joseph Battle, Frank Harary, Yukihiro Kodama, and J.W.T Youngs. Additivity of the genus of a graph. Bulletin of the American Mathematical Society, 68(6):565–568, 1962.
  • [2] Jeremy Berry, Sarah McGinnis, and E. J. Sanchez. Subgroup graphs of non-orientable and oriented genus one. Missouri State REU, 2012.
  • [3] Joseph P. Bohanon and Les Reid. Finite groups with planar subgroup lattices. Journal of Algebraic Combinatorics, 23(3):207–223, 2006.
  • [4] Andrea Lucchini. The genus of the subgroup graph of a finite group. Bulletin of Mathematical Sciences, 11(01):2050010, 2020.
  • [5] Christian Millichap and Fabian Salinas. Embedding grid graphs on surfaces. Graphs and Combinatorics, 38(3), 2022.
  • [6] Colin L. Starr and Galen E. Turner III. Planar groups. Journal of Algebraic Combinatorics, 19(3):283–295, 2004.
  • [7] The Sage Developers. SageMath, the Sage Mathematics Software System (Version 10.0), 2023. https://www.sagemath.org.
  • [8] Arthur T. White. The genus of repeated cartesian products of bipartite graphs. Transactions of the American Mathematical Society, 151(2):393–393, 1970.