Uniform Intersection Numbers and Subgroup-Based Generating Graphs of Groups

Let G be an abelian group of order n=\prod_{i=1}^kp_i^{t_i}. Then G\cong \prod_{i=1}^kS_i, where each S_i is the Sylow subgroup of order p_i^{t_i}. By the fundamental theorem of finite abelian groups, each S_i can be written as

Moreover, every subgroup H of G decomposes as H=\prod_{i=1}^kH_i, where H_i is a subgroup of S_i for each i. We divide the proof into the following three cases.

Case 1. k\geq 3: Let H_1=S_1\times 0\times\cdots\times 0 and H_2=0\times S_2\times 0\times\cdots\times 0. Then H_1,H_2\in V(\mathcal{G}(G)). Since H_1\nsim H_2, we have \operatorname{d}(H_1,H_2)\geq 2. We next show that \operatorname{d}(H_1,H_2)\geq 3. Suppose, for contradiction, that there exists a proper subgroup T=\prod_{i=1}^kT_i of G such that H_1\sim T\sim H_2, where T_i is a subgroup of G_i for i\in [k]. Since \langle H_1,T\rangle=G, we must have T_i=S_i for all i\in \{2,\dots,k\}. Similarly, from \langle T,H_2\rangle=G, it follows that T_1=S_1. Thus T=G, contradicting the assumption that T is a proper subgroup of G. Hence \operatorname{d}(S_1, S_2)\geq 3. By Theorem 3.3, we have \operatorname{diam}(\mathcal{G}(G))=3.

Case 2. k=2: In this case, G\cong S_1\times S_2, where S_1=\prod_{j=1}^{n_1}\mathbb{Z}_{{p_1}^{t_{1j}}} and S_2=\prod_{j=1}^{n_2}\mathbb{Z}_{{p_2}^{t_{2j}}}. First, suppose that n_1=n_2=1. If t_{11}=t_{21}=1, then G\cong \mathbb{Z}_{p_1}\times \mathbb{Z}_{p_2}. By Theorem 3.4, \operatorname{diam}(\mathcal{G}(G))=1.

Now suppose that t_{i1}\geq 2 for some i\in[2]. Then \mathcal{M}(G)=\{m_1=\langle p_1\rangle \times \mathbb{Z}_{p_2}^{t_{21}}, m_2=\mathbb{Z}_{p_1}^{t_{11}}\times \langle p_2\rangle \}. Therefore, \mathcal{G(G)} is a complete bipartite graph with bipartition V_1=\{H\in V(\mathcal{G}(G)): H\leq m_1,\ H\nleq m_2\} and V_2=\{H\in V(\mathcal{G}(G)): H\leq m_2,\ H\nleq m_1\}. Hence \operatorname{diam}(\mathcal{G}(G))=2.

Finally, suppose that n_i\geq 2 for some i\in[2]. Without loss of generality, assume that n_1\geq 2. Let H_1=T_1\times 0 and H_2=0\times S_2 be vertices of V(\mathcal{G}(G)), where T_1 is a proper subgroup of S_1. Since H_1\nsim H_2, \operatorname{d}(H_1,H_2)\geq 2. Suppose, for contradiction, that \operatorname{d}(H_1,H_2)=2. Then there exists L=L_1\times L_2\in V(\mathcal{G}(G)) such that H_1\sim L\sim H_2, where L_i is a subgroup of S_i. Since H_1\sim L, we have L_2=S_2. On the other hand, H_2\sim L implies L_1=S_1. Therefore, L=S_1\times S_2=G, a contradiction. Consequently, \operatorname{d}(H_1,H_2)\geq 3. Since \operatorname{diam}(\mathcal{G}(G))\leq 3, we have \operatorname{diam}(\mathcal{G}(G))=3.

Case 3. k=1: In this case G\cong \prod_{i=1}^{n_1}\mathbb{Z}_{p^{t_{1i}}} and the maximal subgroups are of the form m_i=\mathbb{Z}_{p^{t_{11}}}\times \dots \times \langle p\rangle \times \dots \mathbb{Z}_{p^{t_{1{n_1}}}}. Therefore, \Phi(G)=\langle p\rangle\times \langle p\rangle\times \dots \times \langle p\rangle. Now, if n_1=1, then G has a unique maximal subgroup, and hence \mathcal{G}(G) is a null graph. Thus, assume that n_1\geq 2.

If n_1=2 and t_{11}=t_{12}=1, then G\cong \mathbb Z_p\times\mathbb Z_p and Theorem 3.4 gives \operatorname{diam}(\mathcal{G}(G))=1.

Now suppose that either n_1\geq3, or n_1=2 and t_{11}t_{12}\neq1. By Theorem 3.4, \operatorname{diam}(\mathcal{G}(G))\geq2. Let S_1, S_2\in V(\mathcal{G}(G)) be such that S_1\nsim S_2. Since S_1,S_2\nsubseteq\Phi(G), choose x=(x_1,\ldots,x_{n_1})\in S_1\setminus\Phi(G) and y=(y_1,\ldots,y_{n_1})\in S_2\setminus\Phi(G). Thus, at least one coordinate of each of x and y is not divisible by p. If there exists an index r such that p\nmid x_r and p\nmid y_r, define \varphi:G\longrightarrow\mathbb Z_p by \varphi((a_i)):=a_r\pmod p. Then \varphi(x)\neq 0 and \varphi(y)\neq 0. Otherwise, choose indices r and s such that p\nmid x_r and p\nmid y_s,. Since no coordinate is simultaneously nonzero modulo p for both x and y, we have p\mid y_r and p\mid x_s. Define \varphi :G\longrightarrow\mathbb Z_p by \varphi((a_i))=a_r+a_s\pmod p. Then again \varphi(x)\neq 0 and \varphi(y)\neq 0. In either case, let m=\ker\varphi. Since \varphi is surjective onto \mathbb Z_p, m is a maximal subgroup of G. Moreover, S_1\nsubseteq m and S_2\nsubseteq m. Thus S_1\sim m\sim S_2. Therefore every pair of nonadjacent vertices of \mathcal{G}(G) is at distance 2, and so \operatorname{diam}(\mathcal{G}(G))=2.

If |\mathcal{M}(G)|\geq 3, then Proposition 3.10 yields \operatorname{gr}(\mathcal{G}(G))=3. Hence, assume that |\mathcal{M}(G)|=2. By Lemma 3.12, we have G\cong \mathbb{Z}_{{p_1^{t_1}}{p_2^{t_2}}}. If t_1, t_2\geq 2, then \langle p_1^2\rangle\sim \langle p_2\rangle\sim \langle p_1\rangle\sim \langle p_2^2\rangle\sim \langle p_1^2\rangle is a 4-cycle in \mathcal{G}(G). On the other hand, Proposition 3.10 implies that \mathcal{G}(G) contains no triangles. Therefore, \operatorname{gr}(\mathcal{G}(G)=4.

Finally, suppose that either t_1=1 or t_2=1. Without loss of generality, let t_1=1. Then \mathcal{G}(G)\cong K_{1,|t_2|} and hence \operatorname{gr}(\mathcal{G}(G))=\infty. Combining all of the above cases, we have the desired result.

We show the equivalences (1)\iff(2) and (2)\iff(3), which together complete the proof. Note that (1)\iff(2) is immediate by Corollary 3.16. To show (2)\iff(3), we spilt the proof into several propositions.

If G is isomorphic to one of the groups in (a), (b) or (c), then G has at most four maximal subgroups. Hence by Corollary 3.29, \mathcal{G}(G) is a perfect graph. Now, suppose G is isomorphic to one of the groups in (d), (e), then the uniform intersection number of G is two. Therefore, Corollary 3.29 again implies that \mathcal{G}(G) is perfect. Let G be a group satisfying condition (f). Suppose, to the contrary, that C: S_1\sim S_2\sim S_3\sim \cdots \sim S_{2k+1}\sim S_1 is an induced odd cycle in either \mathcal{G}(G) or \overline{\mathcal{G}(G)}, where k\geq 2. By Corollary 3.28, we have |M(S_i)|\geq 2 for each i. Hence, for each i, there exists p_i\in \mathcal{S}_{P}(G) such that p_1\sim p_2\sim \cdots\sim p_{2k+1}\sim p_1 form an induced odd cycle of length 2k+1 in either \mathcal{G}(G) or \overline{\mathcal{G}(G)}, a contradiction. Consequently \mathcal{G}(G) is a perfect graph.

Conversely, suppose that G is a finite group such that \mathcal{G}(G) is perfect. If G has at most four maximal subgroups, then G is isomorphic to one of groups listed in (a), (b) or (c). Thus, assume that |\mathcal{M}(G)|\geq 5. If \operatorname{i}(G)\geq 3, then by Lemma 3.30, \mathcal{G}(G) is not perfect. Hence \operatorname{i}(G)=2. We now consider three mutually exclusive cases according to the value of \operatorname{ui}(G). Suppose first that \operatorname{ui}(G)=2. Then, by Theorem C, G is isomorphic to one of the groups listed in (d) or (e). Next, suppose that 3\leq \operatorname{ui}(G)\leq 5. If there exist p_1,p_2,\ldots,p_{2k+1}\in\mathcal{S}_P(G), with k\geq2, that form an induced odd cycle in either \mathcal{G}(G) or \overline{\mathcal{G}(G)}, then, by Lemma 3.26, \mathcal{G}(G) is not perfect. Hence, no such induced odd cycle exists, and G is one of the groups described in (f).

Finally we show that \operatorname{ui}(G)\geq 6 is not possible. Indeed, if \operatorname{ui}(G)\geq 6, then there exist five distinct maximal subgroups m_{i_1},m_{i_2},m_{i_3},m_{i_4},m_{i_5} of G such that m_{i_1}\cap m_{i_2}\cap m_{i_3}\cap m_{i_4}\cap m_{i_5} \neq \Phi(G). Consequently, m_{i_1}\cap m_{i_2}\sim m_{i_3}\cap m_{i_4}\sim m_{i_1}\cap m_{i_5}\sim m_{i_2}\cap m_{i_3}\sim m_{i_4}\cap m_{i_5}\sim m_{i_1}\cap m_{i_2} form an induced C_5 in \mathcal{G}(G), a contradiction. This completes the proof.

Let \operatorname{ui}(G_1)=k and \operatorname{ui}(G_2)=l. Then \bigcap_{i=1}^km_i=\Phi(G_1) and \bigcap_{j=1}^ln_j=\Phi(G_2), where m_i\in \mathcal{M}(G_1) for i\in [k] and n_j\in \mathcal{M}(G_2) for j\in[l]. Notice that, for each i\in [k], m_i\times G_2 is a maximal subgroup of G_1\times G_2 and, for each i\in[l], G_1\times n_i is a maximal subgroup of G_1\times G_2. Since \Phi(G_1\times G_2)=\Phi(G_1)\times \Phi(G_2), we have

Thus \operatorname{ui}(G_1\times G_2)\leq \operatorname{ui}(G_1)+\operatorname{ui}(G_2).

Let \operatorname{ui}(G)=k. Since N\subseteq \Phi(G), there is a one to one correspondence between the set of maximal subgroups of G and {G}/{N}. Using Lemma 2.3, we have

for any distinct k-maximal subgroups m_1,\ldots,m_k of G. Also note that there must exist m_1, m_2, \dots , m_{k-1}\in \mathcal{M}(G) such that \bigcap_{i=1}^{k-1}\left({m_i}/{N}\right)\not=\Phi({G}/{N}), otherwise \operatorname{ui}(G)<k, which contradicts our hypothesis. Hence, the result follows.

By [hum, Theorem 5.1], the maximal subgroups of the group \operatorname{Dic}_n are of the form \langle a\rangle and \langle a^{p_j}, a^ix\rangle, where j\in [k] and i\in \{0,1,\dots, p_j-1\}. Thus \Phi(\operatorname{Dic}_n)=\langle a^{p_1\dots p_k}\rangle and |\mathcal{M}(\operatorname{Dic}_n)|=\sum_{i=1}^kp_i+1. For each t\in[k], let C_{p_t} denote the conjugacy class of maximal subgroups of the form

Clearly, |C_{p_t}|=p_t. If k=1, then the intersection of any two distinct maximal subgroups of \operatorname{Dic}_n is equal to the \Phi(\operatorname{Dic}_n) and hence \operatorname{ui}(\operatorname{Dic}_n)=2.

Let k\geq 2. Then

Therefore, \operatorname{ui}(Dic_n)>\sum_{i=1}^{k-1}p_i+1. Now, consider any collection of \sum_{i=1}^{k-1}p_i+2 maximal subgroups of \operatorname{Dic}_n. Then this collection must contain at least one maximal subgroup from each conjugacy class. Hence, the intersection of any \sum_{i=1}^{k-1}p_i+2 distinct maximal subgroups of \operatorname{Dic}_n is \Phi(G). Therefore, \operatorname{ui}(\operatorname{Dic}_n)=\sum_{i=1}^{k-1}p_i+2.

We first assume that k=1. By [hum, Theorem 5.3], the maximal subgroups of \operatorname{D}_{2n} are precisely \langle r \rangle and \langle r^{p_1},sr^j\rangle, where j\in \{0,1,2\dots,{p_1}-1\}. Then \Phi(\operatorname{D}_{2n})=\langle r^{p_1}\rangle and the intersection of two distinct maximal subgroups of \operatorname{D}_{2p_1^{t_1}} is equal to \langle r^{p_1}\rangle. Thus \operatorname{ui}(\operatorname{D}_{2n})=\operatorname{i}(\operatorname{D}_{2n})=2.

We now prove the assertion for k\geq 2. By [hum, Theorem 5.4], the maximal subgroups of \operatorname{D}_{2n} are of the form \langle r \rangle and \langle r^{p_i}, sr^j\rangle, where i\in [k] and j\in \{0,1, \dots, p_{i-1}\}. Thus \Phi(\operatorname{D}_{2n})=\langle r^{p_1p_2\cdots p_k}\rangle. Let C_{p_0}=\{\langle r \rangle\} and, for each i\in [k], C_{p_i}=\{\langle r^{p_i},sr^j\rangle: j\in [p_i-1]\}. Then, by an argument analogous to that used in Theorem Theorem 2.5, we have \operatorname{ui}(\operatorname{D}_{2n})=\sum_{i=1}^{k-1}p_i+2.

Note that \operatorname{QD}_n is a group of order 2^{n} and the maximal subgroups of qdic_n are consist of exactly three isomorphism classes:

1. The cyclic group \mathbb{Z}_{2^{n-1}},

2. The dihedral group \operatorname{D}_{2^{n-1}},

3. The generalized quaternion group \operatorname{Q}_{2^{n-1}}.

Thus, every maximal subgroup of \operatorname{QD}_n has order 2^{n-1}, and \Phi( \operatorname{QD}_n)=\mathbb{Z}_{2^{n-2}}. Moreover, for any two distinct maximal subgroups m_1,m_2\in\mathcal{M}(\operatorname{QD}_n), |m_1\cap m_2|=2^{n-2}=|\Phi(\operatorname{QD}_n)|. Consequently, \operatorname{ui}(\operatorname{QD}_n)=\operatorname{i}(\operatorname{QD}_n)=2.

By the Fundamental theorem of an abelian Group,

Since \operatorname{i}(\mathbb{Z}_{{p_j}^{t_j}})=1 for each j\in [k], by [hum, Theorem 4.7], we have \operatorname{i}(G)=k. Also, k=\operatorname{i}(G)\leq \operatorname{ui}(G)\leq \vert \mathcal{M}(G)\vert=k. Consequently, \operatorname{ui}(G)=k.

Suppose \operatorname{i}(G)=\operatorname{ui}(G)=k. Then, for every k-element subset \Lambda of \mathcal{M}(G), \bigcap\limits_{m_i\in \Lambda}m_i=\Phi(G). Consequently, |\mathcal{S}_P(G)|=\sum_{i=1}^{k-1}\left(^m_i \right)+2.

Conversely, suppose |\mathcal{S}_P(G)|=\sum_{i=1}^{k-1}\left(^m_i \right)+2. Then \operatorname{ui}(G)=k. It remains to show let \operatorname{i}(G)=k. If not, then there exist maximal subgroups m_1, m_2,\dots, m_t of G such that \bigcap\limits_{i=1}^tm_i=\Phi(G), where t\leq k-1. This implies |\mathcal{S}_P(G)|<\sum_{i=1}^{k-1}\left(^m_i \right)+2, a contradiction. Thus \operatorname{i}(G)=\operatorname{ui}(G)=k.

We prove the equivalent contrapositive statement: the two vertices S_1 and S_2 in \overline{\mathcal{G}(G)} are adjacent if and only if M(S_1)\cap M(S_2)\not=\emptyset.

Let S_1\sim S_2 in \overline{\mathcal{G}(G)}. Then \langle S_1, S_2\rangle\not=G. Since G is finitely generated, by Zorn’s lemma, we have \langle S_1, S_2\rangle \subseteq m for some m\in \mathcal{M}(G). Then m\in M(S_1)\cap M(S_2). Thus M(S_1)\cap M(S_2)\not=\emptyset.

To prove the converse, suppose that there exist nontrivial proper subgroups S_1,S_2\leq G, with S_1,S_2\nsubseteq \Phi(G), such that M(S_1)\cap M(S_2)\not=\emptyset. Then there exists m\in \mathcal{M}(G) such that m\in M(S_1)\cap M(S_2). This implies \langle S_1, S_2\rangle \subseteq m\not=G. Hence S_1\sim S_2 in \overline{\mathcal{G}(G)}.

Let \mathcal{G}(G) be a nontrivial connected graph. Since G is finitely generated, we have |\mathcal{M}(G)|\geq 1. If |\mathcal{M}(G)|=1, then \mathcal{G}(G) is a null graph, a contradiction. Consequently, |\mathcal{M}(G)|\geq 2.

Conversely, suppose that |\mathcal{M}(G)|\geq 2, and let S_1, S_2 be two arbitrary vertices of \mathcal{G}(G). If S_1\sim S_2, there is nothing to prove. Thus, assume that S_1\nsim S_2. Then M(S_1)\cap M(S_2)\not=\emptyset. Since S_1, S_2\nsubseteq \Phi(G), there exist m_1, m_2\in \mathcal{M}(G) such that S_1\nsubseteq m_2 and S_2\nsubseteq m_1. Thus, S_1\sim m_2 and m_1\sim S_2. If m_1=m_2=m, then S_1\sim m\sim S_2. On the other hand, if m_1\neq m_2, the maximality of m_1 and m_2 yields \langle m_1,m_2\rangle=G and hence m_1\sim m_2. Consequently, S_1\sim m_2\sim m_1\sim S_2 is a S_1-S_2 path of length at most 3. Therefore, \mathcal{G}(G) is connected and \operatorname{diam}(\mathcal{G}(G))\leq 3.

Since |\mathcal{M}(G)|\geq 2, by Proposition 3.2, \mathcal{G}(G) is connected.

1. Since G is finitely generated, each subgroup of G is contained in some maximal subgroups of G. Therefore, by Lemma 3.1, we have \operatorname{diam}(\mathcal{G}(G))=1 if and only if every vertex of \mathcal{G}(G) is a maximal subgroup of G.

2. Suppose that \operatorname{diam}(\mathcal{G}(G))=2. By the first part, \vert V(\mathcal{G}(G))\setminus \mathcal \mathcal{M}(G)\vert\geq 1. If \vert V(\mathcal{G}(G))\setminus \mathcal \mathcal{M}(G)\vert=1, we are done. Let \vert V(\mathcal{G}(G))\setminus \mathcal \mathcal{M}(G)\vert\geq 2. It remains to show for every pair of nonadjacent vertices S_1, S_2\in V(\mathcal{G}(G)), M(S_1)\cup M(S_2)\subsetneq \mathcal{M}(G). Suppose, to the contrary, that there exist nonadjacent vertices S_1, S_2 in V(\mathcal{G}(G))\setminus \mathcal{M}(G) such that M(S_1)\cup M(S_2)=\mathcal{M}(G). Then there is no vertex H\in V(\mathcal{G}) for which S_1\sim H\sim S_2. Therefore, \operatorname{d}(S_1, S_2)\geq 3, a contradiction. Consequently, M(S_1)\cup M(S_2)\subsetneq \mathcal{M}(G) for every pair of nonadjacent vertices S_1,S_2\in V(\mathcal{G}(G)).

   To prove the converse, first assume that \vert V(\mathcal{G}(G))\setminus \mathcal \mathcal{M}(G)\vert=1, and let A\in V(\mathcal{G}(G))\setminus \mathcal{M}(G). Since G is finitely generated and A\nsubseteq \Phi(G), there exist m_1, m_2\in \mathcal{M}(G) such that A\subsetneq m_1 and A\nsubseteq m_2. Thus A\sim m_2\sim m_1. Since A\subsetneq m_1, the vertices A and m_1 are nonadjacent, and hence \operatorname{d}(A,m_1)=2. Furthermore, any two distinct maximal subgroups of G are adjacent. It follows that \operatorname{diam}(\mathcal{G}(G))=2.

   Now suppose that \vert V(\mathcal{G}(G))\setminus \mathcal \mathcal{M}(G)\vert\geq2. By part (1), \mathcal{G}(G) is not complete. Let S_1, S_2\in V(\mathcal{G}(G)) be two non adjacent vertices. By hypothesis, there exists m\in \mathcal{M}(G) such that m\notin M(S_1)\cup M(S_2). Thus, every pair of nonadjacent vertices of \mathcal{G}(G) is at distance 2, and consequently, \operatorname{diam}(\mathcal{G}(G))=2.

3. This follows immediately from the preceding two cases, together with the fact that \operatorname{diam}(\mathcal{G}(G))\leq 3.

(i). Suppose that \operatorname{i}(G)>\lceil \frac{k+1}{2}\rceil. Define S_1:=m_1\cap m_2\cap \ldots \cap m_{\lceil\frac{k+1}{2}\rceil}, S_2:=m_{\lceil\frac{k+1}{2} \rceil}\cap \cdots \cap m_k, where m_i\in \mathcal{M}(G) for i\in [k]. Then S_1,S_2 \in V(\mathcal{G}(G)) and m_{\lceil\frac{k+1}{2}\rceil}\in M(S_1)\cap M(S_2). Thus, the vertices S_1 and S_2 are not adjacent in \mathcal{G}(G) and M(S_1)\cup M(S_2)=\mathcal{M}(G). Therefore, by Theorem 3.3, \operatorname{diam}(\mathcal{G}(G))=3.

(ii). Since G is not isomorphic to one of the groups mentioned in Lemma 3.4, \operatorname{diam}(\mathcal{G}(G))\geq 2. We now consider two cases. First, we assume that \operatorname{ui}(G)<\lceil \frac{k+1}{2}\rceil. Then for any H\in \mathcal{G}(G), we have |M(H)|\leq \lceil \frac{k+1}{2}\rceil -2, since H \nsubseteq \Phi(G). Therefore, for any H_1, H_2\in V(\mathcal{G}(G)), we have |M(H_1)|+|M(H_2)|\leq 2\lceil \frac{k+1}{2}\rceil -4<k. This implies M(H_1)\cup M(H_2)\subsetneq \mathcal{M}(G) and by Theorem 3.3, \operatorname{diam}(\mathcal{G}(G))=2.

Suppose that \operatorname{i}(G)\leq \operatorname{ui}(G)=\lceil \frac{k+1}{2}\rceil. Then, for any H\in V(\mathcal{G}(G)), we have |M(H)|\leq \lceil\frac{k+1}{2}\rceil-1. Let H_1,H_2\in V(\mathcal{G}(G)) be two nonadjacent vertices. Since H_1\nsim H_2, we have M(H_1)\cap M(H_2)\neq\emptyset. Therefore |M(H_1)|+|M(H_2)|\leq 2\lceil \frac{k+1}{2}\rceil -3<k and hence M(H_1)\cup M(H_2)\subsetneq \mathcal{M}(G). It now follows from Theorem 3.3 that \operatorname{diam}(\mathcal{G}(G))=2.

The result follows immediately from Corollary  2.8 and Proposition  3.5.

Since G is an almost finite simple group, by [burn, Theorem 1(i)], we have \operatorname{i}(G)\leq 4. By hypothesis \vert \mathcal{M}(G)\vert \geq 8. Consequently,

Therefore, by Proposition 3.5, \operatorname{diam}(\mathcal{G}(G))=2.

Let n=3. Then S_n\cong \mathbb{Z}_3\rtimes \mathbb{Z}_2. By Theorem 3.4, \operatorname{diam}(\mathcal{G}(G))=1. Let n\geq 4. By [burn, Theorem 1], \operatorname{i}(S_n)\leq 3. Since for n\geq 4, \vert \mathcal{M}(G)\vert\geq 8 and

where k=\vert \mathcal{M}(G)\vert. One can observe that for n\geq 4, S_n is not isomorphic to any of the groups mentioned in Theorem 3.4. Therefore, from Proposition 3.5, we have \operatorname{diam}(\mathcal{G}(S_n))=2.

Let S_1, S_2\in V(\mathcal{G}(G)) be two nonadjacent vertices. Since G is a finitely generated free abelian group, G\cong \bigoplus_{i\in \Lambda}\mathbb{Z}, for some finite index set \Lambda. If |\Lambda|=1, then it follows from Theorem 3.3 that \operatorname{diam}(\mathcal{G}(G))=2. Hence, assume that |\Lambda|=t\geq 2, and let \{e_i\}_{i\in\Lambda} denote the standard basis of G. Let x=(x_1,x_2,\dots,x_t)\in S_1 and y=(y_1, y_2,\dots,y_t)\in S_2 be two nonzero elements of G. If x_j\not=0 and y_j\not=0 for some j\in [t], then (x)\sim m\sim (y), where m=\langle e_1, e_2,\dots,e_{j-1},(5x_jy_j+1)e_j, e_{j+1},\dots, e_t\rangle \not=G. If not, then there must exist i,j\in [t] such that x_i\not=0 and y_j\not=0. Define m:=\langle e_1,e_2, \dots ,e_{i-1},e_i+x_iy_je_j,e_{i+1},\dots, e_{j-1}, e_j+x_iy_je_i,e_{j+1},\dots, e_t\rangle. We next show that m \neq G. Let A be the t\times t integer matrix whose columns are the generators of m. Then the matrix A differs from the identity matrix only in the i-th and j-th coordinates. Thus \det(A)=1-(x_i y_j)^2. Since x_i y_j\neq0, it follows that \det(A)\neq\pm1. Now, suppose, for contradiction, that m=G. Then the columns of A generate \mathbb{Z}^t , and hence A\in \operatorname{GL}_t(\mathbb{Z}). Consequently, \det(A)=\pm1, a contradiction. Thus m \neq G and one can easily observe that (x)\sim m\sim (y). Hence \operatorname{diam}(\mathcal{G}(G))=2.

Suppose that |\mathcal{M}(G)|\geq 3, and choose three distinct maximal subgroups m_1,m_2,m_3\in\mathcal{M}(G). Since any two distinct maximal subgroups of G generate G, m_1\sim m_2\sim m_3\sim m_1 forms a triangle inn \mathcal{G}(G). Thus \operatorname{gr}(\mathcal{G}(G))=3.

Now, suppose that \mathcal{M}(G)=\{m_1,m_2\} and there exist subgroups S_1, S_2\in V(\mathcal{G}(G)) \setminus \mathcal{M}(G) such that M(S_i)=\{m_i\} for i\in[2]. Then S_1\sim m_2\sim m_1\sim S_2\sim S_1 forms a 4-cycle in \mathcal{G}(G). We claim that \mathcal{G}(G) contains no triangle. Suppose, to the contrary, that H_1\sim H_2\sim H_3\sim H_1 is a triangle. Since H_1\sim H_2, M(H_1)\cap M(H_2)=\emptyset. As \mathcal{M}(G)=\{m_1,m_2\}, we may assume, without loss of generality, that M(H_1)=\{m_1\} and M(H_2)=\{m_2\}. Again, since H_1\sim H_3 and H_2\sim H_3, we must have M(H_1)\cap M(H_3)=\emptyset and M(H_2)\cap M(H_3)=\emptyset. Thus M(H_3)=\emptyset, which is impossible for a vertex corresponding to a proper subgroup of the finitely generated group G. Therefore, \mathcal{G}(G) contains no triangle and we conclude that \operatorname{gr} (\mathcal{G}(G))=4.

Finally, suppose that, for some i\in[2], m_i is the only vertex S\in V(\mathcal{G}(G)) satisfying M(S)=\{m_i\}. Then \mathcal{G}(G) is a star graph, and hence \operatorname{gr}(\mathcal{G}(G))=\infty.

Suppose that \mathcal{G}(G) is a k-partite graph with partite sets V_1,\ldots,V_k. Assume, for contradiction, that |\mathcal{M}(G)|\geq k+1. By the pigeonhole principle, some partite set, say V_t, contains at least two maximal subgroups, say m_1 and m_2. Since m_1\sim m_2 in \mathcal{G}(G), the partite set V_t is not independent, a contradiction. Thus, |\mathcal{M}(G)|\leq k.

We next show that |\mathcal{M}(G)|\geq k. If not, then some partite set, say V_k, contains no maximal subgroup. Let S\in V_k. Since G is finitely generated, S is contained in a maximal subgroup m_j, which lies in some partite set V_j, j\neq k. As V_j is independent and m_j\in V_j, every vertex of V_j is contained in m_j. Hence S may be moved to V_j without destroying independence. Repeating this for every S\in V_k, we obtain a (k-1)-partite representation of \mathcal{G}(G), contradicting the minimality of k. Therefore, |\mathcal{M}(G)|= k.

To prove the converse, suppose that \mathcal{M}(G)=\{m_1,\ldots,m_k\}, where 2\leq k<\infty. Then \mathcal{G}(G) is a connected graph. Define

Clearly, V_i\not=\emptyset for each i\in[k] and \{V_i:i\in [k]\} form a partition of V(\mathcal{G}(G)). Since every V_i is an independent set, \mathcal{G}(G) is a k-partite graph.

We first assume that \mathcal{G}(G) is a complete k-partite graph. Let m_1, m_2\in \mathcal{M}(G) such that m_1\cap m_2\not=\Phi(G). Then m_1\cap m_2\in V(\mathcal{G}(G)) and m_1\cap m_2 must belong to some partite set. Since m_1\sim m_2, m_1 and m_2 belong to different partite sets. Moreover, m_1\cap m_2 is neither adjacent to m_1 nor m_2. This contradicts the fact that \mathcal{G}(G) is a complete k-partite graph. Thus m_1\cap m_2=\Phi(G) for all distinct m_1, m_2\in \mathcal{M}(G).

Conversely, let \mathcal{M}(G)=\{m_1,\ldots,m_k\} and m_i\cap m_j=\Phi(G) for all i\neq j. By Proposition 3.15, \mathcal{G}(G) is a k-partite graph with partite sets V_1, V_2, \dots, V_k. Since any two maximal subgroups of G are adjacent in \mathcal{G}(G), without loss of generality assume that m_i\in V_i for i\in [k]. If a vertex H\in V_i that is not adjacent to some vertex in V_j for i\neq j, then H\subseteq m_i\cap m_j=\Phi(G), which contradicts H\in V(\mathcal{G}(G)). Thus \mathcal{G}(G) is a complete k-partite graph.

Let N be a proper normal subgroup of G, and suppose that \mathcal{G}(G) is complete multipartite. If |\mathcal{M}(G/N)|\leq 1, then G/N is a null graph. Hence, assume that |\mathcal{M}(G/N)|\geq 2 and let {m_1}/{N}, {m_2}/{N}\in \mathcal{M}(G/N) be distinct maximal subgroups. Then m_1, m_2\in \mathcal{M}(G). Since \mathcal{G}(G) is a complete multipartite graph, by Corollary 3.16, m_1\cap m_2=\Phi(G). Moreover, N\subseteq m_1\cap m_2, and hence N\subseteq\Phi(G). Consequently,

By Lemma 2.3, \Phi({G}/{N})={\Phi(G)}/{N}. This implies m_1/N\bigcap m_2/N=\Phi\left(G/N\right). By Corollary 3.16, \mathcal{G}({G}/{N}) is a complete multipartite graph.

Suppose that G\cong \underbrace{\mathbb{Z}_p\times \mathbb{Z}_p\times \dots \times \mathbb{Z}_p}_{r~\text{times}}. Then \Phi(G)=\{0\} and the intersection of the two maximal subgroups m_1=0\times \mathbb{Z}_p\times \mathbb{Z}_p \dots\times \mathbb{Z}_p and m_2=\mathbb{Z}_p\times 0\times\mathbb{Z}_p\times \dots \times \mathbb{Z}_p is not equal to \Phi(G). By Corollary 3.16, \mathcal{G}(G) is not a complete multipartite graph.

Since G be a p-group of order p^n, one of the following holds: (i) G\cong \mathbb{Z}_{p^n}; (ii) G\cong \mathbb{Z}_{p^s} \times \mathbb{Z}_{p^t} with s+t=n; (iii) G is an abelian p-group of rank k\geq 3; (iv) G is a nonabelian p-group. If G\cong \mathbb{Z}_{p^n}, then G is a null graph.

Let G\cong \mathbb{Z}_{p^s} \times \mathbb{Z}_{p^t}, where s+t=n. Then |\mathcal{M}(G)|=p+1. Observe that \Phi(G)\cong \mathbb{Z}_{p^{t-1}}\times\mathbb{Z}_{p^{s-1}}. Furthermore, for any two distinct maximal subgroups m_1,m_2\in\mathcal{M}(G), m_1\cap m_2=\Phi(G). Consequently, \mathcal{G}(G) is a complete (p+1)-partite graph.

Now consider the case where G be an abelian p-group of rank k\geq 3. Then there exists a normal subgroup H of G such that G/H is isomorphic to the elementary abelian group \mathbb{Z}_p\times \mathbb{Z}_p\times \dots \times \mathbb{Z}_p (k times). By Lemma 3.17 and Lemma 3.18, \mathcal{G}(G) is not a complete multipartite graph.

Finally, suppose that G be a non abelian p-group. Let G be 2-generated. Then G/\Phi(G)\cong \mathbb{Z}_p\times \mathbb{Z}_p, which is a 2-dimensional vector space over \mathbb{Z}_p. Hence, there is a one to one correspondence between the maximal subgroups of G and the one dimensional subspaces of G/\Phi(G). Since any two distinct one-dimensional subspaces of G/\Phi(G) intersect trivially, it follows that m_1\cap m_2=\Phi(G) for any two distinct maximal subgroups m_1,m_2 of G. By Corollary 3.16, \mathcal{G}(G) is complete multipartite. On the other hand, if G requires k\geq3 generator, then G/\Phi(G)\cong \mathbb{Z}_{p}^k. Therefore, by Lemma 3.18, \mathcal{G}(G) is not a complete multipartite graph.

First suppose that G\cong \mathbb{Z}_{p_1^{n_1}p_2^{n_2}}. Then \mathcal{M}(G)=\{\langle p_1\rangle, \langle p_2\rangle \}. This implies \mathcal{G}(G) is a complete bipartite graph with bipartition V_1=\{H\in V(\mathcal{G}(G)):H\subseteq \langle p_1 \rangle \} and V_2=\{H\in V(\mathcal{G}(G)): H\subseteq \langle p_2\rangle\}.

Conversely, suppose that \mathcal{G}(G) is complete multipartite. We distinguish two cases according as G is cyclic or noncyclic. Assume first that G is a non-cyclic nilpotent group of order n with k\geq 2. By Lemma 3.21, we have |\pi({G}/{\phi(G)})|=|\pi(G)|\geq 2, which implies {G}/{\phi(G)}\cong \prod_{i=1}^kG_i, where each G_i is an elementary abelian group and k\geq 2. If k\geq 3, then by Lemmas 3.17 and 3.18, \mathcal{G}(G) is not a complete multipartite graph. Let k=2. Since G is non cyclic, one of G_i must be non cyclic. Without loss of generality suppose that G_1 is non cyclic. Then there exist maximal subgroups m_1, m_2 \in \mathcal{M}(G_1) such that m_1\cap m_2 \not=\Phi(G_1). Now consider the maximal subgroups m_1\times G_2 and m_2\times G_2 of G_1\times G_2. Since \Phi (G_1\times G_2)=\Phi(G_1)\times \Phi(G_2), m_1\times G_2\cap m_2\times G_2\neq \Phi (G_1\times G_2). By Proposition 3.15 and Lemma 3.17, \mathcal{G}(G) is not a complete multipartite graph.

We now assume that G is cyclic nilpotent of order n. Then G\cong \mathbb{Z}_n and \mathcal{M}(G)=\{\langle p_i \rangle : i\in [k]\}. If k\geq 3, then \langle p_1 \rangle \cap \langle p_2 \rangle \not=\Phi(G)=\langle p_1 \rangle \cap \langle p_2 \rangle \cap \dots \cap \langle p_k \rangle =\langle p_1p_2\cdots p_k\rangle, contradicting Corollary 3.16. Hence k=2 and G\cong \mathbb{Z}_{p_1^{n_1}p_2^{n_2}}.

Suppose that G\cong P_1\rtimes P_2, where P_1 is an elementary abelian sylow p_1-group of order p_1 and P_2 is a Sylow p_2-subgroup of G of order p_2, and p_2|(p_1-1). Then P_1\cong \mathbb Z_{p_1} and P_2\cong \mathbb Z_{p_2}. Hence G\cong \mathbb Z_{p_1}\rtimes \mathbb Z_{p_2}, where p_1\neq p_2. Since G is non-nilpotent, the action of \mathbb Z_{p_2} on \mathbb Z_{p_1} is nontrivial and \Phi(G)=\{1\}. Thus \mathcal{G}(G) is a complete multipartite graph.

Conversely, suppose that \mathcal{G}(G) be a complete multipartite graph. We divide the remainder of the proof into two cases.

Case 1. \Phi(G) is trivial: Since G is non-nilpotent, there exists a non-normal maximal subgroup m of G. As m\leq N_G(m)\leq G and m is maximal, the non-normality of m implies that m=N_G(m). Hence, for every x\in G\setminus m, we have m^x\neq m. Thus m and m^x are distinct maximal subgroups, and Corollary 3.16 yields m\cap m^x=\Phi(G)=\{1\}. This implies G is a Frobenius group with the Frobenius complement m. Hence G\cong k_G\rtimes m, where k_G is the Frobenius Kernel of G and \operatorname{gcd}(|m|,|k_G|)=1. We next show that the group m contains only one maximal subgroup i.e., |\mathcal{M}(m)|=1. If not, then |\mathcal{M}({G}/{k_G})|=|\mathcal{M}(m)|\geq 2. By Corollary 3.16, k_G is a subgroup of \Phi(G), a contradiction. Therefore, |\mathcal{M}(m)|=1. This implies m\cong \mathbb{Z}_{p_2^t} for some prime p_2 and a natural number t. If t\geq 2, then 1\not=m_{p_2}=\{x\in m:x^{p_2}=1\} is a non trivial proper subgroup of m. Hence, \langle k_G,m_{p_2}\rangle is a nontrivial proper subgroup of G contained in a maximal subgroup m_1\in \mathcal{M}(G)\setminus\{m\}. Since G is a complete multipartite, m_{p_2}\subseteq m\cap m_1=\Phi(G)=\{1\}, which is a contradiction. Therefore, m\cong \mathbb{Z}_{p_2}.

We now show that k_G\cong \mathbb{Z}_{p_1}^r for some prime p_1\not=p_2. Since k_G is nilpotent,it is isomorphic to the direct product product of its Sylow subgroups, i.e. k_G\cong \prod_{i=1}^nS_i. If possible, let n\geq 2. Then m\subsetneq \langle S_1,m\rangle \not=G, which contradicts the maximality of m. Therefore, k_G is a Sylow p_1 subgroup of G for some prime p_1\not=p_2 such that p_2|p_1-1. Also, since G is solvable, \Phi(k_G)=k_G\cap \Phi(G)=\{1\}. Consequently, k_G\cong \mathbb{Z}_{p_1}^r, where r is the rank of r. If r\geq 2, then there exist maximal subgroups k_1\rtimes \mathbb{Z}_{p_2} and k_2\rtimes \mathbb{Z}_{p_2} such that k_1\rtimes \mathbb{Z}_{p_2}\cap k_2\rtimes \mathbb{Z}_{p_2}\not=\{1\}, which contradicts that is complete multipartite graph Consequently, G\cong \mathbb Z_{p_1}\rtimes \mathbb Z_{p_2}.

Case 2. \Phi(G) is non trivial: Since \mathcal{G}(G) be a complete multipartite graph, by Lemma 3.17, {G}/{\Phi(G)} is a complete multipartite graph. If possible, let {G}/{\Phi(G)} be a nilpotent group. Since G is non abelian, {G}/{\Phi(G)} is non cyclic. By Lemma 3.21, every prime divisor of the order of a finite group also divides [G:\Phi(G)]. Thus, from Propositions 3.19 and 3.22), we must have |\pi({G}/{\Phi(G)})|=|\pi(G)|=1, which is a contradiction. Thus {G}/{\Phi(G)} is not a nilpotent group. Similarly as above, we have {G}/{\Phi(G)}\cong {P_1}/{\Phi(P_1)}\rtimes {P_2}/{\Phi(P_2)}\cong \overline{P_1}\times \overline{P_2}, where P_i is a Sylow p_i subgroup of G for i\in \{1,2\}. Thus G=P_1P_2. Since P_1\Phi(G) is a normal subgroup of G, we have G=N(P_1), the normalizer of P_1 in G ([robin], 5.2.14). Hence P_1 is normal in G. Since \overline{P_1}= \frac{P_1\Phi(G)}{\Phi(G)}\cong \frac{P_1}{P_1\cap \Phi(G)}=\frac{P_1}{\Phi(P_1)}\cong \mathbb{Z}_{p_1}, we have P_1\cong \mathbb{Z}_{p_1^s} for some s\in \mathbb{N}. Now, if possible, let |\mathcal{M}(P_1)|\geq 2. Since \frac{G}{P_1}\cong P_2, we have |M(P_2)|\geq 2. Also, since \mathcal{G}(G) is a complete multipartite graph, P_2\subseteq m_1\cap m_2=\Phi(G) (cf. Corollary 3.16), which is a contradiction. Therefore, |\mathcal{M}(P_2)|=1, i.e. P_2\cong \mathbb{Z}_{p_2^t} for some t\in \mathbb{N}.

If possible, let s=t=1. Then \Phi(G)=\{1\}, which contradicts our hypothesis. Let at least one of s and t be greater than one. Then \Phi(G)\cong \mathbb{Z}_{{p_1}^{s-1}}\times Z_{p_2}^{t-1} and one can verify that there exist maximal subgroups whose intersection is not equal to the Frattini subgroup of G. This concludes the proof.

By Lemma 3.26, \mathcal{G}(G), it suffices to show that neither \mathcal{G}(G) nor its complement \overline{\mathcal{G}(G)} contains an induced odd cycle of length at least five.

Suppose, for contradiction, that S:S_1\sim S_2\sim \cdots \sim S_{2k+1}\sim S_1 is an induced cycle in \mathcal{G}(G), where k\geq 2. First, we show that |M(S_t)|\geq 2 for every vertex S_t of S. Suppose |M(S_t)|=1 for some S_t\in S. Then M(S_t)=\{m\} for some m\in \mathcal{M}(G). Since S_t\nsim S_{t+2} in S, m\in M(S_{t+2}). Similarly S_t\nsim S_{t+3} implies m\in M(S_{t+3}). Hence m\in M(S_{t+2})\cap M(S_{t+3}) which implies S_{t+2}\nsim S_{t+3}, a contradiction. Hence M(S_t)\geq 2 for any t\in [2k+1]. By hypothesis, G has at most four maximal subgroups. Let \mathcal{M}(G)=\{m_1,m_2,m_3,m_4\}. Without loss of generality, let M(S_1)=\{m_1, m_2\}. Since S_1\sim S_2, we must have m_3, m_4\in M(S_2). Again, as S_2\sim S_3 in S, we must have m_1, m_2\in M(S_3). Also, S_1\sim S_{2k+1} and therefore, m_3, m_4\in M(S_{2k+1}). Since M(S_3)\cap M(S_{2k+1})=\emptyset, we must have S_3\sim S_{2k+1}, which contradicts that S is an induced cycle. Hence, \mathcal{G}(G) contains no induced odd cycle of length at least 5. By a similar argument, the same conclusion holds when |\mathcal{M}(G)|=2 or 3.

Now, suppose for contradiction that S':S_1'\sim S_2'\sim \cdots \sim S_{2k+1}'\sim S_1' be an induced cycle in \overline{\mathcal{G}(G)}, where k\geq 2. To prove that |M(S'_t)|\geq 2 for any t\in[2k+1], let M(S'_t)=\{m\} for some t\in [2k+1] and m\in \mathcal{M}(G). Since S'_t\sim S_{t+1} in \overline{\mathcal{G}(G)}, we must have m\in M(S'_{t+1}). Again, as S'_t\sim S'_{t-1}, we must have m\in M(S'_{t-1}). Therefore, m\in M(S'_{t+1})\cap M(S'_{t-1}) and therefore S'_{t-1}\sim S'_{t+1} in S'. This contradicts the notion that S' is an induced cycle in \overline{\mathcal{G}(G)}. Hence |M(S'_t)|\geq 2 for any t\in [2k+1]. Let \mathcal{M}(G)=\{m_1,m_2,m_3,m_4\} and without loss of generality, let M(S'_1)=\{m_1,m_2\}. Since S'_1\nsim S'_3 in S', we must have m_3, m_4\in M(S'_3). Again, as S'_3\nsim S'_{2k+1}, we must have m_1, m_2\in M(S'_{2k+1}). In addition, S'_{2k+1}\nsim S'_2 implies m_3, m_4\in M(S'_2). Then M(S'_1)\cap M(S'_2)=\emptyset and hence, they are not adjacent in \overline{\mathcal{G}(G)}, a contradiction. Similarly, we can show that if \mathcal{M}(G)\leq 3, then there does not exist any induced cycle of odd length of at least five in \overline{\mathcal{G}(G)}.

If |\mathcal{M}(G)|\leq 4, then by Proposition \ref{propper}, \mathcal{G}(G) is a perfect graph. Now, assume that |\mathcal{M}(G)|\geq 5 and m_1\cap m_2=\Phi(G) for all distinct m_1, m_2\in \Phi(G). If possible, let C:C_1\sim C_2\sim \dots\sim C_{2k+1}\sim C_1 be an induced cycle in \mathcal{G}(G), where k\geq 2. By an argument similar to that used in the proof of Proposition 3.27, every vertex of C is contained in at least two maximal subgroups of G. Let C_i\in C and write M(C_i)=\{m_1,\ldots,m_t\}, t\geq 2. Then C_i\subseteq \bigcap_{i=1}^tm_i=\Phi(G), contradicting C_i\in V(\mathcal{G}(G)). Hence, the result follows.

Let \mathcal{M}(G)=\{m_1,\ldots,m_t\}, where t\geq 5. Then m_1\cap m_2\sim m_3\cap m_4\sim m_1\cap m_5\sim m_2\cap m_3\sim m_4\cap m_5\cap m_1\cap m_2 is an induced C_5 in \mathcal{G}(G). Thus \mathcal{G}(G) is not a perfect graph.

(i) Let P be a pendant vertex in \mathcal{G}(\mathbb{Z}_n). Since P\in V(\mathcal{G}(\mathbb{Z}_n)), we have P\nleq \Phi(\mathbb{Z}_n). Hence there exists a prime divisor p_i of n such that \langle p_i\rangle \notin M(P). If t_i\geq 2, then both \langle p_i\rangle and \langle p_i^2\rangle are adjacent to P, which is a contradiction. Hence t_i=1.

Conversely, suppose that t_k=1. Then every vertex of the form (\prod_{i=1}^{k-1}p_i^{s_i}) with 1\leq s_i\leq t_i, is adjacent only to \langle p_k\rangle. Thus \langle p_k\rangle is pendent.

(ii) Let t_i\geq 2 for i\in [l] and t_i=1 for i\in \{l+1, \dots,k\}. Then any vertex of the form
\langle p_1^{s_1}p_2^{s_2}\dots p_l^{s_l}p_{l+1}\dots p_{l+i-1}p_{l+i+1}\dots p_{l+r}\rangle is adjacent to \langle p_i\rangle only, where 1\leq i\leq r and l+r=k. Therefore, the number of pendant vertex of \mathcal{G}(\mathbb{Z}_n) is (k-l)\prod_{i=1}^lt_i.

Let k=2 with t_1=t_2=r\geq 2. Then

is a Hamiltonian cycle in \mathcal{G}(\mathbb{Z}_n). Thus \mathcal{G}(\mathbb{Z}_{{p_1^r}{p_2^r}}) is a Hamiltonian graph.

Conversely, suppose that \mathcal{G}(\mathbb{Z}_n) is a Hamiltonian graph. We show that none of the following can occur: k\neq 2, t_1=t_2=1, or t_1\neq t_2, since in each of these cases \mathcal{G}(\mathbb{Z}_n) is not Hamiltonian.

Let k=2. If t_1=t_2=1, then \mathcal{G}(\mathbb{Z}_n)\cong K_2, which is not a Hamiltonian graph. Let t_1\not=t_2. Without loss of generality, assume that t_1> t_2. If t_1-t_2=1, then \mathcal{G}(\mathbb{Z}_n) is a bipartite graph of odd order and hence is not a Hamiltonian Graph. If t_1-t_2\geq 2, then one can observe that the circumference of the complete bipartite graph \mathcal{G}({\mathbb{Z}_n}) is 2t_2<t_1+t_2=\vert \mathcal{G}(\mathbb{Z}_n)\vert and hence the graph \mathcal{G}(\mathbb{Z}_n) is again not Hamiltonian.

Finally, let k\geq 3. If t_i=1 for some i\in [k], then by Corollary 4.2, \mathcal{G}(\mathbb{Z}_n) contains a pendant vertex and hence is not a Hamiltonian graph. Now suppose that t_i\geq 2 for all i\in [k], and let t_k=\min\{t_i:i\in [k]\}. Consider the sets A=\left\{\left\langle\prod_{i=1}^{k-1}p_i^{s_i}\right\rangle:1\leq s_i\leq t_i\right\} and B=\left\{\langle p_k^{s_k}\rangle:1\leq s_k\leq t_k\right\}. Every vertex in A is adjacent to every vertex in B, while no two vertices in A are adjacent and no two vertices in B are adjacent. Thus, the subgraph of \mathcal{G}(\mathbb{Z}_n) induced by A\cup B is complete bipartite. Suppose that \mathcal{G}(\mathbb{Z}n) contains a Hamiltonian cycle. Then this cycle must contain a path whose internal vertices include all the vertices of A and whose two end vertices lie in B. However, Since k\geq 3 and t_i\geq 2 for all i\in [k], we have \vert A\vert=\prod_{i=1}^{k-1}t_i>t_k=\vert B\vert. By the same argument as above, no such path can exist in the subgraph induced by V(A\cup B), a contradiction. This concludes the proof.