1 Introduction

For a finite group G, let \(\textrm{Irr}(G)\) be the set of all complex irreducible characters of G. We write

$$\begin{aligned} \textrm{Irr}_2(G):= \{\chi \in \textrm{Irr}(G)\mid \chi (1) = 1 \,\ \text {or}\,\ 2\mid \chi (1) \} \end{aligned}$$

and

$$\begin{aligned} \textrm{S}^m_2(G):={\sum \limits _{\chi \in \textrm{Irr}_2(G)}\chi (1)^m}/{\sum \limits _{\chi \in \textrm{Irr}_2(G)}\chi (1)^{m-1}}, \end{aligned}$$

where m is a positive integer. Ito–Michler theorem [3, 5] states that if a prime p does not divide the degree of every irreducible character of a finite group G, then G has a normal abelian Sylow p-subgroup. For the prime \(p=2\), the \(\textrm{S}^m_2(G)\)-version of the Ito–Michler theorem asserts that if \(\textrm{S}^m_2(G)=1\), then G has a normal abelian Sylow 2-subgroup. In this paper, we improve this for \(p=2\).

Main theorem. Let G inite group with \(\textrm{S}^m_2(G) < (1+2^{m-1})/(1+2^{m-2})\). Then, G has a normal Sylow 2-subgroup.

We note that \((1+2^{m-1})/(1+2^{m-2})\) is the exactly value of \(\textrm{S}^m_2(G)\) with \(G = S_3\), the symmetric group of degree 3, and \(S_3\) does not have a normal Sylow 2-subgroup. For G, the nonabelian group of order 8 or the alternating group \(A_4\) of degree 4, the value of \(\textrm{S}^m_2(G)\) respectively is \((1+2^{m-2})/(1+2^{m-3})\) and 1, which are both less than \(\textrm{S}^m_2(S_3)\), indeed G has a normal Sylow 2-subgroup.

For \(m=1, 2\), we obtain the following corollary.

Corollary 1.1

Let G be a finite group. If \(\textrm{S}^1_2(G)< 4/3\) or \(\textrm{S}^2_2(G)< 3/2,\) then G has a normal Sylow 2-subgroup.

Proof

See [1, Theorem 1.1] and [6, Theorem A].\(\square \)

In the following of this paper, we prove the Main theorem in Sect. 2, and discuss other variations of the Main theorem in Sect. 3.

2 Proof of Main Theorem

We denote by \(n_k(G)\) the number of irreducible complex characters of degree k of G. If N is a normal subgroup of G and \(\theta \in \textrm{Irr}(N)\), then \(\textrm{Irr}(G|\theta )\) denotes the set of irreducible characters of G that lie over \(\theta \). We write \(I_G(\theta )\) to denote the inertia subgroup of \(\theta \) in G. For an M-invariant subgroup N, we write \(M\ltimes N\) to denote a semidirect of M and N.

The first lemma is the following observation.

Lemma 2.1

Let N be a normal subgroup of a group G contained in the derived subgroup \(G'\) of G. If \(\textrm{S}^m_2(G) \leqslant 2,\) then \(\textrm{S}^m_2(G/N) \leqslant \textrm{S}^m_2(G)\).

Proof

We write \(R=\textrm{S}^m_2(G)\). Then, \(1 \leqslant R \leqslant 2\) and

$$\begin{aligned} R = \frac{n_1(G)+\sum \limits _{2 \mid k}k^m n_k(G)}{n_1(G)+ \sum \limits _{2 \mid k}k^{m-1}n_k(G)}. \end{aligned}$$

It follows that

$$\begin{aligned} \sum \limits _{2\mid k}(k-R)k^{m-1}n_k(G)=(R-1)n_1(G). \end{aligned}$$

Since \(\textrm{Irr}(G/N) \subseteq \textrm{Irr}(G)\) and \(N \subseteq G'\), we have

$$\begin{aligned} \sum \limits _{2\mid k}(k-R)k^{m-1}n_k(G/N) \leqslant \sum \limits _{2\mid k}(k-R)k^{m-1}n_k(G) \end{aligned}$$

and \(n_1(G)=n_1(G/N)\). Hence \(\sum \limits _{2 \mid k}(k-R)k^{m-1}n_k(G/N) \leqslant (R-1)n_1(G/N)\), which is equivalent to \(\textrm{S}^m_2(G/N) \leqslant R = \textrm{S}^m_2(G)\). \(\square \)

Lemma 2.2

For a normal subgroup N of G,  let \(\chi \in \textrm{Irr}(G)-\textrm{Irr}(G/N)\) with degree 2. If \(G/\textrm{ker}(\chi )\) is nonsolvable, then \(\chi \) is a primitive character of \(G/\textrm{ker}(\chi )\).

Proof

We write \(\overline{G}=G/\textrm{ker}(\chi )\), and suppose that \(\chi \) is not a primitive character of \(\overline{G}\). Then, there is a proper subgroup \(\overline{H}\) of \(\overline{G}\) and some character \(\phi \) of \(\overline{H}\) such that \(\chi = \phi ^{\overline{G}}\). It follows that \(2=\chi (1)=|\overline{G}:\overline{H}|\phi (1)\), \(|\overline{G}:\overline{H}|=2\) and \(\phi (1)=1\). Hence, \(\overline{H}\) is normal in \(\overline{G}\) and all the irreducible constituents of \(\chi _{\overline{H}}\) are linear. Then, \([\overline{H},\overline{H}] \leqslant {\ker }(\chi )=1\), and \(\overline{H}\) is abelian. So \(\overline{G}\) is solvable, a contradiction. \(\square \)

The following lemma implies that G in our main theorem is solvable.

Lemma 2.3

Let G be a finite group with \(\textrm{S}^m_2(G) < (1 + 4^m)/(1 + 4^{m-1})\). Then, G is solvable.

Proof

Suppose that G is nonsolvable and let G be a counterexample of minimal order. Let A be a minimal nonsolvable normal subgroup of G, and let N be a minimal normal subgroup of G contained in A. Then, \(N \subseteq A = A' \subseteq G'\). If in addition \([A, \textrm{Rad}(A)] > 1\), then we can choose \(N \subseteq [A, \textrm{Rad}(A)]\), where \(\textrm{Rad}(A)\) is the solvable radical (i.e., the unique largest solvable normal subgroup) of A. We now discuss the following two cases: N is abelian or not.

(1). N is abelian.

Then, G/N is nonsolvable. Since \(|G/N|<|G|\), we have \(\textrm{S}^m_2(G) < (1+4^m)/(1+4^{m-1}) \leqslant \textrm{S}^m_2(G/N)\). Hence, we obtain

$$\begin{aligned} \sum \limits _{2\mid k, k\geqslant 4}(k-1+4^{m-1}k-4^m)k^{m-1}n_k(G) < 3\cdot 4^{m-1}n_1(G) + 2^{m-1}(2\cdot 4^{m-1}-1)n_2(G) \end{aligned}$$

and

$$\begin{aligned}{} & {} 3\cdot 4^{m-1}n_1(G/N) + 2^{m-1}(2\cdot 4^{m-1}-1)n_2(G/N) \\{} & {} \quad \leqslant \sum \limits _{2\mid k, k \geqslant 4}(k-1+4^{m-1}k-4^m)k^{m-1}n_k(G/N). \end{aligned}$$

Since \(n_1(G/N)=n_1(G)\) and \(n_k(G/N) \leqslant n_k(G)\) for \(k \geqslant 2\), we obtain \(n_2(G/N) < n_2(G)\). Now there is \(\chi \in \textrm{Irr}(G)-\textrm{Irr}(G/N)\) such that \(\chi (1)=2\). Since \(N \nsubseteq \textrm{ker}(\chi )\), we know that \(A\textrm{ker}(\chi )/\textrm{ker}(\chi )\) is a nontrivial subgroup of \(G/\textrm{ker}(\chi )\), and thus \(G/\textrm{ker}(\chi )\) is nonsolvable. Now by Lemma 2.2, we know that \(G/\textrm{ker}(\chi )\) is a nonsolvable primitive linear group of degree 2. Write \(C/\textrm{ker}(\chi )=\textrm{Z}(G/\textrm{ker}(\chi ))\). According to the proof in [6, Theorem 3.1], we know that \(G=AC\) is a central product, \(G/C\cong A_5\), \(N=A\cap C \cong \mathbb {Z}_2\), \(A \cong SL(2,5)\), \(n_1(G)=n_1(C/N)\), \(n_2(G)=n_2(C/N)+2n_1(G)\), \(n_4(G) \geqslant 2n_1(G)\), and \(n_6(G) \geqslant n_1(G) + 2n_2(C/N)\). Now

$$\begin{aligned}&\sum \limits _{2\mid k, k\geqslant 4}(k-1+4^{m-1}k-4^m)k^{m-1}n_k(G)\\&\quad \geqslant 3\cdot 4^{m-1}n_4(G) + (5+2\cdot 4^{m-1})\cdot 6^{m-1}n_6(G)\\&\quad \geqslant 6\cdot 4^{m-1}n_1(G) + (5+2\cdot 4^{m-1})6^{m-1}\cdot (n_1(G)+2n_2(C/N))\\&\quad =(6\cdot 4^{m-1}+ 5\cdot 6^{m-1}+2\cdot 24^{m-1})n_1(G) + (10+4^m)6^{m-1}n_2(C/N) \\&\quad =(6\cdot 4^{m-1}+ 5\cdot 6^{m-1}+2\cdot 24^{m-1})n_1(G) + (10+4^m)6^{m-1}(n_2(G)-2n_1(G)) \\&\quad =(6\cdot 4^{m-1}- 15\cdot 6^{m-1} -6 \cdot 24^{m-1})n_1(G) + (10+4^m)6^{m-1}n_2(G) \\&\quad =(6\cdot 4^{m-1}- 15\cdot 6^{m-1} -6 \cdot 24^{m-1})n_1(G) + [(10+4^m)6^{m-1}\\&\qquad -2^{m-1}(2\cdot 4^{m-1}-1)]n_2(G)+2^{m-1}(2\cdot 4^{m-1}-1)n_2(G)\\&\quad \geqslant (6\cdot 4^{m-1}- 15\cdot 6^{m-1} -6 \cdot 24^{m-1})n_1(G)+ 2[(10+4^m)6^{m-1} \\&\qquad -2^{m-1}(2\cdot 4^{m-1}-1)]n_1(G)+2^{m-1}(2\cdot 4^{m-1}-1)n_2(G)\\&\quad =(5\cdot 6^{m-1} + 2\cdot 24^{m-1} + 6 \cdot 4^{m-1} -4\cdot 8^{m-1}+ 2^m)n_1(G) \\&\qquad + 2^{m-1}(2\cdot 4^{m-1}-1)n_2(G) \\&\quad \geqslant 3\cdot 4^{m-1}n_1(G)+2^{m-1}(2\cdot 4^{m-1}-1)n_2(G). \end{aligned}$$

Hence, \(\textrm{S}^m_2(G) \geqslant (1+4^m)/(1+4^{m-1})\), a contradiction.

(2). N is not abelian.

Assume that \(N \ncong A_5\). By [1, Theorem 2.2], there exists \(\phi \in \textrm{Irr}(N)\) of even degree such that \(\phi (1) \geqslant 8\) and \(\phi \) is extendible to \(I:= I_G(\phi )\). By [1, Proposition 2.3], we have \(n_1(G) \leqslant n_d(G)|G: I|\) and \(n_2(G) \leqslant n_{2d}(G)|G: I| + \frac{1}{2}n_d(G)|G: I|\), where \(d=\phi (1)|G: I| \geqslant 8|G: I|\). Now

$$\begin{aligned}&3 \cdot 4^{m-1}n_1(G) + 2^{m-1}(2\cdot 4^{m-1}-1)n_2(G) \\&\quad \leqslant 3\cdot 4^{m-1} n_d(G)|G: I| + 2^{m-1}(2 \cdot 4^{m-1}-1)n_{2d}(G)|G: I| \\&\qquad + 2^{m-2}(2\cdot 4^{m-1}-1)n_d(G)|G: I|\\&\quad = (3\cdot 4^{m-1} + 8^{m-1}-2^{m-2})n_d(G)|G:I| + 2^{m-1}(2 \cdot 4^{m-1}-1)n_{2d}(G)|G:I|\\&\quad \leqslant (3\cdot 4^{m-1} + 8^{m-1}-2^{m-2})n_d(G)\frac{d}{8} + 2^{m-1}(2 \cdot 4^{m-1}-1)n_{2d}(G)\frac{d}{8}\\&\quad = \left( \frac{3}{2}\cdot 4^{m-2} + 8^{m-2}-2^{m-5}\right) dn_d(G) + 2^{m-4}(2 \cdot 4^{m-1}-1)dn_{2d}(G)\\&\quad \leqslant (d-1 + 4^{m-1}\cdot d-4^m)d^{m-1}n_d(G) + (2d-1+4^{m-1}\cdot 2d -4^m)(2d)^{m-1}n_{2d}(G) \\&\quad \leqslant \sum \limits _{2\mid k, k\geqslant 4}(k-1+4^{m-1}k-4^m)k^{m-1}n_k(G), \end{aligned}$$

so we have \(\textrm{S}^m_2(G) \geqslant (1+4^m)/(1+4^{m-1})\), a contradiction.

Assume that \(N \cong A_5\). Since the irreducible character of N of degree 4 is extendible to G, by [1, Proposition 2.3] again, we have \(n_1(G) \leqslant n_4(G)\) and \(n_2(G) \leqslant n_8(G)\). Now

$$\begin{aligned}&3 \cdot 4^{m-1}n_1(G) + 2^{m-1}(2\cdot 4^{m-1}-1)n_2(G) \\&\quad \leqslant 3 \cdot 4^{m-1}n_4(G) + 2^{m-1}(2\cdot 4^{m-1}-1)n_8(G) \\ {}&\quad \leqslant 3 \cdot 4^{m-1}n_4(G) + (7+4^m)8^{m-1}n_8(G)\\&\leqslant \sum \limits _{2\mid k, k\geqslant 4}(k-1+4^{m-1}k-4^m)k^{m-1}n_k(G), \end{aligned}$$

and thus \(\textrm{S}^m_2(G) \geqslant (1+4^m)/(1+4^{m-1})\). This contradiction completes the proof.\(\square \)

Lemma 2.4

Let \(G=M\ltimes N\). If \(\lambda \) is a G-invariant linear character of N,  then \(\lambda \) is extendible to G.

Proof

Clearly, we may assume that \(\textrm{ker}(\lambda )=1\). Then, N is cyclic. Let \(\chi \in \textrm{Irr}(\lambda ^G)\), then \(\chi _N=\chi (1)\lambda \). Hence, \(N \subseteq \textrm{Z}(\chi )=\{g \in G\big | |\chi (g)|=\chi (1)\}\) and \(N\textrm{ker}(\chi )/\textrm{ker}(\chi ) \subseteq \textrm{Z}(\chi )/\textrm{ker}(\chi ) = \textrm{Z}(G/\textrm{ker}(\chi ))\). Then, \([G, N] \subseteq \textrm{ker}(\chi )\). It follows that \([G, N] \subseteq N \cap {\ker }(\chi ) = \textrm{ker}(\lambda ) = 1\). Hence, \(N \subseteq \textrm{Z}(G)\). Now \(G = M \times N\). Let \(\psi =1_m \times \lambda \). Then, \(\psi \in \textrm{Irr}(G)\) and \(\psi _N=\lambda \). So \(\lambda \) is extendible to G. \(\square \)

The following lemma is important in the proof of our Main theorem.

Lemma 2.5

Let \(G=M\ltimes N,\) where \(N \leqslant G'\) is an abelian group. Assume that no nontrivial irreducible character of N is invariant under M. If \(\textrm{S}^m_2(G) < (1+2^{m-1})/(1+2^{m-2}),\) then there is no orbit of even size in the action of M on the set of irreducible characters of N.

Proof

Let \(\{\theta _0=1_N, \theta _1, \cdots , \theta _t\}\) be a set of representatives of M-orbits on \(\textrm{Irr}(N)\). Let \(I_i= I_G(\theta _i)\) for \(i \in [1, t]\). By hypothesis, we have \(I_i < G\) for \(i \geqslant 1\). Suppose that there is some orbit of even size in the action of M on \(\textrm{Irr}(N)\). Then, we can find an integer k such that \(2\mid |G:I_k|\). For \(0 \leqslant i \leqslant t\), we set \(n_{i,1}=n_1(I_i/N)\) and \(\textrm{T}_{i, m} = \sum \limits _{\lambda \in \textrm{Irr}(I_i/N), 2\mid \lambda (1)}\lambda (1)^m\).

By Lemma 2.4, every \(\theta _i\) has an extension \(\psi \) to \(I_i\). By Gallagher theorem [2, Corollary 6.17], we have bijections \(\lambda \mapsto \lambda \psi _i\) from \(\textrm{Irr}(I_i/N)\) to \(\textrm{Irr}(I_i|\theta _i)\). By Clifford correspondence, we have a bijections \(\lambda \psi _i \mapsto (\lambda \psi _i)^G\) from \(\textrm{Irr}(I_i|\theta _i)\) to \(\textrm{Irr}(G|\theta _i)\). Observe that \((\lambda \theta _i)^G(1)=|G:I_i|\lambda (1)\) is even if and only if \(|G:I_i|\) is even or \(|G: I_i|\) is odd and \(\lambda (1)\) is even. Then,

$$\begin{aligned} \sum \limits _{\chi \in \textrm{Irr}_2(G)}\chi (1)^{m-1}&= n_1(G/N) + \sum \limits _{2\mid |G:I_i|}|G:I_i|^{m-1}\sum \limits _{\lambda \in \textrm{Irr}(I_i/N)}\lambda (1)^{m-1}\\&\quad + \sum \limits _{2 \not \mid |G:I_i|}|G:I_i|^{m-1}\textrm{T}_{i, m-1} \end{aligned}$$

and

$$\begin{aligned}&\sum \limits _{\chi \in \textrm{Irr}_2(G)}\chi (1)^{m} = n_1(G/N) + \sum \limits _{2\mid |G:I_i|}|G:I_i|^{m}\sum \limits _{\lambda \in \textrm{Irr}(I_i/N)}\lambda (1)^{m} + \sum \limits _{2 \not \mid |G:I_i|}|G:I_i|^{m}\textrm{T}_{i,m}\\&\quad \geqslant n_1(G/N) + \sum \limits _{2\mid |G:I_i|}|G:I_i|^{m}\sum \limits _{\lambda \in \textrm{Irr}(I_i/N)}\lambda (1)^{m-1} + \sum \limits _{2 \not \mid |G:I_i|}|G:I_i|^{m}\textrm{T}_{i, m-1}. \end{aligned}$$

From \(\sum \limits _{\chi \in \textrm{Irr}_2(G)}\chi (1)^{m}=\textrm{S}^m_2(G)\sum \limits _{\chi \in \textrm{Irr}_2(G)}\chi (1)^{m-1}\), we have

$$\begin{aligned}&\quad \textrm{S}^m_2(G)(n_1(G/N) + \sum \limits _{2\mid |G:I_i|}|G:I_i|^{m-1}\sum \limits _{\lambda \in \textrm{Irr}(I_i/N)}\lambda (1)^{m-1} + \sum \limits _{2 \not \mid |G:I_i|}|G:I_i|^{m-1}\textrm{T}_{i, m-1})\\&\geqslant n_1(G/N) + \sum \limits _{2\mid |G:I_i|}|G:I_i|^{m}\sum \limits _{\lambda \in \textrm{Irr}(I_i/N)}\lambda (1)^{m-1} + \sum \limits _{2 \not \mid |G:I_i|}|G:I_i|^{m}\textrm{T}_{i, m-1}. \end{aligned}$$

Then,

$$\begin{aligned}&\sum \limits _{2 \mid |G:I_i|}(|G:I_i|-\textrm{S}^m_2(G))|G:I_i|^{m-1}\sum \limits _{\lambda \in \textrm{Irr}(I_i/N)}\lambda (1)^{m-1}\\&\qquad + \sum \limits _{2 \not \mid |G:I_i|}(|G:I_i|-\textrm{S}^m_2(G))|G:I_i|^{m-1}\textrm{T}_{i,m-1}\\&\quad \leqslant (\textrm{S}^m_2(G)-1)n_1(G/N). \end{aligned}$$

In particular, for k, we obtain

$$\begin{aligned} (|G:I_k|-\textrm{S}^m_2(G))|G:I_k|^{m-1}\sum \limits _{\lambda \in \textrm{Irr}(I_k/N)}\lambda (1)^{m-1} \leqslant (\textrm{S}^m_2(G)-1)n_1(G/N). \end{aligned}$$

Since \(n_1(G/N)=|G/N:G'/N|\) and \(n_{k,1}=n_1(I_k/N)=|I_k/N, (I_k/N)'|\), so \(n_1(G/N) \leqslant |G:I_k|n_{k,1}\). In addition, \(\sum \limits _{\lambda \in \textrm{Irr}(I_k/N)}\lambda (1)^{m-1} \geqslant n_1(I_k/N)\). Hence, we get that

$$\begin{aligned} (|G:I_k|-\textrm{S}^m_2(G))|G:I_k|^{m-1}n_1(I_k/N) \leqslant (\textrm{S}^m_2(G)-1)|G:I_k|n_1(I_k/N). \end{aligned}$$

It then follows that

$$\begin{aligned} \textrm{S}^m_2(G) \geqslant \frac{|G:I_k|^m + |G:I_k|}{|G:I_k|^{m-1} + |G:I_k|} \geqslant \frac{2^{m-1}+1}{2^{m-2}+1}, \end{aligned}$$

which is a contradiction.\(\square \)

We now prove our main theorem, which is restated.

Theorem 2.6

Let G be a finite group. Suppose that \(\textrm{S}^m_2(G) < (1+2^{m-1})/(1+2^{m-2})\). Then, G has a normal Sylow 2-subgroup.

Proof

Suppose the theorem is false and let G be a counterexample of minimal order. By Lemma 2.3, G is solvable. Let N be a minimal normal subgroup of G contained in the derived subgroup \(G'\) of G. Then, N is elementary abelian. By Lemma 2.1, \(\textrm{S}^m_2(G/N) \leqslant \textrm{S}^m_2(G) < (1+2^{m-1})/(1+2^{m-2})\), and by the choice of G, G/N has a normal Sylow 2-subgroup R/N. If N is a 2-group, then R is a normal Sylow 2-subgroup of G, a contradiction. Hence, N is a \(2'\)-group. By the Schur–Zassenhaus theorem, \(R = P \ltimes N\), where P is a Sylow 2-subgroup of R. By the Frattini argument, \(G=RN_G(P) = NN_G(P)\). Clearly, \(N \nsubseteq N_G(P)\). Since \(N_G(P)\cap N < N\) and \(N_G(P)\cap N \unlhd G= NN_G(P)\), we obtain \(N_G(P)\cap N=1\), and thus \(G=N_G(P)\ltimes N\). If \(N\subseteq Z(G)\), then \(R= P \times N\) and \(P \unlhd G\), a contradiction. Hence, \(N \nsubseteq Z(G)\). Then, \([N,G]=N\). Let \(\lambda \in \textrm{Irr}(N)-1_N\). Assume that \(\lambda \) is G-invariant. Then by Lemma 2.4, \(\lambda \) has an extension \(\chi \) to G. Now we have \(\chi _N = \lambda \), \(N \subseteq \textrm{ker}(\chi )\), and \(\lambda = 1_N\), this contradiction implies that no nontrivial irreducible character of N is G-invariant. By Lemma 2.5, there is no orbit of even size in the action of P on N, that is, P acts trivially on N, hence \(R=P \times N\) and \(P \unlhd G\), this final contradiction completes the proof. \(\square \)

3 Variations of Main Theorem

Recall that a character \(\chi \in \textrm{Irr}(G)\) is real if \(\chi (g) \in \mathbb {R}\) for every element \(g \in G\), and \(\chi \in \textrm{Irr}(G)\) is strongly real if \(\chi \) is afforded by a real representation, or equivalently, its Frobenius–Schur indicator \(v_2(\chi )\) is 1. We denote by \(n_{k,+}\) the number of irreducible strongly real characters of degree k of G. We write

$$\begin{aligned}&\textrm{Irr}_{2,\mathbb {R}}(G):= \{\chi \in \textrm{Irr}_2(G)\mid \chi \,\ \text {is real} \},\\&\textrm{S}^m_{2, \mathbb {R}}(G):={\sum \limits _{\chi \in \textrm{Irr}_{2,\mathbb {R}}(G)}\chi (1)^m}/{\sum \limits _{\chi \in \textrm{Irr}_{2,\mathbb {R}}(G)}\chi (1)^{m-1}},\\&\textrm{Irr}_{2,+}(G):= \{\chi \in \textrm{Irr}_2(G)\mid v_2(\chi )=1 \},\\&\textrm{S}^m_{2, +}(G):={\sum \limits _{\chi \in \textrm{Irr}_{2, +}(G)}\chi (1)^m}/{\sum \limits _{\chi \in \textrm{Irr}_{2, +}(G)}\chi (1)^{m-1}}. \end{aligned}$$

Similar to Lemma 2.1, we have the following observation.

Lemma 3.1

Let N be a normal subgroup of a group G such that \(N \subseteq G'\). If \(\textrm{S}^m_{2, +}(G) \leqslant 2\), then \(\textrm{S}^m_{2,+}(G/N) \leqslant \textrm{S}^m_{2, +}(G)\).

Proof

Similar to Lemma 2.1. \(\square \)

For a normal subgroup with odd index, we have the next lemma.

Lemma 3.2

Let N be a normal subgroup of a group G with G/N odd order. If \(\textrm{S}^m_{2, +}(G) \leqslant 2,\) then \(\textrm{S}^m_{2, +}(N) \leqslant \textrm{S}^m_{2, +}(G)\).

Proof

Firstly, every strongly real linear character of G restricts to a strongly real linear character of N. Secondly, by [4, Lemma 2.1], every strongly real linear character of N lies under a unique strongly real linear character of G. Thus, we obtain \(n_{1, +}(N) = n_{1, +}(G)\). In addition, from [4, Lemma 2.1], we also have

$$\begin{aligned} \sum \limits _{2\mid k, k\geqslant 2}n_{k, +}(G) \geqslant \sum \limits _{2\mid k, k\geqslant 2}n_{k, +}(N). \end{aligned}$$

Denote by \(V = \textrm{S}^m_{2, +}(G)\). Then

$$\begin{aligned} V = \frac{n_{1,+}(G)+\sum \limits _{2\mid k}k^m n_{k, +}(G)}{n_{1,+}(G)+\sum \limits _{2\mid k}k^{m-1} n_{k, +}(G)}, \end{aligned}$$

and we obtain

$$\begin{aligned} (V-1)n_{1,+}(G) = \sum \limits _{2\mid k}(k^m-Vk^{m-1}) n_{k, +}(G). \end{aligned}$$

It then follows that

$$\begin{aligned} (V-1)n_{1,+}(N) = \sum \limits _{2\mid k}(k^m-Vk^{m-1}) n_{k, +}(G) \geqslant \sum \limits _{2\mid k}(k^m-Vk^{m-1}) n_{k, +}(N), \end{aligned}$$

which is equivalent to \(\textrm{S}^m_{2, +}(N) \leqslant V = \textrm{S}^m_{2, +}(G)\), the proof is complete. \(\square \)

Now we give a variation of Main theorem for strongly real character.

Theorem 3.3

Let G be a finite group. Suppose that \(\textrm{S}^m_{2,+}(G) < (1+2^{m-1})/(1+2^{m-2})\). Then, G has a normal Sylow \(2-\)subgroup.

Proof

Suppose that the theorem is not true, and let G be a counterexample of minimal order. Let N be a minimal normal subgroup of G contained in \(G'\). Observe that

$$\begin{aligned} \frac{\sum \limits _{\chi \in \textrm{Irr}_{2,+}(G)}\chi (1)}{|\textrm{Irr}_{2,+}(G)|}&\leqslant \frac{n_{1,+}(G) + \sum \limits _{2 \mid k}k^2n_{k,+}(G)}{n_{1,+}(G) + \sum \limits _{2 \mid k}kn_{k,+}(G)}\\&\leqslant \cdots \\&\leqslant \frac{n_{1,+}(G) + \sum \limits _{2 \mid k}k^mn_{k,+}(G)}{n_{1,+}(G) + \sum \limits _{2 \mid k}k^{m-1}n_{k,+}(G)}\\&= \textrm{S}^m_{2,+}(G)\\&< \frac{1+2^{m-1}}{1+2^{m-2}} < 2, \end{aligned}$$

by [1, Theorem 5.1] and Lemma 3.1, we know that G is solvable, N is an elementary abelian group with odd order, G/N has a normal Sylow 2-subgroup R/N, and \(G = R = P \ltimes N\), where \(P \in \textrm{Syl}_2(G)\).

From the proof of [6, Theorem 5.1], we have that \(n_{1, +}(G) \leqslant |N| - 1 \leqslant \chi _1(1) + \cdots + \chi _s(1)\), where \(\chi _1, \cdots , \chi _s\) are the strongly real irreducible characters of G. Now

$$\begin{aligned} \frac{n_{1,+}(G)+\sum \limits _{i=1}^s\chi _i(1)^mn_{\chi _i(1), +}}{n_{1,+}(G)+\sum \limits _{i=1}^s\chi _i(1)^{m-1}n_{\chi _i(1), +}}&\geqslant \frac{n_{1,+}(G)+\sum \limits _{i=1}^s\chi _i(1)^m}{n_{1,+}(G)+\sum \limits _{i=1}^s\chi _i(1)^{m-1}}\\&\geqslant \frac{n_{1,+}(G)+ 2^{m-1}\sum \limits _{i=1}^s\chi _i(1)}{n_{1,+}(G)+2^{m-2}\sum \limits _{i=1}^s\chi _i(1)}\\&\geqslant \frac{n_{1,+}(G)+ 2^{m-1}n_{1,+}(G)}{n_{1,+}(G)+2^{m-2}n_{1,+}(G)}\\&= \frac{1+2^{m-1}}{1+2^{m-2}}. \end{aligned}$$

Then for any \(\chi \in \textrm{Irr}_{2,+}(G)\) with \(\chi (1) \geqslant 2 \geqslant (1+2^{m-1})/(1+2^{m-2})\), we have

$$\begin{aligned} \textrm{S}^m_{2,+}(G)=\frac{{\sum \limits _{\chi \in \textrm{Irr}_{2, +}(G)}\chi (1)^m}}{{\sum \limits _{\chi \in \textrm{Irr}_{2, +}(G)}\chi (1)^{m-1}}} \geqslant \frac{1+2^{m-1}}{1+2^{m-2}}, \end{aligned}$$

and this contradiction completes the proof. \(\square \)

Note that for any \(\chi \in \textrm{Irr}_{2,\mathbb {R}}(G) - \textrm{Irr}_{2, +}(G)\), we have \(\chi (1) \geqslant 2\). If \(\textrm{S}^m_{2,+}(G) \leqslant 2\), then we have

$$\begin{aligned} \frac{\sum \limits _{\chi \in \textrm{Irr}_{2,+}(G)}\chi (1)^m}{\sum \limits _{\chi \in \textrm{Irr}_{2,+}(G)}\chi (1)^{m-1}} \leqslant \frac{\sum \limits _{\chi \in \textrm{Irr}_{2,\mathbb {R}}(G)}\chi (1)^m-\sum \limits _{\chi \in \textrm{Irr}_{2, +}(G)}\chi (1)^m}{\sum \limits _{\chi \in \textrm{Irr}_{2,\mathbb {R}}(G)}\chi (1)^{m-1}-\sum \limits _{\chi \in \textrm{Irr}_{2, +}(G)}\chi (1)^{m-1}}. \end{aligned}$$

Therefore, we obtain the following corollary, which is a variation of Main theorem for real characters.

Corollary 3.4

Let G be a finite group with \(\textrm{S}^m_{2,\mathbb {R}}(G) < (1+2^{m-1})/(1+2^{m-2})\). Then, G has a normal Sylow \(2-\)subgroup.