1 Introduction

All groups considered in this paper are finite. Let G be a group, and x an element of G. We denote by x G the conjugacy class of G containing x and by | x G | the size of x G .

The relationship between the p-regular conjugacy class sizes and the structure of a group G has been studied by many authors, see, for example, [15]. Let N be a normal subgroup of a group G. Clearly N is the union of some conjugacy classes of G. So, it is interesting to decide the structure of N by some arithmetical properties of the G-conjugacy class contained in N, for example, [3, 6, 7]. Particularly, in [3], we decided the structure of N when N possesses two G-conjugate class sizes. In this paper, the case considered is that N has more than two G-conjugate class sizes.

In a recent paper [1], the authors studied the structure of G under the condition that the largest two p-regular conjugacy class sizes (say, m and n) of G are coprime, where m>n and pn. Notice that, when G=N, the condition n dividing |N/Z(N)| is of course true, so our aim is, by eliminating the assumption pn, to investigate the properties of N under the corresponding condition. More precisely, we prove the following.

Theorem A Let N be a normal subgroup of a p-solvable group G. If m=| b G |>| a G |=n are the two longest sizes of the non-central p-regular G-conjugacy classes of N with (m,n)=1 and n dividing |N/Z(N)|, where a,bN, then either a p-complement of N/Z(N) is a prime power order group or

  1. (i)

    | N | p = | a N | p | b N | p | Z ( N ) | p ;

  2. (ii)

    | x G |=m for any non-central p-regular element x C N (b). Furthermore, C N ( b ) p is abelian;

  3. (iii)

    if d and t are two non-central p-regular elements of N such that | t G |m=| d G |, then ( C N ( t ) C N ( d ) ) p =Z ( N ) p Z(G) and n p divides | t N |.

Based on this, in Section 4, we give our improvement and generalization of [1] by considering the case that p does not divide n.

Let π be a set of some primes; we use x π and x π for the π-component and the π -component of x, respectively. Moreover, G π denotes a Hall π-subgroup of G, G π a Hall π -subgroup of G, n π the π-part of n whenever n is a positive integer. Apart from these, we call an element x non-central if xZ(G), where Z(G) is the center of G. Following [8] a group G is said to be quasi-Frobenius if G/Z(G) is a Frobenius group and then the inverse images of the kernel and a complement of G/Z(G) are called the kernel and complement of G.

2 Preliminaries

We first list some lemmas which are useful in the proof of our main result.

Lemma 2.1 [[9], Lemma 1.1]

Let N be a normal subgroup of a group G and x an element of G. Then:

  1. (a)

    | x N | divides | x G |;

  2. (b)

    | ( N x ) G / N | divides | x G |.

Lemma 2.2 Let N be a p-solvable normal subgroup of a group G and B= b G , C= c G with (|B|,|C|)=1, where b, c are two p-regular elements of N. Then:

  1. (a)

    G= C G (b) C G (c).

  2. (b)

    BC=CB is a p-regular G-conjugacy class of N and |BC| divides |B||C|.

Proof Set G=N in Lemma 1 of [4], the proof is finished. □

Lemma 2.3 Let N be a p-solvable normal subgroup of a group G and B 0 be a non-central p-regular G-conjugacy class of N with the largest size. Then the following properties hold:

  1. (a)

    Let C be a p-regular G-conjugacy class of N with (| B 0 |,|C|)=1, then | C 1 C| divides | B 0 |.

  2. (b)

    Let n,m=| B 0 | be two largest p-regular G-conjugacy class sizes of N with (m,n)=1 and D be a p-regular G-conjugacy class of N with |D|>1. If (|D|,n)=1, then |D|=m.

Proof (a) By Lemma 2.2(b), C B 0 is a p-regular G-conjugacy class of N. Clearly, |C B 0 || B 0 |, so the hypotheses of the lemma imply that |C B 0 |=| B 0 |, from which it follows that C 1 C B 0 = B 0 , and hence C 1 C B 0 = B 0 , consequently | C 1 C| divides | B 0 |.

  1. (b)

    Suppose that A is a p-regular G-conjugacy class and |A|=n. Lemma 2.2(b) implies that DA is a p-regular G-conjugacy class. Also |DA||A|, so |DA|=n or m. If |DA|=n, then D 1 DA is a p-regular G-conjugacy class, and hence D 1 DA=A, which implies that D 1 DA=A. It follows that | D 1 D| divides |A|. On the other hand, D 1 D A 1 A, so | D 1 D| divides | A 1 A|. By (a), we find that | A 1 A| divides | B 0 |, from which it follows that | D 1 D| divides | B 0 |, a contradiction. Consequently |DA|=m, equivalently, | B 0 | divides |A||D|, it follows that |D|=| B 0 | by the hypotheses of the lemma, as wanted. □

Lemma 2.4 Suppose that N is a p-solvable normal subgroup of a group G, Let B 0 be a non-central p-regular G-conjugacy class of N with the largest size. Write

M= D | D  is a  p -regular  G -conjugacy class of  N  with  ( | D | , | B 0 | ) = 1 .

Then M p is abelian, furthermore, if ( Z ( G ) N ) p < M p , then π( M p / ( Z ( G ) N ) p )π( B 0 ).

Proof Write

K= D 1 D | D  is a  p -regular  G -conjugacy class of  N  with  ( | D | , | B 0 | ) = 1 .

By the definition of M and K, we have K=[M,G]. Let dD, where D is a p-regular G-conjugacy class of N with (|D|,| B 0 |)=1. Applying Lemma 2.3(a), we have π(K)π( B 0 ), which implies that (|K|,|D|)=1, hence | d K |=1. It shows that K= C K (d), so KZ(M). Notice that M/KZ(G/K), we find that M is nilpotent, hence M=P× M p , where P Syl p (M). Obviously, ( Z ( G ) N ) p M p . If ( Z ( G ) N ) p < M p , let rπ( M p / ( Z ( G ) N ) p ), R Syl r (M), then RG. Also 1[R,G][M,G]=K, we have rπ(K)π( B 0 ), which implies that π( M p / ( Z ( G ) N ) p )π( B 0 ). Suppose that D is a generating class of M and dD, then | d R | divides |R|, and | d R | divides |D|. The fact that (|R|,|D|)=1 implies that R= C R (d), so RZ(M) by the nilpotence of M, which shows that M p is abelian. □

3 Proof of Theorem A

In this section we are equipped to prove the main result.

Proof of Theorem A Suppose that N/Z(N) is not a prime power order group. We will complete the proof by the following steps:

Step 1 We may assume that N r Z(G) for every p -prime factor r of |N|.

Otherwise, there exists a p -prime factor r of |N| such that N r Z(G), then N= N r × N r . Obviously, N r satisfies the condition of the theorem. Application of the induction hypothesis to |N| shows that the conclusion of the theorem holds, and hence we may assume that N r Z(G) for every p -prime factor r of |N|.

Step 2 If the p-regular element xZ( C G (b))N, then either xZ(G), or C G (x)= C G (b).

Obviously, C G (b) C G (x), which implies that | x G | divides | b G |, it follows that (| x G |,n)=1. If xZ(G), Lemma 2.3 shows that | x G |=m, so C G (x)= C G (b).

Step 3 We may assume that b is a prime power order q-element (qp).

Let q be a prime factor of o(b), b q be the q-component and C G ( b q )G. Notice that C G (b)= C G ( b q b q )= C G ( b q ) C G ( b q ) C G ( b q ), applying Step 2, we have C G ( b q )= C G (b), and this completes the proof by replacing b with b q .

Step 4 C N (b)= P b Q b ×L, where P b is a Sylow p-subgroup of C N (b), Q b is a Sylow q-subgroup of C N (b), L is a { p , q } -Hall subgroup of C N (b) with LZ( C G (b)). Particularly, if LZ(G), then C N ( b ) p Z( C G (b)).

Let x C N (b) be a { p , q } -element. Notice that C G (bx)= C G (b) C G (x) C G (b), we find that | b G | divides | ( b x ) G |, so the maximality of | b G | implies that | b G |=| ( b x ) G |, and hence C G (bx)= C G (b) C G (x), from which it follows that xZ( C G (b)). Consequently, C N (b)= P b Q b ×L, where P b is a Sylow p-subgroup of C N (b), Q b is a Sylow q-subgroup of C N (b), L is a { p , q } -Hall subgroup of C N (b) with LZ( C G (b)).

Particularly, if LZ(G), let yL be a non-central prime power order r-element, then Step 2 implies that | y G |=m, and hence C G (y)= C G (b). By the above argument, we have C N (y)= P y R y × L y , where P y is a Sylow p-subgroup of C N (b), R y is a Sylow r-subgroup of C N (b), L y is a { p , r } -Hall subgroup of C N (b) with L y Z( C G (b)). Clearly, C N (b)= P b ×L L y , so C N ( b ) p =L L y Z( C G (b)).

Step 5 qm.

If q|m, then qn, and of course we have q| a N | by Lemma 2.1. Notice that | a N |=|N: C N (a)|=| C N (b): C N (a) C N (b)|, hence C N (a) C N (b) contains a Sylow q-subgroup of C N (b), which implies that b C N (a) C N (b), therefore a C N (b). We distinguish two cases according to the structure of C N (b).

  1. (1)

    If LZ(G), Step 4 implies that C N ( b ) p Z( C G (b)). By the above a C N (b), we have C G (b) C G (a), which leads to | a G | dividing | b G |, a contradiction.

  2. (2)

    If LZ(G), then we may assume that a is a q-element since a C N (b)= P b Q b ×L. For every { p , q } -element x C N (a), we have C G (ax)= C G (a) C G (x) C G (a), the hypothesis of the theorem shows that C G (ax)= C G (a) C G (x), from which it follows that xZ( C G (a)). Notice that b C N (a), so x C N (b), which implies that xL, consequently a p-complement of N/Z(N) is a prime power order group, a contradiction.

Step 6 We may assume that a is a { p , q } -element.

Let a= a q a q , where a q , a q are the q-component and q -component of a, respectively. Notice that C G (a)= C G ( a q ) C G ( a q ) C G ( a q ); we have a q M, where M is ever defined in Lemma 2.4. If a q Z(G), then, by Lemma 2.4, qπ( M p / ( Z ( G ) N ) p )π(m), in contradiction to Step 5.

Step 7 C N ( a ) q Z ( G ) q .

Suppose that there exists a non-central q-element y C N (a), then C G (ay)= C G (a) C G (y) C G (a), so we have C G (ay)= C G (a) by the hypothesis of the theorem. It follows that ayM, which implies that yM because of aM. So qπ( M p / ( Z ( G ) N ) p )π(m), in contradiction to Step 5. Hence, C N ( a ) q Z ( G ) q , as required.

Step 8

(8.1) ( C N ( a ) C N ( b ) ) p =Z ( N ) p Z(G).

(8.2) | N | p = | a N | p | b N | p | Z ( N ) | p .

(8.1) Our immediate object is to show that ( C N ( a ) C N ( b ) ) p Z(G). Otherwise, then there exists a non-central p-regular element y C N (a) C N (b). In view of Step 4, if LZ(G), then we may assume that y is a q-element. So the q-element y lies in C N (a), in contradiction to Step 7. So another possibility LZ(G) must prevail, which implies that y C N ( b ) p . Keeping in mind that C N ( b ) p Z( C G (b)), we have C G (y)= C G (b) by the hypotheses, from which it follows that a C N (y)= C N (b). Applying C N ( b ) p Z( C G (b)) once again, we have C G (b) C G (a), a contradiction, as wanted. Consequently, ( C N ( a ) C N ( b ) ) p ( Z ( G ) N ) p . Notice that ( Z ( G ) N ) p Z ( N ) p and Z ( N ) p ( C N ( a ) C N ( b ) ) p , so ( C N ( a ) C N ( b ) ) p =Z ( N ) p , as required.

(8.2) Next, the conclusion | N | p = | a N | p | b N | p | Z ( N ) | p is to be dealt with. Obviously, (| a N |,| b N |)=1 by Lemma 2.1 in terms of (| a G |,| b G |)=1, which implies that N= C N (a) C N (b). This leads to |N|=| C N (a)|| C N (b)|/| C N (a) C N (b)|, and hence |N|=| a N || b N || C N (a) C N (b)|. Notice that ( C N ( a ) C N ( b ) ) p =Z ( N ) p , and we have | N | p = | a N | p | b N | p | Z ( N ) | p , as wanted.

Step 9 It followed that n=| a N | p α , where α0. And hence, if LZ(G), | a G | is at most a {p,q}-number.

Notice that |N|=| a N || b N || C N (a) C N (b)| and ( C N ( a ) C N ( b ) ) p =Z ( N ) p , and by the hypothesis that n divides |N/Z(N)|, we have n=| a N | p α , where α0. On the other hand, | a N |=| C N (b)|/| C N (a) C N (b)|. By what has already been proved, we find that, if LZ(G), then | a N | is at most a {p,q}-number, and so is | a G |.

Step 10 | x G |=m for any non-central p-regular element x C N (b), and therefore C N ( b ) p is abelian.

By the structure of C N (b), we distinguish two cases:

  1. (1)

    If LZ(G), then C N ( b ) p Z( C G (b)) by Step 4. Obviously, C N ( b ) p is abelian. In addition, for any non-central p-regular element x C N (b), we find that | x G | divides | b G |, application of Lemma 2.3 yields | x G |=m.

  2. (2)

    If LZ(G), by Step 9, we know | a G | is at most a {p,q}-number. Also, x C N (b)= P b Q b ×L is a non-central p-regular element; we may assume that x Q b Z(G) if necessary by a suitable conjugate.

On the other hand, we know M p is abelian and Z ( G ) q N M p C N (a), by Step 7, we can write M p =S×(Z ( G ) q N) where q|S|. Notice that x acts coprimely on the abelian subgroup S, and by coprime action properties, we have

S= [ S , x ] × C S (x).

Denote by U=[S,x]. As aS, we can write a=uw with uU, w C S (x). Consider the element g=wx; we have

C G (g)= C G (w) C G (x) C G (x).

If | g G |=m, then | x G |=m by Lemma 2.3 since | x G | divides | g G |.

If | g G |=n, notice that | x G | divides | g G |, then x M q . However, M q C N ( a ) q Z ( G ) q , a contradiction.

So we are left with only one alternative: | g G |<n. Keeping in mind that G is a p-solvable group, let T be a Hall {p,q}-subgroup of G, ST is a subgroup of G since S is a normal subgroup of G. Notice that

| S T : C S T ( g ) | | g G | < n = | a G | = | G : C G ( a ) | = | G : C G ( a ) | { p , q } = | T | : | C G ( a ) | { p , q } .
(3.1)

Moreover, since SG, S is abelian, ST=1, and S C G (w) we have

C S T ( g ) = C S T ( w ) C S T ( x ) = S C T ( w ) C S ( x ) C T ( x ) = C S ( x ) [ S C T ( w ) C T ( x ) ] = C S ( x ) [ C T ( w ) C T ( x ) ] .
(3.2)

We denote D= C T (w) C T (x). Combining (3.1) and (3.2), we have

| T | | C G ( a ) | { p , q } > | S | | T | | C S ( x ) | | D | .

This implies that |D|: | C G ( a ) | { p , q } >|S: C S (x)|=|U|.

On the other hand, as D C G (x) and U=[S,x], then D normalizes U. Also,

C D (u)= C G (u)D= C T (u) C T (w) C T (x) C T (a) C T (x),

so

| C D (u)|| C T (a) C T (x)|| C G (a) | { p , q } .

Therefore

| u D |=|D: C D (u)|=|D|:| C D (u)||D|:| C G (a) | { p , q } >|U|,

a contradiction.

So the argument on the above three cases forces | x G |=m.

Next, we show that ( C N ( b ) ) p is abelian when LZ(G). Notice that C N (b)= P b Q b ×L, it is enough to show that Q b is abelian. For any non-central element x Q b , we have C S (x)Z ( G ) p . Otherwise, by the above, we have | x G |=m. Replacing b with x, we have ( C N ( a ) C N ( b ) ) p Z(G), which contradicts Step 8. Therefore Q b / Q b Z(G) acts on the group S/SZ(G) fixed-point-freely, which implies that Q b / Q b Z(G) is cyclic or a generalized quaternion. So, if Q b is not abelian, then q=2 and Q b / Q b Z(G) is a generalized quaternion group. Now bZ( Q b ) but bZ(G), so there exists an element y Q b and yZ( Q b ) such that b= y 2 c where cZ(G) Q b . So C G (y) C G (b), which indicates that yZ( C G (b)), and of course we have yZ( Q b ), a contradiction. Therefore Q b is abelian, as required.

Step 11 If d and t are two non-central p-regular elements of N such that | t G |m=| d G |, then:

(11.1) ( C N ( t ) C N ( d ) ) p =Z ( N ) p Z(G).

(11.2) | a N | p divides | t N |, and hence n p divides | t N |.

(11.1) Firstly, it will be established that ( C N ( t ) C N ( d ) ) p Z(G). Otherwise, there exists a non-central p-regular element y C N (t) C N (d), and we will distinguish two cases as for the structure of C N (d).

  1. (1)

    If LZ(G), then C N ( d ) p Z( C G (d)) by Step 4, which implies that yZ( C G (d)). However, y is non-central, the hypotheses of the theorem shows that C G (y)= C G (d), from which it follows that t C N ( d ) p . Notice that C N ( d ) p Z( C G (d)), we have C G (t)= C G (d) by the hypotheses of the theorem, but this contradicts the fact that | t G |m=| d G |. Therefore ( C N ( t ) C N ( d ) ) p Z(G).

  2. (2)

    Suppose that LZ(G). Our immediate object is to show that t can be assumed to be a q -element. In fact, let t q be the q-component of t. If t q Z(G), Step 10 implies that | t q G |=m. Notice that C G (t) C G ( t q ), we have | t G |=m by Lemma 2.3, against the fact that | t G |m=| d G |. It follows that t q Z(G), so C G (t)= C G ( t q ), where t q is the q -component of t. Thus, without loss of generality, we may assume that t is a q -element.

Also, in this case, we may assume that y is a q-element. Keeping in mind that qm, application of Step 10 once again, we have | y G |=m. Moreover,

C G (ty)= C G (t) C G (y) C G (y),

which implies that | ( t y ) G |=m by the maximality of m. So C G (ty)= C G (y) C G (t), which shows that | t G |=m by Lemma 2.3(b), a contradiction. Thus, ( C N ( d ) C N ( t ) ) p Z(G).

Next, in a similar manner as in Step (8.1), the equality in (11.1) follows.

(11.2) Consider the quotient group ( C N ( b ) / Z ( N ) ) p and the set { t N }. For any x ¯ ( C N ( b ) / Z ( N ) ) p and y{ t N }, without loss of generality, we may assume that x ¯ =xZ(N) where x C N (b), and we set

y x ¯ = y x .

Clearly, ( C N ( b ) / Z ( N ) ) p acts as a group on the set { t N } through the above action. Obviously, t N C N (b)= and ( C N ( t ) C N ( b ) ) p =Z ( N ) p , this shows that the group ( C N ( b ) / Z ( N ) ) p acts on the set { t N } fixed-point-freely. Therefore | C N ( b ) / Z ( N ) | p divides | t N |. Notice that | C N ( b ) / Z ( N ) | p = | a N | p , so we find that | a N | p divides | t N |. By Step 9, n p divides | t N |, which is fairly straightforward.  □

Corollary 1 Suppose that G is a group. Let | b G |=m>n=| a G | be the two longest sizes of the non-central conjugacy classes of G. If (m,n)=1, then G is solvable and the conjugacy class size of the element in G is exactly 1, n or m.

Proof Let p be a prime and p be not a prime factor of |G|, then G is p-solvable. Obviously, n divides |G/Z(G)|. So, in Theorem A, by taking N=G, we have | x G |= n p =n if | x G |m, from which it follows that x M p . On the other hand, | y G |=n for any non-central element y C G (a). In fact, if | y G |=m, we have y C G (a) C G (y), in contradiction to (iii) in Theorem A. It follows that C G (a)= M p is abelian. By Theorem A, C G (b) is abelian, and the solvability of G is obtained. □

4 Application of Theorem A

Based on Theorem A, we consider the case that p does not divide n.

Theorem B Let N be a normal subgroup of a p-solvable group G. If m=| b G |>| a G |=n are the two longest sizes of the non-central p-regular G-conjugacy classes of N with (m,n)=1 and n dividing |N/Z(N)|, where a,bN and pn, then either a p-complement of N/Z(N) is a prime power order group or

  1. (i)

    the p-regular G-conjugacy class sizes of N are exactly 1, n and m.

  2. (ii)

    Let x be a non-central p-regular element of N. If x C N (a), then | x G |=n; if x C N (b), then | x G |=m.

  3. (iii)

    A p-complement of N is a solvable quasi-Frobenius group with abelian kernel and complement.

  4. (iv)

    The conjugacy class sizes of a p-complement of N are exactly 1, | a N |, and | b N | p .

Proof Suppose that a p-complement of N/Z(N) is not a prime power order subgroup, and t, d are two non-central p-regular elements of N with | t G |m=| d G |.

  1. (i)

    Now, we show the G-conjugacy class size of the p-regular element of N is 1, n or m. Obviously, | t G |=n is to be dealt with. Since n is a p -number, by Theorem A, we find that n divides | t G |, forcing | t G |=n, as wanted.

  2. (ii)

    Let x C N (a) be a non-central p-regular element. If | x G |=m, then x ( C N ( a ) C N ( x ) ) p , in contradiction to (iii) in Theorem A. So | x G |=n. By Theorem A once again, the conclusion (ii) is obtained.

  3. (iii)

    Let C N ( a ) p C N ( b ) p =H, then H= M p C N ( b ) p is a p-complement of N. Now, C N ( a ) p (= M p ) and C N ( b ) p are abelian, we find that H is solvable. Taking into account ( C N ( a ) C N ( b ) ) p =Z ( N ) p by Theorem A, we find that Z(H)=Z ( N ) p . For convenience, we employ ‘_’ to work in the factor group modulo Z(H). Notice that | M p ¯ |=| C N ( a ) p ¯ | divides | b N | and | C N ( b ) p ¯ | divides | a N |, so we have (| C N ( b ) p ¯ |,| M p ¯ |)=1.

Now, M p ¯ is an abelian normal subgroup of H ¯ . To prove that H ¯ is a Frobenius group, we are left to show C M p ¯ ( x ¯ )=1 for any non-central element x C N ( b ) p . Otherwise, there exists a non-central element y M p such that 1 y ¯ C M p ¯ ( x ¯ ), then ( x y ¯ ) o ( x ¯ ) = ( x ¯ y ¯ ) o ( x ¯ ) = y ¯ o ( x ¯ ) 1 since (o( x ¯ ),o( y ¯ ))=1. Hence

x y ¯ o ( x ¯ ) = y ¯ o ( x ¯ ) C N ( x ) ¯ C N ( x y ) ¯ = C N ( x ) C N ( x y ) ¯ = C N ( x ) C N ( y ) ¯ ,

so C N (x) C N (y) contains a non-central p-regular element.

Notice that | x G |=m and | y G |=n; by Theorem A, we have ( C N ( x ) C N ( y ) ) p Z(G), in contradiction to the previous paragraph. So a p-complement of N is a solvable quasi-Frobenius group with the abelian kernel M p and complement C N ( b ) p .

  1. (iv)

    Let xH be a non-central element. We have | x G |=n or m by (i). If | x G |=n, then

    | x H |=|H: C H (x)|=| C N ( b ) p /Z ( N ) p |=| C N (b)/ C N (a) C N (b)|=| a N |.

If | x G |=m, then C N ( x ) p M p is also a p-complement of N. In view of the p-solvability of N, replacing H if necessary by a suitable conjugate, we may assume that C N ( x ) p H. Notice that C N ( x ) p M p ( C N ( x ) C N ( a ) ) p =Z ( N ) p , we have | C N ( x ) p |=| C N ( b ) p |. Also,

C N ( b ) p /Z ( N ) p = C N ( b ) p /Z(H)= C N ( b ) p ¯ ,

keeping in mind that (| C N ( b ) p ¯ |,| M p ¯ |)=1, by the solvability of H, we find that C N ( x ) p ¯ and C N ( b ) p ¯ are conjugate in H ¯ . Consequently, we may assume that there exists an element gN such that x g C N (b), so

| x H | = | x g H | = | H : C H ( x g ) | = | C N ( a ) p / Z ( N ) p | = | C N ( a ) / C N ( a ) C N ( b ) | p = | b N | p .

So the conjugacy class sizes of H are 1, | a N |, and | b N | p . □

Corollary 2 [[1], Theorem A]

Suppose that G is a p-solvable group. Let m>n>1 be the two longest sizes of the non-central p-regular conjugacy classes of G. Suppose that (m,n)=1 and p is not a prime divisor of n. Then G is solvable and

  1. (a)

    the p-regular conjugacy class sizes of G are exactly 1, n, and m;

  2. (b)

    a p-complement of G is a quasi-Frobenius group with abelian kernel and complement. Furthermore, its conjugacy class sizes are exactly 1, n, and m p .

Proof Obviously, n divides |G/Z(G)|. So, in Theorem B, by taking N=G, the proof of this corollary is finished. □