1. Introduction
In contrast to the finite case, a celebrated example by Heineken and Mohamed [
1] shows that there exist infinite groups with a trivial centre in which all subgroups are subnormal. On the other hand, a relevant result of Möhres [
2] states that groups with only subnormal subgroups are at least soluble, while Smith [
3] proved that torsion-free groups of this type are nilpotent. The structure of infinite groups which are somehow rich in subnormal subgroups has been investigated by several authors, with special emphasis on the imposition of chain conditions. In particular, groups satisfying the minimal or the maximal condition on non-subnormal subgroups were described in [
4,
5], respectively. A further step was the investigation of groups satisfying the so-called weak minimal and weak maximal condition on non-subnormal subgroups (see [
6,
7]). We consider here the effect of the imposition of a further very general chain restriction on the set of non-subnormal subgroups.
Let 
 be a partially ordered set and let 
 and 
 be non-empty subsets of 
. We say that 
 has 
-
deviation 0 if either 
 or the set 
 satisfies the minimal condition; moreover, if 
 > 0 is any ordinal, we use induction to say that 
 has 
-
deviation  if every descending chain
      
      of elements of 
 there exists a positive integer 
t such that the interval 
 has a 
-deviation strictly smaller than 
 for each 
, and 
 is the smallest ordinal with such a property. In particular, if the set 
 satisfies the 
weak minimal condition (i.e., if for any descending chain of elements of 
 there is a positive integer 
t such that 
 is finite for all 
), then 
 has 
-deviation at most 1.
Let G be a group and let sn(G) be the set of all subnormal subgroups of G. Then a non-empty set  of subgroups of G is said to have a G-subnormal deviation if it has a sn(G)-deviation in the partially ordered set  of all subgroups of G. In particular, if X is a subgroup of G and the set of all subgroups of X has a G-subnormal deviation, we just say that X has a G-subnormal deviation; the group G is said to have a subnormal deviation if it has a G-subnormal deviation, which means that the set  has sn(G)-deviation. Of course, if a group G has a subnormal deviation, then every set of subgroups of G has a G-subnormal deviation.
The aim of this paper is to give a further contribution to the theory of groups with many subnormal subgroups, characterizing periodic groups with a subnormal deviation, at least within the universe of soluble groups. The structure of locally soluble groups with subnormal deviation at most 1 was described in [
8,
9], where it was specifically proved that such groups are soluble. Our main result proves that if a periodic soluble group has a subnormal deviation, then only the extreme (and unavoidable) cases can occur.
Theorem 1. Let G be a periodic soluble group with a subnormal deviation. Then either G is Černikov or all its subgroups are subnormal. In particular, the subnormal deviation of G is 0.
 It follows from the above result that a soluble group with a positive subnormal deviation cannot be periodic; however, the study of soluble non-periodic groups with a subnormal deviation needs different methods and will be the subject of a forthcoming paper.
Recall, finally, that a subgroup 
X of a group 
G is said to be 
pronormal if 
X and 
 are conjugate in 
 for every element 
g of 
G. A group 
G is said to have a 
pronormal deviation if 
 has pn(
G)-
deviation, where pn(
G) is the set of all pronormal subgroups of 
G. The replacement of pn(
G) by the set of all normal subgroups of 
G gives the corresponding concept of normal deviation. Since a subgroup of an arbitrary group is normal if, and only if, it is subnormal and pronormal, the combination of our theorem with the results in [
10] on groups with a pronormal deviation gives the following statement.
Corollary 1. Let G be a periodic soluble group with a normal deviation. Then either G is Černikov or all its subgroups are normal. In particular, the normal deviation of G is 0.
 Our notation is mostly standard and can be found in [
11], and we refer to [
12] for results concerning subnormal subgroups.
  2. Proof of the Theorem
The proof of our main result will be accomplished in a series of lemmas. We recall that, if 
G is a group and 
X and 
Y are subnormal subgroups of 
G, then also 
 is subnormal in 
G (see [
12], Proposition 1.1.2), while 
 need not be subnormal in general; however, this is the case if 
 (see [
12], Theorem 1.2.5).
Lemma 1. Let G be a group and let  be subgroups of G such that Y is normal in X and  is the direct product of an infinite collection  of non-trivial subgroups. If the interval  has G-subnormal deviation, then every  is subnormal in G.
 Proof.  Let 
 be an arbitrary element of 
, and let 
 and 
 be infinite subsets of 
, such that 
 and 
. Choose in 
 an infinite descending chain of subsets
        
        such that 
 is infinite for all 
n, and write
        
        for each non-negative integer 
n. Similarly, let
        
        be an infinite descending chain of subsets of 
, such that 
 is infinite for all 
n, and put
        
        for each non-negative integer 
n. Then
        
        and
        
        are infinite descending chains of elements of 
.
Let  be the G-subnormal deviation of the interval  and assume first , so that  satisfies the minimal condition on subgroups which are not subnormal in G. Then there exist non-negative integers h and k, such that  and  are subnormal in G. It follows that  is likewise subnormal in G, and so the statement is proved when .
Suppose now 
. By definition there exists a non-negative integer 
r, such that the interval 
 has a 
G-subnormal deviation strictly smaller than 
 for each 
. Since
        
        is the direct product of an infinite collection of non-trivial subgroups, by induction on 
 we have that 
 is subnormal in 
G for each 
. The same argument shows that there exists a non-negative integer 
s, such that 
 is subnormal in 
G for all 
 in 
. Then also 
 is subnormal in 
G and the proof is complete.    □
 Lemma 2. Let G be a group with a subnormal deviation and let X be a subgroup of G. If X contains a normal subgroup Y, such that  is the direct product of infinitely many non-trivial subgroups, then X is subnormal in G.
 Proof.  Let
        
        where the set 
 is infinite and each 
 is a non-trivial subgroup of 
. Obviously, we may consider two infinite subsets 
 and 
 of 
, such that 
 and 
. Put
        
Application of Lemma 1 to the direct products
        
        yields that the subgroups 
V and 
W are subnormal in 
G. Since 
, it follows that also 
 is subnormal in 
G.    □
 Lemma 3. Let A be a periodic abelian group which is not Černikov. Then, A contains a subgroup B such that  is the direct product of infinitely many non-trivial subgroups.
 Proof.  Of course, it can be assumed that the set  is finite, so that there exists a prime number p, such that the p-component  of A is not Černikov. Write , where  is divisible and  is reduced. Since  is a direct factor of A, we may also suppose that  is the direct product of only finitely many Prüfer subgroups. Then  is infinite and so also  is infinite. It follows that  is the direct product of infinitely many non-trivial subgroups.    □
 Corollary 2. Let G be a group with a subnormal deviation and let X be a subgroup of G. If X contains a normal subgroup Y such that  is periodic abelian but not Černikov, then X is subnormal in G.
 Proof.  It follows from Lemma 3 that  contains a subgroup , such that  is the direct product of infinitely many non-trivial subgroups. Then X is subnormal in G by Lemma 2.    □
 Lemma 4. Let G be a group and let  be subgroups of G, such that  and Y is normal in ,  is the direct product of an infinite collection  of non-trivial Z-invariant subgroups and . If the interval  has a G-subnormal deviation, then the subgroup  is subnormal in G.
 Proof.  We can repeat the argument of the proof of Lemma 1 with a few slight modifications. Use the same notation of that proof, but take the infinite subsets 
 and 
 of 
 in such a way that 
 and 
. For each positive integer 
n, put
        
        and
        
The proof can now be completed as in Lemma 1, just replacing the interval  by the interval .    □
 The following easy result is well known.
Lemma 5. Let G be a group and let A be an abelian normal subgroup of G of prime exponent. If  is finite, then there exists a collection  of finite non-trivial G-invariant subgroups of A, such that .
 If G is any group, the Baer radical of G is the subgroup generated by all abelian subnormal subgroups of G, and G is called a Baer group if it coincides with its Baer radical. Thus, any Baer group is locally nilpotent and all its finitely generated subgroups are subnormal. Our next lemma shows that if a group G has a subnormal deviation and contains a periodic abelian subgroup which is not Černikov, then all the elements of finite order of G belong to the Baer radical.
Lemma 6. Let G be a group with a subnormal deviation. If G contains a periodic abelian subgroup A which is not Černikov, then all finite cyclic subgroups of G are subnormal.
 Proof.  The subgroup 
A is subnormal in 
G by Corollary 2 and so it is contained in the Baer radical 
B of 
G. Thus, the subgroup 
T of all elements of finite order of 
B is not Černikov. Let 
g be any element of finite order of 
G. Then 
 is a periodic locally soluble group and, hence, it follows from a result of Zaicev [
13] that it contains an abelian 
-invariant subgroup 
C which is not Černikov. Of course, 
C can be chosen either of prime exponent or with infinitely many non-trivial primary components, and so by Lemma 5 in both cases there exists an infinite collection 
 of non-trivial 
-invariant subgroups of 
C, such that
        
        and 
. We can now apply Lemma 4 for 
, 
, and 
 to obtain that the subgroup 
 is subnormal in 
G.    □
 It is well known that any divisible abelian normal subgroup of a periodic nilpotent group 
G lies in the centre 
 of 
G (see for instance [
11] Part 1, Lemma 3.13). For our purposes, we need the following slight generalization of this result.
Lemma 7. Let G be a periodic Baer group and let D be a a divisible abelian subnormal subgroup of G. Then .
 Proof.  Let 
g be any element of 
G. Since 
 is subnormal in 
G and the normal closure 
 is abelian (see [
11] Part 1, Lemma 4.46), the subgroup 
 is nilpotent. Thus, 
 and hence 
. Therefore, 
.    □
 The following result is crucial for our purposes.
Lemma 8. Let G be a group with a subnormal deviation, D a subgroup of G which is the direct product of finitely many Prüfer subgroups and A a periodic abelian D-invariant subgroup of G. If A is not Černikov, then D is subnormal in G.
 Proof.  Assume for a contradiction that the subgroup D is not subnormal in G. Then A cannot be the direct product of an infinite collection of non-trivial D-invariant subgroups, since otherwise D would be subnormal in G as an application of Lemma 4 for , , and . In particular, the set of primes  must be finite and, hence, there is a prime number p, such that the p-component  of A is not Černikov; clearly, the socle S of  is likewise infinite and so the replacement of A by S allows us to suppose that A is an elementary abelian p-group.
The subgroup  is a Baer group by Lemma 6. Assume that D is subnormal in , so that  by Lemma 7. Then  is abelian and, hence, it is subnormal in G by Corollary 2, which is clearly impossible because D is not subnormal in G. Therefore, D cannot be subnormal in , so that also  is a counterexample to the statement and, hence, we may suppose without loss of generality that . Then G is a Baer group and, in particular, it is locally nilpotent. The intersection  is a finite normal subgroup of G, and we may be replace G by the counterexample  and so assume . Moreover,  is not subnormal in G, while it follows from Lemma 4 that the centralizer  is finite, so that we may further replace G by  and assume .
Of course, 
D can be written as the union of an ascending chain
        
        of finite subgroups. If 
a is any non-trivial element of 
A, the finite subgroup 
 is nilpotent and, hence, 
. It follows that 
 is a non-trivial 
D-invariant subgroup for each 
n. Consider now the descending chain
        
        and note that
        
         In particular, the above chain cannot stop after finitely many steps and so each 
 is infinite.
Among all counterexamples choose the group 
G and its non-subnormal subgroup 
D in such a way that the set of all subgroups of 
G containing 
D has a minimal 
G-subnormal deviation 
. Suppose first 
. Then the set 
 satisfies the minimal condition on its members which are not subnormal in 
G and, hence, there is a positive integer 
m, such that 
 is subnormal in 
G for all 
. Application of Lemma 7 yields that 
 is contained in the centre of 
 for all 
, so that
        
        which is, of course, impossible. Therefore, 
 and there exists a positive integer 
t such that, for each 
, the interval 
 has a 
G-subnormal deviation strictly smaller than 
. It follows that 
 is subnormal in 
 for each 
. In fact, if 
 is finite, we have that 
 has finite index in 
, while if 
 is infinite the same conclusion follows from the minimal assumption on 
. Thus, 
 is subnormal in 
 for each positive integer 
k and hence 
 by Lemma 7. Thus,
        
        a final contradiction that completes the proof.    □
 Lemma 9. Let G be a group with a subnormal deviation and let N be a periodic soluble normal subgroup of G. If N is contained in the Baer radical of G, then all subgroups of N are subnormal in G.
 Proof.  Assume for a contradiction that the statement is false, and among all counterexamples for which N has smallest derived length choose one such that N contains a non-subnormal subgroup X whose derived length is minimal possible. Let A be the smallest non-trivial term of the derived series of N. It follows from the minimality of the derived length of N that  is subnormal in G, so X cannot be subnormal in . Moreover, the minimality of the derived length of X yields that the subgroup  is subnormal in G of, say, defect k. The intersection  is a normal subgroup of  and the factor group  is still a minimal counterexample, so that without loss of generality we may suppose  and ; in particular, G is a Baer group.
Put 
 and
        
        for each positive integer 
. Clearly every 
 is a normal subgroup of 
G and 
, so that
        
        is a chain running from 
G to 
X. It follows that there exists a positive integer 
, such that 
 is not subnormal in 
. Of course, the subgroup 
 is contained in 
 and normal in 
, whence the factor group 
 is likewise a minimal counterexample. Thus, we can replace 
G by 
 and 
X by 
, which allows us to suppose that 
X is abelian.
Since X is not subnormal in G, it follows from Corollary 2 that X is a Černikov group. Let J be the largest divisible subgroup of X and let E be a finite subgroup, such that . Clearly, E is subnormal in G, because G is a Baer group, so that J is not subnormal in G and, hence, not even in the normal subgroup  of G. It follows that  is not Černikov, so neither A is a Černikov group. However, in this case J is subnormal in G by Lemma 8, and this contradiction completes the proof.   □
 We can now prove our main result.
Proof of Theorem 1. Suppose that 
G is not a Černikov group, so that it contains an abelian subgroup 
A which is not Černikov (see [
11] Part 1, Theorem 3.45). Then, it follows from Lemma 6 that 
G is a Baer group and, hence, all its subgroups are subnormal by Lemma 9.    □