C O L L O Q U I U M M A T H E M A T I C U M
VOL. LXIV 1993 FASC. 2
A CERTAIN PROPERTY OF ABELIAN GROUPS
BY
WITOLD S E R E D Y ´ N S K I
ANDJACEK ´ S W I A ¸ T K O W S K I (WROC LAW)
1. Statement of the result
1.1. Definition. A subset A of an abelian group G is said to cover not less than 1/n of the group if there exist g
1, . . . , g
n∈ G such that S
ni=1
g
iA=G.
1.2. R e m a r k. g
n= e may be assumed in the above definition since G = g
−1nG = S
ni=1
g
−1ng
iA .
1.3. The Main Theorem. Let G be an abelian group and A
1, . . . , A
nits disjoint subsets, each covering not less than 1/n of the group. Then S
ni=1
A
i= G.
1.4. R e m a r k. For obvious arithmetical reasons the theorem holds for all finite (not necessarily abelian) groups. On the other hand, it is easy to construct a counterexample in a free group with two generators. The question arises about the class of groups for which the theorem holds, but we do not give any ideas for the answer in this paper.
2. Reduction of the main theorem to a special case. In the following only abelian groups will be considered and therefore we will use the additive notation.
2.1. Lemma. Let G be abelian and let sets A
icover not less than 1/n of it. Assume that A
i∪ S
n−1k=1
(g
ik+ A
i) = G for i = 1, . . . , n (compare 1.2).
If Theorem 1.3 holds for the subgroup H of G generated by all elements g
ikand for the sets A
0i= A
i∩ H then it holds for G.
P r o o f. The action of H on its cosets in G is the same as on itself, so the theorem holds for each coset, and consequently for the whole G.
2.2. R e m a r k. Let H be as in 2.1 and A
0i= A
i∩ H. Let Z
n(n−1)be the free abelian group with generators e
ik, i = 1, . . . , n, k = 1, . . . , n − 1, and let φ : Z
n(n−1)→ H be a homomorphism such that φ(e
ik) = g
ik. Put B
i= φ
−1(A
0i). Then B
i∪ S
n−1k=1
(e
ik+ B
i) = Z
n(n−1)for i = 1, . . . , n, thus each
156
W. S E R E D Y ´N S K I AND J. ´S W I A¸ T K O W S K IB
icovers not less than 1/n of Z
n(n−1). If we prove that S
ni=1
B
i= Z
n(n−1), then by surjectivity of φ we get S
ni=1
A
0i= H.
According to 2.1 and 2.2, in order to prove the main theorem, it is enough to prove it in the following special case.
2.3. Theorem (The special case of the main theorem). Let Z
n(n−1)be the free abelian group with generators e
ik, i = 1, . . . , n, k = 1, . . . , n − 1, and let B
ibe its disjoint subsets with B
i∪ S
n−1k=1
(e
ik+ B
i) = Z
n(n−1)for 1, . . . , n. Then S
ni=1
B
i= Z
n(n−1).
3. Proof of the special case (Theorem 2.3)
3.1. Let Z
in−1be the subgroup of Z
n(n−1)generated by e
i1, e
i2, . . . . . . , e
i,n−1. Then Z
n(n−1)= L
ni=1
Z
in−1(L denotes the free abelian prod- uct) and therefore we may identify Z
n(n−1)with the cartesian product Z
1n−1× Z
2n−1× . . . × Z
nn−1.
3.2. Definition. For a ∈ Z
n(n−1)let K
ai= {a, a − e
i1, a − e
i2, . . . , a − e
i,n−1}.
3.3. R e m a r k. Note that for each a ∈ Z
n(n−1)we can find in K
aian element which belongs to B
i. If it is not a then, since B
i∪ S
n−1k=1
(e
ik+ B
i) = Z
n(n−1), a ∈ e
ik+ B
ifor some k, and then a − e
ik∈ B
i.
3.4. Definition. Let a ∈ Z
n(n−1)and let a = a
1+ . . . + a
nbe its expansion with respect to the free abelian product of 3.1, i.e. a
i∈ Z
in−i. Then define K
a= K
a11× . . . × K
ann.
3.5. Lemma. For each i ∈ {1, . . . , n} and each a ∈ Z
n(n−1), K
ais a disjoint sum of n
n−1sets of the form K
xi.
P r o o f. Assume i = n. Then
K
a= (K
a11× . . . × K
an−1n−1) × K
ann= [
y∈Ka11 ×...×Kan−1n−1
{y} × K
ann,
which is a disjoint sum of sets of the desired form. For other i the proof is the same.
3.6. R e m a r k. According to 3.3 the sets K
aand B
ihave at least n
n−1elements in common. Since B
iare disjoint an arithmetical argument proves that K
a⊂ S
ni=1
B
i.
3.7. Lemma. Z
n(n−1)is a disjoint sum of sets of the form K
a.
P r o o f. It will be proved in the next section that Z
in−1is a disjoint sum of sets of the form K
xi, i.e. Z
n−1= S
x∈Ii
K
xifor each i and some I
i⊂ Z
in−1.
ABELIAN GROUPS
157
Thus
Z
n(n−1)= [
(x1,...,xn)∈I1×...×In
K
x11× . . . × K
xnn
and the lemma follows.
Theorem 2.3 now follows from 3.6 and 3.7.
4. Combinatorial fact used in the proof of Lemma 3.7
4.1. Proposition (used in 3.7). Let Z
n−1be a free abelian group with generators e
1, . . . , e
n−1. For x ∈ Z
n−1put K
x= {x, x − e
1. . . . , x − e
n−1}.
Then Z
n−1is a disjoint sum of sets of the form K
x.
Before proving the proposition we give some notations and lemmas.
4.2. Notation. Let Z
nbe the group of integers (mod n), i.e. a cyclic group of n elements. Denote by a
i, i ∈ {0, 1, . . . , n − 1}, the generators in (Z
n)
n−1of the form (1, 0, . . . , 0), (0, 1, 0, . . . , 0), . . . , (0, . . . , 0, 1). Put x
i= a
i+ ia
n−1for i = 1, . . . , n − 2, and denote by V the subgroup of (Z
n)
n−1generated by x
1, . . . , x
n−2.
4.3. Lemma. The elements x
1, . . . , x
n−2are free abelian generators of V over Z
n.
P r o o f. Suppose P
n−2i=1
k
ix
i= 0, 0 ≤ k
i< n. Then P
n−2i=1
k
i(a
i+ ia
n−1) = P
n−2i=1
k
ia
i+ P
n−2i=1
ik
ia
n−1= 0, which implies k
i= 0 for i = 1, . . . , n − 2 and the lemma follows.
4.4. R e m a r k. Lemma 4.3 implies that V consists of n
n−2elements.
4.5. Lemma. For 1 ≤ i ≤ n − 1, a
i6∈ V . P r o o f. Suppose a
i= P
n−2l=1
k
lx
l= P
n−2l=1
k
la
l+ P
n−2l=1
lk
la
n−1. If i 6= n − 1 then k
i= 1 and k
l= 0 for l 6= i, but then P
n−2l=1
lk
l= i 6= 0 (mod n), a contradiction.
If i = n − 1 then k
l= 0 for each l, which contradicts P
n−2l=1
lk
l= 1 (mod n).
4.6. Lemma. For 1 ≤ i, j ≤ n, i 6= j, a
i− a
j6∈ V .
P r o o f. C a s e 1: i 6= j, i 6= n − 1 and j 6= n − 1. Suppose a
i− a
j= P
n−2l=1
k
lx
l= P
n−2l=1
k
la
l+ P
n−2l=1
lk
la
n−1. Then k
i= 1, k
j= −1 (mod n), k
l= 0 for l 6= i, j and consequently P
n−2l=1
lk
l= i − j 6= 0 (mod n), a contradiction.
C a s e 2: i 6= j, i = n − 1. By the argument of the previous case we get k
j= −1 (mod n), k
l= 0 for l 6= j, and consequently P
n−2l=1
lk
l= −j (mod n), which contradicts P
n−2l=1
lk
l= 1 (mod n) since 1 ≤ j < n − 1.
This completes the proof.
158
W. S E R E D Y ´N S K I AND J. ´S W I A¸ T K O W S K I4.7. Lemma. For x ∈ (Z
n)
n−1define K
x= {x, x − a
1, . . . , x − a
n−1}.
Then {K
x: x ∈ V } is a disjoint family and covers (Z
n)
n−1.
P r o o f. Suppose that K
x1∩ K
x26= ∅ for some x
1, x
2∈ V , x
16= x
2. Then either x
1= x
2− a
i, x
2= x
1− a
ior x
1− a
i= x
2− a
jfor some i, j ∈ {1, . . . , n − 1}. The first two cases imply a
i∈ V and the third implies i 6= j (since x
16= x
2) and a
i−a
j∈ V , but none of these is possible according to 4.5 and 4.6. It follows from 4.4 that S{K
x: x ∈ V } consists of n
n−1elements and thus equals (Z
n)
n−1.
4.8. P r o o f o f 4.1. Let φ : Z
n−1→ (Z
n)
n−1be the canonical homo- morphism satisfying φ(e
i) = a
i, and V
0= φ
−1(V ). Then from 4.7 it follows that the family {K
x: x ∈ V
0} is a disjoint cover of Z
n−1, which completes the proof.
Witold Seredy´ nski Jacek ´ Swi¸ atkowski
INSTITUTE OF MATHEMATICS INSTITUTE OF MATHEMATICS
TECHNICAL UNIVERSITY OF WROC LAW WROC LAW UNIVERSITY
WYBRZE ˙ZE WYSPIA ´NSKIEGO 27 PL. GRUNWALDZKI 2/4
50-370 WROC LAW, POLAND 50-384 WROC LAW, POLAND