Zero divisors of support size 3 in group algebras and trinomials divided by irreducible polynomials over GF(2)
††footnotetext: 2010 Mathematics Subject Classification. 20C07; 20K15; 16S34.
Keywords and Phrases. Group algebra, torsion-free, zero divisor, support size, trinomial.
Alireza Abdollahi and Zahra Taheri
Abstract.
A famous conjecture about group algebras of torsion-free groups states that there is no zero divisor in such group algebras. A recent approach to settle the conjecture is to show the non-existence of zero divisors with respect to the length of possible ones, where by the length we mean the size of the support of an element of the group algebra. The case length 2 cannot be happen. The first unsettled case is the existence of zero divisors of length 3.
Here we study possible length 3 zero divisors in the rational group algebras and in the group algebras over the field Fp with p elements for some prime p.
As a consequence we prove that the rational group algebras of torsion-free groups which are residually finite p-group for some prime p=3 have no zero divisor of length 3. We note that the determination of all zero divisors of length 3 in group algebras over F2 of cyclic groups is equivalent to find all trinomials (polynomials with 3 non-zero terms) divided by irreducible polynomials over F2. The latter is a subject studied in coding theory and we add here some results, e.g. we show that 1+x+x2 is a zero divisor in the group algebra over F2 for some element x of the group if and only if x is of finite order divided by 3 and we find all β in the group algebra of the shortest length such that (1+x+x2)β=0; and 1+x2+x3 or 1+x+x3 is a zero divisor in the group algebra over F2 for some element x of the group if and only if x is of finite order divided by 7.
1. Introduction
Let F be a field and G a group. We say that a non-zero element α in the group algebra F[G] is a zero divisor if αβ=0 for some non-zero β∈F[G]. A famous conjecture about group algebras states that there is no zero divisor in F[G] whenever G is torsion-free (see [4, 5]). The support of an element α=∑g∈Gαgg∈F[G], denoted by supp(α) is the set {g∈G∣αg=0}. The length of an element of F[G] is defined as the size of its support.
A recent approach to settle the conjecture is to show the non-existence of zero divisors with respect to the length of possible ones (see [1, 2, 7, 10]). The case length 2 is not so hard, however non-trivial, and has been done (e.g. [7, Theorem 2.1]). The first unsettled case is the existence of zero divisors of length 3.
Here we study possible length 3 zero divisors in the rational group algebras. We first note that
Theorem 1.1**.**
Let G be a torsion-free group and α=α1h1+α2h2+α3h3 be a zero divisor with the support size 3 in Q[G]. Then either h1+h2+h3 or h1−h2+h3 is a zero divisor in Z[G].
We then prove that the rational group algebras of torsion-free groups which are residually finite p-groups for some prime p=3 have no zero divisor of length 3.
Theorem 1.2**.**
Let G be a residually finite p-group for some prime number p=3. Then Q[G] has no zero divisor whose support size is 3. Furthermore, if G is a residually finite 3-group then there exist no zero divisor of the form h1−h2+h3 in Q[G].
We note that the determination of all zero divisors of length 3 in group algebras over F2 of cyclic groups is equivalent to find all trinomials (polynomials with 3 non-zero terms) divided by irreducible polynomials over F2.
The latter is a subject studied in coding theory [3] and we add here some results.
Proposition 1.3**.**
Let G be a group, x∈G, n:=o(x)>2. Suppose that α∈F[G], 1∈supp(α)⊆⟨x⟩ is a zero divisor so that αβ=0 for some non-zero β∈F[G] where 1∈supp(β) and ∣supp(β)∣ is minimum with respect to the property αβ=0. Then, there exists at least one irreducible factor of Xn−1 in F[X] which divides α(X). Also, there exists at least one irreducible factor of Xn−1 in F[X] which divides β(X).
Theorem 1.4**.**
Let G be a group, x∈G, n:=o(x)>2, α∈{1+x+x−1,1+x+x2}⊂F2[G] and αβ=0 for some non-zero β∈F2[G] where 1∈supp(β) and ∣supp(β)∣ is minimum with respect to the property αβ=0. Then β=β′ or β=β′x−1 where β′:=∑i=0n−2xsi such that s0=0, s1=1 and si=si−2+3, for all i∈{2,3,…,k}.
Theorem 1.5**.**
Let G be a group, x∈G, n:=o(x)>2, α∈{1+x+x−1,1+x+x2}⊂F2[G] and αβ=0 for some non-zero β∈F2[G] where 1∈supp(β) and ∣supp(β)∣ is minimum with respect to the property αβ=0. Then n must be a multiple of 3 and ∣supp(β)∣=2n/3.
Theorem 1.6**.**
Let G be a group, x∈G, n:=o(x)>2, α∈{1+x+x3,1+x2+x3}⊂F2[G] be a zero divisor. Then n must be a multiple of 7.
2. Zero divisors with odd support sizes
Throughout this paper, let Fq denote the finite field of size q. The following result is well-known, we mention its proof for the reader’s convenience.
Proposition 2.1**.**
Let G be a finite p-group and F be a field of characteristic p. If α:=∑g∈Gαgg in F[G] is such that ∑g∈Gαg=0, then α is invertible. In particular, α is not a zero divisor of F[G].
Proof.
Let λ:=∑g∈Gαg. Then ι=λ⋅1G−α∈I(G), where I(G) is the augmentation ideal of F[G].
Since I(G) is nilpotent (see e.g. [9, Exercise 6 (b), page 226]), ιm=0 for some positive integer m and so (1-\lambda^{-1}\iota)\big{(}\sum_{i=0}^{m-1}(\lambda^{-1}\iota)^{i}\big{)}=1. Thus α=λ(1−λ−1ι) is invertible. This completes the proof.
∎
Theorem 2.2**.**
Let G be a residually finite 2-group. Then F2[G] has no zero divisor whose support size is odd.
Proof.
Suppose, for a contradiction, that αβ=0 for some non-zero α,β∈F2[G] such that ∣supp(α)∣ is odd. Let A:={x−1y∣x,y∈supp(α) or x,y∈supp(β) and x=y}. So, there exists a normal subgroup N of G such that A∩N=∅ and NG is a finite 2-group. It follows that αβ=0, where :F2[G]→F2[NG] is the natural ring epimorphism such that x=xN for all x∈G. Since A∩N=∅, ∣supp(β)∣=∣supp(β)∣ and ∣supp(α)∣=∣supp(α)∣. The latter contradicts Proposition 2.1. This completes the proof.
∎
Proof of Theorem 1.1.
We may assume α is a zero divisor in Z[G] and gcd(α1,α2,α3)=1. Suppose, for a contradiction, that p∣α1 but p∤α2 or p∤α3 for some prime p. So α′=α2h2+α3h3 is a zero divisor with the support size 1 or 2 in the group algebra of G over the finite field of size p, a contradiction, since for any field F, F[G] does not contain a zero divisor whose support is of size at most 2 (see [7, Theorem 2.1]). Therefore ∣α1∣=∣α2∣=∣α3∣. This completes the proof.
∎
Proof of Theorem 1.2.
If p=2 then by Theorem 2.2 and [7, Theorem 2.1 and Lemma 2.2], the statement is obviously true. Let G be a residually finite p-group for some prime number p=2 and α=α1h1+α2h2+α3h3 be a zero divisor with the support size 3 in Q[G]. Then by Theorem 1.1, there exists a zero divisor of the form h1+h2+h3 or h1−h2+h3 in Q[G]. So, there exists a zero divisor of the form h1+h2+h3 or h1−h2+h3 in Fp[G]. Therefore, if p>3 then there is a contradiction by Proposition 2.1 and so Q[G] has no zero divisor whose support size is 3. On the other hand, if p=3 then by Proposition 2.1, h1−h2+h3 can not be a zero divisor in Fp[G] and so there exist no zero divisor of the form h1−h2+h3 in Q[G]. This completes the proof.
∎
3. Zero divisors of the form 1+xi+xj in group algebras over F2
Let G be a group, x∈G, n:=o(x)>2. Suppose that α∈F2[G], 1∈supp(α)⊆⟨x⟩ is a zero divisor so that αβ=0 for some non-zero β∈F2[G]. We may assume that 1∈supp(β) and by [1, Lemma 2.5], if we choose β of minimum support size with respect to the property αβ=0, then supp(β)⊆⟨supp(α)⟩⊆⟨x⟩.
In this section, we want to study possible values of n with the following property: There exist distinct i,j∈{1,2,…,n−1} such that
α=1+xi+xj.
Let F[X] denote the polynomial ring in the indeterminate X. In the following we use the ring isomorphism F[⟨x⟩]≅⟨Xn−1⟩F[X], where ⟨Xn−1⟩ is the ideal generated by Xn−1 in F[X].
Actually the map x↦X can be extended to an epimorphism from F[⟨x⟩] to F[X] whose kernel is ⟨Xn−1⟩. We denote by α(X) the image of α∈F[⟨x⟩] under the latter ring epimorphism. Note that αβ=0 in F[⟨x⟩] is equivalent to α(X)β(X)∈⟨Xn−1⟩.
Lemma 3.1**.**
Let α(X)∈F[X] and α(X)∈⟨Xn−1⟩ for some positive integer n. Then, α(X)β(X)∈⟨Xn−1⟩ for some β(X)∈F[X] such that β(X)∈⟨Xn−1⟩ if and only if there exists at least one irreducible factor of Xn−1 in F[X] which divides α(X).
Proof.
Suppose, for a contradiction, that f(X)∤α(X) for each irreducible polynomial f(X)∈F[X] which is a factor of Xn−1. Since α(X)β(X)∈⟨Xn−1⟩, it follows that β(X)∈⟨Xn−1⟩, a contradiction.
Now suppose that α(X)=f(X)r(X) for some irreducible factor f(X) of Xn−1 and some r(X)∈F[X]. Assume that Xn−1=f(X)β(X) for some β(X)∈F[X]. It follows that β(X)∈⟨Xn−1⟩ and α(X)β(X)∈⟨Xn−1⟩. This completes the proof.
∎
Proof of Proposition 1.3.
It follows from Lemma 3.1.
∎
Definition 3.2**.**
If f∈Fq[X] is a polynomial such that f(0)=0, then the least positive integer t for which f divides Xt−1 is called the order of f.
Lemma 3.3**.**
[6, Corollary 3.4]**
If f∈Fq[X] is an irreducible polynomial over Fq of degree m, then the order of f divides qm−1.
Definition 3.4**.**
For any field F, a polynomial f in F[X] is called a trinomial whenever f has only three non-zero terms i.e., f=α1Xi+α2Xj+α3Xk, where α1,α2,α3 are non-zero and i,j,k are pairwise distinct non-negative integers.
Remark 3.5**.**
- (1)
If f∈Fq[X] is an irreducible polynomial such that f(0)=0, then the order of f is t if and only if f divides the cyclotomic polynomial Qt(X)=∏d∣t(1−Xt/d)μ(d), where μ(d) is the Möbius function. Also, any monic irreducible factor of Qt has the same degree [6].
2. (2)
Let t be an odd positive integer and f,g∈F2[X] be two distinct monic irreducible factors of Qt. Then f divides a trinomial if and only if g divides a trinomial. Furthermore, if an irreducible polynomial in F2[X] of order t divides a trinomial, then any irreducible polynomial in F2[X] of order lt divides a trinomial, for all odd positive integers l [3].
3. (3)
Let G be a group, x∈G, n:=o(x)>2, α=1+xi+xj∈F2[G] for some distinct i,j∈{1,2,…,n−1} and αβ=0 for some non-zero β∈F2[G], where 1∈supp(β) and ∣supp(β)∣ is minimum with respect to the property αβ=0. Then by Proposition 1.3, there exists at least one irreducible factor of Xn+1 in F2[X] which divides the trinomial α(X). If n=2kl for some odd integer l>1 and some positive integer k, then by Proposition 1.3, there exists at least one irreducible factor of Xl+1 in F2[X] which divides the trinomial α(X), since Xn+1=(Xl+1)2k in F2[X].
Theorem 3.6**.**
[3, Theorem 7 (Welch’s Criterion)]**
For any odd positive integer t, any irreducible polynomial in F2[X] of order t divides a trinomial if and only if gcd(1+Xt,1+(1+X)t) has degree greater than 1.
Corollary 3.7**.**
There exists a zero divisor α in the group algebra F2[Zn] such that ∣supp(α)∣=3 if and only if there exists an odd positive integer t such that t∣n and gcd(1+Xt,1+(1+X)t) has degree greater than 1 in the ring F2[X].
Remark 3.8**.**
By Corollary 3.7, if α∈F2[Zn], ∣supp(α)∣=3 and there exists an irreducible polynomial f∈F2[X] of order t, for some odd positive integer t∣n, which divides α(X), then α is a zero divisor in the group algebra F2[Zn].
**
Example 3.9**.**
Let α=x3+x+1 be an element of the group algebra F2[Z7], where Z7=⟨x⟩. It is easy to see that α(X)=X3+X+1 is an irreducible polynomial of order 7 in F2[X] which is already a trinomial. So, α is a zero divisor of F2[Z7]. Also, αβ=0 for β=(x3+x2+1)(x+1)∈F2[Z7] because α(X)β(X)=X7+1.
Example 3.10**.**
Let α=x16+x+1 be an element of the group algebra F2[Z85] where Z85=⟨x⟩. It can be seen that α(X)=X16+X+1=(X8+X6+X5+X3+1)(X8+X6+X5+X4+X3+X+1) and f(X)=X8+X6+X5+X4+X3+X+1 is an irreducible polynomial of order 85. So, α is a zero divisor of F2[Z85].
Remark 3.11**.**
- (1)
Let M be the set of all positive integers t such that any irreducible polynomial in F2[X] of order t divides a trinomial but no irreducible polynomial of order d divides a trinomial, for all d∣t such that d=t. It is easy to see that if t∈M then t is an odd positive integer greater than 1. Also we note that over any finite field, t∣n if and only if (Xt−1)∣(Xn−1), for all positive integers t and n.
2. (2)
Let G be a group, x∈G, n:=o(x)>2, α=1+xi+xj∈F[G], for some distinct i,j∈{1,2,…,n−1}, and αβ=0 for some non-zero β∈F2[G] where 1∈supp(β) and ∣supp(β)∣ is minimum with respect to the property αβ=0. Then by part (1) and Remark 3.5, the set of all possible values of n is the set of all odd multiples of elements of M.
In the following, some results about the elements of M which are obtained in [3] are given.
Theorem 3.12** (See [3]).**
Let M be the set of all positive integers t such that any irreducible polynomial in F2[X] of order t divides a trinomial but no irreducible polynomial of order d divides a trinomial, for all d∣t such that d=t. Then,
- (1)
All Mersenne primes i.e., prime numbers of the form 2m−1 where m∈N, are in M.
2. (2)
By computer search among non-Mersenne prime numbers smaller than 3000000, there exist only five elements 73,121369,178481,262657 and 599479 in M .
3. (3)
By computer search among non-prime numbers smaller than 1000000, there exist only ten elements 85,2047,3133,4369,11275,49981,60787,76627,140911 and 486737 in M.
4. (4)
Eight other larger non-prime elements of M which are currently known are
1826203,2304167, 2528921,8727391,14709241,15732721,16843009 and 23828017.
4. Zero divisors of the form 1+xi+xj∈F2[⟨x⟩] for some special values of (i,j)
Let G be a group, x∈G, n:=o(x)>2. Suppose that α∈F2[G], 1∈supp(α)⊆⟨x⟩ is a zero divisor so that αβ=0 for some non-zero β∈F2[G]. We may assume that 1∈supp(β) and by [1, Lemma 2.5], if we choose β of minimum support size with respect to the property αβ=0, then supp(β)⊆⟨supp(α)⟩⊆⟨x⟩.
In this section, we focus on the cases that (i,j)∈{(1,−1),(1,2)}∪{(1,3),(2,3),(2,n−1),(n−3,n−2),(1,n−2),(n−3,n−1)}. Firstly for the case (i,j)∈{(1,−1),(1,2)}, we show that n must be a multiple of 3 and β=β′ or β=β′x−1 where β′:=∑i=0n−2xsi such that s0=0, s1=1 and si=si−2+3, for all i∈{2,3,…,k}. Secondly for the case (i,j)∈{(1,3),(2,3),(2,n−1),(n−3,n−2),(1,n−2),(n−3,n−1)}, we show that n must be a multiple of 7.
Throughout this section, let A+:=A+1, A2+:=A+2 and A−:=A−1, where n>1 is a positive integer and A is a subset of Zn.
Lemma 4.1**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that t0=0, 1≤t1<t2<⋯<tk≤n−1 and the multiplicity of each element in A∪A+∪A2+ is 2. Then for all i∈{1,2,…,k}, either ti=ti−1+1 or ti=ti−1+2.
Proof.
Since t0=0 and 1≤t1<t2<⋯<tk≤n−1, the following inequalities are satisfied:
[TABLE]
Let i∈{1,2,…,k}. Since the multiplicity of each element in A∪A+∪A2+ is 2, either ti=tl+1(modn) or ti=tl+2(modn), for some l∈{0,1,…,k}.
(1) Let ti=tl+1(modn). If i≤l then 1≤ti≤tl≨tl+1≤n, that is a contradiction with ti=tl+1(modn). Therefore, i>l and so tl≨ti and tl≨n−1. Since ti≤n−1 and tl+1≤n−1, we have ti=tl+1. Therefore by 4.1, l=i−1 and ti=ti−1+1.
(2) Let ti=tl+2(modn). Suppose, for a contradiction, that i≤l. Then ti≤tl≨tl+2≤n+1. If tl+2≨n+1, then 1≤ti≨tl+2≤n, that is a contradiction with ti=tl+2(modn). Therefore, tl+2=n+1 and so ti=1, tl=n−1 and (0=)t0=tl+1(modn). Hence, ti=t0+1=tl+2(modn) i.e., the multiplicity of an element in A∪A+∪A2+ is greater than 2, a contradiction. Therefore, i>l and so tl≨ti and l<k. Since tl≨ti≤n−1, we have tl+2≤n. If tl+2=n, then ti=0(modn), that is a contradiction with 1≤ti≤n−1. Therefore, tl+2≤n−1 and so ti=tl+2 because 1≤ti≤n−1 and ti=tl+2(modn).
Since l<i we have l≤i−1. In the following we show that l=i−1. Suppose, for a contradiction, that l<i−2. By 4.1, ti−3≨ti−3+1≤ti−2≨ti−2+1≤ti−1. So, ti−3+2≤ti−1. On the other hand, l≤i−3 and so tl≤ti−3. Therefore, ti−1≨ti=tl+2≤ti−3+2, that is a contradiction with ti−3+2≤ti−1. Thus i−2≤l≤i−1. Now suppose, for a contradiction, that l=i−2. Then ti=ti−2+2. By 4.1, ti−2≨ti−2+1≤ti−1≨ti=ti−2+2 and so ti−1=ti−2+1. Therefore, ti−2+2=ti=ti−1+1 i.e., the multiplicity of an element in A∪A+∪A2+ is greater than 2, a contradiction. So, l=i−1 because i−2≨l≤i−1. Therefore, ti=ti−1+2.
∎
Lemma 4.2**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that the multiplicity of each element in A∪A+∪A2+ is 2. Then ti+1∈A or ti+2∈A, for all i∈{0,1,…,k}.
Proof.
Suppose, for a contradiction, that there are distinct l,s∈{0,1,…,k} such that tl=ti+1(modn) and ts=ti+2(modn). Then ts=tl+1=ti+2(modn) i.e., the multiplicity of an element in A∪A+∪A2+ is greater than 2, a contradiction. This completes the proof.
∎
Corollary 4.3**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that t0=0, 1≤t1<t2<⋯<tk≤n−1 and the multiplicity of each element in A∪A+∪A2+ is 2. Then either ti=ti−1+1 and ti−1+2∈A or ti=ti−1+2 and ti−1+1∈A, for all i∈{1,2,…,k}.
Proof.
Lemmas 4.1 and 4.2 complete the proof.
∎
Lemma 4.4**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that t0=0, 1≤t1<t2<⋯<tk≤n−1 and the multiplicity of each element in A∪A+∪A2+ is 2. If ti−1∈A for some i∈{1,2,…,k}, then ti+1(modn)∈A. Furthermore if i<k, then ti+1=ti+1.
Proof.
Suppose that ti−1∈A for some i∈{1,2,…,k}. Then ti−1≨ti−1 because ti−1+1≤ti. Therefore, ti>ti−1+1 and so by Lemma 4.1, ti=ti−1+2. Since the multiplicity of each element in A∪A+∪A2+ is 2, there must be an l∈{0,1,…,k}∖{i} such that either ti+1=tl+2(modn) or ti+1=tl(modn). If ti+1=tl+2(modn), then ti−1+2=ti=tl+1(modn) i.e., the multiplicity of an element in A∪A+∪A2+ is greater than 2, a contradiction. So, tl=ti+1(modn) for some l∈{0,1,…,k}∖{i}. Now if i<k, then
[TABLE]
By 4.2, ti+1=ti+1.
∎
Lemma 4.5**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that t0=0, 1≤t1<t2<⋯<tk≤n−1 and the multiplicity of each element in A∪A+∪A2+ is 2. If 0∈A+ then tk=n−1 and 1∈A. Also if 0∈A2+ then tk=n−2 and 1∈A.
Proof.
Since the multiplicity of each element in A∪A+∪A2+ is 2 and (0=)t0∈A, either 0=tl+1(modn) or 0=tl+2(modn), for some l∈{1,2,…,k}.
(1) Let 0=tl+1(modn) i.e., 0∈A+. Therefore because 1≤tl≤n−1, tl+1=n and so tl=tk=n−1 and 1=tk+2(modn). If 1∈A, then t1=1 and t1=t0+1=tk+2(modn) i.e., the multiplicity of an element in A∪A+∪A2+ is greater than 2, a contradiction. So, 1∈A.
(2) Let 0=tl+2(modn) i.e., 0∈A2+. Therefore because 1≤tl≤n−1, we have tl+2=n and so tl=n−2. So, l=k−1 or l=k because 1≤tl≤tk≤n−1. If l=k−1 then tk=n−1, tk−1=n−2 and t0=tk−1+2=tk+1(modn) i.e., the multiplicity of an element in A∪A+∪A2+ is greater than 2, a contradiction. Therefore, l=k and tk=n−2. If 1∈A, then there exist no i∈{1,2,…,k} such that t0+1=ti(modn) or t0+1=ti+2(modn) because 2≤ti≤n−2 and 4≤ti+2≤n i.e., the multiplicity of t0+1∈A∪A+∪A2+ is 1, a contradiction. So, 1∈A.
∎
Corollary 4.6**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that t0=0, 1≤t1<t2<⋯<tk≤n−1 and the multiplicity of each element in A∪A+∪A2+ is 2. If 1∈A then tk=n−2 and otherwise tk=n−1.
Proof.
If 1∈A, then by Lemma 4.5, 0∈A2+ and tk=n−2. If 1∈A, then by Lemma 4.5, 0∈A+ and tk=n−1.
∎
Theorem 4.7**.**
Let A:={t0,t1,…,tk} be a subset of Zn such that t0=0, 1≤t1<t2<⋯<tk≤n−1 and the multiplicity of each element in A∪A+∪A2+ is 2. Then for all i∈{2,3,…,k}, ti=ti−2+3 and one of the following cases is satisfied:
- (1)
If 1∈A, then t1=1, tk=n−2 and we have
[TABLE]
2. (2)
If 1∈A, then t1=2, tk=n−1 and we have
[TABLE]
Proof.
(1) Let 1∈A. By 1≤t1<t2<⋯<tk≤n−1, we have t1=1. Also by Corollary 4.6, tk=n−2. We argue by induction on i. For i=2, it follows from t1=t0+1 and Lemma 4.2 that 2=t0+2=t1+1∈A. So by Lemma 4.1, t2=t1+2=3. Now assume inductively that for i<k, the statement is true.
- (a)
If i is odd, then ti=ti−2+3=ti−1+1 and so by Corollary 4.3, ti+1=ti−1+2∈A. Therefore by Lemma 4.1, ti+1=ti+2=ti−1+3. So, i+1 is even and ti+1=t(i+1)−1+2=t(i+1)−2+3.
2. (b)
If i is even, then ti=ti−2+3=ti−1+2 and so by Corollary 4.3, ti−1=ti−1+1∈A. Therefore by Lemma 4.4, ti+1=ti+1=ti−1+3. So, i+1 is odd and ti+1=t(i+1)−1+1=t(i+1)−2+3.
(2) Let 1∈A. By Lemma 4.1, t1=t0+2=2 because 1=t0+1∈A. Also by Corollary 4.6, tk=n−1. We argue by induction on i. For i=2, it follows from Lemma 4.4 that t2=t1+1=3 because 1=t1−1∈A. Now assume inductively that for i<k, the statement is true.
- (a)
If i is even, then ti=ti−2+3=ti−1+1 and so by Corollary 4.3, ti+1=ti−1+2∈A. Therefore by Lemma 4.1, ti+1=ti+2=ti−1+3. So, i+1 is odd and ti+1=t(i+1)−1+2=t(i+1)−2+3.
2. (b)
If i is odd, then ti=ti−2+3=ti−1+2 and so by Corollary 4.3, ti−1=ti−1+1∈A. Therefore by Lemma 4.4, ti+1=ti+1=ti−1+3. So, i+1 is even and ti+1=t(i+1)−1+1=t(i+1)−2+3.
This completes the proof.
∎
Proof of Theorem 1.4.
Note that n:=o(x) is finite because the group algebra of an infinite cyclic group has no zero-divisors (see [8, Theorem 26.2]). By [1, Lemma 2.5], if we choose β of minimum support size with respect to the property αβ=0, then supp(β)⊆⟨supp(α)⟩⊆⟨x⟩. Let β=∑i=0kxti such that t0=0 and 1≤t1<t2<⋯<tk≤n−1. If α=1+x+x−1 and α′=1+x+x2 then α′=xα. Also αγ=0 if and only if α′γ=0, for some γ∈F2[G]. Let A:={t0,t1,…,tk} be a subset of Zn. Since (1+x+x2)β=0, the multiplicity of each element in A∪A+∪A2+ is 2. So by Lemma 4.7, ti=ti−2+3 for all i∈{2,3,…,k}. Also, if 1∈A then t1=1 and tk=n−2, and if 1∈A then t1=2 and tk=n−1. Therefore, if x∈supp(β) then β=β′ and otherwise β=β′x−1 because 1=x0=xn. This completes the proof.
∎
Lemma 4.8**.**
Let A be a subset of Zn such that the multiplicity of each element in A∪A+∪A− is at least 2. Then Zn=A∪A+=A−∪A=A+∪A− and so ∣A∩A−∣=∣A∩A+∣=∣A+∩A−∣. Furthermore, if the multiplicity of each element in A∪A+∪A− is exactly 2, then n must be a multiple of 3, ∣A∣=2n/3 and ∣A∩A−∣=∣A∩A+∣=∣A+∩A−∣=n/3=∣A∣/2.
Proof.
By the hypothesis on the multiplicities, we have the following three inclusions:
A⊆A+∪A−, A+⊆A−∪A and A−⊆A∪A+.
For the first part, by induction on i, we prove that A+i⊆A∪A+ for all i∈N. It is clear for i=0,1. For i=2, it follows from the second inclusion that A2+⊆A∪A+. Now assume inductively that A+i⊆A∪A+. It follows that
A+(i+1)⊆A+∪A2+⊆A+∪(A∪A+)=A∪A+.
This completes the induction. Now as Zn=⋃i=1nA+i, Zn=A∪A+ and the above first three inclusions imply that Zn=A∪A−=A+∪A−. Therefore n=2∣A∣−∣A∩A−∣=2∣A∣−∣A∩A+∣=2∣A∣−∣A+∩A−∣ and so
t:=∣A∩A−∣=∣A∩A+∣=∣A+∩A−∣.
For the second part, suppose that the multiplicity of each element in A∪A+∪A− is exactly 2. Thus, A∩A+∩A−=∅. Therefore n=3∣A∣−3t=3(∣A∣−t) and so n is a multiple of 3. Also we have n=3∣A∣−3t=2∣A∣−t. Thus t=∣A∣/2 and so ∣A∣=2n/3. This completes the proof.
∎
Proof of Theorem 1.5.
Note that n:=o(x) is finite because the group algebra of an infinite cyclic group has no zero-divisors (see [8, Theorem 26.2]). By [1, Lemma 2.5], if we choose β of minimum support size with respect to the property αβ=0, then supp(β)⊆⟨supp(α)⟩⊆⟨x⟩. Let β=∑i=0kxti such that t0=0 and 1≤t1<t2<⋯<tk≤n−1. If α=1+x+x−1 and α′=1+x+x2 then α′=xα. Also αγ=0 if and only if α′γ=0, for some γ∈F2[G]. Let A:={t0,t1,…,tk} be a subset of Zn. Since (1+x+x−1)β=0, the multiplicity of each element in A∪A+∪A− is 2. So by Lemma 4.8, n must be a multiple of 3. Also, ∣supp(β)∣=∣A∣=2n/3.
∎
Proof of Theorem 1.6.
Note that n:=o(x) is finite because the group algebra of an infinite cyclic group has no zero-divisors (see [8, Theorem 26.2]). By [1, Lemma 2.5], if we choose β of minimum support size with respect to the property αβ=0, then supp(β)⊆⟨supp(α)⟩⊆⟨x⟩. For all γ=∑k=0n−1akxk in the group algebra F2[⟨x⟩], a polynomial γ(x):=∑k=0n−1akxk from the ring F2[x] can be corresponded to γ. So, α(x),β(x)∈⟨xn−1⟩ and α(x)β(x)∈⟨xn−1⟩. Let α=1+x+x3. So, α(x)=1+x+x3 is an irreducible polynomial in F2[⟨x⟩] and α(x)∣xn−1, because α(x)β(x)∈⟨xn−1⟩ and α(x),β(x)∈⟨xn−1⟩. Therefore, the reciprocal polynomial of α(x) i.e., α∗(x)=xdeg(α(x))α(x1)=1+x2+x3 divides xn−1, too. Also, it is easy to see that α(x) and α∗(x) do not divide x−1. So, x7−1=α(x)α∗(x)(x−1) divides xn−1. Furthermore, it is easy to see that xa−1∣xb−1 if and only if a∣b. Therefore, n must be a multiple of 7 because x7−1 divides xn−1. The same discussion for the case that α=1+x2+x3 completes the proof because the reciprocal polynomial of α(x) is α∗(x)=1+x+x3.
∎
Proposition 4.9**.**
Let G be a group, x∈G, n:=o(x)>2, α∈{1+x2+xn−1,1+xn−3+xn−2,1+x+xn−2,1+xn−3+xn−1}⊂F2[G] be a zero divisor. Then n must be a multiple of 7.
Proof.
Note that n:=o(x) is finite because the group algebra of an infinite cyclic group has no zero-divisors (see [8, Theorem 26.2]). By [1, Lemma 2.5], if we choose β of minimum support size with respect to the property αβ=0, then supp(β)⊆⟨supp(α)⟩⊆⟨x⟩. If α is equal to 1+x2+xn−1, 1+xn−3+xn−2, 1+x+xn−2 or 1+xn−3+xn−1, then α′β=0 where α′ is equal to xα=1+x+x3, x3α=1+x+x3, x2α=1+x2+x3 or x3α=1+x2+x3, respectively. So by Theorem 1.6, n must be a multiple of 7. This completes the proof.
∎