4.8 Proofs of the results of Section 2.8
4.8.2 Proofs for Section 2.8.2
We need the next general result regarding the subgroups of the group GˆZq, where pG,`q is an arbitrary finite Abelian group. For a subgroup H of G (notation H ď G) consider the congruence relation%H on Gdefined for x, x1 PG byx%Hx1 if x´x1 PH.
Lemma 4.8.1. For a finite Abelian group pG,`q and q PN let
IG,q :“ tpH, α, dq:H ďG, αPSH, d|q and pq{dqαPHu,
where SH is a complete system of representants of the equivalence classes determined by
%H. For pH, α, dq PIG,q define
VH,α,d :“ tpkα`β, kdq: 0ďk ďq{d´1, β PHu.
Then VH,α,d is a subgroup of order pq{dq#H of GˆZq and the map pH, α, dq ÞÑ VH,α,d is a bijection between the set IG,q and the set of subgroups of GˆZq.
Proof. LetV be a subgroup ofGˆZq. Consider the natural projectionπ2 :GˆZqÑZq
given by π2px, yq “y. Thenπ2pVqis a subgroup of Zq and there is a unique divisor d of q such that π2pVq “ xdy :“ tkd: 0ďk ďq{d´1u. Let αP Gsuch that pα, dq PV.
Furthermore, consider the natural inclusion ι1 :G ÑGˆZq given by ι1pxq “ px,0q. Then ι´11 pVq “H is a subgroup of G. We show that V “ tpkα`β, kdq:k PZ, β PHu.
Indeed, for every k P Z and β P H, pkα`β, kdq “ kpα, dq ` pβ,0q P V. On the other hand, for every pu, vq PV one has v P π2pVq and hence there is k P Z such that v “kd.
We obtainpu´kα,0q “ pu, vq ´kpα, dq PV, thus β :“u´kαPι´11 pVq “ H.
Here a necessary condition is thatpq{dqα PH (obtained fork “q{d, β“0). Clearly, if this is verified, then for the above representation of V it is enough to take the values 0ďk ďq{d´1.
Conversely, every pH, α, dq P IG,q generates a subgroup VH,α,d of order pq{dq#H of GˆZq. Furthermore, for fixed H ďG and d | q we have VH,α,d “HH,α1,d if and only if α%Hα1. This completes the proof.
In the caseG“Zm (and with q“n) Lemma 4.8.1 can be stated as follows:
Lemma 4.8.2. For every m, nP N let
Im,n :“ tpa, b, sq PN2ˆNY t0u:a |m, b|n,0ďsďa´1 and a| pn{bqsu and for pa, b, sq PIm,n define
Va,b,s :“ xpa,0q,pb, sqy (4.32)
“ tpia`js, jbq: 0ďiďm{a´1,0ďj ďn{b´1u.
Then Va,b,s is a subgroup of order mnab of Zm ˆZn and the map pa, b, sq ÞÑ Va,b,s is a bijection between the set Im,n and the set of subgroups of ZmˆZn.
Note thata| pn{bqsholds if and only ifa{gcdpa, n{bq |s. That is, for sPIm,n we have s“ at
A, 0ďtďA´1, (4.33)
where A “ gcdpa, n{bq, notation given in Theorem 2.8.5. This quickly leads to formula (2.44) regarding the number spm, nq of subgroups of ZmˆZn, namely
spm, nq “ ÿ
a|m,b|n
ÿ
0ďtďA´1
1“ ÿ
a|m,b|n
gcdpa, bq.
deduced by different arguments.
Proof of Theorem 2.8.5. Apply Lemma 4.8.1 for G “ Zm ˆZn and with q “ r. For the subgroups V “ Va,b,s given by Lemma 4.8.2 a complete system of representants of the equivalence classes determined by %V is Sa,b “ t0,1, . . . , a´1u ˆ t0,1, . . . , b´1u.
Indeed, the elements of Sa,b are pairwise incongruent with respect to V, and for every px, yq P ZmˆZn there is a unique px1, y1q P Sa,b such that px, yq ´ px1, y1q PV. Namely, let
px1, y1q “ px, yq ´ty{bups, bq, px1, y1q “ px1, y1q ´tx1{aupa,0q. We obtain that the subgroups of ZmˆZnˆZr are of the form
U “UH,α,c “ tpkα`β, kcq: 0ďk ďr{c´1, β PVu, wherec|r and α“ pu, vq PSa,b such that pr{cqαPV.
Now using (4.32) we deduce
U “Ua,b,s,α,c
“ tpia`js`ku, jb`kv, kcq: 0ďiďn{a´1,0ďj ďn{b´1,0ďk ďn{c´1u, where (4.33) holds and pr{cqpu, vq P V. From the latter condition we deduce that there are i0, j0 such that
pr{cqu“i0a`j0s, pr{cqv “j0b. (4.34)
The second condition of (4.34) holds if b| pr{cqv, that is b{gcdpb, r{cq |v. Let v “ bv1
B , 0ďv1 ďB´1, (4.35)
where B “ gcdpb, r{cq. Also, j0 “ rv{pbcq and inserting this into the first equation of (4.34) we obtain pr{cqu ” rvs{pbcq (mod a). This linear congruence in u has a solution u0 if and only if
gcdpa, r{cq | rvs
bc (4.36)
and its all solutions areu“u0`az{C with 0ďz ďC´1 with C “gcdpa, r{cq.
Substituting (4.35) and (4.33) into (4.36) we obtain gcdpa, r{cq | rab
gcdpab, nqgcdpbc, rqv1t, that is
gcdpab, nqgcdpac, rqgcdpbc, rq |abcrv1t, equivalent to
gcdpab, nqgcdpac, rqgcdpbc, rq
gcdpabcr,gcdpab, nqgcdpac, rqgcdpbc, rqq |v1t, and to
X |v1t, (4.37)
where X is defined in the statement of Theorem 2.8.5. Note that X | B (indeed, A | a, C| pr{cqand the property follows from X “B{gcdppa{Aqpr{cq{Cq, B).
Lett be fixed. We obtain from (4.37) thatv1 is of the formv1 “Xw{gcdpt, Xq, where 0 ď w ď Bgcdpt, Xq{X ´1. Also, from (4.35), v “ bXw{Bgcdpt, Xq. Collecting the conditions ona, b, c, t, w, z in terms ofA, B, C, X finishes the proof.
Proof of Theorem 2.8.6. According to Theorem 2.8.5, the number of subgroups of Γ is spm, n, rq “ ÿ
a|m,b|n,c|r
ÿ
0ďtďA´1
ÿ
0ďwďBgcdpt,Xq{X´1
ÿ
0ďzďC´1
1
“ ÿ
a|m,b|n,c|r
C ÿ
0ďtďA´1
B
X gcdpt, Xq “ ÿ
a|m,b|n,c|r
BC X
ÿ
1ďtďA
gcdpt, Xq.
Here X |A (similar to X |B shown above), hence the inner sum is pA{XqPpXq and we obtain formula (2.47).
Proof of Theorem 2.8.7. Letspnq “spn, n, nq. According to (2.48), the function spnq can
gcdpapr{cq, ABCqPpXq{X ď ÿ
a,b,c|n From the Euler product formula
Hpzq “ź
Furthermore, by partial summation we obtain from (1.1) that ÿ and inserting (4.41) we get
ÿ
. This gives the asymptotic formula (2.49).
Chapter 5
Proofs of the results of Chapter 3
5.1 Proofs of the results of Section 3.1
Proof of Theorem 3.1.1. We use the polynomial identity (proved in our paper [111])
k´1 right hand side shows how it can be written as a polynomial of the elementary symmetric polynomials.
“ζps1q ¨ ¨ ¨ζpsrq
The function ψr,k is also multiplicative, symmetric in the variables and for any prime powers pν1, . . . , pνr,
From (5.2) we deduce ÿ
where the first sum is overu1, . . . , ur P t0,1usuch that at least oneui is 0. Let u1, . . . , ur be fixed and assume that ur “0. Since px{diqui ďx{di for every i, we have
A:“xu1`¨¨¨`ur ÿ
d1,...,drďx
|ψr,kpd1, . . . , drq|
du11¨ ¨ ¨durr ďxr´1 ÿ
d1,...,drďx
|ψr,kpd1, . . . , drq|
d1¨ ¨ ¨dr´1 Assume that k ě3. Then
Aďxr´1
8
ÿ
d1,...,dr“1
|ψr,kpd1, . . . , drq|
d1¨ ¨ ¨dr´1 !xr´1,
since the series Dr,kp1, . . . ,1,0q is absolutely convergent for k ě3 by Theorem 3.1.1. We obtain that
Qr,kpxq !xr´1 pk ě3q. (5.5)
Ifk “2, then
Aďxr´1ź
pďx 8
ÿ
ν1,...,νr“0
|ψr,2ppν1, . . . , pνrq|
pν1`¨¨¨`νr´1 (5.6)
“xr´1ź
pďx
ˆ
1`r´1 p ` c2
p2 ` ¨ ¨ ¨ ` cr´1 pr´1
˙ ,
by (5.3), where c2, . . . , cr´1 are certain positive integers, using also that we have p in the denominator if and only if νr “ 1 and exactly one of ν1, . . . , νr´1 is 1, the rest being 0, which occursr´1 times. We deduce that
A!xr´1ź
pďx
ˆ 1`1
p
˙r´1
!xr´1plogxqr´1 by Mertens’ formula. This shows that
Qr,2pxq !xr´1plogxqr´1. (5.7) Furthermore, for the main term of (5.4) we have
ÿ
d1,...,drďx
ψr,kpd1, . . . , drq d1¨ ¨ ¨dr
“
8
ÿ
d1,...,dr“1
ψr,kpd1, . . . , drq d1¨ ¨ ¨dr ´
ÿ
H‰IĎt1,...,ru
ÿ
diąx, iPI djďx, jRI
ψr,kpd1, . . . , drq
d1¨ ¨ ¨dr , (5.8) where the series is convergent by Theorem 3.1.1 and its sum isDr,kp1, . . . ,1q “Ar,k, given by (3.4).
LetI be fixed and assume thatI “ t1,2, . . . , tu, that isd1, . . . , dtąxanddt`1, . . . , dr ď x, wheretě1. We estimate the sum
B :“ ÿ
d1,...,dtąx dt`1,...,drďx
|ψr,kpd1, . . . , drq|
d1¨ ¨ ¨dr by distinguishing the following cases:
Case i) kě3,t ě1:
B ă 1 x
8
ÿ
d1,...,dr“1
|ψr,kpd1, . . . , drq|
d2¨ ¨ ¨dr
! 1 x, since the series is convergent by Theorem 3.1.1.
Case ii)k “2,t ě3: if 0 ăεă1{2, then B “
ÿ
d1,...,dtąx dt`1,...,drďx
|ψr,2pd1, . . . , drq|dε´1{21 ¨ ¨ ¨dε´1{2t d1{2`ε1 ¨ ¨ ¨d1{2`εt dt`1¨ ¨ ¨dr
ăxtpε´1{2q
8
ÿ
d1,...,dr“1
|ψr,2pd1, . . . , drq|
d1{2`ε1 ¨ ¨ ¨d1{2`εt dt`1¨ ¨ ¨dr !xtpε´1{2q,
since the series is convergent (fortě1). Using thattpε´1{2q ă ´1 for 0ăεă pt´2q{p2tq, here we needt ě3, we obtain B ! 1x.
Case iii) k “ 2, t “ 1: Let d1 ą x, d2, . . . , dr ď x and consider a prime p. If p | di
for an i P t2, . . . , ru, then p ď x. If p | d1 and p ą x, then p - di for every i P t2, . . . , ru and ψr,2pd1, . . . , drq “0 by its definition (5.3). Hence it is enough to consider the primes pďx. We deduce
B ă 1 x
ÿ
d1ąx d2,...,drďx
|ψr,2pd1, . . . , drq|
d2¨ ¨ ¨dr
ď 1 x
ź
pďx 8
ÿ
ν1,...,νr“0
|ψr,2ppν1, . . . , pνrq|
pν2`¨¨¨`νr ! 1
xplogxqr´1, similar to the estimate of (5.6).
Case iv) k “2, t“2: We split the sum B into two sums, namely
B “ ÿ
d1ąx,d2ąx d3,...,drďx
|ψr,2pd1, . . . , drq|
d1¨ ¨ ¨dr
“ ÿ
d1ąx3{2,d2ąx d3,...,drďx
|ψr,2pd1, . . . , drq|
d1¨ ¨ ¨dr ` ÿ
x3{2ěd1ąx,d2ąx d3,...,drďx
|ψr,2pd1, . . . , drq|
d1¨ ¨ ¨dr “:B1`B2, say, where
B1 “
ÿ
d1ąx3{2,d2ąx d3,...,drďx
|ψr,2pd1, . . . , drq|
d1{31 d2¨ ¨ ¨dr
1 d2{31
ă 1 x
8
ÿ
d1,...,dr“1
|ψr,2pd1, . . . , drq|
d1{31 d2¨ ¨ ¨dr ! 1 x, since the series is convergent. Furthermore,
B2 ă 1 x
ÿ
x3{2ěd1,d2ąx d3,...,drďx
|ψr,2pd1, . . . , drq|
d1d3¨ ¨ ¨dr ,
where d1 ď x3{2, d2 ą x, d3, . . . , dr ď x. Consider a prime p. If p | di for an i P t1,3, . . . , ru, then pďx3{2. Ifp|d2 and pąx3{2, thenp-di for everyiP t1,3, . . . , ruand ψr,2pd1, . . . , drq “ 0 by its definition. Hence it is enough to consider the primes pďx3{2. We deduce, cf. the estimate of (5.6),
B2 ă 1 x
ź
pďx3{2 8
ÿ
ν1,...,νr“0
|ψr,2ppν1, . . . , pνrq|
pν1`ν3`¨¨¨`νr ! 1
xplogx3{2qr´1 ! 1
xplogxqr´1.
Hence, given any t ě 1, we have B ! 1x for k ě 3 and B ! 1xplogxqr´1 for k “ 2.
Therefore, by (5.8),
ÿ
d1,...,drďx
ψr,kpd1, . . . , drq
d1¨ ¨ ¨dr “Ar,k`OpRr,kpxqq (5.9) with the notation (3.4) and (3.5).
The proof is complete by putting together (5.4), (5.5), (5.7) and (5.9).