• Aucun résultat trouvé

On the continued fraction expansions of (1

N/A
N/A
Protected

Academic year: 2023

Partager "On the continued fraction expansions of (1"

Copied!
6
0
0

Texte intégral

(1)

Comptes Rendus

Mathématique

Stéphane R. Louboutin

On the continued fraction expansions of (1

+pp q

)/2 and

pp q

Volume 359, issue 9 (2021), p. 1201-1205 Published online: 25 November 2021 https://doi.org/10.5802/crmath.266

This article is licensed under the

Creative Commons Attribution4.0International License.

http://creativecommons.org/licenses/by/4.0/

LesComptes Rendus. Mathématiquesont membres du Centre Mersenne pour l’édition scientifique ouverte

www.centre-mersenne.org e-ISSN : 1778-3569

(2)

2021, 359, n9, p. 1201-1205

https://doi.org/10.5802/crmath.266

Number theory /Théorie des nombres

On the continued fraction expansions of (1 + p

p q )/2 and p p q

Stéphane R. Louboutina

aAix Marseille Université, CNRS, Centrale Marseille, I2M, Marseille, France E-mail:[email protected]

Abstract.The evenness and the values modulo 4 of the lengths of the periods of the continued fraction expansions ofpp andp

2pforp3 (mod 4) a prime are known. Here we prove similar results for the continued fraction expansion ofpp q, wherep,q3 (mod 4) are distinct primes.

Mathematical subject classification (2010).11A55, 11R11.

Manuscript received 21st June 2021, accepted 27th August 2021.

1. Introduction

Letαbe a real quadratic irrational number. Its continued fraction expansionα=[a0,a1,a2, . . . ] is periodic, i.e. there existsk ≥0 andl ≥1 such thatai+l =ai forik. In that case we write α=[a0, . . . ,ak−1,ak, . . . ,ak+l−1]. The least suchlis called the length of the period of the periodic continued fraction expansion ofα. The evenness of the length of the period of the continued fraction expansion ofppforp≡3 (mod 4) a prime is well known. In [8] we determined its value modulo 4 and gave a similar result forp

2p:

Theorem 1. Take d=p or d=2p, where p≡3 (mod 4)is a prime integer. Let l≥1be the length of the period of the periodic continued fraction expansionp

d=[a0,a1, . . . ,al]. Then, (i) a0= bp

dc, al=2a0and ak=al−kfor1≤kl−1, (ii) l=2L is even and L is even if and only if p≡7 (mod 8), (iii) al/2=aLis the integer in{a0−1,a0}of the same parity as d .

This behavior in the case ofd=phad already been proved in [3, Corollary 2 p. 2071]. Our proof was different and applied both tod=pandd=2p. It was based on the arithmetic of quadratic number fields and their ideal class groups in the narrow sense (as in [6] and [7]). LetI be an integral ideal of the ring of algebraic integersZK of a real quadratic number fieldK. Recall that Iis principal if and only if there existsα∈ZKsuch thatI=αZK, whereasI is principal in the narrow sense if there exists a totally positive elementα∈ZK such thatI=αZK. Here, bearing on a similar approach, we prove:

(3)

1202 Stéphane R. Louboutin

Theorem 2. Let p,q be two prime integers equal to3 (mod 4), with3≤p<q. Let l≥1be the length of the period of the periodic continued fraction expansion(1+pp q)/2=[a0,a1, . . . ,al].

Then,

(i) al=2a0−1and ak=alkfor1≤kl−1, (ii) l=2L is even and(−1)Lp

q

´

(Legendre’s symbol), (iii) al/2=aLis the unique odd integer in{bp

q/pc −1,bp q/pc}.

Theorem 3. Let p,q be two prime integers equal to3 (mod 4), with3≤p<q. Let l≥1be the length of the period of the periodic continued fraction expansionpp q=[a0,a1, . . . ,al]. Then,

(i) a0= bppqc, al=2a0and ak=al−kfor1≤kl−1, (ii) l=2L is even and(−1)L=

³p

q

´

(Legendre’s symbol), (iii) al/2=aL=2bp

q/pcis even.

Part of Theorem 3 was proved in [10, Corollary 1], [2, Theorem 2] and [1], but notice that point (iii) of Theorem 3 is much more precise than [1, Theorem 1.2].

2. On the continued fraction expansions of some real quadratic irrational numbers

(i). Letωbe a real quadratic irrational number. Henceω=(P+p

d)/Q for some non-square integerd>1, someP∈Zand someQ∈Z\ {0} dividingdP2. Thenωis calledreducedifω>1 and−1/ω0>1, whereω0=(P−p

d)/Qis the conjugate ofωinQ(p

d). Hence,ωis reduced if and only ifP+p

d>Q>p

dP>0, which implies 0<Q<2p

d,|P| <p d, 2p

d/Q−1<ω<2p d/Q andbωc ∈{b2p

d/Qc −1,b2p d/Qc}.

(ii). Thecontinued fraction expansionω0=[a0,a1, . . . ] ofω0=(P0+p

d)/Q0withP0,Q0∈Z, and Q06=0 dividingdP02, can be computed inductively by writingωk=[ak, . . . ] asωk=(Pk+p

d)/Qk, where thePk,Qk∈ZwithQk6=0 dividingdPk2are inductively computed, usingak= bωkcand ωk=ak+1/ωk+1, byPk+1=akQkPkandQk+1=(d−Pk+12 )/Qk=(d−Pk2)/Qk+2akPkak2Qk. (HenceQ1is a non-zero rational integer,Qk+1=Qk−1+2akPkak2Qkfork≥1 and theQk’s are non-zero rational integers, by induction onk.)

(iii). Assume thatω0=(P0+p

d)/Q0is reduced. Usingωk=ak+1/ωk+1, we obtain that all the ωk’s are reduced, by induction. Hence 0<Qk <2p

d and|Pk| <p

d fork ≥0 and there are only finitely many pairwise distinctωk’s. It follows thatωm =ωn for somem>n ≥0, which impliesωk+l =ωkandak+l=akforkb, wherel:=mn≥1. Hence, the continued fraction expansion ofω0isl-periodic. In fact is purely periodic, which we writeω0=[a0, . . . ,al−1], i.e.

ωk+l=ωkandak+l=akfork≥0, wherel:=mn≥1. (Notice thatωk+l=ωkandk≥1 imply ωk−1ak−1=1/ωk=1/ωk+l=ωk+l−1ak+l−1, hence implyωk+l−1ωk−1=ak+l−1ak−1∈Z andωk+l1−ωk1=ω0k+l1ω0k1∈(−1, 1)∩Z, hence implyωk+l1=ωk1.) The least suchl≥1 is calledthe lengthof the purely periodic continued fraction expansion of the reduced quadratic irrational numberω0.

In that case−1/ω00=[al−1, . . . ,a0] (e.g. see [4, XV page 311]).

(iv). If ω0 =[a0,a1, . . . ,al−1]∈ Q(p

d) is reduced, using ωk = ak+1/ωk+1 we obtain Mk := Z+Zωk=Z+Zω−1k+1=ω−1k+1Mk+1andM0=ω−11 M1=ω−11 ω−12 M2= · · · =ε−1Ml =ε−1M0, where ε=ω1ω2. . .ωl=ω0ω1. . .ωl−1. Therefore,εis a unit of normN(ε)=Ql−1

k=0(ωkω0k)=(−1)l of the Z-moduleM0=Z+Zω0⊆Q(p

d) (asωk>1 and−1/ω0k>1).

(v). See [4, p. 305–322], [5, Chapter 10] and [9] for more information on continued fractions.

C. R. Mathématique2021, 359, n9, 1201-1205

(4)

3. Proof of Theorem 2

Letd≡1 (mod 4) be a non-square integer, withd≥5. Letg0≥1 be the unique odd integer in [p

d−2,p

d). Thenω0=(P0+p

d)/Q0=(g0+p

d)/2 is reduced. Its continued fraction expansion ω0=[g0,a1, . . . ,al−1] is purely periodic andω1=[a1, . . . ,al−1,g0]=1/(ω0g0)=2/(p

dg0)=

−1/ω00=[al1, . . . ,a1,g0]. Hence,ak=alkfor 1≤kl−1. UsingQ0=2, the oddness ofP0=g0, the evenness ofQ1=(d−P20)//Q0=(d−g02)/2 and the identitiesQk+1=Qk−1+2akPkak2Qkfor k≥1 andPk+1=akQkPkfork≥0, we obtain that theQk’s are even and thePk’s are odd for k≥0. Consequently, ifdis square-free thenM0is equal to the ring of algebraic integersZKof the real quadratic number fieldK=Q(p

d) and theZ-modules Ik:=(Qk/2)Mk=(Qk/2)Z+Pk+p

d

2 Z=(Qk/2)ω1. . .ωkM0=αkZK

are primitive, principal, integral ideals of normsQk/2 of the real quadratic number fieldQ(p d), whereαk=(Qk/2)ω1. . .ωk∈Ik ⊆ZK is an algebraic integer of norm (−1)k(Qk/2) (recall that ωk>1 and−1/ω0k>1 fork≥0). Hence,Ikis principal in the narrow sense if and only ifkis even.

Now, assume that d is divisible by a prime p ≡ 3 (mod 4). Since the congruence x2d y2 ≡ −4 (modp) has no solution in rational integers, any algebraic unit of Q(p

d) has norm +1. The algebraic unit ε = ω0ω1. . .ωl−1 := Z+Zω0 being of norm (−1)l, l =2L is even, ω0 = [g0,a1, . . . ,aL1,aL,aL1, . . . ,a1], ωL = [aL, . . . ,a1,g0,a1, . . . ,aL1] and ωL+1 = [aL−1, . . . ,a1,g0,a1, . . . ,aL] = −1/ω0L. Hence, −1 = ωL+1ω0L = PL+1+

pd QL+1

PLp d

QL , which implies PL+1=PL. SincePL+1=aLQLPL, we havePL+1=aL(QL/2) andaLis odd. Moreover,

dPL2+1=da2L(QL/2)2=4(QL/2)(QL+1/2).

Hence,QL/2 dividesd. Finally,ωL=(PL+p

d)/QL being reduced, we have 1<QL<2p d and aL= bωLc ∈{b2p

d/QLc,b2p

d/QLc −1}, and we obtain the following Proposition and Corollary from which Theorem 2 follows:

Proposition 4. Let d ≡1 (mod 4) be a square-free integer, with d ≥5 such that at least one prime p≡3 (mod 4)divides d . Let g0≥1be the unique odd integer in the interval[p

d−2,p d).

Setω0=(g0+p

d)/2. l ≥1be the length of the period of the purely periodic continued fraction expansionω0=[g0,a1, . . . ,al−1]. Then

(i) ak=al−kfor1≤kl−1;

(ii) l=2L is even;

(iii) QL/2divides d and1<QL/2<p d ; (iv) aLis odd and aL= bωLc ∈{b2p

d/QLc,b2p

d/QLc −1};

(v) The integral idealIL=(QL/2)Z+PL+2pdZof norm QL/2is principal and L is even if and onlyILis principal in the narrow sense.

Corollary 5. Let p,q be two prime integers equal to3 (mod 4), with3≤p<q. Take d=p q≡1 (mod 4). Then QL/2=p. Hence,ILis the prime ramified idealP of norm p of the ring of algebraic integers of the real quadratic fieldQ(p

d)and aLis the unique odd integer in{bp

q/pc,bp

q/pc−1}.

Moreover,P is principal in the narrow sense if and only if³p

q

´

= +1.

Proof. LetP andQbe the prime ideals above p andq, respectively. HenceP =I =(α) is principal andP Q=(p

d) is also clearly principal. Sincep

d∈P Q⊆P =(α), we havep d=αβ for some algebraic integerβ. HenceQ=(β) is also principal. SinceN(α)N(β)=N(αβ)=N(p

d)=

d<0, only one of the two principal idealsP orQis principal in the narrow sense. IfP =(α) is principal in the narrow sense, withα=(x+yp

d)/2 such thatp=N(α)=(x2p q y2)/4, then pdividesx=p X, 4=p X2q y2and³p

q

´

= +1. IfP is not principal in the narrow sense, then Q=(β) is principal in the narrow sense, withβ=(x+yp

d)/2 such thatq=N(β)=(x2p q y2)/4.

Hence,qdividesx=q X, 4=q X2p y2and³−p

q

´

= −

³p

q

´

= +1.

(5)

1204 Stéphane R. Louboutin

4. Proof of Theorem 3

Letd≡1 (mod 4) be a non-square integer, withd≥5. Setg= bp

dc. Thenω0=(P0+p

d)/Q0= g+p

dis reduced. SinceM0=Z[ω0]=Z[p

d] is not the ring of algebraic integers ofQ(p d), the proof of Theorem 3 is a little more tricky than the one of Theorem 2. Here again, the continued fraction expansionω0=[2g,a1, . . . ,al−1] is purely periodic andω1=[a1, . . . ,al−1, 2g]=1/(ω0− 2g)=1/(p

dg)= −1/ω00=[al−1, . . . ,a1, 2g]. Hence,ak =al−k for 1≤kl−1. Suppose that we hadQn ≡2 (mod 4) for somen≥0. ThenPn+1would be odd andQn+1would be even, as QnQn+1=dPn2+1. Therefore, all theQk’s would be even forkn, asQk+1=Qk1+2akPka2kQk fork≥1, hence fork≥0, by pure periodicity of the continued fraction expansion ofω0. SinceQ0 is odd, we deduce thatQk6≡2 (mod 4) fork≥0.

Now, assume thatdis divisible by a primep≡3 (mod 4). As above,l=2Lis even and 2PL+1=aLQL, and 4d−a2LQ2L=4QLQL+1.

Hence,QL divides 4d and 4 does not divideQL and we obtain the following Proposition from which Theorem 3 follows, by Corollary 5:

Proposition 6. Let d≡1 (mod 4)be a square-free integer, with d≥5such that at least one prime p≡3 (mod 4)divides d .

Setω0=g+p

d , where g= bp

dc. Let l ≥1be the length of the period of the purely periodic continued fraction expansionω0=[2g,a1, . . . ,al1]. Then

(i) ak=al−kfor1≤kl−1;

(ii) l=2L is even;

(iii) QLdivides2d and1<QL<2p d ; (iv) aL= bωLc ∈{b2p

d/QLc,b2p

d/QLc −1};

(v) if d =p q, where p,q are prime numbers equal to3 modulo4with p <q, then aL = 2bp

q/pc, QL =p, the prime idealP of norm p of the ring of algebraic integers of the real quadratic fieldQ(p

d)is principal and L is even if and only ifP is principal in the narrow sense.

Proof. It remains to prove point (v). SinceQLdivides 2p q,QL6≡2 (mod 4) andQL<2pp q, we haveQL∈{p,q}. SinceQL=qwould yield the contradiction 4qQL+1=4d−a2LQ2L≤4p q−4q2<0, we haveQL=p. Hence,aLis even, as 2PL=aLQL, andaL∈{b2xc,b2xc −1}, wherex=p

d/QL. Sinceb2xc ∈{2bxc, 2bxc +1} forxreal, we have thataL is even andaL∈{2bxc −1, 2bxc, 2bxc +1}.

Therefore,aL=2bxc =2bp

d/QLc =2bp q/pc. Finally, setβL=QLω1. . .ωL. Then

JL:=βLZ[p

d]=βLM0=QLω1. . .ωLM0=QLML=QLZ+(PL+p

d)Z⊆Z[p d].

Hence,βL=x+yp d∈Z[p

d] and {QL,PL+p

d} and {βL,βL

pd}={x+yp

d,d y+xp

d} are two Z-bases ofJLand the change of basis matrix

A= ÃxyP

L QL

d yxPL QL

y x

!

is in M2(Z) and of determinant ±1, i.e. ±1 =(x2d y2)/QL =N(βL)/p. Therefore, N(βL)= (−1)Lp, with βL ∈Z[p

d]. It follows that the prime ideal P of the ring of algebraic integers ZK=Z[(1+p

d)/2] ofK=Q(p

d) lying abovep is principal and equal to (βL) and thatP it is

principal in the narrow sense if and only ifLis even.

C. R. Mathématique2021, 359, n9, 1201-1205

(6)

References

[1] S. Das, D. Chakraborty, A. Saikia, “On the period of the continued fraction ofppq”,Acta Arith.196(2020), no. 3, p. 291-302.

[2] C. Friesen, “Legendre symbols and continued fractions”,Acta Arith.59(1991), no. 4, p. 365-379.

[3] E. P. Golubeva, “Quadratic irrationals with fixed period length in the continued fraction expansion”,J. Math. Sci., New York70(1994), no. 6, p. 2059-2076.

[4] H. Hasse,Vorlesungen über Zahlentheorie, Grundlehren der Mathematischen Wissenschaften, vol. 59, Springer, 1964.

[5] L. K. Hua,Introduction to number theory, Springer, 1982, translated from the Chinese by Peter Shiu.

[6] S. R. Louboutin, “Continued fractions and real quadratic fields”,J. Number Theory30(1988), no. 2, p. 167-176.

[7] ——— , “Groupes des classes d’idéaux triviaux”,Acta Arith.54(1989), no. 1, p. 61-74.

[8] ——— , “On the continued fraction expansions ofppandp

2pfor primesp3 (mod 4)”, inClass groups of Number fields and related topics, Springer, 2020, p. 175-178.

[9] O. Perron,Die Lehre von den Kettenbrüchen. Band I. 3. erweiterte und verbesserte Aufl., Teubner, 1954.

[10] A. J. van der Poorten, P. G. Walsh, “A note on Jacobi symbols and continued fractions”,Am. Math. Mon.106(1999), no. 1, p. 52-56.

Références

Documents relatifs

Kraaikamp, Metric properties of Denjoy's canonical continued frac- tion expansion.. Ito, Algorithms with mediant convergents and their

In this section we apply the continuous fraction method to calculate response functions in infinite symmetric matter for a realistic case, using the Gogny effective interaction D1

GOLDFELD, An asymptotic formula relating the Siegel zero and the class number of quadratic fields, Annali Scuola Normale Superiore, Serie IV, 2

On the length of the continued fraction for values of quotients of power sums.. par PIETRO CORVAJA et UMBERTO

In 1968 Shanks [18] proposed an algorithm relying on the baby-step giant-step method in order to compute the structure of the ideal class group of an imaginary quadratic number field

Then, coming back to the Euclidean Dy- namical System, we compute (in polynomial time) three im- portant constants which play a central rˆ ole in the Euclidean algorithm: (i)

The expansion (10) can also be found independently and directly by using Euler's general continued fraction expansion [8]... Take the logarithm on both sides of (16) and expand the

It is proved that with respect to γ, the Hausdorff dimension (denoted by dim H ) of E ϕ is analytic, increasing from 0 to 1, and tends to 1 when γ goes to infinity.. In this paper,