Diophantine equations

From testwiki
Jump to navigation Jump to search

Sequences, series and numbers generated by diophantine equations and their applications by Jamel Ghanouchi

Abstract

Our purpose in this article which has been published is to show how much diophantine equations are rich in analytic applications. Effectively, those equations allow to build amazing sequences, series and numbers. The question of the proof of some theorems remains of course, we will see it in this communication. We will make also an allusion to the very known Fermat numbers (22n). We will see how this problem of the proof is actual and how it can be solved using amazing sequences and series.

The Ghanouchi's sequences

Let us begin by Fermat equation (E), it is

Un=Xn+Yn

with GCD(X,Y)=1

We pose

u=U2n

x=UnXn

y=UnYn

z=XnYn

then

u=U2n=Un(Xn+Yn)=x+y(1)

and

1z=1XnYn=U2nUnXnUnYn=uxy=x+yxy=1x+1y(2)

If U, X, Y are integers verifying equation (E), then u, x, y, z as defined verify

LEMMA 1

u=x+y(1)

1z=1x+1y(2)

Firstly

z=XnYn

u=(Xn+Yn)2

x=Xn(Xn+Yn)

y=Yn(Xn+Yn)

We pose

x1=x

and

y1=y

but

x1,y1

z1

verifying

1z1=1x1+1y1

and

z1=xyx+y=z

then

(x1+y1)z1=x1y1

or

x1(y1z1)=z1y1

we pose

y2=y1z1=z1y1x1

and

y1(x1z1)=z1x1

also

x2=x1z1=z1x1y1

and

x2y2=z12

which means

x1=x2+z1=x2+x2y2

and

y1=y2+z1=y2+x2y2

and

u1=u=(x1+y1)=(x2+y2)2>x2+y2>0

(x1+y1) is an integer

x1=x2(x2+y2)>x2>0

x1 is an integer

y1=y2(x2+y2)>y2>0

y1 is an integer

z1=x1y1x1+y1=XaYb=x2y2>z2=x2y2x2+y2>0

z2 is rational, because

x2,y2 rationals

z2 rational verifying

1z2=1x2+1y2

until infinity. For i

(xi+yi)=(xi+1+yi+1)2>xi+1+yi+1>0

xi+yi is rational for i>1

xi=xi+1(xi+1+yi+1)>xi+1>0

xi is rational for i>1

yi=yi+1(xi+1+yi+1)>yi+1>0

yi is rational for i>1

zi=xiyixi+yi=xi+1yi+1>zi+1=xi+1yi+1xi+1+yi+1>0

zi is rational for i>1, and

1zi+1=1xi+1+1yi+1

(xi+yi)=(xi+1+yi+1)2

LEMMA 2

xi and yi have expressions

xi=x2i1j=0j=i2(x2j+y2j)1(H)

yi=y2i1j=0j=i2(x2j+y2j)1(H)

Proof of lemma 2

By induction

x=x2(x2+y2)=x2(x+y)12

x2=x2x+y

also

y2=y2x+y

it is verified for i=2. We suppose (H) and (H') true for i, then

xi=xi+1(xi+1+yi+1)=xi+1(xi+yi)12

and

xi+1=xi2(xi+yi)1

but (H) and (H'), then

xi+1=x2ij=0j=i2(x2j+y2j)2(x2i1+y2i1)1j=0j=i2(x2j+y2j)

=x2ij=0j=i1(x2j+y2j)1

and it is true for i+1, also for yi

xi and yi can be written as it follows

xi=x2i1j=0j=i2(x2j+y2j)1

yi=y2i1j=0j=i2(x2j+y2j)1

i>1

but, a,b

ab=(a2i1b2i1)j=0j=i2(a2j+b2j)1(E)

xy=(x2i1y2i1)j=0j=i2(x2j+y2j)1(E)

j=0j=i2(x2j+y2j)=(x2i1y2i1)xy

the expressions of the sequences become, for

xy

or

xiyi

LEMMA 3

xi=x2i1x2i1y2i1(xy)

=UnXn2i1Xn2i1Yn2i1(XnYn)

and

yi=y2i1x2i1y2i1(xy)

=UnYn2i1Xn2i1Yn2i1(XnYn)

ui=xi+yi=UnXn2i1+Yn2i1Xn2i1Yn2i1(XnYn)


LEMMA 4

The equations (1) have (2) a constant

xiyi=xy

THE GHANOUCHI'S THEOREM

The only solution of equations

u=x+y,((1))

and

1z=1x+1y,(2)

is

xy(xy)=0

Proof of Ghanouchi's theorem

As

xi=x2i1x2i1y2i1(xy)

and

yi=y2i1x2i1y2i1(xy)

if

x>ylimi(xi)=xy,limi(yi)=0

and

y>xlimi(yi)=yx,limi(xi)=0

We will give several proofs that xy(xy)=0 is the solution with the series. We recapitulate

xi>xi+1,yi>yi+1x=y

xi=xi+1=xi+1+xi+1yi+1xy=0

yi=yi+1=yi+1+xi+1yi+1xy=0

xy(xy)=0

x and y are not différent, the initial hypothesis is false there the only solution is

xy(xy)=0

The proofs : we have

yi2xi2=(yx)(xi+yi)=(yi+1xi+1)(xi+1+yi+1+2xi+1yi+1)

=yi+12xi+12+2xi+1yi+1(yi+1xi+1)

yi+1=Z is solution of the following equation

Z4+2xi+1Z32xi+13Zxi+12+xi2yi2=0

Also xi+1=Z is solution of

Z4+2yi+1Z32yi+13Zyi+12+yi2xi2=0

And

Z4+2ZZ32Z3ZZ4+xi2yi2=0

Also

Z4+2ZZ32Z3ZZ4+yi2xi2=0

Let

Z=uZ2

But

(uZ2)4+2Z(uZ2)32Z3(uZ2)Z4+xi2yi2=0

Hence

u4+Z4162u3Z+32u2Z2+

12uZ3+2Zu33u2Z2+32uZ314Z4+

2Z3u+Z4Z4+xi2yi2=0

=u4(32Z2)u2Z3u316Z4+xi2yi2=0

We deduce

(u234Z2)234Z4Z3u+xi2yi2=0

Also Z=vZ2 leads to

(vZ2)4+2Z(vZ2)32Z3(vZ2)Z4+yi2xi2=0

Hence

v4+Z4162v3Z+32v2Z2+

12vZ3+2Zv33v2Z2+32vZ314Z4+

2Z3v+Z4Z4+yi2xi2=0

=v4(32Z2)v2Z3v316Z4+yi2xi2=0

We deduce

(v234Z2)234Z4Z3v+yi2xi2=0

If we add

(u234Z2)2+(v234Z2)234(Z4+Z4)Z3uZ3v=0

=(Z2+Z24+ZZ34Z2)2+(Z2+Z24+ZZ34Z2)234(Z4+Z4)ZZ3Z3ZZ4+Z42=0

=(Z2+Z24+ZZ)2+(Z2+Z24+ZZ)254(Z4+Z4)ZZ(Z2+Z2)=0

=Z4+Z4+2Z2Z28+2Z2Z2+ZZ(Z2+Z2)54(Z4+Z4)ZZ(Z2+Z2)=0

=98(Z4+Z4)+94Z2Z2=0

=(Z2Z2)98=0

The solution is

Z2Z2=yx=0

Another proof : we have

xi1+xi=xi1xixi1xi=xiyixi1xi

=1xi1xiyixixiyi=1xi1+yi1xiyi1yi

=1xi1+yi14xi4xi4yi=xi4yixi1+yi14xi1xi1+yi14

=xi4yi(xi1+yi14)xi1+yi1xi1xi4yixi1+yi14yi+14

Because

xi1+yi1xi1=yi1xi1+yi1+xi1yi+14

yi1=yixi1+yi1yi+14(xi1+yi1+xi1)=yixi+yi4(xi1+yi1+xi1)

And

xi+yi4xi1+yi1xi1+yi1+xi1

Because

xi+yi4=xi12+yi124xi1+yi141+xi1xi1+yi1

And

xi12+yi124xi1+yi14+xi1+yi14xi1xi1+yi1

We deduce

xi1+xixi4yiyi+14=xi4yi+14(xi+yi4)yi+14=xi4(xi+yi4)

In the infinity

0<xylimi(xi1+xi)=2xylimi(xi4xi+yi4)=xy

Therefore

xy=2xy=0

Another proof : let

xi=aixi+1=xi+1+xi+1yi+1ai=1+yi+1xi+1=1+yixi

Also

yi=biyi+1=yi+1+xi+1yi+1bi=1+xi+1yi+1=1+xiyi

We deduce

ai=1+aibi=1+yixi

And

bi=1+biai=1+xiyi

Thus

aibi=ai+bi

And

xi=ciyi+1=xi+1+xi+1yi+1ci=bi(bi1)

Also

yi=dixi+1=yi+1+xi+1yi+1di=ai(ai1)

We have

xi=bixi+1yi+1,yi=aixi+1yi+1

Let also

xi=iαi,yi=iβi

And

αi=a'iαi+1,βi=b'iβi+1

We deduce

aibi=yixi=βiαi=a'ib'i

Because

iαi=xi=xi+1+xi+1yi+1=(i+1)(αi+1+αi+1βi+1)

iβi=yi=yi+1+xi+1yi+1=(i+1)(βi+1+αi+1βi+1)

Or

αiβi=αi+1βi+1=a'iαi+1b'iβi+1

And

b'i=i+1ibi2,a'i=i+1iai2

But

b'i2=αiαi+1βi+12=a'iαiβi2

=(a'i2)xiyi+2(xiyi1)

And

a'i2=βiαi+1βi+12=b'iβiαi2

=(b'i2)yixi+2(yixi1)

Thus

(a'i2)2xiyi+2(xiyi1)(a'i2)

=(b'i2)2yixi+2(yixi1)(b'i2)

=((a'i2)xiyi+2(xiyi1))2yixi+2(yixi1)((a'i2)xiyi+2(yixi1))

=(a'i2)2xiyi+4(xiyi1)2yixi+4(xiyi1)(a'i2)+2(1xiyi)(a'i2)+4(yixi1)2

=(a'i2)2xiyi+4(xiyi1)2yixi+2(xiyi1)(a'i2)+4(yixi1)2

Hence

4(xiyi1)2yixi+4(yixi1)2=0

=4((xy)2yi2)yixi+4((xy)2xi2)=0

=4(xy)2(1xiyi+1xi2)=0

(xy)2=0

Another proof : We have

cidi=aibi

Let

ai1bi1=fiaibi

We have

fi=ai1bi1aibi=xi1yi1xi+1yi+1xi2yi2

=xiyi(xi+yi)2xi+1yi+1xi+1yi+1(xi+yi)21

aibi=aibi+biai+2=(aibi+biai)2=(ai1bi12)2

And

aibi=ei=(fiaibi2)2

It means

ei2fi2(4fi+1)ei+4=0

And

ei=aibi=4fi+1+8fi+12fi2

But

aibi=fiaibi2fiaibiaibi2=0

Thus

ei=aibi=1+1+8fi2fi

And

ei2=1fi+1+8fi3fi2fi=1fi+1fi2+8fi(1fi)1+8fi+3fi2fi

=(1fi)1+1+9fi8fi1+3fi2fi

And

ei4=aibi4=(ei2)(ei+2)=(ei2)fiaibi=(1fi)fiaibi

1fi2+4fi(1fi)+1fi2+8fi(1fi)8fi+1+3fi22fi2

=(1fi)1+5fi+1+9fi8fi+1+3fi22fi2

(1fi)(fiaibi1+5fi+1+9fi1+8fi+3fi22fi2)=0

Or

(1fi)(aibi1+5fi+1+9fi1+fi+3fi22fi3)=0

=(1fi)(fi(1+4fi)+fi8fi+115fi1+9fi1+8fi+3fi22fi3)=0

The expression between the parenthesis is not equal to zero, we deduce

fi=1

Else

aibi=2fiaibi1

aibi2=2(2fiaibifiaibi)=2(2(1fi)+fi(2aibi)fiaibi)

We deduce

(aibi2)(1+2aibi)=(aibi2)(aibi+2aibi)

=(aibi2)(fiaibiaibi)=(aibi2)fiaibi=41fifiaibi

Thus

(1+1+8fi2fi2)fiaibi=(1fi)(1+1+9fi1+8fi+3fi2fi)=(1fi)4fiaibi

The expressions between the parenthesis are not equal, therefore

fi=1

Else

ai4=(ei2)(ei+2)=(1fi)(1+1+9fi1+8fi+3fi2fi)(1+8fi+1+8fi2fi)

Hence

(1fi)(1+5fi+1+9fi1+8fi+3fi22fi2(1+1+9fi1+8fi+3fi2fi)(1+8fi+8fi+12fi))=0

Or

(1fi)(2+10fi+21+9fi1+8fi+3fi2(1+8fi+(1+9fi)8fi+18fi++3fi+8fi+1+(1+8fi)(1+9fi)8fi+1+3fi))=0

The expression between the parenthesis is not equal to zero, we deduce

fi=1xiyi(aibi4)=(xy)2=0

And, else

2fi2ei=4fi+1+8fi+1

(2fi2ei4fi1)2=8fi+1=4fi4ei2+16fi216fi3ei4fi2ei+8fi+1

fi2ei2+44fieiei=0

(fi2ei4fi1)ei=4

ei=41+4fifi2ei

ei4=16fi+4fi2ei1+4fifi2ei

=16fi+16fi2+4fi2(ei4)1+4fifi2ei

(ei4)(14fi21+4fifi2ei)

=(fi1)16fi1+4fifi2ei=4fiei(fi1)

=(ei4)(1fi2ei)

ei=4fi+1+8fi+12fi2

ei4=4fi(1fi)+1fi2+8fi+13fi22fi2

=4fi(1fi)+(1fi)(1+fi)+8fi8fi4+1fi48fi+1+3fi22fi2

=(1fi)(5fi+1+9fi3+9fi2+9fi+18fi+1+3fi22fi2)

(fi1)(4fiei+(1fi2ei)(5fi+1+9fi3+9fi2+9fi+18fi+1+3fi22fi2))=0

=(fi1)(4(4fi+1+8fi+1)2fi+(14fi+1+8fi+12)(5fi+1+9fi3+9fi2+9fi+18fi+1+3fi22fi2))=0

=(fi1)(5fi+1+9fi3+9fi2+9fi+18fi+1+3fi22fi2(4fi+1+8fi+12)(3fi+1+9fi3+9fi2+9fi+18fi+1+3fi22fi2))=0

And the expression between the parenthesis is not equal to zero, it means that

fi=1aibi=ei=4

ai2bi24aibi=(aibi)2=0=(yixi)2xi+1yi+1

And

(aibi)xi+1yi+1=yx=0

Another proof :

xi and yi are particular cases of vi and wi which follow

vi=xixiyi(xy),wi=yixiyi(xy)

xi=v2i1,yi=w2i1

Also

ei=vi+wiwi,fi=vi+wivi

ei=ei+fifi,fi=ei+fiei

bi=e2i1,ai=f2i1

So, we have

viwi=xy

And

eifi=ei+fi

But

b1b2...bi=1+e1f1+...+e2i1f2i1

Let

1+e1f1+e2f2+...+eifi=1+v1w1+v2w2+...+viwi

=1+xy+x2y2+...+xiyi=1x(i+1)y(i+1)1xy=2ei+12e1

Thus

ei+12=ei+1fi+1fi+1=vi+1wi+1wi+1=xywi+1

=(e12)(1+(e11)+(e21)+...+(ei1))=(e12)(1+i+(e12)+(e22)+...+(ei2))

=(e1f1f1)(1+i+e1f1f1+e2f2f2+...+eififi)

=(v1w1w1)(1+i+v1w1w1+v2w2w2+...+viwiwi)

Or

(xy)(1wi+11w1(1+i+xyw1+xyw2+...+xywi))=0

And

1wi+1(1+i)1w1

is not equal to zero for xy=0, and it is the case, of course, of

1w1(xy)(1w1+...+1wi)

therefore the expression between the parenthesis is not equal to zero. We gave the fourth proof that the only solution is

xy(xy)=0

And there are others (see the series).

The Ghanouchi's series

As seen

xiyi=yi1yi=xi1xi

we deduce

xixi+1=xi+1yi+1

xi1xi=xiyi


x1x2=xx2=x2y2

then

j=2j=i+1(xjyj)=xx2+x2x3+...+xixi+1=xxi+1

and

j=2j=(xjyj)=limi(xxi+1)

if x>y

j=2j=(xjyj)=limi(xxi+1)=x(xy)=y

if x<y

j=2j=(xjyj)=limi(xxi+1)=x

The applications of the Ghanouchi's series

j=2j=i((1)jxjyj)=xx2(x2x3)+(x3x4)...+(1)i(xi1xi)

=x2x2+2x3...+2(1)i1xi1+(1)i+1xi

=2j=2j=i1((1)j+1xj)+x+(1)i+1xi

=2j=1j=i((1)j+1xj)x(1)i+1xi

=j=2j=i1((1)j+1xj)+j=1j=i((1)j+1xj)

=2j=2j=i1((1)j+1yj)+y+(1)i+1yi

=2j=1j=i((1)j+1yj)y(1)i+1yi

=j=2j=i1((1)j+1yj)+j=1j=i((1)j+1yj)

or

2j=1j=i((1)j+1xj)=j=2j=i((1)jxjyj)+x+(1)i+1xi

and

2j=1j=i((1)j+1yj)=j=2j=i((1)jxjyj)+y+(1)i+1yi

As we do not know the limit of (1)i+1xi, then

j=1j=((1)jxj)

can be not convergent. But

j=2j=((1)jxjyj)

is convergent.

Also, knowing that yi tends to zero in the infinity, we can say

j=1j=((1)jyj)

is convergent. The limit of

j=2j=i((1)jxjyj)=2j=1j=i((1)j+1yj)y(1)i+1yi

=2j=1j=i((1)j+1xj)x(1)i+1xi

exists and the series are convergent. It means that for x and y integers and for conditions on the exponents like n3 for Fermat equation :

limi(xi)=xy=0

It is confirmed by the fact that the général term of the series tends to zero. Let us prove it. We give two proofs. We must remark that we prove firstly that the following series are convergent, we do not present the proof, here. Let


k=1k=2m((1)k+1xkek2m)

=xe12mx2e22m+x3e32m...+(1)2m+1x2me2m2m

=xe22m+x(e12me22m)x2e22m+x3e42m+x3(e32me42m)x4e42m+...x2me2m2m

=xe22m(e12m1)+x3e42m(e12m1)+...+x2m1e2m2m(e12m1)+

+(xx2)e22m+(x3x4)e42m+...+(x2m1x2m)e2m2m

=(e12m1)(xe22m+x3e42m+...+x2m1e2m)+(x2y2e22m+x4y4e42m+...+x2my2me2m2m)

=(e12m1)k=1k=m(x2k1e2k2m)+k=1k=m(x2ky2ke2k2m)

Also

k=1k=2m((1)k+1ykek2m)

=ye12my2e22m+y3e32m...+(1)2m+1y2me2m2m

=(e12m1)k=1k=m(y2k1e2k2m)+k=1k=m(x2ky2ke2k2m)

But

(e12m1)k=1k=m(y2k1e2k2m)=S

We have

(e12m1)k=1k=m(y2k1e2k+12m)<S<(e12m1)k=1k=m(y2k1e2k12m)

(e12m1)e32mk=1k=m(y2k1e2k22m)<S<(e12m1)e12mk=1k=m(y2k1e2k22m)

Thus

limm((e12m1)k=1k=m(y2k1e2k22m))=limm((e12m1)y+(e12m1)k=2k=m(y2k1e2k22m))

=limm((e12m1)k=2k=m(y2k1e2k22m))=limm(S)

We have

(e12m1)k=pk=m(y2k1e2kp2m)=A

And

(e12m1)k=pk=m(y2k1e2kp+12m)<A<(e12m1)k=pk=m(y2k1e2kp12m)

Or

(e12m1)e22mk=pk=m(y2k1e2kp12m)<A<(e12m1)k=pk=m(y2k1e2kp12m)

Hance

limm((e12m1)k=pk=m(y2k1e2kp12m))

=limm((e12m1)y2p1ep12m+(e12m1)k=p+1k=m(y2k1e2kp12m))

=limm((e12m1)k=p+1k=m(y2k1e2kp12m))=limm(A)=limm(S)

And

p+1=mlimm(A)=limm(S)=limm((e12m1)y2m1em2m)=0

Consequently

limm((e12m1)k=1k=m(y2k1e2k2m))=0

Also

(e12m1)k=1k=m(x2k1e2k2m)=S

We have

(e12m1)k=1k=m(x2k1e2k+12m)<S<(e12m1)k=1k=m(x2k1e2k12m)

(e12m1)e32mk=1k=m(x2k1e2k22m)<S<(e12m1)e12mk=1k=m(x2k1e2k22m)

Thus

limm((e12m1)k=1k=m(x2k1e2k22m))=limm((e12m1)x+(e12m1)k=2k=m(x2k1e2k22m))

=limm((e12m1)k=2k=m(x2k1e2k22m))=limm(S)

And

(e12m1)k=pk=m(x2k1e2kp2m)=A

And

(e12m1)k=pk=m(x2k1e2kp+12m)<A<(e12m1)k=pk=m(x2k1e2kp12m)

Or

(e12m1)e22mk=pk=m(x2k1e2kp12m)<A<(e12m1)k=pk=m(x2k1e2kp12m)

Hence

limm((e12m1)k=pk=m(x2k1e2kp12m))

=limm((e12m1)x2p1ep12m+(e12m1)k=p+1k=m(x2k1e2kp12m))

=limm((e12m1)k=p+1k=m(x2k1e2kp12m))=limm(A)=limm(S)

Or

p+1=mlimm(A)=limm(S)=limm((e12m1)x2m1em2m)=0

Consequently

limm((e12m1)k=1k=m(x2k1e2k2m))=0

We deduce

0<limm(k=1k=2m((1)k+1xkek2m))

=limm(k=1k=2m((1)k+1xkek2m))

=limm(k=1k=m(x2ky2ke2k2m))<limm(k=1k=m(x2ky2k))

<limm(k=1k=2m(xkyk))=y

Thus

limm(k=1k=2m((1)k+1(xkyk)ek2m))=0

=limm((xy)k=1k=2m(ek2m))=limm((xy)e12m1e2m1+e12m)=xy2=0

And

xy=0

the only solution is xy(xy)=0, xy=0 if at least one of the sequences xi or yi is constant. And the second proof. Let

xi(t)=k=2k=i1(xkekt)=k=2k=i(xkekt)xieit=xi(t)xieit

yi(t)=k=2k=i1(ykekt)=k=2k=i(ykekt)yieit=yi(t)yieit

ui(t)=k=2k=i(xkykekt)=k=2k=i1((xk1xk)ekt)

=xe2t+x2(e2t+e3t)+x3(e3t+e4t)+...+xi1(e(i1)t+eit)xieit

=xe2t+x2e2t(1+et)+x3e3t(1+et)+...+xi1e(i1)t(1+et)xieit

=xe2txieit+(1+et)(x2e2t+x3e3t+...+xie(i1)t)

=xe2txieit+(1+et)xi(t)=xe2txieit+(1+et)xi(t)(1+et)xieit=xe2txie(i+1)t+(1+et)xi(t)

let

t=1ilimi(ui(1i))=limi(k=2k=i(xkykeki))=limi(k=2k=i(xkyk))=y

=x+limi((1+e1i)k=2k=i1(xkeki)xiei)=x

=y+limi((1+e1i)k=2k=i1(ykeki)yiei)=y

We Recapitulate

xi>xi+1,yi>yi+1x=y

xi=xi+1=xi+1+xi+1yi+1xy=0

yi=yi+1=yi+1+xi+1yi+1xy=0

xy(xy)=0

is the only solution of (1) and (2). This result is paradoxal, we remark that we have not put any condition on n, because there are solutions for n2. The answer is related to Matiasevic theorem which claims that dos not exist an algorithm to prove theorems related to diophantine equations and we gave one : The approach must conduct then to an impossibility. We confirm Matiasevic theorem and prove it because our algorithm is available for n=1. The approach is more important than it appears, it is an answer to problems more general than Fermat theorem or Beal or Fermat-Catalan conjectures. We will try to prove them. The series become

k=2k=((1)kxkyk)=k=2k=((1)kxk)=k=2k=((1)kyk)

=x2x3+x4x5+...

=2k=1k=((1)k+1xk)x=x2x2+2x32x4+...

=2k=1k=((1)k+1yk)y=y2y2+2y32y4+...

3(x2x3+x4x5+...)=3k=2k=((1)kxk)=3k=2k=((1)kyk)=x=y

x2x3+x4x5+...=k=2k=((1)kxk)=k=2k=((1)kyk)=x3=y3

and

k=2k=(xkyk)=k=2k=(xk)=k=2k=i(yk)

=x2+x3+x4+x5+...=y2+y3+y4+y5+...=y=x

and

xi=x2i1j=0j=i2(x2j+y2j)1=yi

=x2i1j=0j=i2(2x2j)1=x2i12i1x2i11

=xi=x2i1=yi=y2i1

This development is in fact a test of impossibility. The sequences and series are a consequence of Fermat equation and of other diophantine equations (as we will see). The question now is : why are there solutions for n=2 ? The answer is in the formulas, as seen. It is important to note that for n=1, there are trivial solutions. But, for n=1, lemma 3 allows to write

xi=x2i1x2i1y2i1(xy)=UX2i1X2i1Y2i1(XY)

=UX22i2X22i2Y22i2(XY)

and

yi=y2i1x2i1y2i1(xy)=UY2i1X2i1Y2i1(XY)

=UY22i2X22i2Y22i2(XY)

and

ui=xi+yi=UX22i2+Y22i2X22i2Y22i2(XY)=X22i2+Y22i2X22i2Y22i2(X2Y2)

It is the expression of ui1, the ui1 exponent 2.

The case n=1 conducts to the case n=2 and as there are solutions for n=1, it will be the same for n=2 !

For n=4

ui=xi+yi=X42i3+Y42i3X42i3Y42i3U(XY)

the case n=4 is different, in this formula the exponent i-3 does not guarantee the existence of the sequences if

i=2.

Then, the case n=2 is the only exception. The only solution for n>2 is xy(x-y)=0, there is no solution.

Another application is Beal equation. It is

Uc=Xa+Yb

GCD(X,Y)=1

We pose

u=U2c

x=UcXa

y=UcYb

z=XaYb

and

u=U2c=Uc(Xa+Yb)=x+y

and

1z=1XaYb=U2cUcXaUcYb=uxy=x+yxy=1x+1y

it is lemma 1 and, with

Uc=Xa+Yb

the solutions are

xy(xy)=U2cXaYb(UcXaUcYb)=0

or impossible solutions for a>2 et b>2 et c>2.

Another application is the following equation.

Un=X1n1+X2n2+...+Xini

GCD(Xj,Xk)=1;j,k,jk

We conjecture and prove that there is no solutions for n>i(i-1) and nj>i(i1), j{1,2,...,i}. We can not know when there are solutions as proved by Matiasevic when one of the exponent is less or equal to i(i-1).

LEMMA 6

The solution k2 is

Xk=0


Proof of lemma 6

We pose

u=U2n

x=UnXknk

y=Un(UnXknk)

z=Xknk(UnXknk)

or

u=x+y

and

1z=1x+1y

it is lemma 1. Its solution is

U=Xk=0;k{1,2,...,i}

But, why are not they solutions for n>i(i-1) and nj>i(i1) ?

We will generalize the definition of the sequences.

We will define general sequences.

Our goal is to prove that if Xi,

(i2), ni, U, n, PGCD(X1,X2,...,Xi)=1

are positive integers, then

Xa=Xb=0;a,b=1,2,...,i

nk>i(i1),k=1,2,...,i,n>i(i1)

for the equation

X1n1+X2n2+...+Xini=Un(e)

When ni(i1),nki(i1),k=1,2,...,i

there are solutions, for example :

i=2 has 32+42=52

and i=3 has 33+43+53=63

and

958004+2175194+4145604=4224814

and i=4 has 275+845+1105+1335=1445

etc...

We suppose (e) verified and that GCD(Xk)=1

k{1,2,...,i};i2, soit

xk=U(i1)nXknk

k{1,2,...,i}

and

u=Uin

and

v=X1n1X2n2...Xini

LEMMA 7

xk, k=1,2,...i, u, v verify

x1+x2+...+xi=U(i1)n(X1n1+X2n2+...+Xini)=Uin=u(3)

and

1v=1X1n1X2n2...Xini=Ui(i1)nU(i1)nX1n1U(i1)nX2n2...U(i1)nXini=u(i1)x1x2...xi(4)

We pose

xk,0=xk

u0=u

v0=v

and

xk,1=xki(x1+x2+...+xi)(i1)

k=1,2,...,i

it implies

u=x1+x2+...+xi=(x1,11i+x2,11i+...+xi,11i)i>u1>1

and

xk,0=xk=xk,11i(x1+x2+...+xi)(i1)i

=xk,11i(x1,11i+x2,11i+...+xi,11i)(i1)>xk,1>0

and

v=x1,0x2,0...xi,0u(i1)=x1,11ix2,11i...xi,11i>v1=x1,1x2,1...xi,1u1(i1)>0

The reasoning is available until infinity. Then

uj=x1,j+x2,j+...+xi,j=(x1,j+11i+x2,j+11i+...+xi,j+11i)i>uj+1>0

and

xk,j=xk,j+11i(x1,j+11i+x2,j+11i+...+xi,j+11i)(i1)>xk,j+1>0

and

k=1,2,...,i

and

vj=x1,jx2,j...xi,juj(i1)=x1,j+11ix2,j+11i...xi,j+11i>vj+1=x1,j+1x2,j+1...xi,j+1uj+1(i1)>0

xk,j,vj,uj are positive j>1, k=1,2,...,i.

LEMMA 8

(P) is the expression :

xk,j=xkij(l=0l=j1x1il+x2il+...+xiil)(i1)


Proof of lemma 8

for j=1, it is verified because of the definition of xk,1,u1,v1,

we suppose (P) true and the expression of uj implies, with (P), that

xk,j=xk,j+11i(x1,j+11i+x2,j+113+...+xi,j+11i)(i1)

or

xk,j+11i=xk,j(x1,j+11i+x2,j+11i+...+xi,j+11i)(i1)

and

xk,j+1=xk,ji(x1,j+1i+x2,j+11i+...+xi,j+11i)(i1)i=xk,ji(x1,j+x2,j+...+xi,j)(i1)

=xkij+1l=0l=j1(x1il+x2il+...+xiil)i(i1)(x1ij+x2ij+...+xiij)(i1)l=0l=j1(x1il+x2il+...+xiil)(i1)2

=xkij+1l=0l=j(x1il+x2il+...+xiil)(i1)

then

xk,j=xkijl=0l=j1(x1il+x2il+...+xiil)(i1)

Why are not they solutions for n>i(i1),nk>i(i1) ?

We suppose

n=nk=i(i1)

the formula becomes

xk,j=xkij(l=0l=j1x1il+x2il+...+xiil)(i1)

=UnijXknkij(l=0l=j1(UnilX1n1il+UnilX2n2il+...+UnilXiniil))(i1)

=Ui(i1)ijXki(i1)ij(l=0l=j1(Ui(i1)ilX1i(i1)il+...+Ui(i1)ilXii(i1)il))(i1)

=U(i1)ij+1Xk(i1)ij+1(l=0l=j1(U(i1)il+1x1(i1)il+1+...+U(i1)il+1Xi(i1)il+1))(i1)

It is the expression of xk,j+1 of the exponent i-1.

If we suppose that exist solutions for the exponent i-1,

there will exist solutions for an exposant not greater than i(i-1).

Some times, we must make attention to the initial change of the data. For example, let the following equationd

kUn=Xn+Yn

for some k integers like 7, there are solutions, for others like 2, there is no solution. It is too easy toi pose

u=(kUn)2

x=kUnXn

y=kUnYn

z=XnYn

the lemma 1 is satisfied

u=kUn(Xn+Yn)=x+y

1z=k2U2nkUnXnkUnYn=uxy=1x+1y

The correct solution is to pose

u=U2n

x=UnXn

y=UnYn

z=XnYn

Like this

u=U2n=kUn(Xn+yn)k2=x+yk

and

1z=U2nUnXnUnYn=uxy=x+ykxy=1kx+1ky

Conclusion

The conclusion is that Ghanouchi's sequences, series and numbers have several applications in all diophantine equations, we saw some of them and there are many others like Pilai equation, Smarandache equation, the Catalan equation, etc...