On certain functional equations defining polynomials

Abstract

Authors

Keywords

?

Paper coordinates

T. Popoviciu, Sur certaines équations fonctionnelles definissant des polynômes, Mathematica, 10 (1935), pp. 194-208 (in French).

PDF

About this paper

Journal

Mathematica

Publisher Name
DOI
Print ISSN
Online ISSN

google scholar link

??

Paper (preprint) in HTML form

ON CERTAIN FUNCTIONAL EQUATIONS DEFINING POLYNOMIES.

By

Tiberiu Popoviciu
in Cluj.

Received on July 20, 1934.

I.

Functional equations in one variable.

  1. 1.
    • Eitherf(x)f(x)a function defined in the closed interval (has,ba,b),has<ba<b, and verifying the linear functional equation

i=0nhasif(x+ih)=0\sum_{i=0}^{n}a_{i}f(x+ih)=0 (1)

for everythingxxand for everythingh>0h>0such ashasx,x+nhba ≤ x, x + nh ≤ b, THEhasia_{i}being constants.

To the functional equation (1) we attach the characteristic equation

F(z)=has0+has1z++hasnzn=0.\mathrm{F}(z)=a_{0}+a_{1}z+\cdots+a_{n}z^{n}=0.

We can assume, without restricting the generality, that the polynomialF(z)F(z)not be rational with respect to a positive integer power ofzz.

We can assumehas00,hasn0a_{0}\neq 0,a_{n}\neq 0These conditions can be expressed by saying that equation (1) is of degree n. It is easy to see, by appropriately adding relations of the form (1), that if a functionf(x)f(x)verifies a functional equation of degreennit also satisfies an equation of degreen+mn+m, whatever the positive integermm.

We will also say that equation (1) is of orderkkwhen 1 is a root of orderkkof multiplicity of the characteristic equation.

Iff(x)f(x)verifies an equation of degreennorder numberkkit also verifies an equation of degreen+mn+mand of an order at least equal tokk, whatever the positive integermm2.
- For equation (1) to be of orderkkit is necessary and sufficient that

F(1)=F(1)==F(k1)(1)=0,F(k)(1)0.F(1)=F^{\prime}(1)=\ldots=F^{(k-1)}(1)=0,\quad F^{(k)}(1)\neq 0.

The functionxmx^{m}is a solution of equation (1) ifm=0.1,k1m=0,1,\ldots k-1, butnnit's not one formkm\geq ksince

i=0nhas1ik=[(zddz)(k)F(z)]z=1=F(k)(1)0\sum_{i=0}^{n}a_{1}i^{k}=\left[\left(z\frac{d}{dz}\right)^{(k)}\mathrm{F}(z)\right]_{z=1}=\mathrm{F}^{(k)}(1)\neq 0

Therefore,
the functional equation (1) is satisfied by any polynomial of degree k-1, but not by a polynomial whose effective degree exceedsk1k-1.

3 - Let's consider the polynomialGp(z)G_{p}(z)which is obtained by the formula

Gp(zp)=F(z)F(αz)F(αp1z)\mathrm{G}_{p}\left(z^{p}\right)=\mathrm{F}(z)\cdot\mathrm{F}(\alpha z)\ldots\mathrm{F}\left(\alpha^{p-1}z\right)

Orα\alphais a primitive root of orderppof unity.
Suppose that the polynomialF(z)F(z)verifies identity

G2(z)has0F(z), Or F(z)F(z)has0F(z2).\mathrm{G}_{2}(z)\equiv a_{0}\cdot\mathrm{\penalty 10000\ F}(z),\quad\text{ ou }\quad\mathrm{F}(z)\cdot\mathrm{F}(-z)\equiv a_{0}\cdot\mathrm{\penalty 10000\ F}\left(z^{2}\right). (2)

We then haveF(z)=has0(1z)ki=1nk(1ρ1z)F(z)=a_{0}(1-z)^{k}\prod_{i=1}^{n-k}\left(1-\rho_{1}z\right), Orρi\rho_{i}are roots, different from 1, of unity. Ifp1p_{1}is root of orderppFrom the unit, one can immediately see thatGρ(z)\mathrm{G}_{\rho}(z)cannot be identical tohas0p1.F(z)a_{0}^{p-1}.\mathrm{F}(z).

It follows that a number can always be determinedppso that we have
(3)
G.(z)has0p1.F(z)\mathrm{G}.(z)\equiv\equiv a_{0}^{p-1}.\mathrm{F}(z)
except in the case whereF(z)=has0(1z)n\mathrm{F}(z)=a_{0}(1-z)^{n}
The general form of polynomials F(z)\mathrm{F}(z)satisfying (2), or the more general relationGp(z)=has0p1.F(z)\mathrm{G}_{p}(z)=a_{0}^{p-1}.\mathrm{F}(z), can be obtained easily, but we don't need it later.
4. - Let's return to equation (1) and assume that the orderkkbe less than n. The functionf(x)f(x)also check the functional equation whose characteristic equation isGp(z)=0\mathrm{G}_{p}(z)=0the numberppbeing determined
such that we have relation (3). The function also satisfies the functional equation whose characteristic equation isGp(z)+λF(z)=0\mathrm{G}_{p}(z)+\lambda\mathrm{F}(z)=0,λ\lambdabeing a constant.

We can determine the constantλ[Gρ(k)(1)+λF(k)(1)=0]\lambda\left[G_{\rho}^{(k)}(1)+\lambda F^{(k)}(1)=0\right]in such a way that this last functional equation is of order at least equal tok+1k+1We therefore deduce that

If the functionf(x)f(x)verifies equation (11), of degreennand orderkk : it also satisfies an equation of degreennof at least equal order.has`k+1\grave{a}k+1.

The property relating to degree follows from the remark made at the end of No. 1.5.
- It follows that if the functionf(x)f(x)verifies equation (1); it also verifies a functional equation of ordernnand degreenn.

An equation (1) of degree and ordernnis of the form

i=0n(1)i(ni)f(x+ih)=0\sum_{i=0}^{n}(-1)^{i}\binom{n}{i}f(x+ih)=0

and we know that, iff(x)f(x)is assumed to be continuous, the general solution of this equation is an arbitrary polynomial of degreen1(1)n-1(1).

We therefore deduce the following final property:
The general continuous solution of equation (1) of order k is an arbitrary polynomial of degreek1k-1.

II.

Functional equations of two variables.

  1. 6.
    • Now consider a functionf(x,y)f(x,y)of two variablesxxAndyy, defined in the rectanglehasxb,cyda\leq x\leq b,c\leq y\leq dand verifying the linear functional equation

i=0nI=0mhasi,If(x+ih,y+ih)=0\sum_{i=0}^{n}\sum_{j=0}^{m}a_{i,j}f\left(x+ih,y+ih^{\prime}\right)=0 (4)

for everythingx,yx,yand for everythingh>0,h>0h>0,h^{\prime}>0such ashasx,x+nhb ; a\leq x,x+nh\leq b_{\text{\penalty 10000\thinspace; }}.cy,y+mhdc\leq y,y+mh^{\prime}\leq d, THEhasi,I,a_{i,j,}being constants.
(1) We can subject the function to more general conditions than: continuity. See in my Thesis (Mathematica t. VIII sp. p. 57.) the generalization of a theorem of MW Sierrinski.

To this equation we attach the characteristic equation

F(z,t)=i=0nI=0mhasi,Izit=I=0mtfI(z)=0FI(z)==0nhasi,Izi\begin{gathered}\mathrm{F}(z,t)=\sum_{i=0}^{n}\sum_{j=0}^{m}a_{i,j}z^{i}t^{\prime}=\sum_{j=0}^{m}t^{\prime}f_{j}(z)=0\\ \mathrm{\penalty 10000\ F}_{j}(z)=\sum_{=0}^{n}a_{i,j}z^{i}\end{gathered}

Let us first define the degree and order of the functional equation (4). We say that equation (4) is of degree(n,m)(n,m)if we have

i=0n|hasi,0|0,i=0n|hasi,m|0,I=0m|has0,I|0,I=0m|hasn,I|0.\sum_{i=0}^{n}\left|a_{i,0}\right|\neq 0,\sum_{i=0}^{n}\left|a_{i,m}\right|\neq 0,\sum_{j=0}^{m}\left|a_{0,j}\right|\neq 0,\sum_{j=0}^{m}\left|a_{n,j}\right|\neq 0.

We will say that equation (4) is of order (k,kk,k^{\prime}) if the equationliF(z,t)=0\mathrm{li}\mathrm{F}(z,t)=0admits the rootz=1z=1of orderkkof multiplicity, identically inttand the roott=1t=1of orderkk^{\prime}of multiplicity, identically inzz. The numbertktkis therefore such that we have

F(1,t)=Fz(1,t)==Fz(k1)(1,t)=0F(1,t)=F_{z}^{\prime}(1,t)=\ldots\ldots=F_{z}^{(k-1)}(1,t)=0

identically inttAnd

Fzk(k)(1,t)0\mathrm{F}_{z^{k}}^{(k)}(1,t)\neq 0

for at least one value ofttIn other wordskkis the number of times the polynomialsFf(z)F_{f}(z)they all have in common the factor (1z1-z).

Iff(x,y)f(x,y)verifies equation (4) of degree (n,mn,m) and order (k,kk,k^{\prime}) it also satisfies an equation of degree(n+n1,m+m1)\left(n+n_{1},m+m_{1}\right)and of at least equal order(k,k)()2\left(k,k^{\prime}\right)\left({}^{2}\right), regardless of whether the integers are positive or zeron1n_{1}Andm1m_{1}7.
- If the functionf(x,y)f(x,y)checks equation (4) of order (k,kk,k^{\prime}) it also satisfies the functional equation whose characteristic equation isGp(z,t)=0\mathrm{G}_{p}(z,t)=0Or

Gp(zp,t)=F(z,t)F(αz,t)F(αp1z,t)\mathrm{G}_{p}\left(z^{p},t\right)=\mathrm{F}(z,t)\cdot\mathrm{F}(\alpha z,t)\ldots\mathrm{F}\left(\alpha^{p-1}z,t\right)

"α\alphabeing still a primitive root of orderppof the unit.
The function will also satisfy the functional equation whose characteristic equation isGp(z,t)+H(t)F(z,t)=0,H(t)\mathrm{G}_{p}(z,t)+\mathrm{H}(t)\mathrm{F}(z,t)=0,\mathrm{H}(t)being a polynomial intt.

We can choose the numberppso that

Gp(z,t)(I=1mhas0,ItI)p1F(z,t)\mathrm{G}_{p}(z,t)\equiv\left(\sum_{j=1}^{m}a_{0,j}t^{j}\right)^{p-1}\mathrm{\penalty 10000\ F}(z,t)

(2) The order (k1,k1k_{1},k_{1}^{\prime}) is at least equal to the order (k,kk,k^{\prime}) ifk1k,k1kk_{1}\geq k,k_{1}\geq k^{\prime}and then
we can determine the polynomialH(t)\mathrm{H}(t)such that the new functional equation obtained is of order at least equal to (k+1,kk+1,k^{\prime}a. The degree of this equation is of the form(n,m)\left(n^{\prime},m^{\prime}\right)Ornnn^{\prime}\leq n.

We therefore deduce the following property:
If the functionf(x,y)f(x,y)verifies equation (4) of degree (n,m˙n,\dot{m}) and: of order (k,kk,k^{\prime}) it also satisfies an equation of degree (n,mn,m^{\prime}) and: of order at least equal to (k+1,kk+1,k^{\prime}8.
- We deduce thatf(x,y)f(x,y)verifies an equation of degree (n,mn,m^{\prime}) and of order at least equal to (n,kn,k^{\prime}).

Gρ(z,t)=(1z)nQ(t)Q(αt)Q(αp=1t);\mathrm{G}_{\rho}(z,t)=(1-z)^{n}\cdot Q(t)\cdot Q(\alpha t)\ldots Q\left(\alpha^{p=1}t\right);

We arrive at the property:
If the functionf(x,y)f(x,y)verifies equation (4) of degree (n,mn,m) it: also satisfies an equation of degree (n,mn,m^{\prime}) and order (n,mn,m^{\prime}).

We show in exactly the same way thatf(x,y)f(x,y)check. an equation of degree (nn',mm) and order (nn',mm).

We don't need to specify the numbers:nn^{\prime}Andmm^{\prime}9.
- An equation (4) of degree(n,m)(n,m)and order(n,m)(n,m)is of: the form

i=0nI=0m(1)i+1(ni)(mI)f(x+ih,y+Ik)=0\sum_{i=0}^{n}\sum_{j=0}^{m}(-1)^{i+1}\binom{n}{i}\binom{m}{j}f\left(x+ih,y+jk^{\prime}\right)=0

and we know that the general continuous solution of this equation is a pseudo-polynomial of order (n1,m1n-1,m-1), that is to say a function of the form

i=0n1xiHASi(y)+I=0m1yiBI(x)()3\sum_{i=0}^{n-1}x^{i}\mathrm{\penalty 10000\ A}_{i}(y)+\sum_{j=0}^{m-1}y^{i}\mathrm{\penalty 10000\ B}_{j}(x)\left({}^{3}\right)

THEHASi(y)A_{i}(y)being continuous functions ofyty\in tTHEBi(x)B_{i}(x)continuous functions ofxxalone.

It follows that the general continuum solution of the functional equation (4) is a pseudo-polynomial.
(3) See: A. Marchaud, "On the derivatives and differences of functions of real variables." See also my thesis (Mathematica, vol. VIII), p. 65.
10. - So thatxrysx^{r}y^{s}to be a solution of equation (4), it is necessary that

[β+γF(z,t)zβtγ]z=t=1=0,βr,γs.\left[\frac{\partial^{\beta+\gamma\mathrm{F}(z,t)}}{\partial z^{\beta}\partial t^{\gamma}}\right]_{z=t=1}=0,\quad\beta\leq r,\quad\gamma\leq s.

If (k,kk,k^{\prime}) is the order of the equation; we can immediately see that:xkymx^{k}y^{m}Andxnykx^{n}y^{k}cannot be solutions.

If we are looking for the condition for whichxrHAS(y)x^{r}\mathrm{\penalty 10000\ A}(y)one solution is found to be the functionΛ(y)\Lambda(y)must check the equations

I=0m[(zddz)(β)FI(z)]z=1HAS(y+Ih)=0β=0,1,2,,r\begin{gathered}\sum_{j=0}^{m}\left[\left(z\frac{d}{dz}\right)^{(\beta)}\mathrm{F}_{j}(z)\right]_{z=1}\mathrm{\penalty 10000\ A}\left(y+jh^{\prime}\right)=0\\ \beta=0,1,2,\ldots,r\end{gathered}

IfrkHAS(y)r\geq k\mathrm{\penalty 10000\ A}(y)is necessarily a polynomial of degree at most equal tom1m-1(or it is identically zero) and ifrnHAS(y)r\geq n\mathrm{\penalty 10000\ A}(y)is a polynomial of degree at most equal tok1k^{\prime}-1. Whenr=0,1,2,k1r=0,1,2,\ldots k-1the functionHAS(y)\mathrm{A}(y)can be arbitrary. Solutions of the formysB(x)y^{s}\mathrm{\penalty 10000\ B}(x)enjoy similar properties. Therefore,

The general continuous solution of the functional equation(4) of order (k,kk,k^{\prime}) is the sum of a pseudo-polynomial of order (k1,k1k-1,k^{\prime}-1) and a certain polynomial of degree at most equal to (n1,m1n-1,m-1) ( 4 ).

It easily follows from the above that the necessary and sufficient condition for equation (4) to admit only polynomials as continuous solutions is that this equation be of order (0,0).
11. - Let us determine in particular the equations of degree(n,m)(n,m)whose continuous general solution is an arbitrary polynomial of degree(n1,m1)(n-1,m-1).

It is immediately apparent that these equations have the characteristic equation

F(z,t)=(1t)mF(z)+(1z)nG(t)=0\mathrm{F}(z,t)=(1-t)^{m}\mathrm{\penalty 10000\ F}(z)+(1-z)^{n}\mathrm{G}(t)=0

OrF(z)F(z)is a polynomial of degreenninzzAndG(t)G(t)a polynomial of degreemminttThese two polynomials must satisfy the inequalitiesF(i)(1)0,i=0,1,2,,n1,G(I)(1)0,I=0,1,2,,m1(1)mm!F(n)(1)+(1)nn!G(m)(1)0\mathrm{F}^{(i)}(1)\neq 0,\quad i=0,1,2,\ldots,n-1,\mathrm{G}^{(j)}(1)\neq 0,j=0,1,2,\ldots,m-1(-1)^{m}m!\mathrm{F}^{(n)}(1)+(-1)^{n}n!\mathrm{G}^{(m)}(1)\neq 0
( 4 ) We say that a polynomial is of degree (n,mn,m) if it is of degreenninxxand degreemminyyA polynomial has a degree at most equal to (n,mn,m) if it is of degreen\leq ninxxand degreem\leq minyy.

InequalitiesF(1)0,G(1)0\mathrm{F}(1)\neq 0,\mathrm{G}(1)\neq 0express precisely that the equation is of order (0,0).

The simplest of these equations is the one with the characteristic equation

F(z,t)=12{(1t)m(1+z)n+(1z)n(1+t)m}=0F(z,t)=\frac{1}{2}\left\{(1-t)^{m}(1+z)^{n}+(1-z)^{n}(1+t)^{m}\right\}=0

It therefore follows that the general continuous solution of the functional egation

12i=0nI=0m{(1)i+(1)I}(ni)(mI)f(x+ih,y+Ih)=0\frac{1}{2}\sum_{i=0}^{n}\sum_{j=0}^{m}\left\{(-1)^{i}+(-1)^{j}\right\}\binom{n}{i}\binom{m}{j}f\left(x+ih,y+jh^{\prime}\right)=0

is any polynomial of degree (n1,m1n-1,m-1In particular ,
the general solution of the equation

f(x,y)+f(x+2h,y)f(x,y+2h)+f(x+2h,y+2h)=4f(x+h,y+h)f(x,y)+f(x+2h,y)-f\left(x,y+2h^{\prime}\right)+f\left(x+2h,y+2h^{\prime}\right)=4f^{\prime}\left(x+h,y+h^{\prime}\right)
has+bx+cy+dxya+bx+cy+dxy

has,b,c,da,b,c,dbeing constants.
We also see that the general continuous solution of the equation

i=0nL](n2i)f(x+2ih,y)=I=0[n12](m2I+1)f[x+(2I+1)h,y+h]\left.\sum_{i=0}^{\left\lfloor\frac{n}{L}\right.}\right]\binom{n}{2i}f(x+2ih,y)=\sum_{j=0}^{\left[\frac{n-1}{2}\right]}\binom{m}{2j+1}f\left[x+(2j+1)h,y+h^{\prime}\right]

is an arbitrary polynomial of degreen1n-1compared toxxalone. This result can also be obtained very simply in a direct way.

11 bis. - We can generally find equations (4) whose continuous solution is a function ofxxalone. We can see that such an equation must be of order(1.0)(1,0)at most and not admit solutions of the formxryx^{r}y12.
Let us further seek the equations (4) whose general solution is a polynomial of degreen1n-1inxxAndyyIt is only worthwhile to look for symmetric equations, that is, equations whose characteristic equation exhibits a certain symmetry with respect tozzAndttIt is easily verified that the smallest admissible degree is (n,nn,n). The
characteristic equation is then of the form

F(z,t)=i=0n(1t)i(1z)niQi(z)=0F(z,t)=\sum_{i=0}^{n}(1-t)^{i}(1-z)^{n-i}Q_{i}(z)=0

OrQi(z)Q_{i}(z)is a polynomial of degreeiiinzzThese polynomials must satisfy the inequalities

Qi(1)0,i=0,1,2,nQ_{i}(1)\neq 0,\quad i=0,1,2,\ldots n

and the symmetry conditions.Qn(1)0Q_{n}(1)\neq 0expresses precisely that equation (4) is of order (0,0).

By specializing, we obtain various equations of the form (4) whose general continuous solution is an arbitrary polynomial of degree n-1. Thus, for example

i=0n(1)i(ni)f[x+ih,y+(ni)h]=0i=0n(1)i2ni(ni){I=0i(Ii)f[x+ih,y+(iI)h]}=0i=0nI=0ni(1)i+I{r=0niI(nIri)(I+rI)}f(x+ih,y+Ih)=0\begin{gathered}\sum_{i=0}^{n}(-1)^{i}\binom{n}{i}f\left[x+ih,y+(n-i)h^{\prime}\right]=0\\ \sum_{i=0}^{n}(-1)^{i}2^{n-i}\binom{n}{i}\left\{\sum_{j=0}^{i}\binom{j}{i}f\left[x+ih,y+(i-j)h^{\prime}\right]\right\}=0\\ \sum_{i=0}^{n}\sum_{j=0}^{n-i}(-1)^{i+j}\left\{\sum_{r=0}^{n-i-j}\binom{n-j-r}{i}\binom{j+r}{j}\right\}f\left(x+ih,y+jh^{\prime}\right)=0\end{gathered}

which correspond to the cases

F(z,t)=(tz)n\displaystyle\mathrm{F}(z,t)=(t-z)^{n}
F(z,t)=[2(t+z)]n\displaystyle\mathrm{\penalty 10000\ F}(z,t)=[2-(t+z)]^{n}
F(z,t)=ı=0n(1t)i(1z)ni\displaystyle\mathrm{\penalty 10000\ F}(z,t)=\sum_{\imath=0}^{n}(1-t)^{i}(1-z)^{n-i}

Forn=2n=2we find that the general continuous solution of the equations

f(x,y+2h)+f(x+2h,y)=2f(x+h,y+h)\displaystyle f\left(x,y+2h^{\prime}\right)+f(x+2h,y)=2f\left(x+h,y+h^{\prime}\right)
f(x,y+2h)+2f(x+h,y+h)+f(x+2h,y)==4[f(x+h,y)+f(x,y+h)f(x,y)]\displaystyle\begin{aligned} f\left(x,y+2h^{\prime}\right)+2f\left(x+h,y+h^{\prime}\right)+f(x+2h,y)=\\ =4\left[f(x+h,y)+f\left(x,y+h^{\prime}\right)-f(x,y)\right]\end{aligned}
f(x,z+2h)+f(x+h,y+h)+f(x+2h,y)==3[f(x+h,y)+f(x,y+h)f(x,y)]\displaystyle\begin{aligned} f\left(x,z+2h^{\prime}\right)+f(x+h,y&\left.+h^{\prime}\right)+f(x+2h,y)=\\ &=3\left[f(x+h,y)+f\left(x,y+h^{\prime}\right)-f(x,y)\right]\end{aligned}

osthas+bx+cya+bx+cyOrhas,b,ca,b,care constants.
Among all these equations, it appears that the simplest one
corresponds to the caseF(z,t)=i=0n(1)i(1t)i(1z)niF(z,t)=\sum_{i=0}^{n}(-1)^{i}(1-t)^{i}(1-z)^{n-i}. Forn==2.3n==2,3This gives us the equations

f(x,y+2h)+f(x,y)+f(x+2h,y)=\displaystyle f\left(x,y+2h^{\prime}\right)+f(x,y)+f(x+2h,y)=
=f(x+h,y)+f(x+h,y+h)+f(x,y+h)\displaystyle\quad=f(x+h,y)+f\left(x+h,y+h^{\prime}\right)+f\left(x,y+h^{\prime}\right)
2f(x+h,y)+f(x+3h,y)+2f(x,y+2h)+f(x+h,y+2h)=\displaystyle\quad 2f(x+h,y)+f(x+3h,y)+2f\left(x,y+2h^{\prime}\right)+f\left(x+h,y+2h^{\prime}\right)=
=2f(x,y+h)+f(x,y+3h)+2f(x+2h,y)+f(x+2h,y+h)\displaystyle\quad=2f\left(x,y+h^{\prime}\right)+f\left(x,y+3h^{\prime}\right)+2f(x+2h,y)+f\left(x+2h,y+h^{\prime}\right)\ldots

III.

Sar is a problem of MD Pompeii.

  1. 13.
    • MD Pompéru set out to find the continuous functionsf(x)f(x)defined in (has,ba,b) and verifying equality ( 5 ).

1yxxyf(t)dt=f(x+y2)\frac{1}{y-x}\int_{x}^{y}f(t)dt=f\left(\frac{x+y}{2}\right)

for everythingxx, Andyyincluded in(has,b)(a,b)The solution is a first-degree polynomial.

We propose to generalize this problem.
Consider an increasing sequence of given numbers.λ0,λ1,\lambda_{0},\lambda_{1},\ldotshasλn\lambda_{n}between 0 and 1 and let's form the Lagrange polynomial

P(t)=i=0n(tt0)(tt1).(tti1)(tti+1).(ttn)(tit0)(t1t1)(titi1)(titi+1)(titn)f(ti)\mathrm{P}(t)=\sum_{i=0}^{n}\frac{\left(t-t_{0}\right)\left(t-t_{1}\right)\ldots.\left(t-t_{i-1}\right)\left(t-t_{i+1}\right)\ldots.\left(t-t_{n}\right)}{\left(t_{i}-t_{0}\right)\left(t_{1}-t_{1}\right)\ldots\left(t_{i}-t_{i-1}\right)\left(t_{i}-t_{i+1}\right)\ldots\left(t_{i}-t_{n}\right)}f\left(t_{i}\right)

by takingti=x+λi(yx)t_{i}=x+\lambda_{i}(y-x)
Let's write the equality between the average values .

1yxxyf(t)dt=1yxxyP(t)dt\frac{1}{y-x}\int_{x}^{y}f(t)dt=\frac{1}{y-x}\int_{x}^{y}\mathrm{P}(t)dt

or, after a simple transformation,

1yxxyf(t)dt=i=0nμif[x+λi(yx)]\frac{1}{y-x}\int_{x}^{y}f(t)dt=\sum_{i=0}^{n}\mu_{i}f\left[x+\lambda_{i}(y-x)\right] (5)

(5) D. Pompeto: “On a functional equation that is introduced into a mean problem” GR t. 190 p. 1107.
where

μi=01(tλ0)(tλ1)(tλi1)(tλi+1)(tλn)(λiλ0)(λIλ1)(λiλi1)(λIλI+1)(λIλn)dti=0,1,2,,n\begin{gathered}\mu_{i}=\int_{0}^{1}\frac{\left(t-\lambda_{0}\right)\left(t-\lambda_{1}\right)\ldots\left(t-\lambda_{i-1}\right)\left(t-\lambda_{i+1}\right)\ldots\left(t-\lambda_{n}\right)}{\left(\lambda_{i}-\lambda_{0}\right)\left(\lambda_{I}-\lambda_{1}\right)\ldots\left(\lambda_{i}-\lambda_{i-1}\right)\left(\lambda_{I}-\lambda_{I+1}\right)\ldots\left(\lambda_{I}-\lambda_{n}\right)}dt\\ i=0,1,2,\ldots,n\end{gathered}

Let us now consider the functional equation (5). By construction, this equation is satisfied by a polynomial of degreenn.

We will now assume that the numbersλ0,λ1,,λ\lambda_{0},\lambda_{1},\ldots,\lambdadivide the interval rationally(0.1)(0,1).

We can immediately see that the functionf(x)f(x)must verify the functional equation
i=0nμi{f(x+λiyx2)2f[x+λi(yx)]+f[x+y2+λiyx2]}=0has.\sum_{i=0}^{n}\mu_{i}\left\{f\left(x+\lambda_{i}\frac{y-x}{2}\right)-2f\left[x+\lambda_{i}(y-x)\right]+f\left[\frac{x+y}{2}+\lambda_{i}\frac{y-x}{2}\right]\right\}=0_{\mathrm{a}}.We
deduce that the general continuous solution of equation (5) is a polynomial.
14. - The solution of equation (5) is in general a polynomial of degreenn, but it can happen that it is a polynomial of higher degree. For example ifn=0n=0we can determine the constantλ0\lambda_{0}...so that the solution is a polynomial of degree one. This leads to the equation of MD Pomperu. Ifn=1n=1Equation (5) may have a polynomial of degree 2 as a solution, but then either it is not in symmetric form or, even if symmetry is respected, the constantsλ0,λ1\lambda_{0},\lambda_{1}are not rational. On the contraryn=2n=2we have the very simple equation

1yxxyf(t)dt=16{f(x)+4f(x+y2)+f(y)}\frac{1}{y-x}\int_{x}^{y}f(t)dt=\frac{1}{6}\left\{f(x)+4f\left(\frac{x+y}{2}\right)+f(y)\right\}

which has as its solution an arbitrary polynomial of degree 3.
In general, in the caseλL=in,i=0,1,2,,n\lambda_{l}=\frac{i}{n},i=0,1,2,\ldots,nequation (5) has as a continuous solution a polynomial of degreennifnnis odd and a polynomial of degreen+1n+1ifnnis even.

Using the formulas that give the coefficientsμi\mu_{i}We can easily find the conditions for equation (5) to have as its general continuous solution a polynomial of degreen+sn+s. If we pose

R(t)=(tλ0)(tλ1)(tλn)R(t)=\left(t-\lambda_{0}\right)\left(t-\lambda_{1}\right)\ldots\left(t-\lambda_{n}\right)

These conditions are written

01\displaystyle\int_{0}^{1} =0,\displaystyle=0, I=0.1,,s1\displaystyle j=1,\ldots,s-1
tR(t)dt\displaystyle t^{\prime}\mathrm{R}(t)dt 0,\displaystyle\neq 0, I=s\displaystyle j=s
  1. 15.
    • The equation of MD Pompeii admits the following generalization

1(x2x1)(y2y1)x1x2y1y2f(t,u)dtdu=f[x1+x22,y1+y22]\frac{1}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\int_{x_{1}}^{x_{2}}\int_{y_{1}}^{y_{2}}f(t,u)dtdu=f\left[\frac{x_{1}+x_{2}}{2},\frac{y_{1}+y_{2}}{2}\right] (6)

for a functionf(x,y)f(x,y)of two variablesxxAndyy. defined and continuous within the rectanglehasxb,cyda\leq x\leq b,c\leq y\leq dWe immediately verify thatf(x,y)f(x,y)satisfies the functional equation

f(x1,y1)+f(x1,y2)+(x2,y1)+f(x2,y2)=4f[x1+x22,y1+y22]f\left(x_{1},y_{1}\right)+f\left(x_{1},y_{2}\right)+\left(x_{2},y_{1}\right)+f\left(x_{2},y_{2}\right)=4f\left[\frac{x_{1}+x_{2}}{2},\frac{y_{1}+y_{2}}{2}\right]

From what we have demonstrated in Nr. 11 it follows that the general solution of equation (6) is a polynomial of degree (1,1).

Let us consider the interpolation polynomial

P(t,u)=i=0nI=0mHAS(t)B(u)HAS(ti)B(uI)(tti)(uuI)f(ti,uI)\mathrm{P}(t,u)=\sum_{i=0}^{n}\sum_{j=0}^{m}\frac{\mathrm{\penalty 10000\ A}(t)\mathrm{B}(u)}{\mathrm{A}^{\prime}\left(t_{i}\right)\mathrm{B}^{\prime}\left(u_{j}\right)\left(t-t_{i}\right)\left(u-u_{j}\right)}f\left(t_{i},u_{j}\right)

Or

HAS(t)=i=0n(tti),B(u)=I=0m(uuI)ti=x1+λi(x2x1),uI=y1+λI(y2y1)\begin{gathered}\mathrm{A}(t)=\prod_{i=0}^{n}\left(t-t_{i}\right),\quad\mathrm{B}(u)=\prod_{j=0}^{m}\left(u-u_{j}\right)\\ t_{i}=-x_{1}+\lambda_{i}\left(x_{2}-x_{1}\right),\quad u_{j}=y_{1}+\lambda_{j}^{\prime}\left(y_{2}-y_{1}\right)\end{gathered}

λ0,λ1,,λn;μ1,μ2,,μm\lambda_{0},\lambda_{1},\ldots,\lambda_{n};\mu_{1},\mu_{2},\ldots,\mu_{m}being two increasing sequences of numbers between 0 and 1.

Equality between average values
1(x2x1)(y2y1)x1x2yty2f(t,u)dtdu=1(x2x1)(y2y1)x1x2y1y2P(t,u)dtdu\frac{1}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\int_{x_{1}}^{x_{2}}\int_{y_{t}}^{y_{2}}f(t,u)dtdu=\frac{1}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\int_{x_{1}}^{x_{2}}\int_{y_{1}}^{y_{2}}\mathrm{P}(t,u)dtdu
gives us the functional equation
(7)1(x2x1)(y2y1)x1x2y1y2f(t,u)dtdu=\frac{1}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\int_{x_{1}}^{x_{2}}\int_{y_{1}}^{y_{2}}f(t,u)dtdu=

=i=0nI=0mμi,If[x1+λi(x2x1),y1+λI(y2y1)]=\sum_{i=0}^{n}\sum_{j=0}^{m}\mu_{i,j}f\left[x_{1}+\lambda_{i}\left(x_{2}-x_{1}\right),y_{1}+\lambda_{j}^{\prime}\left(y_{2}-y_{1}\right)\right]

Or

μLII=01R(t)R(λL)(tλL)dt×01S(u)S(λI)(uλI)dui=0.1,,n,I=0.1,,mR(t)=i=0n(tλL),S(u)=I=0m(uλI)\begin{gathered}\mu_{l_{j}j}=\int_{0}^{1}\frac{\mathrm{R}(t)}{\mathrm{R}^{\prime}\left(\lambda_{l}\right)\left(t-\lambda_{l}\right)}dt\times\int_{0}^{1}\frac{\mathrm{\penalty 10000\ S}(u)}{\mathrm{S}^{\prime}\left(\lambda_{j}^{\prime}\right)\left(u-\lambda_{j}^{\prime}\right)}du\\ i=0,1,\ldots,n,\quad j=0,1,\ldots,m\\ \mathrm{R}(t)=\prod_{i=0}^{n}\left(t-\lambda_{l}\right),\quad\mathrm{S}(u)=\prod_{j=0}^{m}\left(u-\lambda_{j}^{\prime}\right)\end{gathered}

Equation (7) is satisfied by a polynomial of degree(n,m)(n,m)We are still assuming that the numbersλi,λI\lambda_{i},\lambda_{j}^{\prime}are rational. The functionf(x,y)f(x,y)must satisfy the functional equation

i=0nI=0mμi,I{f{x1+λix2x12,y1+λIy2y12}+\displaystyle\sum_{i=0}^{n}\sum_{j=0}^{m}\mu_{i,j}\left\{f\left\{x_{1}+\lambda_{i}\frac{x_{2}-x_{1}}{2},y_{1}+\lambda_{j}^{\prime}\frac{y_{2}-y_{1}}{2}\right\}+\right.
+f[x1+x22+λ1x2x22,y1+λIy2y12|+\displaystyle\quad+f\left[\frac{x_{1}+x_{2}}{2}+\lambda_{1}\frac{x_{2}-x_{2}}{2},\left.y_{1}+\lambda_{j}^{\prime}\frac{y_{2}-y_{1}}{2}\right\rvert\,+\right.
+f[x1+λix2x12,y1+y22+λIy2y12]+\displaystyle+f\left[x_{1}+\lambda_{i}\frac{x_{2}-x_{1}}{2},\frac{y_{1}+y_{2}}{2}+\lambda_{j}^{\prime}\frac{y_{2}-y_{1}}{2}\right]+
+f|x1+x22+λix2x12,y1+y22+λiy2y12|\displaystyle\quad+f\left|\frac{x_{1}+x_{2}}{2}+\lambda_{i}\frac{x_{2}-x_{1}}{2},\frac{y_{1}+y_{2}}{2}+\lambda_{i}^{\prime}\frac{y_{2}-y_{1}}{2}\right|-
4f[x1+λi(x2x1),y1+λI(y2y1)]=0\displaystyle\quad-4f\left[x_{1}+\lambda_{i}\left(x_{2}-x_{1}\right),y_{1}+\lambda_{j}^{\prime}\left(y_{2}-y_{1}\right)\right]=0

This equation is of the form (4). It is of order (0,0). In general, this last property results from the fact that all quantities .

i=0nμi,I,I=0.1,,m\sum_{i=0}^{n}\mu_{i,j},\quad j=0,1,\ldots,m

and all quantities

I=0mμi,I,i=0.2,,n\sum_{j=0}^{m}\mu_{i,j},\quad i=0,2,\ldots,n

they may be zero. It can happen that, for certain particular distributions of numbersλi,λf\lambda_{i},\lambda_{f}^{\prime}The condition is expressed in another way. It can be shown that in all cases these conditions amount to the previous ones ( 6 ).

We can therefore say that the general continuous solution of equation (7) is a polynomial.
16. - The solution of equation (7) is in general of degree (n,mn,m) but it can be of a higher degree. Such is, for example, equation (6). In general, for the solution to be of degree (n+s,m+sn+s,m+s^{\prime}) it is necessary and sufficient that.

01thasR(t)dt\displaystyle\int_{0}^{1}t^{a}\mathrm{R}(t)dt =0,\displaystyle=0, α=0.1,,s1\displaystyle\alpha=1,\ldots,s-1
0,\displaystyle\neq 0, α=s\displaystyle\alpha=s
=0,\displaystyle=0, α=0.1,,s1\displaystyle\alpha=1,\ldots,s^{\prime}-1
01uhasS(u)du\displaystyle\int_{0}^{1}u^{a}\mathrm{\penalty 10000\ S}(u)du 0,\displaystyle\neq 0, α=s\displaystyle\alpha=s^{\prime}

Thus in the caseλI=in,λ/=Im\lambda_{I}=\frac{i}{n},\lambda_{/}^{\prime}=\frac{j}{m}the general solution is of degree(2n2+1.2m2+1)\left(2\left\lceil\frac{n}{2}\right\rfloor+1,2\left\lfloor\frac{m}{2}\right\rfloor+1\right).

For example, the general solution of the equation

1(x2x1)(y2y1)x1x2y1y2f(t,u)dtdu=\displaystyle\frac{1}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\int_{x_{1}}^{x_{2}}\int_{y_{1}}^{y_{2}}f(t,u)dtdu=
=136[f(x1,y1)+f(x1,y2)+f(x2,y1)+f(x2,y2)]+\displaystyle\quad=\frac{1}{36}\left[f\left(x_{1},y_{1}\right)+f\left(x_{1},y_{2}\right)+f\left(x_{2},y_{1}\right)+f\left(x_{2},y_{2}\right)\right]+
+19[f(x1+x22,y1)+f(x1+x22,y2)+f(x1,y1+y22)f+(x2,y1+y22)]+\displaystyle+\frac{1}{9}\left[f\left(\frac{x_{1}+x_{2}}{2},y_{1}\right)+f\left(\frac{x_{1}+x_{2}}{2},y_{2}\right)+f\left(x_{1},\frac{y_{1}+y_{2}}{2}\right)f+\left(x_{2},\frac{y_{1}+y_{2}}{2}\right)\right]+
+49f(x1+x22,y1+y22)\displaystyle\quad+\frac{4}{9}f\left(\frac{x_{1}+x_{2}}{2},\frac{y_{1}+y_{2}}{2}\right)

is an arbitrary polynomial of degree(3.3)(3,3).

(6) The property results from the fact that the polynomial
(1+zm)[i=0rαiz2ki]2i=0rαiz2kL,0k0<k4<<kr1<krm\displaystyle\left(1+z^{m}\right)\left[\sum_{i=0}^{r}\alpha_{i}z2k_{i}\right]-2\sum_{i=0}^{r}\alpha_{i}z2k_{l},\quad 0\leq k_{0}<k_{4}<\ldots<k_{r-1}<k_{r}\leq m

cannot be identically canceled out, without the coefficientsαi\alpha_{i}not all null.
17. - We will conclude by giving yet another extension of MD Pompetu's problem. Let us now consider the interpolation polynomial of degreenninttAnduu

Q(t,u)=i=0nI=0ni{α=0i1ttαtitαα=0I1uuαuIuαhas=i+I+1nDα(t,u)Dα(ti,uI)f(ti,ui)}Q(t,u)=\sum_{i=0}^{n}\sum_{j=0}^{n-i}\left\{\prod_{\alpha=0}^{i-1}\frac{t-t_{\alpha}}{t_{i}-t_{\alpha}}\cdot\prod_{\alpha=0}^{j-1}\frac{u-u_{\alpha}}{u_{j}-u_{\alpha}}\cdot\prod_{a=i+j+1}^{n}\frac{\mathrm{D}_{\alpha}(t,u)}{\mathrm{D}_{\alpha}\left(t_{i},u_{j}\right)}\cdot f\left(t_{i},u_{i}\right)\right\}

Or

Dhas(t,u)=|1tu1thasu01t0uhas|,α=1.2,,nti=x1+λi(x2x1),uI=y1+λI(y2y1)\begin{gathered}\mathrm{D}_{a}(t,u)=\left|\begin{array}[]{ccc}1&t&u\\ 1&t_{a}&u_{0}\\ 1&t_{0}&u_{a}\end{array}\right|,\alpha=1,2,\ldots,n\\ t_{i}=x_{1}+\lambda_{i}\left(x_{2}-x_{1}\right),\quad u_{j}=y_{1}+\lambda_{j}^{\prime}\left(y_{2}-y_{1}\right)\end{gathered}

fixed constantsλi,λI\lambda_{i},\lambda_{j}^{\prime}having the same meaning as before.

If we write the equality between the average values ​​of the function and its interpolation polynomialQ(t,u)Q(t,u)we find the functional equation
(8)2(x2x1)(y2y1)(T)f(t,u)dtdu=i=0nI=0nivi,ff[x1+λi(x2x1),y1+λI(y2y1)]\frac{2}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\iint_{(\mathrm{T})}f(t,u)dtdu=\sum_{i=0}^{n}\sum_{j=0}^{n-i}v_{i,f}f\left[x_{1}+\lambda_{i}\left(x_{2}-x_{1}\right),y_{1}+\lambda_{j}^{\prime}\left(y_{2}-y_{1}\right)\right]
the integral being extended to the triangle (T) formed by the points (1,y1\infty_{1},y_{1}),(x2,y1),(x1,y2)\left(x_{2},y_{1}\right),\left(x_{1},y_{2}\right)and the coefficientsvi,Iv_{i,j}are given by the formulas

vi,I=2t0,u0t+u1i1λhasλhastλhasλLλhas=0I1uλhasλIλhashas=i+I+1nDhas(t,u)Dhas(λi,λL)dtduv_{i,j}=2\iint_{\begin{subarray}{c}t\geq 0,u\geq 0\\ t+u\leq 1\end{subarray}}^{\frac{i-1}{\mid\lambda_{a}-\lambda_{a}}}\frac{t-\lambda_{a}}{\lambda_{l}-\lambda_{a=0}^{j-1}}\frac{u-\lambda_{a}^{\prime}}{\lambda_{j}^{\prime}-\lambda_{a}^{\prime}}\cdot\prod_{a=i+j+1}^{n}\frac{\mathrm{D}_{a}^{\prime}(t,u)}{\mathrm{D}_{a}^{\prime}\left(\lambda_{i},\lambda_{l}^{\prime}\right)}dtdu

Or

Dhas(t,u)=|1tu1λhasλ01λ0λhas|,α=1.2,,n.\mathrm{D}_{a}^{\prime}(t,u)=\left|\begin{array}[]{ccc}1&t&u\\ 1&\lambda_{a}&\lambda_{0}^{\prime}\\ 1&\lambda_{0}&\lambda_{a}^{\prime}\end{array}\right|,\alpha=1,2,\ldots,n.

Equation (8) is satisfied by a polynomial of degreenn
The constants λi,λ\lambda_{i},\lambda^{\prime}{}^{\prime}Since they are rational, we can immediately see, as in No. 15, that the function satisfies a functional equation of the type (4). We can easily conclude from this that:

The general continuous solution of equation (9) is a polynomial.
18. - Equation (8) can be verified by a polynomial of degree greater thannnFor the general solution to be a polynomial of degreen+sn+sit is necessary and sufficient that

(t0,u0tαuβ(tλ0)(tλ1)(tλi1)(uλ0)(uλ1)(uλni)dtdu\displaystyle\iint_{(t\geq 0,u\leq 0}t^{\alpha}u^{\beta}\left(t-\lambda_{0}\right)\left(t-\lambda_{1}\right)\ldots\left(t-\lambda_{i-1}\right)\left(u-\lambda_{0}^{\prime}\right)\left(u-\lambda_{1}^{\prime}\right)\ldots\left(u-\lambda_{n-i}^{\prime}\right)dtdu
=0,α+β=0,1,2,,s10,α+β=si=0.1,,n+1\displaystyle\qquad\begin{array}[]{cl}=0,&\alpha+\beta=0,1,2,\ldots,s-1\\ \neq 0,&\alpha+\beta=s\\ i=0,1,\ldots,n+1\end{array}

Finally, let's give an example. The general solution of the functional equation

2(x2x1)(y2y1)(Υ)f(t,u)dtdu=13[f(x2,y1)+f(x2,y1)+f(x1,y2)]\frac{2}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\iint_{(\Upsilon)}f(t,u)dtdu=\frac{1}{3}\left[f\left(x_{2},y_{1}\right)+f\left(x_{2},y_{1}\right)+f\left(x_{1},y_{2}\right)\right]

Easthas+bx+cy,has,b,ca+bx+cy,a,b,cbeing constants.

SUMMARY.

The general continuous solution of the equationi=0nhasif(x+ih)=0\sum_{i=0}^{n}a_{i}f(x+ih)=0is a polynomial. We are examining the equation with two variables

i=0nI=0mhasi,If(x+ih1y+Ih)=0\sum_{i=0}^{n}\sum_{j=0}^{m}a_{i,j}f\left(x+ih_{1}y+jh^{\prime}\right)=0

and we determine those whose general continuous solution is a polynomial *. In the third part, we point out equations of the form

1yxxyf(t)dt=i=0nμif[x+λi(yx)]1(x2x1)(y2y1)x1x2y1y1f(t,u)dtdu=i=0ni=0mμi,If[x1+λi(x2x1),y1+λI(y2y1)]2(x2x1)(y2y1)(T)f(t,u)dtdu=i=0nI=0ntvi,If[x1+λL(x2x1),y1+λ(y2y1)]\begin{gathered}\frac{1}{y-x}\int_{x}^{y}f(t)dt=\sum_{i=0}^{n}\mu_{i}f\left[x+\lambda_{i}(y-x)\right]\\ \frac{1}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\int_{x_{1}}^{x_{2}}\int_{y_{1}}^{y_{1}}f(t,u)dtdu=\sum_{i=0}^{n}\sum_{i=0}^{m}\mu_{i,j}f\left[x_{1}+\lambda_{i}\left(x_{2}-x_{1}\right),y_{1}+\lambda_{j}^{\prime}\left(y_{2}-y_{1}\right)\right]\\ \frac{2}{\left(x_{2}-x_{1}\right)\left(y_{2}-y_{1}\right)}\iint_{(\mathrm{T})}f(t,u)dtdu=\sum_{i=0}^{n}\sum_{j=0}^{n-t}v_{i,j}f\left[x_{1}+\lambda_{l}\left(x_{2}-x_{1}\right),y_{1}+\lambda^{\prime}\left(y_{2}-y_{1}\right)\right]\end{gathered}

(T) being the triangle formed by the points(x1,y1),(x2,y1),(x1,y2)\left(x_{1},y_{1}\right),\left(x_{2},y_{1}\right),\left(x_{1},y_{2}\right)whose general continuous solution is a polynomial.

1935

Related Posts