• 検索結果がありません。

OPERATIONAL CALCULUS FOR THE CONTINUOUS LEGENDRE TRANSFORM WITH APPLICATIONS

N/A
N/A
Protected

Academic year: 2022

シェア "OPERATIONAL CALCULUS FOR THE CONTINUOUS LEGENDRE TRANSFORM WITH APPLICATIONS"

Copied!
8
0
0

読み込み中.... (全文を見る)

全文

(1)

Internat. J. Math. & Math. Sci.

VOL. 12 NO. 2 (1989) 355-362

355

OPERATIONAL CALCULUS FOR THE CONTINUOUS LEGENDRE TRANSFORM WITH APPLICATIONS

E.Y.DEEBA

Department

of Applied MatheInatical Sciences Universityof

Houston-Downtown

Houston, Texas

77002 E.L. KOH

Department

ofMathematicsand Statistics University ofRegina

Regina, Saskatchewan, Canada,$4S 0A2

(Received

June 13, 1988 andin revised formSeptember

1988)

ABSTRACT.Thispaper developsanoperationalcalculus for thecontinuousLegendretransform introducedand studiedby

Butzer,

Stens and Wehrens

[1].

It isanextension ofthe work done by Churchill et al

[2], [31

for the discrete case.

In

particular, adifferentiation theorem and a convolution theorem are proved and the results are appliedto the solution ofsome boundary valueproblems.

KEY

WORDSAND PHRASES. Continuous Legendre Transform, Operational Calculus, Convo- lution, Boundary Value Problems.

1980AMS SUBJECT CLASSIFICATION CODES: 44A15, 33A30,

1. INTRODUCTION. For a given function

f

belonging toan appropriate function space, the continuousLegendretransformisdefined by

1

J Px(x)f(x)dx (1)

(Tf)(A)

-

where

P(x)is

the Legendre function and

A >_ -].

This transform has been introduced and studiedby

Butzer,

Stens and Wehrens

[1].

The discreteanalogofthe transformill

(1)

has been

studied by Churchill

[2]

and Churchill and Dolph

[3].

The object of this paperis to develop

anoperational calculusfor the transforin whichisuseful insolving paxtial differential equations whoseunderlyingdifferentialforlnisgiven by

D=xx (1-x)xx (2)

Ill section 2wepresentthebackgroundmaterialneeded in thesequel. Insection3, vederive theoperationalcalculus for

(1)

includingaconvolutiontheorem andatable oftransforms ofsome

functions.

In

thelastsectionweapplytheresults to solvingsomeboundary value problems.

(2)

2.

PRELIMINARIES.

Werecallbasic propertiesof the transform

(Tf)(A) (see [1])

anditnportant coutiguousrelationsthat hold for theLegendrefunction.

The Legendrefunction

P(x)

isgiven

_

by

(-)(+) -

P(x) F(-A, A+I;I;-,2 (k’) (’

2

x)’ xe(-1, 1].

P(x)

satisfies thedifferential Since

P_(x) P_(x),

itsufficies to consider the ce A -5"

equation

Dy + A(A +

1)y 0

where

D

is givenin

(2).

Fher,i satisfies tim relations

P (1) P (1)

lira.__

, (1+

)(.)

0d

lh.__,.( + .)V(.)

Thefollowingcontiguousrelations

(see [4])

will beusefulin he derivationof he calculusfor

(Tf)(A).

and

(2A + 1)xP(x) (A + l)P+(x) + AP_(x) ()

(1 x2)P’(x) -AxP:(x) + APA_(x). (4)

From

(3)

and

(4)

weobtainthe relation

(1 x2)P’(x) _A(A + 1) (PA+I(X)- P_,(x)).

2A+l (5)

The addition formula for theLegendrefunctions

(see [4])

isgivenby

F(A

m

+ 1)p,(cosa)p,(cosl3)cosm

7

(6)

P(o )p(o Z)

p(o

)-

r(A +

where

P"(-)

is the associated Legendre functionand cosy cosacos

fl +

sin

asinflcos7

vith

0

_<

a, /3

_<

r, a

+ fl <

a’, 7real. Formula

(6)

willbeusefulinderiving the convolution theorem.

Another usefulrelationinvolving the Legendrefunctions is sinrA sin

P()P.(-)d ()

,(A_ )(A + + )

A#,

A++#0.

Th

Lnd

tom

(TI)(A)

i i ieg tfom fom

Z(-1,11

into

ta

p

C0(-1,1] L(-I, 1].

For

feL(-1, I],

itwshow in

[1]

that

(Tf)(A) 0(A-I) A

and

(Tf)(A ])eC0(-1,1] L(-1,1].

Fther,it w shown that if

feL(-1,1] C(-1, 1]

and if

(Tf)(A )(+),

th

tae

iiofomigiv by

f(x) T-((Tf)(A))

4

(Tf)(A- )Px_(-x)AsinAdA.

3. BASIC OPERATIONALPROPERTIES FOR

(Tf)(A).

In this sectio, we, .el,all .’

operationalcalculus for thecontinuous Legendre transform

(Tf)(.\)

thus exl,,u,lig tl,, .!cl- obtainedbyChurchill

[2]

andChurchill andDolph

[3]

for the discretecase.

the Legendre transform ofsomefunctions.

Thefirstpropertyin this directioninvolvesthe Dasgivenin

(2).

(3)

CONTINUOUS LEGENDRE TRANSFORMS 357

Theorem 3.1. Let

f

be a function such that

(i) f(JeC(-1,1]

t3

L2(-1,1]

(ii) lim_:t,(1 x2)f(z) lim._,(1 x2)f’(x)

0and

(iii) (Tf)(A)

exists. Then

k 0,1

(T(Df))(A) -A( + 1)(Tf)(1).

P_Lg..Q_[. From

(1)

togetherwith successiveintegrationbyparts,weobtain

Px(x)Df(x)dx (T(Df))(A)

-

-2 Px(x)-x (1-x ) f(x)

dx

Px(x)(1 x)f’(x)-

-

-i+

--A(A+l)f_’aP(z)f(z)dx.

The result followsfromthe facts that

Px(1)

1,

P(1) a{+), lim__,+(1 + x)Px(x)

0and

lim__,+(1 + x)P(x) ,m

togetherwith the hypothesis

(ii).

Ts

bic operationM property reduces agivendifferentiMequation whichinvolvestheoper- ator

D

into Mgebrcone orintoadifferentiM equationwithoneless independentriable.

Remk3.1.

(a)

If,inThrem3.1,

Df D-(Df)

d

f()

satisfy thesamehypotheses, then

T((Df(x)))(A) (-1)A(A + 1)(Tf)(A),

k 1,2,....

(b)

Wenotethat

(9)

cbect into theform

(rI(l- r((Ill(l ( + P(rI(l. (o

The second operational propertyinvolves the relationslfip between the transform ofagiven function

f

and thefunction

g(x) fl f(t)dt.

Theorem3.2. If

f

isapiecewisecontinuousfunction definedon

(-1,1)

and

g(x) if_, f(t)dt

and if

(Tf)(A)

exists,then

(Tg)(A) (T/)(A + 1)- (Tf)(A- 1) 2A+l (11)

Proof. Since

D(PA(x)) -A(A + 1)PA(x),

itfollows that

(Tg)(A)

-2A(A + 1 xx (1

x

2) PA(x) g(x)dx

2A(A--+ 1)(1 x)P(z)9(x)l _ + 2A(A + 1) (1 x2)P(z)f(x)dx.

Since

P,(1)

and

g(1)

are defined,

g(-1)

0 andlimx__,_,,(1

+ x)P(x)

,i,,,___A the first

identically zero. Thus

(Tg)(A)

2\(A + I) (1 x)P(x)f(x)dx.

The contiguousrelation

(5)

will then imply that

(Tg)(A)

2A(A +

1) 2A

+

1

(4)

Equivalently,

(Tg)(A) (Tf)(A + 1)- (Tf)(A- 1) 2+1

Remark3.2. Similar difference relations to that of

(11)

can be obtained in the following situation.

(a)

Ifg(x)

xf(x)

andif

(Tf)(A)

exists, then underappropriate conditionsonf,oneobtains

(Tg)(A) (A

/

1)(Tf)(A

/

1)/ A(Tf)(A 1) 2+I (12)

This willfollow

Dy

applying the contiguousrelation

(3).

() u () L’,(- )()

i

(TI)()

xi,

th, . oi

oiion

on

f,

thecontiguousrelation

(5)

and Theorem3.2yields

(Tg)(A) (TI)(A + 2) 2(T/)(A) + (TI)(A 2)

(2A

/

1) (13)

The next operational property that wewill derive involves the inverse of thedifferential op- erator D. We define the inverse of

D,

denoted by

D-’,

by

D-’(f(x)) g(x)

if and only if

D(g(x)) f(x).

If

(Tf)(A)

is

known,

thenwewant to relate

T((D-Xf))(A)

tothe transform of

f.

If,foragivenfunction

f(x), D(g(x)) f(x),

thenonintegrating twice, weobtain

x /"__ f(a)dadt

-I-c

forsomeconstant c. If

f(x)

is in additionanevenfunctionon

(-1,1),

thenone canshowby employingacontinuityargumentthat

lim,:._,+,(1-z2)g(z) lim,._.+t(1-z)g’(x)

0. Theorem 3.1 will thenimply that

(Tf)(A) T((Dg))(A) -A(A + 1)(Tg)(A).

Equivalently,

Thus

1

(Tf)($)

(Tg)(,)

( + 1---T((Dg))()= -A(A + 1--’--’--"

T(D-’I)(A)

1

-$(A + 1)(Tf)(A).

(TI)()

We

thus This last relation implies that

D-f

isthe inverse Legendretransform of-

have

Theorem.3.3.

If

f(x)

issuch that

f(x)

isevenon

(-1,1), feL(-1,1] C(-1,1], (Tf)()

existsand (T/)(,x)

eL(a+),

then

D-I(f(x))=T-(-(Tf)(A))A(A + 1) (14)

where theinversetransform

T

-a isgiven by

(8).

Weshall finallydevelopaconvolutionproperty fortheLegendretransforin. Inparticular, we willshow

(5)

CONTINUOUS LEGENDRE TRANSFORMS 359 Theorem3.4. If

f(x)

and

g(x)

aregivenfunctionsforwhich

(Tf)(A)

and

(Tg)(A)

respectively exist, thentheirproduct

(Tf)(A)(Tg)(A)

isthe transform ofthe function

h(x) ](x),g(x)

where

h(x)

is givenby

h(cos v)=

-

f(cos)g(cosB)sinododO

where cos/ cosacost

+

sinsintcos8 with 0

<

c, t

<

r, a

+

r,

<

r and O is real. The variables a, t,astd maybe interpretedasthesidesofaspherical triangleonthe unithemisphere and istheanglebetween the sidesaandu

(see

Figure

1).

Proof. From (1),

wehave

1 tt r

P(u)9(y)dy.

Setx cosaand y cos

B.

Then

I(o)m (o)p(o)(o)mae,.

headditionformulaforgleLegendrefunction

(6)

will yield uponaningegrationwih

respecg

to from0 to

1

P (cos .)d

wherecos cos cos

+

sin sin cos

(s

figure

1).

Figure 1 Thus

(T.f)(A)(Tg)(A) f(cosa)sina

P.(cosu)g(cosfl)sinfld-rd/3d,.

In

the spherical triangle

PQR,

wehave

cos cosccos v

+

sincsinvcos8.

Usingthis relationalongwith thesinelaw and transfornation of co-ordinates, thedouble i,tegral canbewritten as:

(6)

Hence,

fo" fo" (o .)g(o 3)

i,,

(Tf)(A)(Tg)(A)

- P,(cosv)sinv

The expressionin thebracketis afunctionofvandwethenwrite

1 f(cosa)g(cosfl)sinoMadt?

(5)

This maybe interpretedas aconvolutionproductof

f

andgand

(Th(cos v))(A) (Tf)(A)(Tg)(A).

This provesTheorem3.4.

Geometrically, the expression

(15)

isthemeanvalue of

f(cosa)g(cos)

over theunit hemi- sphere

x2+y2+z

1, z

>

O. Toseethis,wenotethat the element surfaceareaisdS sinadadS.

This isclear ifweidentify the coordinate transformationin Figure1by

X COS

sintsin0 sinc cos0

Thus

(15)

reads

l

fsff(cosc,)g(cos3)dS.

h(o 1

Wewill nowevaluatetheLegendretransforn ofsome fmmtions.

1.

f(x)

constant k

k sinn

(Tf)(A)

k 0

2./’() P,,().

Thby

(2.)

whve,

o.

0,,2,...,

3.

,() log( ).

1

]_ P (x) log(

1

x)dx

(Tf)(A)

1+ 1) /_, [

d

(I

d

P(x) log(1 x)dx

sinrA 1 1

/ P

x d =. d

(o)(a+)-(+-a(a+ (1-)o(1-)a.

log(I-x)]

Observe thatD(log(1- 1. Thus

sin

A

sin A

(Tf)(A) (og 2)( + (A + 1 A=( + 1)=

4.

f(A) f dt.

Byusing and3above,weoltain

(7)

CONTINUOUS LEGBNDRE TRANSFORMS 361

1 sin

rA

(Tf)(A) +

( + ) ( + )"

5.

f(x) (1

2tx

+ x2)-1/2 .,__otnP,(x),

-1

< <

1.

From (2)

above

(Tf)()

sinr

2

(- .)( + . + )

Wefiny retook that for

A equM

toa non-negativeinteger,the results ofthis sectionyield those obtned in

[2]

and

[3].

4.

APPLICATIONS. In

this sectionwe consider mineapplications ofthe Legendretransform.

Weconsiderproblearisinginheat conductionandinpotentiM theory.

A. Heat ConductionProblem. Consideranon-homogeneousbwith extremities at x 1 mdis iulaed at these end points. Let

u(x,t)

be thetemperatureof the b at positionz at timet. Theonedimensiondheat equationwithprescribediNtidtemperatureisgiven by

0 (’ oN(’

u(x,O) g(x),

-l <z <1

wherek, pd c ephysical constt.s representing therm conductivity, density xd specific het rpective]y. Wesumethat the thermal conductivity kisgiven byk

(1 z),

being

re

comfit. he above equation reads

0

( (1-x) 0u(z,t)) pcOu

a

Ot(x’t)

(,o) ()

-1

< < .

ff

U(A,t) T(u(x,t))(A)

d

G(A) (Tu(x,O))(A),

then, by Theorem 3.1, weobtnupon the application of the transform

Thesolution isgivenby

-U(A,t) ---A(A-I-

pc

1)V(A,t)

v(, o) G().

w(,t) G(),-

Now

u(x, t)

canbe obtained byeitheremploying theinversionformula

(8)

ortheconvolution theorem.

By

employingtheinversionformula and under the assumption that

u(z, t)eC(-1,1]

t3

L2(-1,1)

and

V U(A ,t)LI(R+),

oneobtains

1

_(2_{)tp(_x)AsinAd A

(,t)

4

G(- )

On the other hand theconvolutionproperty

(Theorem 3.4)

willyield

u(cosu,) (cosa)f(cos)sindcdO

where a, /, 0are as in Figure 1 and cos cos a cos u

+

sin sinu cos0 and

f

is the inverse ransform ofe (a

-l.

Tha is, by

(8)

I() ’

,(

t-)am ,xaa.

(8)

B. Dirichlet Problem for theUnitSphere

(see[2])

Consider the problem of determining the potential

v(r, cos0)

in the interiorofaunitsphere witha prescribedpotential

f(cos0)

on r

1, 0

<

0

<

r. The Laplaceequation definingthispotentialis r r sinO

(sin vo)o

O.

Ifz cos

O,

then the equation reduces to

r(rv)rr + ((1-x2)v)

0

(,) f(), - _< _< .

if

v(, )

-,d

F()

d,,oteepiy thLegdet,fomof

(, )

d

f(),

the,upon

applying the transformto the underlying equation,weobtain

-Tj(v(,,))-

d

( + 1))v(,a)

0,

v(,) ().

,The

solution ofthisequationisgiven by

V(r, A) cr + cr

-+).

L

o,d, to

pp

th i,io fo

(s)

w ed to h

o(,)L(-,] C(-,]

d

V(,)L(,+).

ThiwiUimp that 0 d

(, ) F()

wiUipy that

, F().

Hencethe lutionisgivenby

(, ) F()

d

v(r,)

r

F(I- )r-IP(-)IsinIdl.

CKNOWLEDGNMEN. he work ofghe firg euflmr wpgiy supported by

rese gr

from the Uversiy ofHousgon-Downown. Thesecond authorwptiMlysupportedby NSERCof Cede under Grog -7184.

REFERENCES

1.

Butzer, P.L.,

R.L. Stemand M. Wehrens, Thecontinuous Legendretransform, its inverse transform and applications,Internat. J. Math and Math. Sciences, Vol.

3_z

No. 1

(1980),

47-67.

2. Churclfill,

R.V.,

The operational calculus of the Legendre transform, J. Math. Phys.,

( 94),

6- 77.

3. Churchill, R.V. and C.L. Dolph, Inverse transforms of product of Legemlre ttzusform., Proc. Amer. Math.

Soc., _5(1954),

93-100.

4. Erdelyi,

A.,

et al, HigherTranscendental Function,s, Vol. 1,hIcG,w Itill,

:.Y.,

1053.

参照

関連したドキュメント

First, is there a combinatorial significance to the fact that essentially all studied sequences listed in the EIS [5] that have the Hankel transform {1, 1, 1, 1,…} and are related

Gamma function; Beta function; Riemann-Liouville Fractional deriva- tive; Hypergeometric functions; Fox H-function; Generating functions; Mellin transform; Integral representations..

This Calculus agrees with that of the Quaternions, in three important respects: namely, 1st, that its three chief symbols, ι, κ, λ, are (as above suggested) roots of unity, as i, j,

We study tractability in the worst case setting of tensor product linear operators defined over weighted tensor product Hilbert spaces.. Tract- ability means that the minimal number

The results presented in this section aim to demonstrate the main ideas of the proof of Theorem 3.1 in a more simpler problem.. Corollaries have an

The advantage of the method is that the quantity which has here been called the argument of probability, depends in general more simply than does the probability itself on

This paper will study the relation between the coefficients of Legendre expansions of the two- dimensional function and those for the derived function, and gives the normic

Given an extension of untyped λ-calculus, what semantic property of the extension validates the call-by-value