SOLUTIONS TO PROBLEMS AND ADDITIONAL LESSONS

FOR

PAR T I A L D IFF ERE N T I ALE QUA T ION S FOR SCI E N TIS T SAN DEN GIN E E R S

STANLEY J. FARLOW

University of Maine at Orono

JOHN WILEY & SONS New York Chichester Brisbane Toronto Singapore

'1"1LL·; mF.lte:'lal Dey be ror.:::ooducf::d ::Qr' tH:~tiIlt: or ,"" t r-uc t Lona L ;lurpcsl'8 ~y prcople l;S i ng th" text.

lS3N 0 u~l OS~R7-G

?rl:·~·t~j j n t"hs ~}ni"~2(3. Sti"ll~e~ o f l~nleric.a

10 9 8 7 G 5 ~ 3 2

Le5son

TABr,c or CONTENTS

1 2 3

Introduction to Partial Differ"ntial Equations Diffus ion Type Prob.l erns (rdr'"bo~1 j c Equations) Boundar-y Conditions for Diffu,-,iOll TYjle Problems Derivation of the Heat Equation

Separation of Var'i"bles

Transforming Nonhomogeneous BC into Ilomogcnecus Ones Sol v Lng )·lore Complicated Problems ))y Sep,ll'a.t ion of Variables

4 5 6 7

8 9 10 11 12 13

'I'r-ar: s formin g Hard Equa t ion s i nt 0 EilS ier One 13

SolvinG Nonhomogen"oll:; PDE (Eigenfunction expansions) Integral Transforms (Sine aTlcl Co:oinc Transforms)

'l'hc Fcur-Lsir- Scr-i c s and 'j'ra1Ls form

2_lf

The fourIer Trans[o1'm and It" Applicat:lon to PDP The )"lplace Tr'an.'<j'orm

Duhamel's Principle

The Conv" c t i OIl Term (ll ) .i n D i f I'us ion x

'I'hc OIl" Dim"'Tls i ona I Wave f,qua t i on (HypedJO 1 ic equations)

1 S

1&

1'/

The Dr A.l ember't ',oJ lit ion of the Wav(, Cqu,]": ion More OIl tlie Dr A 1 er.lDRrt So l.uti on

1. ~1

Boundapy Cond i t ions P.s",or.j_,rteo with til'" ,lave E'lu"tion

The rj n I te VIbl'at ill!_!. String (Standing \iQV8S) '[he VIk'at i.ng Boa:1L (F'our-th O1'd" 1"' PDE)

22

Dlm(~nsiunle~::~ Problcmd

2J

Cla''ic,lf icat ion ~)f' "'Dr: (Canonical Form of Uw HYf-"e"'DOl ic EqUill ion)

The Wave Equat i.on in Two and Thr-ee ;J; mensi orrs (Free Spilce)

24

Tr_e f'ini te ':"'ouri.el' T:r:ans:fD:nn~~ (Sino and Cos i.ri2 Tr'Clnsfol'TIL'-' )

iii

1 2

6 7

11 12

11·1

17 18 20

73 2lf

26 "27

28 29

3D

33

3S

37

188s0n

26 27 28

Superposition (Thc Backbone of Linear Systoms) Fir''>t OI'uer equations (M",thod of Char-ac t cr-Ls't i c s ) Nonlinear' First Order Equdtions (Conserviltion Equations)

Sy,.,tems of PDC

The Vibrating Dr'ulnbead (Wave Equation in Po l.arCoor-d Lna t s s )

The Laplacian (un intuitive deFcriptionl GCllrcval Nature of Boundary Value Problems Interiot' Dirichlet Problem for a Circle The Dirichlet Problem in an Annulus

29

30

31 37 33

35

Laplace's Equation in Spherical Coordinates (Bphcr-Lca l Har-mon i.cs )

i\ Nonhomogeneous Dir'iehlet Prob l.ern (Green's hmctions) Numcr-Lca I So Lirt io"s (Ellipt ic PT'ob18m.~)

An Exp.l icit Finite DlfferoncQ Method

lin Impl i c it I'Lni te DifferRnce c!ethod (CT'ank-N Lc oLs ot: ~'Jetllod )

Ana t yti,c vs Nurno rIc Solutions

3fi 37 31l

3'l

40

III Cl<lssific.:J'I'ien of rDE (Par-abo l ic and F.lli_p t ic Equa t Lon.s )

42

)·fonte Carle '-!ethous (lin Irrt r-oduct Lon )

43 Monte Carlo Solut.lon.s of Pa r-ti 31 Dif feo!"ent j 0.1 Squat ions

)15

Tho Calculus of Variations (r;ul"l'-Lagrangc Equu t Lous ) Variational ~Bthod8 for Solving PDE (Method of RlLz) Perturbation Mothod foT' Solving PD~

Con[0T'ma1 Mappin8 Solut~on of PD~

46 47

iv

Pi'lge

41 43

44

47 48

51 52

56

S7

61.

62

63

67

Related Topics to !\ccompany Text

Page

!\ B

Control Pr-ob Lerns ill PDt: (Control Theor-y ) Potent La I Tlwory (Rcwr-I t ing BVPe; as Int_egral

Equat ions) -

Integral t~uationG

69

c

78 Gq

Solutions to Le~sons A, B, C

97

v

Lesson 1

1. a ) Th is is a ,; econd 01'08 r , 1 i near, nonhomogc neous , parahol i c yde with nons t an t coefficients and hiD independent variah18s.

b) ':'bis if. il second 01'001', linear, nonhomogeneous parabolic pde with constant coefficients with two independent varian] c s,

c) This is 11 second ordc~, linear, nonhomogeneous, hyperboJic pde wi til com" t arrt coe fficicnts with two independent

v iU' i ab Ie s .

d) Th is i" a fourth order, nonlinear pde in two indcpendent variables (nonlincar "d" aI'" g"ncrally not classifiRd uS bcing par abo I ic , hyperbolic, or elliptic or whethel' they ell'e Ilotnogenpous 0-1:' have constant coefficients).

2. One of the int('t'C'sT ing things about PDE as cont r-as t e d to OJE .i.0 t.he t on" gcnor-a l Lv doosn t t find all the so Lut Lono to J. PDE li\e one does in DUE:, To find all the s o Lution s to "t = uxx ;re-

Cjuir0,3 a digL'C' "sion Ln t o corop l e x var i ab Le s . Ho',,'(-lVfCl' if one

ax t bt

t r i e s -;:0 find so Luti ons of the For-m Q one ",e",3 that

b = a 2. Th'-1t is, ally exponential of th_.' form:

2 ilx tat e

i" "$o.llltiot). OnL' may cry var-Ious values of ,I, especially complex nUDhcl's li.ke i to at'rive at int81'esting solutions. AftC'_r that t ty adding these so Lu tions together to wet still mo::::.'e s o Lut l ons .

3. The answel' is yes i" G '" 0 and it is veY·y easy to prove.

II. U X

o implies 1.1 (x,y) is allY function of y alone. Try

,-oom" Ii kc ut x .y ) = si.n Cv}, u(x,y) = c.

5 .

u ; 0 implies u(x,y) xy

plu3 any function of y.

u(x,y) ~ cos(y).

is lhe ~um of any function of x

Tha-'.: is, like

2 u(x,y) = sin(x) + y ,

1

?

Lesson 2

1. By using a little intuition Olle may guess at 11 ternpel'atut'e;

2

\l(x,t) = c-a t 8in(.'lx).

Ii: it; Cilc<y enough to vc r Lfv that this function SdC sfie.~ the heat equation 2nd ths giv"n Be.

2. It should be clear from OIlPS i~tuition that this temperature u( x , r ) wi}l "vcHlt\l<Jlly ar-r i v e at some temper<Jture t;lil': won t t change in time (at least as t ~ 0). Hence if we 58t ut = 0

we iirri VC, at an oreillary d5.ff('ren'lial equat ion w ir h boundary cond i tions in the steady "tate t~'mperature uh);

1 2 (j,

u(D)

o

u(l)

1

which h~B t~e solution:

u( x ) = - -l [x2 ~ x ] + Y. •

2:/

~. Sotiinl\ ut ~ 0 '.-if' ar-r Lve LIt the DC)unda,'y value prob Lern for the st"J.cly S"_:<1t'2 U( x ) :

o

u(o) ~

uO)

1

wh i ch In" '1:11" '-:0] uri on :

(z) + dnh.,l;;;1 (J-x)J.

Ho-t" ~_}ld L r_eat wi}l br., f.lowinG iIJ a L ,-,aeh end of the' rod and

uut +hc .1 atp.r~l :=,i dos , fiowcvf..:r n.t ct:J.ch ~}QJ.nt along +ho "['oct ~_r~(1

ne t: t'lo'"' in and out ~i' Z('l'O. The "PeH">!' ·,h0111d !;T''-'ph th~~, function for vdr_!_"q.C, valuer; of ,. ctr;d (-.;.

3

Diagmm of S'teady Stilt a Hear Flow

t] •

2

(J. u xx

o < x < 1

o <: t <: DO

BCs

u(o,t) uO,t)

o 10

o <: t <: co

Ie

u(x,O)

sir.(3I1x)

O<x<l.

Noh, that tnp. PD~ .i.e; de f i ne d only jn:=:ine the rp"ion 0 < X <: 1,
0 < t <: oa ~311lr.e the' Uf:'t'i.v("Jti vp.~-) \1t and 11 Qonrt make
xl'.
sense. on tiw h01)"'\ ,wi f' , •. [Il,w the ncs (:lI'O defiJ:('d for t gT'eater than "era or eJ ,·,e tb' BCs and Ie ·,mulct ~:..,::('!2_ '·'l',.,cit'y "the tpmpppai:u"," ilt th'" po l nt s (t ~ 0, x ~ 0) and

(t = 0, x = 1).

4

Lesson 3

l.

I

-.

I I I I

"2.00

Note that the above -';;emperature profiles are dl'a,,'r. so that tho BCs 0:':' the p!:'oble:n are satisfied for all time.

2.

:loat f:::_o'''s into 'tile rod i.ltl:he right hand side at; il corrs+arrt l''-'te whi.l.el:ho len hand side is kept at z er-o degJ'ees (zer-o can stand i'm' any tempGl'atuJ'e). Keeping the le~t hand of the rod at zer-o actually means it '",ill je a heat sink (heat flows into t"'""t point.

5

3.

1

Both ends are illSU1.ill:",_cl and so -the total heat energy in'"icle the rod will pventually b"""Hle distributod uniformly.

PDe )
'I. "t " (j u 0 <: x <: 1 0 < t <: m
xx
'.](O,t) ~ SO
BCs 0 < t <
u (l,t) h luC:,t) 30]
k -
x
Ie u(x,o) :::'0 0 < x < 1 1. each term has wlits dCE/sec. Tho coefficient B is a rate constant for' the flow of ho ar acr-os s the later',.,l boundary arid has un i t s sec-1

:~. Same here--eacn te:rm 11dS units deg/ sec.

3. The r c is not much change in t h i s derivation f r-orn "the ,,1. ruat i on \4h0r0 k is il constant.

u . TLe de"i vat i on is similar to the one in the Lc s son .

7

Lesson ~

1. Straigbtforwcwd sub s t i t u t ion.

2. Straightforward substitution.

3,

1

E A s i nf nnx ) n

.imp Lie s

A n

1

2 J s Lnf mrx ) dx

°

[0 n

n~r n == l~3.,5~ 4 ••

Tbe>refore ]_

4 [sin(7Ix) 'F 13 sin(371X) + -51 sinUi'lIx) ". J. 'II

U 1

x

II SiIle<o the 1." u(x,O) = c_ is expaJlded as

u(X,O) = ~ ls i.nf nx ) -I' -31 s,in(:1nx) + ". .l 7T

the sc Iut: Lon i '" :

4 _.( II)? t 2

',,(x,t) = [c, si.nt rrx ) + -31 ",-On) t sil1(31!x) + ". J.

)1

u(x,t)

-(21i)2t

e' sin(?~x)

J -(4','1)2t

+ '3 e ,-;in ('f;rx)

1 _(IiJT)2t

+ '".5 C~ si_r;U)'I~"X)~

~I

6.

u(x,O) ~

2 x - x

as a sine

A n

s i.n f nn x )

1

A ~ 7 J (x- x2) s Lnt nnx ) dx ,

n

One can €.1. t he» _i.ntr.:~g['(lt;

o

thi.s ir.tp.:.e_;ral by parts

or else ~ubstitute

y ~ nrx

and gel. <'I

iWW i.l~tegral wh i ch OClCO can look up f r-orn t he t ab Le s:

J y dnCy) dy

J / s Ln ly) ely .

In O::_li.thC2l.' C.:=!.S8:

A n

1;[1 - cos(n:r)] (n'l'}3

2 _ 8 ['.'inCTfx) sin(3ux) + s.inC'i~x) + ••.

x - x -:3 1 +.- 27 125

71

and hence t he sollltioll ie,:

9

Le',;::-,on :3

u(x,c) ~ A(t) x + B(t) (l-x) + U(x,t)

S(x,t)

r

stccady stJ.te

t

transient

and suhsti tate S( x , t ) into the ncs to find !\( r ) and 8( L) . Doing this giv<eu

nc t ) 1

.fI.(t) 2/0.-11)

Henes thc solution u(x,t) is

u I x ,t)

2x/(1-h) t (I-x) t U(x,t)

1 - (lth/l-h)x t Uex,t)

Sc.bst;_ t--.lting u( x , ~) .in to t;,e TiWP we gei: tr_e tra;-]sforned problc~ tor U(x,~)

PlJE 1\ ~ o.? r:
xx
U(O'"') 0
3Cs
uel,t) o
IC U(x,O) 8in(211x) t(l+h/l-h)x - 1 w'lieh has the so Lu t i.on (by sepa:~ation of variables) t. ¢(x) -(um)"t .

Lac 8l~(nrrx)

n

n=L

U(x,"':) ~

Hellee tLe s01utior. of the o-riginal preblem is

u(x,t) = 1 - (l+h/l-h)x + U(x,t)

2. The steady state is obvious ly S (x , t ) ~ X and so we 'ATi t.e

u( x , t) = x + u( x , -::;) . Substituting this into the problem gives

PJE Ut ~ U
xx
U(O,c) " 0
DCs U(I,t) ~ 0
U(x,O) ~
Ie " x - x 10

Th i.s ;p, the ';3",Q TIl'!P as Pr-ob Lorn [j, Lesson -5 (exeep't that the IC have opposite signs wh i.ch means t:1H"l the two answers have oppo,;i',,, signs). III n;ilt le"c;cn we sc.lvod the pl'oblem by se)aratiQTl of v ariah Le s , Using t.ha I: l)P'Slllt"~

.incl ~icnce ~

steady stat"

tl:1ansicnt

3.

'1'r'y u{x,t) ~ S(x,l:) of- U(x,t) S(x,l) = ACt) x of- 8(t) (1-x).

BC ~'" Wc'c'

"VII) ,,=l "t'c

Sllb~t i_ t u t Lng

sex,,)

into the

I\(tl - B(d _. 0
[I\( t ) - B(t)] of- h 1\(1:) 1
f)olv1.~£: for lI.(t), B(T) gl vo s .
ACt) 3(t) = lih
S( x , 1:) x 1 (l - x ) 1/'1
and hence = h + ~
h 11

Lesson 7

a n

:' I x :',in(nllx/l) 'h . o

(summed over ,he odd Lnt.eger-s )

x(o) G2 o
xr ( ~) ~J IA· "06/\ 0 -.. coS /i = 0

x (x ) = s;n(nl[x/2) Il

n = J,3,

3.

x (x) co ('()dr(~x) 11

11 .:.: 0,.1/2, +. ~

\l(x,t)

I

n=O

a II

l/2 1

2 f ~{cus(r>1!x) dx o

"

(l11T)~

11

1,2, ....

TlEll .i s :

2

12

1. Sub,stituting u(x,t) IBVP gives:

(){-t/2 )/2

<' ,,,(x,',) .i n to tho or-Lg Lna I

PDS

,; w
1: xX
w( O,t) °
,,,(l,t) D
'HCx,o) 1 Ie

',;hic;h has tllp solution (b8 S8paT'at ion 0 f VCll'iQbl".s);

w{x,t)

'i

11 "1

-, C n 2,4,
a 2 r giu(n1\x} dx t;
n !
0 ~
n 1,3, ... J)811CP'

'f c -h-t/2}/2 q(x,t) ~

'"

1\

L

n=l

[1/(2n-1)J .sin[(2n-l)nx] .

2. Let uex, t ) 8ivp,;:

x + U( x , t ) and by subs t i t ut; i or: i rrt o the THVP

PDt.: U = U - \J
t xx
ur o ,T) 0
llCs
U(l,t) 0
Ie U(x,O) -x NOH .2.0 l U(x, t ) -t w(x,t)
c 1:0 gC't:
PDE w '"
c xx
BCs N(O,t) °
w(l,t) 0
Ie .,,( x,O) -K
Wl1.ich can be solved hy g8paratioTl of variables: ,~ II

Ll n

-} J x si~(~~x) (j% 'J

n

1,), ".

ru:

l:(x,L)

.. j'

'2 L: .

11

,;in(3n:) - .•. l .

Or.c could (...:.~y.::-(~L -:::hi:=:; (~n:-=J·d(::~:--' tu s~p. .1:- .1 t :3~ti:-lf 1.R~:~ a 1.1 fn1 . .::.r'

COIld 1 L ~on" (the "PDr:, t'A'" llC "TId ''"Ve Ie). 'few n.aj or- c; f F;,,,Ui. ty

wl t h t.h l s ~(_'l-" i U':"!, l~'; t:L1 t i._ l.: ccn\lp.:::',f~/_).s v~:,ry slcT/,Tly ~

I '.-,C x , '-,_-) ~ - (;;) 2 t: . (Ir' 'J ' Lvi +,' , 1

w icr= ........ t~~ S 1.. n . Y,). : f~l~CP. n y s.o vt. Jig l n l.S Fll(_) [") _':~lr_

dLC(:~c.tly hV ~e.pL.!.r~atiun of v.:.:;.:ci dbl'2:~ one shou Ld ac!::'..!'_vC' at:

l,e.'ison 9

x (x) = 5 in ( nnx ) n

ar.d JO: {:
n 3
f (·t)
n i- 3
rl To finu

u(x,t)

I:

n=l

'!' (l) s i nt n x ) n

'P' 2
'} l- T Tl (1 ~~
.. Tl eli -71 t
- o
T 1 (0) _. 1
T' 1- ( 3'11)) T3 .1. ~.,. ?
3 .. '~3 (f ) 1 [ 1 p -C,·,I) l'3l
--------:;- -
T3(O) = 0 On)
iT' -I' (n71) 2 'r 0
'[ n n .... T (Q) Q :2 ~)I ~ S ...
:1 ...
IT (0) ~ 0 !I
, n 2 0

Hence u(x,t) ~ (.,-11 t s i ntnx ) t __ l [1 _ e-(3Ir)-"J ,.,'-n(3rr}1).

13'1)2

The tempcratuI'u profiles .J,oak Jib, e<i:>(llld w~wn ,~(1 and gr'dd'HJ] y Ioojc mor-e and more lii(':

1 . (J )

--"2 s i.n ,'1IX

(3n)

1

2 [l

71

2 -11 '1-

e 'J

2. Heve

and so

ll(X,t)

15

3. liel'c

1

?

'," + (no;!') ,

Il '"'

c

II ::

T(OJ~'l

11 L"L

j_

\'

L

Tl=1

,'1at l~'

(~ n ~ "2 ~ / I ~ ...
a
n 1, J .s ,
n
nIT :2 -jr +

2

\ [j_ ~ ,;-e 1-[

'I' (L:) ~ 4-

n ::-i':r

ri

][ene0'

u(x,t)

4- 2 1 2
-IT -r [1 -'j., t]l _"irL(lrx)
J_ c: + - ---
l·11' 2
'II" :2
'f I 1 ,,-(rm) t s i nt n nx )
'1- -J!
:1~3 n I:.

Her'R t he e Lgenf'unc t ion,,, a.t'c X (x) sinO. x) Wh(-~IS_' A ":.lr'C"
Tl Il n
th., roo t s of tan(J.,} t J., - O. Hence ~'1 ( I.) 1. , f (t:) J
n n =

'-1{::,t) =

-J. Let ,_l(x,t) = S(x,t) + U(x,t) wher"

S(x,t) = M the + B(t) (1- x). Sub"Lj_tlJ_l_;_r,~ S(x,t) Lnt o "::lw

se gives B(t) ~ 0, ACt) = cosCt) and Bu

uC x , t) ~ x cos C -;c) + LJ( x,t) . Stcbsti-;CllTing :hh .i nt.o the ovigina]_ prohlem gives.

16

I'DC U ~ U _,_ x 'line t )
t xx
:J(O,t) 0
Bes
lJ( i, t) 0
Ie U(x,O) 0 cmJ cxpand i.r-g t ho rip;!":t Ctc~;-].j roi.de

f (tl

n

l_

'2 J x ",ide) Gin(nl'ITx) C:X

()

n

1, '2,3, ".

U(x,t)

I

n=l

[\'~I' \ n11/ Tn

r CO) = n n

Solviug this gives:

ilnd thus:

T Ctl

Il

C __ ,)n _2_ II _(ntr}2( l-,) ()

n'l! '2 Gin T d T

o

u(x,t)

x cos Ct ) t

I

[1"1

';' (t ) SiCl(nlTX) • n

17

2, Hu.l t: '-ply 2iJell siue of the e qua ciOD by ,he bteU'd lion

JZ_, )t: -u(,) r- ,,! ',-- -

und l:~cn I n i..:(-lgrd.te t}JD cquat Lon ,

3 r Tak i TIl="; t he co ~:~ i ne trt.::lIi_S rov:n ("t he ~~ inc t r-a n s +o rm dOi~3 n t ~

wor-k :;"0'(' tilis pI'oblcm '::'ccausc of ,11" de;~lv2.ti.ve SC) cf tiw -/Bi/~' '"'0 have

dU dt

o:2l-3. U (G ,c) - «/U(I:)J Ii X .

u{al

2

J H{l-x) cos (IDX) d x II 0

-,
? I "as (.OJx) dx
11 0
? sin OJ
TIW He~ce We must solve

dtJ + u2(21) " G dt

U (0) " 2 sir.fll/'rlt.:

which hQS the solution

2 O(t) "

IT(.....\

s i rm

and so -::he Rolu,iOD is

2 u I x jt) = 11'

One must; E' i t her- .:.'i.nd t11e inversE' cosl ne -1:I'd"',.s:'or'lIL oc: 1)( I;) from t-he t2h10s or 21.s2 evaluate -the integr,,1 foX' u(:-:,t) numerically an a compute::",

18

a L

Q

Il

f(x) -r- ~:' r",ld'ix/L) - 1- Sil,(~lTX!T,) + ~ dn(37Tx/1..) - ... [.

Jr:fQl' t.unur o Iy t hi S cqua cion doe an ' t ffiL!:Ze Llny ;'('L"'" be cauae

f'(x) (",,_;_c:g t he ticrj_vill,iv'J of c;;C' 821'.,tOG"Lh c.my,,) io Z!2l"U ex;;e_nt at u,'" _i ump po i.rrt.s ",h2.::'8 -the 1,-i_e-n"t \la-:.Id s l.de (th", cc s Lne

5c~11(~s) ·W i.L.' not cor..v~~T"f,E: "to any~~:r,.i.l1g. ILl ut;hf2r lilorC.~ it is

p(Js~.;inlp. "(0 di~ fC:::'~t:.:~tiL!tc un in£ Ln.i h~ s~.~~,:-,ir:::=; wh ic~) tlcpresents -?~ fU~j(._! tIQL t p'pJf~ ~ ~l "t~~T':1l ;:] nd ~e t ':..i. np.v-.r 0(-31"'_; c:::: tr.at dc(~ ~H~ t l: l'('-""('''Jc;nt til" c"-'l'.i.vat :'.<n' J" (x ) .

2. a)
en
Ii n
0 :l. 2 3 If S ~ f7 g
b)
en 1,9

c)

.S

'_--L-~----_L---,.r-~.r-~.r---_'. n

o 'l. 3 $' 5 ,

4.

s. r( f,)

20

Le~oon 12

I. Straightforward integra~ion.

dll ? 2

d l - -C( r; lJ

1 -((/2}')

--p

-,!'f

u(o}

u(X,l)

1)lx)

~~i. nL'.(~ U(X) n:UClT I,,, b~)~ .. Hld0d (nr C!2_s.::...~ u(x,j~ ) h'LII !:>e
unbornd« d .:....:..3 !;;I "l ltJC liav~ -1 ~ - 0 a:-trl ~::.o
1 U(;;)

s:'_n(x) ---2

2. + (;/,

s'.dx)

,

-('J..c

"

dl! _ sU _ C

Jx;!

ll( 0)

1.

-'-2-~

" +::_

~.,hich I~af] tlu.J 801ut. i on:

\.'(x,s}

1 - xl;;

:;;---- "

,; + 1

2?

1'uK .l)lg the? Inver-so Lapli.lce transfol'ln (US.l'lg l:hi' convo lul ion pr'lnci.j)le) ;

.,

( ) • () 0(.:<: "_,-a~:<:/Lft

l\ x,t ~ :0111 t 01(: r,'

~) ,IlrI'-~

?J

1. Proof i~~ cxac t ly L i_];.:e the one in t he lB0:=:on cxccp t one lec_ves out tlF~ Gtep 0:' mu.Lt ipLvi.ng an". dividing by s ,

'2. 1 t. is clear:: h'l'. WhRTI '..I'" rl i f rcr'l'''L:atc t'w ~"l'!.,:'c, t er-m ~ly term 'di t.h C'."" peel lu t 'd" g"": u soCl'iee HhCJ,~e t0rm', 1',', I: 1 il.rp;pY' <l:ld ] dr'f,eC'

J. j;uhc:u"][(~l' :.-; "PY'j_rH:jp1.f~ sc~ys I r,\t(x,t) .:_s I:hcL -rP~~pow~P. due: 1~{_1 "Rr u{l,-::) ~ O{e)]'

t.

ut x , t ) ~ I HeX, t-T) [(T) de o

1,1(.:', T ) car. Of.) n

d;,pl'oximdtd hy some numcrLca l r ul e (Si'iY the trapezoidal r-u l c ) . Tr_d"L 1;;:

1: n

u( . 5, 1:1)

f

o

I -1) f(T) d: n

L

1 r,,- r-(I) '1- ';>'d .f(t) + '" + :!1-' f(-c )

'] n 1 rr-J, '2 ? n-1

+ H f(t }].

1 H

t

u\x,L) "J W(X,'>T) cos(T} dt o

w(X,t) -- X +

Lesson IS

1. The t~aIlsformati.(lrl t

u«(,o)

wbid ha.", ~()lution \l( ~, r ) = ,,-;: "in( r,) . i!"Lcot/,e' solution 1:0 our pr·oblem 1S!

u(x,l:) " p-t sin(x - 2t:)

L.

"Tfl8 t r-ans f o rma t i.on ll(X,t)" ,,(X-I:) "dx,t) p:iv"s t ue ne'd p roo Lem:

v1 = W

t )I.X

·",(x,O )

s inf x )

which I~~ the sDlution:

w(x,t) = c-t ~in(x) .

2 "'.-( x-2t) 3. u( x , t ) "

4. r:vC'.ludt" the integral in t t»- hint awl. subs t i tuto E.

I ~ t. One gets:

x - vt ,

u(x,t)

2S

Lesson 16

1. D.:.f±'eren [iat", the Y.'-rst Acua-:: ion of (Hi.:O) '~'i rh respect to 't ~nd the sAconJ witb rO"',,,'cL to x . ~1u.lti.1)~y; ng tile first by L a ncl ~"Ltnlctin~ :::;1 vo s :

v - ~ i

xx x

.. v

subst I tu·~e In+c tilt above GOUc,tiGIi. ':':'is is lVP.:-:'

result:

x

t:l'_' de',.; i r-xl

v - L:~ v + ()i.e I LG) v + Rr:;" o.

XX t t r

The wave i n it i a I ~y looks cue cO the frLct ion ,prm

~jke

-ll •

l'

3. Tilis pr cb Lern df!:<crih8S the mo tim, 0" a v ibr-at i np; ~tl'ini', (inihally at res':) tha t is being givc)n c;lmpl'O harm()Tl~c motior: at the r i ght hand :old",. The mor ion b '"0]110'; cat cd nul One can imag I no 1:11'0 '""V8:: mov Lng elm-,rr, the sering [1'0:11 ri p;ht' to lcft··_ DOllIleiIl1': of" -::hA Lcft hend 'end (H'nich is f ix"r' 'It Zl:!:'o) J.nc:', coming bd<;k (only to collicl'2 ',".''-~:'l '",ilVR'; com i nf, from ttw other direct iml). One ca'l so'. ve --:hb prob1.p.m analy t:_lcally 'lY u i tlwr ,he Lap.l ace ~ril.nsfur", ur Duknr"d'" pri"'2i.plA.

4.

Substi.tuting

u{x,t) ~

ax+b t "

II XX

one hilS

and

u(x,d

af."1~ solu Lions for any cons t an t 3.. The -:Jll!n is aLso a soloti.on and so we have more !-;o :.ution:::::

u(x,t) " c1

Later we Hill seA tnClt ,11" ",avo e qua t Lon ha.s more 80111tiom~ than this.

26

Lesson 17

1. Straightforwarrl subs I, i t ut; ion. Tho r"ader should us,, the ge ner-a l formula for d ;,fJ'eront La t i on '..IhQr<"_ the l i.m' t5 ar-e vilI'_Llble.s :

db I x ) [L( )'] -dx-g .i x

3.

u(x,t)

?

1 -(x-t) .

;:; [e

Apply eyuat i.o-:', (1.8.8) bthe 1:8x-[ with f( x )

27

2

xe-x ~(x)=-:J.

2. '["ri'':'" lleul Tt.,raVd (x :> J) dna Li"i~ ir.F)gi.:l.JL"'\' hT'lVP. (;-:: < J) j..,,~"il] ca.n-

CI7~ ~ p ... ,c.:h 0t~1p.r (_:.8 dJE-_'Y I--\q~.s t~J-J""GQ.gh f.:'-3L:.h otln=:'{' ~ TlH-=:- :::~(~~ l

[-;-~t(..~Y.'·/rc.r:'.lt j on 5.8 tha 1: "thp. iAla·ole· h i tt::.-::g t:lP. L01..!JJd.-:l"r~y e t x -: G T.<liJi. rp_.I-.1 ec.t .::ind ~5nce~. wi. l h L:":"jP, j-n('.oulng "Y4"~<VC r

J.

4.

dx,':)

[~(x+ct) + fest-x)]

r

x < CL

D

o

1

Lesson 19

The reader can draw his./helt' own f';'-'Llpl"ls for -n, is 1 ".sson. The on.lv thing to check is +h.rt alI. the boundary corid.it Lonc of th" f,rdphsagree .,,,i t h the prob.Lem and t ha t the s t r Lng .should hijV'" a ,]owm.rarc. force ltJhp.IleV(-~~ tho concav i. ty i s ne~a t i v e ,

29

!.e~SOIl 20

Oo:;erve that thc' sec"nd te-rm vi bra \.1''' 't hr-cc tim,,:; filstel' than tbe fi,~~.:;t ter.m. The ~olut.i(Jn Is pC.:::1iodic ~.J~ i.h pE!ri.od

( f unrla~lcTlt"'l fn,,'! uency) '[' ~ 2J"

[_I:_] sin(3Irx/;,) :;in(~lwljT,) 3TIN

3,

POl!' A > 0 if We' sub s ti t ut e X( x ) in to L he Be

xo» = X(1) ~ 0 "iP lnvc-! X(x) ic1 ;_,l,,"tically zo r-o , not'" that the ,",olutloTl 'j'( L) if', no l pe r-Lod i c in tim" A > 0 \;hich nus t b" t:le Cil8C in our problem.

Also if

a n

i(-llI>+l 211 '2 (n.'!)

o

TJ = 1,3,5, .. ,

5,

n 2 ~ 11 , G ~ •••

h n

all

" .

Hence u( x , t )

4~ ~,in(lTx) c:odlrt) - ~h2 8i,,(3'IIX) "in(3'/Tt)

TI 9"

4h

+ -,-'2- si:1( SeX) "in( Slit) + ... •

16-,'1

N,,1;e e'at the h i ghcr- f""qdcncie,,; ar'" mu LtLp Le s of t.ho fundamen tal Fr-o qucncy . Th i rs Hill give some i nd Lca lion of "hat thL' st.r i ng 100]1'; like loY' t > 0 .

30

L,osson 21

whi~h have solutions:

T(t) A Bin(~t) t B ~oG(~tl

Substituting these into the Be gIves.

ll(X,t)

x (x) [a sin(A t) t h cos(~ t)1

n n n [1 n

x (x ) n

C sin(.ff. x ) + D cost Ii. x ) 1 G s i nh ( 15: xl + r co',h( fA x )

n n n n n J1 n n

and

e n

E n

D - - F

n n

[sin( 15: ) + .'dnh(,!A ) J

n II

. and ~n arc the roots of tho 0qua'~ioil.

IC08~ X t co" x si n» x t siIl xl
xl ()
sinh x - s i » x CO.:3~~ x + cos ThA \, cons r i r ut e a sequence of nUlnhe'rs CO[]VlJ:'gj ng to infidty (f'ound numcricaJly). F'Lna.ll y to find the_, coo f f i c i e uts an and bn "'" "ubsti tute u (,;:, t ) ;,nto the I:, to I':f't·

u(x,o) f(x) I b X (x)
n"J. n n
if>
ut(x,O) g(K} I .\ a X ( x )
n=1 1': n n 31

'I'he Sturm- Llo\Jvi 11.P +he or-y (F.!xtonlled to fourth or-dcr ) says that the X (x) aI'P orthogonill and satisfy·

n -

/ X (x) X (x) d x [ ~ 0 8 m n l~o

m f n

m n

Hencp wp can ~{nd tho coefficie~ts an' bn f~om the e~uations:

1 1 1
f g(x) X (x) dx f f'{x ) X (x) dx
A 1\ n
n Q b 8
a ·~·l-'·~---
n 1. D
J 2 f X2{x)
X (x) ell( dx
D D
0 0 'I'hrs s e , of Goursp, would have to}"" Found numerically with a computer.

2.

( ) . ( ) [. (2) 1 . ( 2 )J

u x , t; ~ c; r n 'I x cas If t .1- ---:2 s z n TI" t

'Ir

3.

u{ x , t )

!

n~l

:0

n

2

cost nir ) t s Ln Crrnx )

wher'c'

b n

1 2 J o

2

(I - x ) s i nf n nx ) dx .

It. T:1C nu t.ur-aL '::requPIlc lOS are the eigenvalues of :

Xliv) _ .\2X = (}
X(O) 0
X' (0) 0
x(.l) ~ 0
x"Cl) " 0 Th . f Xliv)

P genera 1. sol ut i on 0 ..

X( x) = A sin( /.\x) T B GOs( I.\x) + C sinh (I.\x) T D cosh( I.\x)

and substitution into the Ie gives:

32

I sin(if) - ~lnHif) -[sine /A) + s i.nhf if)]

cos( ,1;\) - cosh( if) I -I c.",; (if.) '" cosh ( fA) ]

o .

H(~ncr~ the na tur-a L Fr'e cue nc i.€8 ane tlUj roots 0 f the :tbove Ro_uation.

(Cj/~)t

and

idl"' lL.Jv(-~.

II L t ..:.

( .. " ) / .

'j./" .! I I

-', E, - (liL) lle' ..

I~ X

- (, / :'»

_', r, "U,

/_

11 - U. U

·~t xx

rw

awl tho'

I-I"

(l/L) x (_'J/L)~ +

['DC

~ uu::

U(C,T) - 1

() -: [, <
o -: T <
I) < ( .: .1 U( J ,T) o

1C

u(["o)

q. lith respect to ~h0 new t~me scal~ :hC! W::..LVC is on2~

5. L'ry [, kx. .Ic nce u Ut; r: __ . k ,\. It' Ide p i.c'c k I!v
x x
Wf' gpt the ucw "''lUatlCl".:
u + vu ut + '1 ur: u + u~ ... °
I: X V t (; 34-

Lesson 23

1. a) ellipLr;

b) hype rbo.l ic

c) e.l Li p t Lc ri) elliptic e) ellipt ~c

3.

lj)

y

x L

n 'f - 3x

I~ • Tho canonical (~Clun.ti.on i .. ~ ur,n = 0
5. Tlw a.ltP.!:"'Tlnte cL.ltlonl~L.il ~~4uLl·tlon _u.:. u - '\elf:l 0
:1::( 35

1. If 'ole differen t ia to

x+ct

2c f

x-ct

~Ij" it h 1:2 ~=.;pcct to

t,

us .i ng J ,p ibn ~ tz ',"~le (" "" 9l'oG Lem 7) ,0/" g.:, t u(x,~) = t r~(xtat) + ~(x-ct)].

t.: • l:1tcrc:hiln[';it~g the Ie, the solution tel

\1(X,O) a ut(x,O) ~ x

Lr:; u( x , t ) =

1 xtt
2" f i; ell; - xC lIence if ( • ..fP. dif(,eren tlClte thi,o
x·-:
to t t~'G w~t U = x ',olhien is the solution to ,,(x,O) = x ut(x,O) 0

:J. Sinee t~.e so Lut.Lon i.(; l\C x, y, z , r) c t 0 one can sp:e t1,,-' -i.nltial distLn'bance T'f'gi.,:m Cwhic;'n b the un Lt sph"p,,) giv,,'i

:t' 1. s", to an out var-d.l y propaBa t .i.ng a nnu '[ us. Sac h po i.n t (x , y , z ) QUL;id,,, t'lU unit (;phC'r8 'Hill suddenly eXDericEC(; -:l distpibu-

t 10ll--11111 DI, in <1 o i.st ur'b ed stilte for a length of tiT:lc 2/x-., then U(X,y,2,t) ~ o.

trailIng edge of the disturbance

initial disturbance reg 1 on (i nit i a I explosion)

outwardly moving annulus

leading edge of the disturbance

36

4. In this case each point (x,y) outside the unit circlR will tluduconly exp"r·icnee a shock and aR timo get:s 1 drgrc t he "0.1 utl.on u(x,y,t) grauui111y go"s to ZerO. The snlut~OIl would look mor'e Or' lc s s like u c Lr-cu Lar- rtlaVQ i"lfCf=!l' dropping a pebb.l.e in a 1,,)«,.

and has been ,",ulved in r.,,~"o[] 18.

6. Imagine" ,.;In,,,:l,, .i.n it i aj d Lst ur-hancc dt tl]" or'igiI\ (0,0).

f) Lnce HUYfJ0-n 1 s p ri.nc i.plp. doesr~ T t hold in ll.>:o C:it\p.nsi[r:-·I~J t!~:'s m""lTIS I;l,at af t.er- a 1':iven length ot tim" any o cher po Ln t (x,y) ,,1:.1 c-"lJd"uly oX?,'riDll(;ll a ,"hock due to t n i s cl iet ur-Lunco and

the" BTadually die out ilS ti.mo gets Lar-ge . 'j"'" reClSOIl t;,c

sncck gradually goP_" to ~erO i.~ becau'le Uw or Lg i.n in two d.imerisLons can phyni.c al Ly be interpr"t0d as an infinite' lIne l'er:.l0ndicular to I:he xy ~i.J_ne Lind t he uo i nt s in t.he xy p Lune ar-o h r out i ty being dist\l:rbod by all thc-l poi rrt s a l.cng this Li nc . I\s time incr':!dces all po Ln t s a i.ong this Ll ne will be di.,;turbi.ng H.e point (x,y) and as time incrailses i t l~, the points fur thor away r ho t ilre doing H,,, dl~ rur'hine.

7. Le j bn itz', pule t o i.Ls how to d if fcr-errt La te an Irrt agr-a I \·15 t h r-e spe ct to t whure tho limits ar.d the in t egr-and c,u, depend on t. Direct app Li.ca r i.on of the f'or-mul.a gives:

x+~t

d J d t

,p( s ) d s

~ [,p(xt"t) + ¢(x-"tll .

x-c t

37

1. Solve for A 11

cluJ from.

u n "- 0,1
[\ (O ) n - L~4~!j~G~
J1
. ~ 11 J .\,{t) A1(1) A (Ll

11

1

o ::1:":: 2,.L~,.~·)~6~ +."

]_ _C1;r)2: - ';i C

'2 LL(X~t) ~ 1 j. (:-~ t

cm;(:rrx)

r (1) n

f (0) 0

11

R (t) = n

2 t 2

". ( 1111'0:) I: f ( niTO:) S () d'.. .

('. e ["n co

o

Hence the '<0·1 lit" ion .i s :

u(x,Ll" I:

n= .l

B (t) ,;In(mTx) . n

4. {' n -r- 1
B eo ,. n 3
n . ~
illl ot!wr nls 311

S. A n

1

2 J x cod n nx ) dx o

6. Solve:

dn (t)

n t (ml)2 8 (-t)

dl' n

f 871 t 1

1 (_Untl (2))71)

n ~l

n f 3

{ ~ n - 1
n (0) 0,
n ~ ?,J,
11
for' 13 (t) and then subs t i tu lc inlo:
n uf x , t )

L

ri=]

Fl (t) Si"(,,IIX) . n

39

LP..'380Jl LO

1. Direct SllhstitulJ.OIl.

1. The problem '".i.tl' t:-'" nonhomogeneous l'nE (ilnd homogeneous IC') call be i30J vo d by U", f.i n it e Sinp t r-ansf orm to g8l:

u1(X,I) = - ~. LC03(Jllt) 1]. {3"11 1

ThE' so l u t i on of Uw homogeneous PDJ:: (rlnd nonhomogeneous Tel "an be "01,,e(1 by S~·I';;~;;+-i0\1-~Vin'iablp.s to get:

7

u)x,t) '" ",-If 1: sin(7!x)

Hence the .-o.Lur i on u{ x, t ) of our- pI'ub 1 em is U'E' -.um or theRe two funcl:ioil8+

3. d) Ul·1 \11 ls il s o Lu lion (the "quilt:ion is Ii near and homogeneous) .

b) ul + 112 io, nul a .solulioll (t1w e{jUatiull i.<{ no L llOllL(}genpOli~-l) .

u -t u

J 2

is a ~olu lion (1.1,,-, "'1"a ti on i <; 1. i ne ar- ane!

hOlnog"Il cou ~ ) .

::1) "1 + u1 is nol u s.ol u t i on (lbe equaLeQn is no l lineJr).

II. In each of tL,:, LOllI' ,,;ubpl'ol>lc,m,·, let one of ·the f'out- equation" be n011hom()p.:p.n~:.:-nll:=-, a.nd t ~'1E~ otlJPf' t hr-c e 11()mog,8n(~OUF; 4

to gel. the .s o l u t j o n :

G. Yc':.:.-- -- Lhi~-; Cd!) L~:!" VP.c1 fJ cd by Jlr'pcT nu·b~Jti tu t I on .

1. The solution is a moving cosine cur-ve u( x , t )

2. \l(x,t) = '~:T s inC x/ t )
+
3. u(x,y,t) -I (x_at/c)2 + (y-Lt/c12J - dt/c
e -
'+' u(x,t) r(x-"L) _tLIL
e
!>. u(x,t} f(2x-y, 2x-y} -2(y .. x)
e COe,(X -,).

esson 28

Ln it i a l p01IlI:

x . o

o

a} if x < G, then ~h~ characteristic curve is o

X - e( tI)t ~ x - c. T'o f illd the cons rarrt Len: I he cur-ve

X. !lence the

c,

j,'j'

."' = x

To f,i nd t he cons I'dllt tOI' the

x - g(ult = X - X t = c o

curve sLdrLillg al

(x , 0) ,.jp let + o

o and find

c - x ~ llerice the c:h;:lT'dctel..,j_~,"t i c curve u Li..tr'1:in~ fr-om

o

x is.

o

I, "

1

(X - x ) ()

x

(I

2, feu)

1 3 3 u

'[:'

o

3, The probIcm we must sol vr- -' s :

TC u(x,O) = ~(x)

wh i eh hns Lh(~ ~-;o tu t Lori :

u(x,l) = ,pex - ~)

-kL e

'I, The problem ,le TIn,,, I soI ve is:

rm:

_I_

U I' u + - I_)

x t t t _L

Ie \l(x,uJ = ¢(x)

,,(x,t) = +Lnf t + 1) + 'i)(x -. t) .

~. (;ivf"n u = 1)( x - p;( u I t ) d i ffe"'-'" I Ia tr- eilch soule of the e,["" ti on with r-e spcct to t elm] X

ut -- <p'(-g g u LJ u = ¢' 11 - !"_l u t.]
u L x x
S"bsti,tuti ng i rrt o 1..1 + r:(ulux '.-j(~ r,(":!t"
+
"t + g(u)u 1)' I .g ';" u ., + G 1': r:l.l u 1:'J
x -r x
- -1- (~ , gu IUt + g u l = o
x 43

Lesson 2'1
--~~-.--~
1. [: 0 :] [: 0 :] [: -I :]
1\ 0 n ., 0 C 0
~
0 -c 0 -3
? • -. :; Xl = [1 ] A = -·1 X2 [-: ]
:-> :-> 3.

U -I" Au "0

t x

into

u

WhAT'e:

[1 lJ

) -)

rJ OJ

LOi

;)V ()vi
.I. J 0 ..... vl.{x,t) ~(x- 3L)
<1t + dX -t
ClV) avo
L- ° _. v2{X,L) ljJ(x ~ t )
- "51:.- 1 oX = UJ_(x,t) - rp{x - 3,) + IjJrx,. L) u2{x,tl ~ L'rp{x-- 31-) - )1jJ{x t 1:)

4. Direct <ub s t i t.ut ion .

S. Direct bubstitution.

.1. Dipcc( c;LlD6titut.\on,

? Ilipce t suhs t I I.u Lien.

3. The solution is'

I

m~l

A J (k r) "eR(k t)

llL n OJTI om

T,~There

A ru

1 '2 7

'7j dl-c)J(k r)dr!.T(k)

() om 1 om

o

Th~~ S(~ in t. egt'(-! 1. S Ji"TouJ d hav= t 0 ho (_)V;:~ 1. U<3 t 8d nume r i C ill 1. Y ern Cl

cornpu Ler , The val,,,, of J~(kom) C.<][\ be found ill r ab.Lc s •

1"01' e xamp Le J2(k ) JL1' (7.'1) ,·,h;c.h Cell] be I'ounrl in most

J 01

t.ab Ie". Or,,, \'IOU](1 not ililVf' to l ook ",",roy hard 1.0 fLnd dn

~~xiG"l.i_ng cornpurcr: Pl'0R~-'~m to (~Vd.lUo.:1tof' lrH?sP c:of_~fti cI €Irts.

[1001: comput e r 1 i_Dl'dJ,i '-,,) '~',)uld COf\ta ill cl f)['op'am tu f l"d tile CXpFlllS.lOn:

f(r) ~ ),

n~l

1\ ,i (k r)

Tl 0 0[[

for> nrbit.rdl"'~~· runcti.Qn,cJ fer) 0 -; l' < 1. The r'p.ddr.'{' r::~Jn CUllsu l t hi~,!h·?_l.; COli"l.pU\_CI' ~:"i?nlc~r' e.rHJ i.n'.p.~ir(~ ;,.Lul)t. a;)pliL'u

I!. U (r, t ) 'C ,J (l ( :;> • 4 r) cO(; ('2 , 4 L )

J (?,4r) cos(~,4L} o

,5 ,T (H.GSr) "od8.b51.) o

+ .'2'1 J (1'1.93r) co,-,(l4.C)lL} o

1.

45

.1.

t 2. u ]

[' r

/.

t l uJ

l' l'

3. rj?u uu
;> + ili;
de r-
It 11'18 l"t v -

d\l ell'

t.h is e quat.Lcn

b ecome s

dv 1. dI' + :l'

v = iJ

Wllich

has the ao Iu t l,.", v Cr) _ co 1

r

to Ld.pJaG8ts equation js

S. Sinc" U I U ~ 0 r·01" a LI X and Y, ('dch po.l ni. on til"

xx yy

8urofdcP_ ,,( x,y) ~ T)' 1,., "quLil to tI,e average of 1.Ts nc.lghborE

on a cil'clc arounu (x,y).

111

L8GSOn 32

1. 'I'he r- dcpc ndc ncc in the so l.u t i.on i" 1 i W'in' and the solution LUP1W Olll 1:0 be ll(r,O} ~ r dn(IJ}. Ther-e is no pedSOTI to bel ieve il pRl'SOn s hou l.d have this kind of ill tui ti on but thinking what t hc so l u t ion shoul,1 look l.Lk e i'; u sc Fu l , 'l'Iic l'e<luet' uhou l d Lmag i nc vhar t hi s sur-Face look" 1 i k e .

2. No s incR the uc t tLux is not Le1'O.

3. Just 1'emembl'1' +h« so l u ti cn u doe",,' t have lo L", :;_~~~ ·to

s i n( e) on the bOlmrlary but if u ie,] R"" (gr'eat<=r) than sin(O) heat will flow in (out) 80 that the steady >lt~~B solutio" will t.c nd Lo fullO',.J si.n( 0). If 1, is bip: the solution u on t h f' boundary '.J i 11 for il 1 1 pra c t Lea I 1" mposes be equal

to dn(O).

~. S~ttinp: uL ~nd Utt equal to ze~n w~ g~t.

d2"
--2" + u 0
dx
nCO) - 0
,,0) - 0 wh i ch ILc'" the bolu lion u( x ) - o. Thi s b no I: sur pr-I sine since the PDC d~,,".Cri.bes a v i hr-a ting st r i ng llllfler' t.he i.n f l ueric e of cl I'r-Lcr Lon r errn -ut. The so Lu t l on LO t.h i S pr-ob Lem ,;tal't.~.

:'. Ca n W! find 'Olll'face" u(r,O) and eonsta::tG A so that at each poi.nt (r-, 8) the Lap l acLan ([]lhlSUres tlw <legr'oo of

cur-ve t.uro of the surFace ) Ls propor t.Lona I to -u? The anf.wer is yes if we pick the proportionality ). appr-opri at e l y (root s ot R es s~ 1 fll,wt inn,·; ) . 1 n fClC t ,~e sa", what those sur-f ac e s

Umn (I', e) looked 11k" l/ho11 we studied the circular vibrating

memb r-ano (Le S80n J 0) •

6. ','Ie are lr'V lug t o f l nd the s t o ady state so LutLon (,,"y temperatur~) i ns ide a sqUiIl'e t ha t is Lnau La to d OIl t he two Lat cr-a L

s ides and the tal' is fixed a t: Olle ',.Jlll1e the hottom I,; Ln contact wilh a modium-ottwo de[,;reo.s (01' \ol\ntever). One "an imagino thR rlirection of the flow of heat (from bottom to top).

48

j. Dirf'ct comiJutation.

,,) u(r,U) - 1 o;in(8) l' """UI)
+ r' +
L
b) ll(r,O) /
c ) H( f', G ) .: r ~: Ln( (j )
ell u(r,O) ,3 ",IIl(3il)
I' 3. u(r,ll)

% o;in(O) ::'

u(r,O) = r sin(20) If

).

u(r,0)

I

[l-U

i'l 1/;(
')
r [l _l ,?J,5::-
':j -2
1\ ?,4~6~
'--'2-- n ..
n I n -1) i'" 11 ~ 1
t.)
Il
0 n f i H(;ln(~p:

u(r,O)

It ,,1111 be he Lpf u I t.o U[,P th" For-mu tas .

dn(A) ,;i])(n)

) r~in(A.R) + sin(A-B)J 1

'2 ["os(A-Il) - ""s(fI+Illl .

5.

POISSON ]<[:RNEL ~ R2

2

r' --._.-2

._--( • O-a) + I"

Lei-< (;OG ,

P

Cr,8)

to (l ,et) .

(\ o()

J

Hence

Ke.,...ne.\

o

II

into tenfO problem. roblem. into the p

.r .

. of

S\II,S t i_ l U t Lon Dlrcoct

~UbSlitutioll of Di.:ect .")

u

i •

1.

2. a) u(r,B) 1

b) u(Y',8)

+ J.i cos( JB) r

d u(r,O)

;. sin(B) + 1] cos(30) r

d)

u(r,B)

2. + 3. ,l ,:;1.fl ( 0 )

2 11 '"

1 . ("(ll J • (58) 1

t- -5 ::;1 n oJ + ~5 s i.n + ~ ~ -.

3,.-. 5,'

3. u(:r,8) -: .. .!: s i nt n )

T'

'I. Dit'E'ct ,.;uLstitul.i.on.

S1

1. Direct substitution.

?

u,,,,, Uw c hai n r-u l.e

d1> d¢

x

cost ¢).

3. Direct subs t it u t l on ,

]

'I. em; ( 3 ¢) ~ 'I "OS' «p) - :-1 cos ( 'p)

1 ~

- (8/5) ~ [5 cos (¢)

Hence
ll(r,¢) 8 3 P:/ cos ¢) 3r P1(cos ¢)
- r -
5 5
1\ 3 L~ 3 (¢) cos( 'p) J 3I'
-I' cos - 3 - 5 coster)
~~
u(r,cl>l L II (,,,-,s ¢)
S. a r P
II n
n'-'O a n

2nl ;I

11/2

f

o

p (cus ¢) sin(¢) d¢

II

Tlwf;e .integrals can be nva·IUil.ted but one mighL "'ant to USc a c orupu 1:,,[, t o ev,,-lu,,- te them numcr i ca l Ly ,

6. u(r,¢) = r + _~ co,;(¢), r

52

l. Th" t ot a L ouLw"Y'd flux acr-oss tlw co."hc'rc oC radius t' 1::,
? i 2
-41Tr' u H.'llce r WP. sel- --4Ilr u = q lo-JC have
l' I' u r:

and ;-;0 v[(~ h n. vi:

ued

---.3__

ilfIT'

1 }II 11'.(-::-) R

I (R) 111 n R .

J.

u(X,y)

-k

J' f

)11

o

I (_!::_) jr

_. n R (C, dn "

,'J PI""" pOC,l t i vi: "hilY',';e"~ at U;, n) und (-;,-1"1) dnd Ilegat_lvc orie s at. (-f"I)) and ([',-'1) so that Cree,,'s f'unc t i on ""en be:

G(x,y,s,n)

S3

Try U (1') ~ Ar2. p

l_ 2

u (1') ~ - c. If WR nnw let

p 'I

1 2

u">'-lLi'f'

w

"ill c;"tlr;fy the probl"m: ",2" _ 0

und hence w(r,O) '-

~' + r '; i n( ()) . lIRn"e l.h« .sol" Lion'

1 2

u(r,e) ""4 (I' -1) + I' sin(O).

J,(J s son 37

h2 f(x+h) ~ rex) + h f'(x)'~ ? f"(x)

I';

11 f ' (x) + ) I" ( x ) •

f( x )

i\Cldine tIH':;(J 1''0'0 cqu.i t.Lon-s and solving 101' f"(x) w" gc't;

f"(:.<1 ~ ) ff(xlh) - '2 rex) ~ rex-·ll)] , t,

2.

o

o

0 I.)
0 ,lIJ8 , LOP, IJ
0 .3'2 . 3~' 0
get, . Rf;G
fl r-s t ;_ terd,tion 0 U
0 ,107 • J 07 0
() .3)3 .323 0
• m,d'} , flf),)
.~~cr.ond iteration I.) .11(i .Ll6 0

o .216 .21G 0

,eGG . Sbf)

~ ni. L.i ali ZC .i nt-prior' grid points to be

;:Jv(~r.:Igc of the! Be

Tl)C~ numbe r s dr~e r.or~ver'ging +ai r-Ly rast--Dut rm= '~houl0 rpi"11·17.C~ they ayo" [10'- eonver·gi.,g to t lic c,oluLion o f til" rm~ at tl10,·,e

po i n ra buttu the t,olulioll 01 Uy, ~1 __ i.__fTer·e~ 2..'-juatio.::_ t ha r ,'1'...£E.?xi rna l",; the PDt, If we \;"Tll. a lYe etc'" tJolu tLon we must use

mor-e grid points. Tt pould he " ~,ood c xpe r i mon r to double ti,e

[lumber of gr Ld po i.rits and ~;pe 1 r the num~cI'~ convEn'~p to the 3ame va l ue s ilS t hcy do t hcre .

3.

?

l[ t- I + 1-2____j'

Ui,j = 2" "_]_t1,j Ui_l,j'ilL,jt1 "I,i-l 4- .i,j

',1",:er'e f , . = I(x,.y).

1,:1

4-. U ••

.l,J

--J I.u.. + ll.. 1 u , . ,I u , • ] .

2(h .. ";'i) 1,J+l 1,:1 1 H1,'1 1--1,J

5. Use the usuaI. rep luCentE'" l formula £01' V2u ~ 0 to (>', lima tc t.hc Lut cr-Lor erirl po i.n ra and LhL' f'_]_"i.L" diffel'("JlC'-e Lor-mu La :

u, l,l"J

u , h g.

i"n-l 1

to f i rul L I LL' bound ar'y jlDt"rrl- iu 1 s ,

Lesson ,]8

.1.. Straip;hlfot'wilpd computation.

2

2. u(x,l} = ",-rr t sin(rrx).

J. More or' less the ""m" "c< thE' p"",,,holi,, flow di.ilf'pam.

4 . U lr,'c t comput a t i on. The n""" ,.;e;aT'Y cq ua t i01' s loa ve be" II dcr-I vcd in tho lot3-sun.

~j 6

1 , Stra j gh t fon! a l'd c.omput a11 on.

?, The finite di f"Pr'81ll'P formul<l few t.h. P1"n:: ·"i.l.l be tr,e same c~

i I! the le~~~"ofl. The Be LH')COID!)8:

(1

·U. _. 11,

l.. ,tl 1. ,.n-1. + u ,

-'-'--h-'-'-- L,n

g. . -'

e q_ ua t _i or: ~3 for U ~_ ~ ')' 11 i_ ~:.)' ..• ~ II i. ::0 n

wouLd gi v e- us an <~xtr.:.J. eq_\la t lon ~

Thee lie at 'I - 1

3, SilTll(' as i.he pT'CFJl.c·m in r he les'lOJ) ('xcPj't I.h-Jt the d i rferpnce P(,lliHi.oJ) ',IOU).([ I", ,;]jgll"i:ly G..ifL»'~'~l.

L+ • The moll' c U] Jf' f orm i ~ .

0)

~t:'sson 40

Not", th"l "'.; tiwe increases we can approximate the ,;olution fc!5rly accur a+e 1 y from U,,, fi r st two term~:

l.l(X,t)

He uce if x -= .!_i WP. havc :

u(.~,t)

2 [ -(110:)\ <3

11

1: .: ( 3 'I\(( ) 2 t; J 3 - .

1/e "Lin now find l~ tlL'" t ,.ji 11 fit this CU:rV8 to t.h« o]"-,f't'vr-'ll rlald point". This won1d be a Le aat. 3'[ucl'I'p'; I'l'oblpm (sep

pnob ?em 2). Th)"" ar-e computc!' pl'op;!'ams ava i Lab Le tha l ri t cur-ve= 1 iko t.h is t.o arl-, lt r-ary dLita poi Ill>'.

7.

3 011" "all d""ign all e xpe ri ment to meas.ur-c 1.11," con""ntra lion "t difh'l'ent va l ue s of l' a ft.e r it has come to cl st0il(ly stale (c-;ome ,;rwl: of optical device). Wr~ H\0Jl fit (le'-'~t SqU21Pl'S) l:llP

t heor-e t 'ica L ,-;oJ u1:ion u( r- ,"') (w),i ch is k nown ) to th is ,lilt" by

p Lck ink: the peoper ehoi Cf' of D. We then ,-,o.'Lvp ,I'm' the molocul.ar- w'Oignt dl[,;'_'br,;5",UJy ft'om this va l uo of D.

58

Lessor, 41

L, a) par-abo I i.c--cilnoIlicill

b) hypcrLolic--nol: canonical

c) eJ J i p t i c= -uo t cFlnoni.cal

d) j)crrabo1ic--<>HlonicFll

2. The new coopa i nat e s ,wO' (-,- y - 3x and ~I - Y and "thO' h'iln s tCPTIled C q ua t ion i ,", :

lfi u -I- l\ L.

WI

,-,ooPclinC!tes S 2
3. The nf~H arr:~ '( and 11 xj') and t he t r-anu
farm,--,o ('quat i on 1::: ;
'-\f, + t ~ -11
u u L'
TlTI ?n 11 L. Appl.y t\w t cchn iqu« i 11 1.h(,· 1 ",;son dire" Ll v to lbe hn)cl.: or:

r( xl = ",Sill (x ) (no L(. t hLs fun" t i.on is \-'0'; lt I v e f,,~' r< < x < .1). Tlw JrrE' thod ill Li1.i.S J e"son d8sumcs .r( x ) > (1 bu+ it r.ar' b,.·

~. "iI-I ~ 3I'i. + II mod 7. The seque'ncR 01 r-andom integp,.,s C,lilrt:.i.ng

r r-orn z~ro t~-; U, I~., 2,3 .s , 1) 0 -~ 1'I3p~·)~ll::.:i..

:1.

60

Onl? can chdIlgC' L he va 1 til' of M -md f l.nd n~Jr."l ('~-~t ilTIdtf':==', f ('I::' ,,::::-- •• ~

in I. "'1'ra1. Tho va Iuo of ll,c 1 n L('fT,l'J1 ",j 1.1 s Lab i Li z'" a~ '~ gl't'< 1dcgc,r' d'ld 1 iJY'iC/'Y'.

~; o -: r < .25
X .)~ < r < .7',
. 7~i < l' .' 0 -: l' < I ilJ)d 0 < e < 11/2 ,mn
1 -
,~ = l' + ',iIJ( 0) > 1 in thE
L 2 Compu (0 t Iw frac t ien or (Oo'"('~' p line and l-}1EU c,I:.;t..lmc1U~ H .fl"'uJn t}7~'

I hell. LiH' )W,'c\J,' '2

(1 rpF-d: .~ on ;- ::: f>.

2

7J = -;::

p

llne

---- - - -------------~------- ._-_- .. ----

J

1 i rh~

b_l

S,. T":!~ f:nd.tp. differenc~ app1")()~_j.m[l1 ion b(-~com(~~:!. (after sol vlng for Ili,:i ):

u ..

"d

u. . 1 + 11. • 1 + x ~ (1). 1 . + u , 1 .:\

J ,.1" , , .1·· 1.1 I ,J 1-, J

~ ----- -~-·-~-2--··------------

2 [1 I- x.J J

,,'Once t he problem of r,oinE; from u , . to one of j t.e four , ,j

neir,hbors is the cOF,.['fic:l"nt,; of t.he (:0rre"pondinr, IYnP1s. Hote rhs t th-E~ cooF.ficieots ar-e oonneeativp. cH~d ~Hm tu uno.

...

Th" HCs

u (l,t) ~ 0 x

f,i vl"l \lC; -li1(, difFerence cquat lOO!

~. \J = 0

1.;11 i,.n~l

stat,tiny, all j ,'11 sft0u1cl be "qual TO the avel'age l'ewal'd

start inp, J t (i, n- 1.). 'fllis givQS t.ue st:rate"y of a reElect] 'dg boundary. Tbat .i s , i_i· the willo hits t.he ] inc X" 1 he auto1Mti",-,11y bounces hlCk.

I- I'll Ui•J 1 p~ '-\,ill

wlx'r" the P: in'p_ numbe r-s c:omputpn from lhA c:00ffi cif);YLs in

l

th" PDE anrl rep-rec;",nt t.h'" fll"obaGilitias of r;oing to til<' ;:,or'--

re'''[landi.nG DolntG. OItf' couLd mod i fy thf' BC~ of ·ll,l" I'Fob.1 I'm

to s o Iv c pr·oblems Hlth 1"_1.nx acr os s t.hc hcundarics

(d,,,·, vill iv," ll(;).

i,:i 1

i, j+ 1

Pl(~

:i +1, j

"

i+l,l'f-J

61

1. r(x,y,y')

r y

d

]:'

elx y'

_i_ __ 1_' __ dx r----?

V'l + s "

o ,

Hl'nee, ';01 vIn~ t l: i s ()DL a Tong with th,' BCs y( 0) ~ 0 and

y ( 1) = 1 we get y ( x ) = x , Since ,j [y ] i", the tu nc t i a ns 1 t·OT' tbp length of the CUI've bCIW"Cil two points uud s Lnc e UK' cnd I,oi nt s her-o an.' (0,0) and (1, L) this ,l.ns'.;c,·' itl not SU,"pri'lilli',. Th". f\_l1\Cti_OTldl cva Iua ted ill. 1111s m i n irn i z i ng C\JrVe ,·,hou] d bp h-.

2. m y + k y ~ 0

J. The Eu l et'- LaSl"HL?,l' c qua t ion l,et'E" 1 'l ·-2y - 'Jv" 0 ",h.i.ell

si mp Ll fie s to s" + y = 0 and ala])!,: ,"i1-h r ho Be" .,,(0) = 0

and y( IT /2) ~ I I"C ge1: y( x)- 5 i n( x ) . Tlrc f unc ti onal evaJudted at th i s mi n Im i zi ru; cur-vr is ,J [s i n x 1 .: 0,

Jj < Dir-cc r 1 y il.""10,,,OIl':' 1.0 thp ts ingll' .l n Logl'O 1 pr-ob Lsm ill tllP

Le s son ~ Now, lluwevp.p ~ ill} f unc l inIJ::', depend o n l.Tr"IC indppp.nclen (_ v<:lrL~b] c~.:;. 4

63

J.

1 1 ,)ruJ ~ f f o 0

2 2

I_u + 11 + 2u:i dx <1y. x Y

2. L"t z(ll" (I,xl y(1). He nr;e z ' ~ -y + O-x) y' aud so we .ri.nd z tha l min imizos tl,,, tr'ail"fol'med inlegcal:

Zl

+ --- + J - x

and then .l ct Y

1 - x

Th" point is t.ha t w" can 11"e t he

mo t hod of Ri tz cu t ho fund: ion L:( t ) sin"" it sa t Is f i E'S z(U) = 7.(U = O.

4. 'J'h i s pr-oof La exactly ILko the proof in tho last Lesson axc ep t ThaL single integT'als ace T'eJ-'laced by double integrals.

o , 'l'o solve The Diriol,lct p r-ob Lem by the fin i_ t" s i n> t.r-an s for-m (on xl 1'10 geT:

:/ <1y

2

(rin ) lJ (y)

II

n = 1

n 7,3,

U (0) - U n

u (1) _, 0

n

A oily + TI

1 - 1,2

u (0) = 0

n

oJ = 2,1,

and so taking the Lnver-sc ~,i ne rr-ansf or-m We have:

Now that wre know The ,,;olu r Lon He", eel" r I nd the po te n t La 1 l'[l('l'!':Y r'r om the e qua t Lun:

1. l ,Jlu] "C J J o 0

tee v,)

j,

Set t l.e c(wffjci",nL of L, c, both s i d8S or- tlw PUr. FInd TIC,

:>

c "

"'qua 1. to ('aec h 0 til'" r on

J, If I-If' s uhs t l t u tu

u o

+ u

.1

r eos(O)

Into

'2 2

V U tll

and

u{ 1,0)

'2 2

r;; U 1 u

w'" wi _l 1 <;{'e t hill u ( 1 , e) "0 and

l'lla I. lh" Tunc t ion

eIpclu 0 < r < 1,

TllU tf'rlil

is v"'ry small .i ns Ldo t he unit

,,2 I.i 11 <,ontai" "j [1e U:.:rms Lu I.

4, To 'oolvc t he BVP

'2

V u = 0

o < r < 1

u(1 ,(I) " - sinCe) <'os(B)

o < B < '2 If

',,'" !'i.T"ot tr'c1Ilsfor'm this problem Lo One that h",,, 7,cro IlC hy making the followin!,: Tl'anhfQrmation:

u(r,U) = - sinCO) <,os(6) + U(r,8) .

Ilencc the ne" fum: ti on U( r , B) will satisfy u," BV!':

\72u 1 ,-,os(26)
.- 2' 0 < r' < l (wCl've
used a
P trig identity
u(J,O ) = 0 0 < 0 < 2'11 here) 'rk noW solve this p'ubl"m by ,.rri ti ng
lI(r,B) - UI}r),0) l U (r,0) whe11p' Uh(p,(Jl Is a g,muI'al form
p V~u
uf the s o Li: t ion of th", homop;eneous equation - 0 and 11 (r, I}) p

tion. A general

equu t Lon ) is:

is apart Lcu Lar: sul"l, ion of the nOllhomo!l,meous cqua-

form of the homogeneous eq ua lion (Laplace T"

L

n~O

n r:

[a cos(nO) + b sin(nO)],

n n

66

One can for Flny U (r,e)

p

cas Lry veri f v choice of th" ',-Ie try'

L:la I t h Ls sum Gil L ;.sf ie3 [,"p.Lace' '> ('.;:.:acow; t an t s "s and h". To rir:c

II n

lJ (r,e) - 1\ "0,,(10) p

(we gcneC'ally

t he rO',lCI' of ,. by t.wo ) and spp

U P

I n Lo t h" P(lUrl r ion

IJ = Uj + U WI' hav~:

1 p

U(I',O)

X

n=O

cn Lil cos(nll) + h :;in(1l8)]

n n

and by substituting this into thc BC il2 = 1/11 and all t he ollwr a,/' J.ml

have t he <o l ut lou ,J{ 1',8) :

U(1,8) = 0 we ~ee

b" "re zero. Hence \If n

2

ll(I',O) " - sinCe) L'us((-1) + l' c.o,:(/8) I,

* COf,(?O).

Hence the ri-rC"',t two lC11m-::; in the appr-ox i me t i on aloe one Fm.~:,,:-h of the above lerm p Ius r cos (0) . Olle wou l rl have to usc' a computer to ,-,ep hOl-' '"e 1.1 t h i" f unc+ i on a.pproxi",,--, le s the dR'formed boundar-y problp[]l

2 'l II

o

o < Y'

uel + ~ ~in(8). 0)

co,,;! 8) .

67

c.w 2

- ~ ~-- and hcnc.e -Lbo mapping 50; vmfm-omal f'xc:ept aL d;, ,? _ 1

z s: ;, 1. If one wr5 tf'f_ the mappil1f!; in r",,} form:

U

17, ~ J I log I-~. z -j- 1

one can shO\. that the up[Jcr Lal f p Lsne map" into tbe g i.ven rezj on ~

3 [ i8-13 3 3iO

"",itf' W" 7. = .r- '" = r- c and So the t Lr-e t cJ1wdrilnt

•.. the z plane rua pc; j nto ttle f Lr st t.hrc c quadrvrn l,o; of the

w pLJ[w.

3. r~.c tra[\sfQnn"d pr-ob Lem is'

" e :!ich has tlw_ ao l u t, rOT):

u(u,V) ~ 1 tan-1 11

f LV }

) 2

<J + V - J_

!<riting Lhc' t r-anuf'or-ma t Lon " z? in real ~o1'm u·~ x2

v = 2xy one has thf' ,-;oJut ion of tho orig1 nal pro101 em.

2 Y ,

1 -J

Ijl(x,y) ~ z: tan

.. 11

f 4xy )

? ?) ~) .

(x -y 1 + [fX Y - 1

G8

'I. TILe mapping w '- 10[;( 7.) '""P" I.hu \;pc]p:p in (JUl' [Jroblf'm into the rpgion:

e=o//'f

¢ ::: I rt'/'f

I{:( !:~ ~<L = a

~///////;,

<I> = I ,,jll. <1>,--::"

.;;lcp -:::0'

;//;//1///,

x

¢ :::0

o

It is c l par t ha r tile tl',J.f!sfot'lllCU (lulu Lion 1" ,;() if we >I['.i1_2 1.11(> lransfonfielticm h' = loge;;)

'IV

'p(x,v) ~ 11 emu

in T'CJl form

-J

tell) -Cy/x) ~ e Wp gc't the solutioIl:

69

Additional Lesson A:

Control Problems )n POE

PURPOSE OF LESSON; TO SHOW HOW CONTROL TIIEORY ENTERS INTO POE

THEORY AND, IN PARTICULAR, TO SHOW HOW THE TEMPERATURE OF AN INSULATED ROD CAN BE CONTROLLED BY PROPER TEMPERATURE CONTROL OF ONE OF THE ENDS OF THE ROO. MORE SPECIFICALLY, WE WILL FIND THE BOUNDARY CONTROL f(t) IN THE ISVP

POE LI ~ LI o < X < I O<t<oo
t xx
LI (0, t ) 0 (left end insulated)
Be x
u (1 ,t) f(t) (right end controlled)
IC u(x,O) o < x < 1 THAT MAKES THE TEMPERATURE u(x,t) GO TO ZERO IN MINIMUM TIME. WE DO THIS BY USING THE LAPLACE TRANSFORM ANO A FEW IOEAS FROM CONTROL THEORY.

Lesson A;

Control Problems in POE

One of the recent developments in POE theory has been determining boundary conditions in initial boundary~value problems so that

the solution of the problem performs In some desirable manner.

One should refer to reference 2 of the recommended reading list for additional Information. Typical problems of thIs kind arise in the steel industry where the boundary temperature of the ingots is con tro 11 ed so that the s tee 1 ins i de the ingots r shea ted rapidly but uniformly. Another example would be instilling, by some means, small vibrations in a telescopic mirror so that the effect of atmospheric disturbances on incoming light is decreased.

Minimum Time Problem;

One interesting problem that we present

in this lesson is determining the boundary temperature of a rod so

70

that the initial temperature goes to zero In mInImum tIme. Mathema tical Iy, til i s corresponds to find i ng the func t ion f (t)

(cant ro j rune t i 011) tha t makes the tempera t ure u (x, t), descr j bed

by

POE ut u o <: x <: 1 0 < t <: co
xx
u (O, r ) 0 (1 eft end is insulated)
(I) Be x
u (I, t ) f (t ) (right end i 5 COO t r o lied)
Ie u (x ,0) 0 < x < go to zero

[u(x,t) ;0; oj ill minimum time.

The solution of this problem is found in reference 3 of the recommended readings, and the following dIscussion continues along the lines of that article.

We start by transforming the problem into an ODE by means of the laplace transform

U(x,s)

l[u(x,t)]

J

o

u(x,t) e-st d t

This transform was dIscussed In lesson 13 and converts the or l q l na l problem (1) to the new 8VP

ODE d~U sU - I o < x < I
dx2
(z)
dUrO) a
Bes ~
u (1) F(s) This problem can be solved very easily; the reader can verify that

the solution is given by

(3)

U{s,x)

coshfS + [sF{s) ~ 11 cosh{5 x s cosh.,lS

Generally, the next stop 1s to find the inverse Laplace transform and, hence, the solution. However, since our goal here Is to find the boundary function f(t), a short digression follows.

CONTROL THEORY

The fUnction fIt) is called the control function and Is always restricted to a given class of functions (like smooth functions, integrable functions, piecewise continuous functions and so forth. In this problem, suppose we pick the control from the class of ball9~bang controls. That is, we assume the control function fIt) can take on only two extreme values (which we pick as +1 and -I). III other words, our control ac t l on will look something like the graph in Figure 1.

f(t) boundary temperature

+1

,....--.

t 1 t t3 ...
2
'-- t

~ I

Example of bang-bang control Figure \

72

Therefore) in order to control the rod temperature 50 that it goes to zero in minimum time, "Ie must find the switching times tl, t2, t3, ... tn ~f the bang-bang control. This will tell us when the bounda ry temperature s hou I d change f rom + I to -1 (or vice versa) .

The idea for finding tl, t2, ... tn is based on a simple principle of the Laplace transform which we now present.

Property of the Laplace Transform

If g(t), 0 <: t <: cc is a func t I on that Is zero for all t greater than some fixed number T, then the Laplace transform

G (s) of g(t) Is a bounded funct ion of s in the complex plane.
Examples 1. g (r) = 1 0 < t < 00 Gh) = I/s
r1 ,Y
[not zero for t2_T] [not bounded) {: o < t < 1 I +s
2. g (t) G(s) ,_ ~ [l-e )
s
< t < 00
r" '"
[zero for t > TJ [G (s ) is bounded) By using this property one can present the following argument. Since we want the solution u(x,t) to be zero for t greater than some T (T as small as possible), one picks the switching times

t1, t2, so t hs t U(x,s) is bounded. The interesting point

he r e 1 s tha t the denom i nil tor s cosh IS has zeros a t the values

k

I ,Z ,3, ."

and, hence, in o rde r for U (x , s ) to r ema in bounded, the nume r a tor must also be zero at these same points. That is,

73

(4)

o

k " 1,2, ".

which give us necessary equations for the Infinite number of switching times. We approximate this control f'unc t l on f(tl by finding the first n swItches tl, t2, ", , tn and then turnIng the control f(t) to zero. The temperature u(x,t) wi 11 hopefully be close to zero if we pick n large.

To find the expression for F(s), remember the control f(t)

is assumed to be bang-bang (it's obvious the control starts at -1 when t" 0 for our problem), and, hence

Ht) -E(t) + 2 E(t-t\) - 2 E{t-t2} + ... + (-1)n E(t-tn_\)

+ (-1) n+ \ E ( t - t )

n

where E(t) Is the simple off-on switch at zero

= [°1 E(t) l

t < °

t > 0

See Figure 2 ,

+1

f (t) n "
r--
last
cant
t 1 tz t3 t4 4 ill this example swl tch tu r ns off rol

--+ t

-\

Graph of a TypIcal Bang-Bang Control Figure 2.

74

Since the Laplace transform of E(t-a) is

-sa e

we have

(5)

F (s )

:rJfJ

- ..!.. + l e -HI _ l. e -st2 +

5 S S

n 2 -st

+ (-\) - e n-\

s

SubstitutIng this expressIon in equatIon (4), we get the desired set of equations for t1, t2, ••. , tn

(6)

The aUlhors in reference 3 of Other Reading have solved this system of equatIons numerIcally for various values of n, and some of the results are given in Table 1

t 1 t t3 t ts t6
2 4
n
0.281
2 0.315 0.346
3 0.318 0.362 0.372
4 0.319 0.365 0.382 0.387
5 0.319 0.366 0.385 0.393 0.395
6 0.319 0.367 0.386 0.395 O.~OO 0,402 Optimal switching times for a different number of swItches Table \

IS

The tem?erature corres?onding to these controls can be found

either by finding the inverse laplace transform of ulx,s) or by solving the original problem numerically using our optimal

control. The authors in reference 3 of the recommended reading found the analytic solution from a table of inverse Laplace transforms, and the results are given in Figure 3 for the simple control with n ~ 2 swItches, Figure 3 gives the temperature for various values of time.

u

t=O

x

switches at

t] .315, t2 = .346

Controlled Solution for Two Switches Figure 3

One can surmise from Table 1 and the above Figure 3 that the bang-bang control with switches at tl = .319, t2 - ,367,

t6 = .402 will do a good job In driving the solution to zero, The final time t6 .402 corresponds to the time when the right hand control f{t) is set to zero. It is also true that the values of the switches, tn' get closer together as we add more and more switches.

Notes

1. I f the right end of the rod were he I d at temperatu r'e zero, the rod would al50 go to zero, but not nearly so fast. Note, too, that after t he las t swi tch , the r i gh t end of the rod is

76

held at zero.

2. We might try to use thIs same approach to control the vibrations of a vIbrating string. That is, vary the posilicn of the right end of the string so that the string comes to res~ in minimum time.

3. The reader shouldn't get the Impression that the minimum time problem Is the only problem we try to solve (or that the control Is always bang-bang). Reference I of the recommended reading I ist gIves many other objective functions that occur in appl ications.

4. We cool d a I so change the con t ro I prob I em that we solved i n this lesson, 50 that the bou~dary co~dition on the right-hand side of the rod (x=l) was

u (I,t) = fit)

x

In this way, we w l l l be controlling the amount of heat flux at x= 1 instead of the ac tua I tempe ra ture. The au thor 5 in reference 3 have also solved this problem.

Problems

1. Graph the opt Irna I bounda ry-cont rc 1 u (t) for n:4 us i ng the data of Table I, Draw rough graphs of the controlled solution for t = 0, 0.1, 0.25,0.37,0.40

2. Show that the sol u t i on of the BVP (2) i ~ given by

U (x ,5)

cosh,{s + [sr(s) - lJ cosilY'S 5 cosh Js'

3. Solve the system of equations (6) for t c tl when n=l. Does this answer agree with Table I 7 What would the graph of this control function fit) look like

77

4, What system of algebraic equa t l ons would you solve to find

the aanq-banq control f(t) that sends the solution to the following problem to zero in minimum time 1

PDE

o < x < I

Q<t<oo

BCs

u (0, t)

o

o < t < ""

u (I,t) Ht) x

Ie

u (x ,0)

x

a < x < I

Other Read i ng

1, Operational Mathematics by R. V. Chwrchll1. McGraw Hill, 1972. One of the best applied books on integral transforms, conta i ns many worked examp 1 es .

4. Distributed Parameter Systems edited by W. H. Ray and D. G. Leiniotis. Dekker, 1978. A comprehensIve review of modern control thpory in PDE.

3. Time Optimal Control of a Linear Diffusion Process by

R. M. Goldwyn, K. P. Sriham, and M. Graham. SocIety of Industri~l and App 11 ed Mathemat i cs , Jour nil I of Control (5) No.2, 1967,

P 295.

HUD

78

Additional Lesson B

Potential Theory (rewritIng BVPs as integral equations)

PURPOSE OF LESSON

TO SHOW HOW THE FOLLOWING BVPs:

(I)

172u 0 Inside a region 0 in 2 dimensions
u ~ f on the boundary C of 0 (Dirichlet Proble:1l
172u " 0 inside a regioll 0 111 2 dimensions
3u the boundary (Neumann P rob I em)
"5il 9 on C of 0 (2)

CAN BE SOLVED IN FAIRLY GENERAL DOMAINS 0 BY LOOKING FOR THE SOLUT ION I N TERMS OF TilE SURFACE POTENT I AL S :

(3) W(><.,y) f )1 (5) (l I (double-layer potential
~. (In -) ds
C an r
(4) V(x,y) J o(s) In.!. ds (sll1gle-layer potential)
C r THE IDEA IN SOLVING PROBLEMS 1 AND 2 IS THAT BOTH 3 AND 4 SATISFY ~2U ~ 0, BUT NOW WE MUST FIND THE SURFACE POTENTIAL DENSITIES

J-I (s ) AND 0 (5) ( i n the in teg ra 15 3 and it) so THAT THE Be 5 u '" f,

auldn ~ 9 ARE SATISFIED. BY SUBSTITUTING 3 AND it IN I AND 2, RESPECTIVELY, WE ARRiVE AT TWO FREDHOLM INTEGRAL EQUATIONS FROM WHICH WE CAN SOLVE FOR ~(s) AND 0(5) •

lesson B: Potential Theory (rewriting BVPs as Integral equations)

in previous lessons, we solved laplace's equation with boundary conditions

u

f

on the boundary

(Dirichlet BC)

(1 )

9 all the boundary

(Neumann BC)

;~ req l ons D that were very simple (inside a circle and 111 the upper half plane). There are methods for solving these problems in more genera I reg ions of space; chapter 7 in reference 2 of

the recommended read i rlg d i 5CU5ses the m" thod of Ba I ayage, the ~eron-Remak method, i rlteg r a 1 equat ions, D i rich I et pr i nc I p l e ,

f;.~; t~-J i fference method, and conf orrna I mapp i ng. In th Is took, we have already discussed conformal mappi~g, fInIte differences, aM the D ; ri ch 1 e t p r i nc i p 1 e . 'We w 11 1 nO';J dis cu sst he met hod 0 f ,ntegral equations that consists of converting Dirichlet and Neumann problems into integral equatIons (of Fredholm type)

and then solvIng these problems. The actual solution of our integral equations will be left to the next lesson (lesson C).

To begin, however, we must introduce the concept of surface potentials and discuss their propertIes. It turns out that the solution to the Neumann Problem will be a single-layer potential.

Kinds of Surface Potentials

Surface potentials come in many forms. 'We will discuss two of them, the single-layer and the double-layer (most common kind). Firs t 1 the sing 1 e -) aye r pot e n t i a I •

Si~9Ie-Layer Surface Potential Suppose we have a single charge (we restrict ourselves to two dimensions) +q at some point

From what we already know, the potential at any

other point P ~ (x,y) due to this s\ngle charge is gIven by

u(x,y)

\ In (-) r

(logarithmic potentIal)

Now, suppose instead of a single charge located at one poInt, we have a continuum of charges with charge den5ity 0(5) along a

curve C

(Figure 1).

80

o (s ) cha rge ilt s

C {(F;(S),fl(S)):O<s::.\)

Single-Layer Surface Potential figure I

We can Interpret (centered ass)

0(5) as the cha rge ina very srna II line seg~nt divided by the length of the segment. If we

pass to the I lmi t by let t i ng the I ength of the l i ne segment go to zero, we would a r r l v e at the exact value of 0(5). Now, to find

the potent i a I at any point (x , y) due to the ellt i re cha rge on C,

we me re 1 y sum t he paten t i a 15 due to each charge on the curve to get

(2:)

V(x,y)

See Figure 1 . This is the single-layer potential due to the charge dens i ty on C, and \ ater we wi \ I see how the 501 ut i on to ,~e Neumann problem

o

in the region 0

g

on the boundary of D

can be represented in terms of this potential. The general idea is that the single-layer potential satisfies Laplace's equation

for any o(s), but we must fInd a(s) that satisfIes the 8e

au/~n

g. If we substitute the single-layer potentIal in the

81

~e~mann Be, we will arrive at a Fredholm Integral equation from which we can so I v e ror (J (5) . Now we wi J J d j s cus s the second surface potential.

Oouble-Layer Potential The double-layer surface potential

'5 a little more complicated. We start with two charges, a

f!OS it i ve cha rge +q a nd a nega t i ve. charge -q I oca ted at a distance d apart. See Figure 2

(O,d/2)

(x,y)

-q (O,-d/2)

Two Charges GivIng Rise to a Dipole Potentia!

Figure 2

If we want to find the potential field due to these two charges and if we are located far from the charges In c~~parlson to

the distance d between the charges and calling ~ - qd (dipole moment), we can write

W(x,y) ! 1m 1 1
q [In(- ) In(- )1
d+O r2 rl
].J 1 I
d 11m [In(- ) In(- ) 1
r2 r1
(j I (definition of
V;)i1 I n(-)
r n=O derIvative) 112

We can find an alternative form of the dipole potential by evaluating this derivative to get

W(x,y)

\.1 cos 0 r

where r Is now the distance from the dipole center (the origin

in FIgure 2 to the point pl.

Tha tis

r

(XL + y2) 1/2

The reader may find it useful to graph this dipole potential in three dimensions.

We now place dipoles along a curve C and allow the dipole moment density jJ(s) to vary. See Figure 3. The dipole density \.1(s) can be interpreted a5 placing dipoles along the curve, each with its 0\1n dipole moment, and then finding the density of these dipoles in a rnanne r similar to the singlelayer density. We will let the sign of j.J(s) be positive if the inner charge is positive and negative if the inner charge is negat i ve. Hence, a un I form dens i ty of ,I (s ) "Ion the boundary means a II I nnar cha rges are +1 and a II au te r charges a r e -I.

dipole density

negative 4 -

at this,__. ~

point ~ .......

#--

dipole density '\.. positive at this poInt

s inc reases In the counter-

c l ockwl se direction

Potential Field due to a Double-layer Surface Charge Figure 3

In

To find the potential due to this double-layer density along C, ~e merely sum the indIvidual potentials due to each dipole to get

W(x,y) J u (s) a In{l) ds
C an r
J \J{s) cos e d5
C r Thi5 completes our Introduction to single and double-layer potent i a 15. We wi 11 now (w i t hout any proofs) \ is t SOI1\e propert i e s of these two potentials and descrIbe what these surfaces look like. We then show how these properties can be used to convert the Dirichlet and Neumann problems to Integral equations.

Properties of the Single-Layer Potential

V(x,y) "" J 0(5) !roll) tis

C r

Suppose ~Je have a smoo th curve C conta I n I ng a reg ion D and suppose we have a single-layer density 0(5) distributed along C. The fol101-Jing are a few propertIes of the single-layer potential

V (x ,V) :

V(x,y) satisfies Laplace's equation ~2U outside the curve C.

o inside and

2. The potential surface V(x,y) is contInuous across C.

3. The normal derivative of V(x,y) has a Jump dIscontinuity of 211G(X,y) across C at the po l nt (x.,y)

~. The property

11m

av(x,y) + aV(;,i)

dn an

as (x,y) + (~,y), holds for points (x,y) approaching boundary points (x,y) from the inside.

84

Properties of the Double-Layer Potential W(x,y)

f ).J(s)t-In(_i__)ds

C an r

We now I i s t four lrnoor tan t properties of the double-layer potential.

1.

w (x , y) sa tis fie sLap I a ce I s equ at i on v? u outside the curve C.

o inside and

2. The potential surface W(x,y) has 11 jump discontinuity of 2~).J(i,~) across C at the point (i,;).

3. The normal derIvative of W(x,y) is continuous across the curve C.

4. The property

as (x,y) + (i,~) holds for points (x,y) approaching any boundary point (i,y) from the inside.

For example, in the simple case, if the double-layer density is W(t;,ll) = I, then the double-layer potential is

211

(x,y) inside C (x,y) on C

I{(x,y)

1T

o (x,y) outside C

g i v I ng a j urnp of 21\. We now get to the rna] or task of the 1 e s son , converting BVPs to integral equations.

Converting the Dirichlet Problem to an Integral EquatIon

Suppose we wish to solve the avp

inside D

u

f

on the boundary of D

let us try to find the solution as a double-layer potential

u(x,y)

W(x,y)

S (\ 1

\lls) -;;- InH ds

C on r

Since we know this expression satisfies laplace's equation for any density \lIs), we try to find the density u(s) that will satisfy the Be

W(x,y)

f(x,y) on the boundary.

We must be careful here due to the fact that W(x,y) has a jump discontinuIty across the boundary C (Figure 4).

inside C

outside C

Slice of W(x,y) Across the Boundary Figure 4

Since we want the solution u(x,y) ~ W(x,y) to be a continuous ~~ out to the boundary we replace our Be

u(x,y) "" f(x,y)

by

I I m W(x,y) (x,y)-+(x,y)

fIx,:;;)

or, from Figure 4, w(x,:;;) + TIU(x,y)

flG

Now, replacing W(;,~) by it's integral form we have

J

c

u I s ) cos 0 d s r

+ 11))(X,y) =

which is a Fredholm integral equation of the second kind (Figure 5).

fixed poInt

s increases in the counter clockwIse direction

c

D

r

n

(Cn)

moving point of integration

Variables in the Fredholm Integral Equation Figure 5

'We can now be assured that when we solve this integral equation for u(x,y) and substitute It Into

u (x ,y) J l-l (s ) ~ I
3i1 lnH ds
C r
We wi 11 get a continuolJs function u(x,y) that satisfIes:
i) 1,72u = 0 inside C
I I) u(x,y) ~ f(x,y) on C 87

This completes our discussion of the Dirichlet problem and how We can convert It to an Integral equation. The technique of finding the solution of the Neumann problem as a single-layer potential is carried out in a similar manner. In this case, we use the

property

1 i m ~~ (x,y)

(x ,y)+(x ,V)

dV - - - -

3i1 (x, y) - 110 (x, y)

to arrive at the Fredholm integral equation

I

J o Is ) In(-) ds

C r

g(x.y)

from which we can solve for the single-layer potential a(s). We will discuss the general nature of integral equations in Lesson C.

Notes

1. An integral equation is an equation where the unknown function occurs in the Integrand (possibly other places too).

2. This method leaves a lot to be desired since It is difficult

to find analytic solutions to integral equations for the general domain. We may think that all we've done is convert one hard problem (involving derivatives) to another hard problem (Involving integrals). This is true In a 5ense, but there is another side

to the story. It turns out that Integral equations are better

to work wi th f rom a t heor e t I ca I po I nt of v I ew, Suppose we

aren't so interested in finding the analytic solution (maybe we can't find it) as we are in knowing if there is a solution and what kind of assumptions of the boundary condition and boundary

we must make in order for a solution to exIst. It's certainly

not true that all BVPs even have solutions, and many times, we would like to know what physical conditions we must assume in order to get solutions to our equation. Integral equations are better to work with when these questions about existence and uniqueness of solutions have to be answered.

Problems

1. Prove the alternative form of the double-layer potential can be written

W(x,y)

~ cos 8 r

given that we start with the first form (as given in the text).

2. Plot the dipole potential due to two charges +q and -q at

a distance d apart.

3. Derive the Fredholm Integral equation for o(s) in the Neumann problem. Draw a picture describing the relevant variables.

Other Reading

1. Partial DIfferential Equations by I. G. Petrovsky.

Interscience Publishers, 1957. Contains a nice section on potential theory, describing the single and double-layer potentials.

2. 1962.

Partial Differential Equations by B. Epstein. McGraw-Hill, Chapter 7 of this text discusses many techniques for

solving the Dirichlet problem in general domains.

89

b

4>(x) - II f K(x,s) 4>(s) d s

a

f(x) a <; X < b

Additional Lesson C: Integral Equations

PURPOSE OF LESSON: TO INTRODUCE THE BASIC FREDHOLM INTEGRAL EQUATION

WHERE THE UNKNOWN FUNCTION ¢(x) OCCURS UNDER THE INTEGRAL SIGN AND TO SHOW HOW THIS EQUATION CAN BE SOLVED FOR CERTAIN KERNEL FUNCTIONS K(x,s). IN PARTICULAR, IF K(x,s) CAN BE WRITTEN IN THE FORM (or approximated by)

K(x,s}

n

r al(x) b.(s)

;=1 I

THEN THE INTEGRAL EQUATION CAN BE REDUCED TO A SYSTEM OF LINEAR ALGEBRAIC EQUATIONS.

FINALLY, A SPECIFIC DIRICHLET PROBLEM IS SOLVED BY CONVERTING IT TO AN INTEGRAL EQUATION BY MEANS OF THE POTENTIAL-THEORY TECHNIQUES DISCUSSED IN LESSON C.

Lesson C:

Integral Equations

Linear integral equations can be interpreted as extensions of systems of linear algebraic equations to an infinite continuum. For example, the linear system of n equations with n unknowns Xl :

n

l: a'J xJ j=1 I

b.

I

1,2, ... n

has the continuous analog

b

f A(x,s) f(s) cis ~ b(x) a

a<)«b

where now the unknown f(s) is a function, and the continuous

'30

variables x and 5 take the place of the dIscrete variables I and j. The equation where the functIons A(x,s) and b(x) are as sume d known and f (5) Is unknown is ca II ed all i nteg ra I equation, and It 15 an example of the general Fredholm Integral equation

(I)

b

¢(x) - \ J K(x,s) ¢(s) ds

a

f [x )

a < x < b

The general names for the variables are:

f{x) = right-hand side of the integral equation K(x,s) - kernel of the integral equation

\ - known constant (orlen an eigenvalue)

¢(s) ~ unknown functIon (everything else Is known)

and a couple of examples are

$(X) -

1

J s l n txs ) q,(s) ds o

o < x <

¢(X) __ ( t~~) d s

sin x

-c..:) < x < iXt

The la5t example is a sillgular-integral equation, since th~ kernel and domain are unbounded.

We will now find the solution to integral equation (I) for those cases where the kernel K(x,s) can be approximated by an expression of the form

K(x,s)

n E j=1

a. (x) b. (s )

J J

(degenera te kerne I) .

Solving Integral Equations with Degenerate Kernels Suppose the kernel K(X,5) in the Fredholm equation

'11

b

~(x) - f K(x,s) ~(5) ds a

fix}

a < x < b

ca~ be written (or approximated) as

K(x,s)

n

l: a. (x) b. (5)

J~l J J

Substituting this expression In the Integral equation, we get

~

<jJ (x) - Ea. (x) j~l J

b

f b.(s) <p(s) cis ~ f(x) a J

or

(2)

n

<p(x) - ~ c. a.(x) = fix) j=l J J

where

c. j

b

f b. (5) <p(s) cis a J

If we now multiply equation (2) by each of the functions b1(x), b2(x), .. ,' \(x) and integrate the resulting equation with respect to x, we get the following system of algebraic equa t ions:

b n b b
(3) f <P(s) b I (s ) d 5 - l: c. f b. (5) a.(s)cis f f(s) b. (5) ds
j= 1 J 1 J 1
a a a
{i=i ,2,. ,. n}
or in more compact form n
(4) c. l: k .. c, f. i'" 1 ,2, ... n
1 j=l I) J 1
where
b
k .. f b. (s) a. Is) cis
IJ 1 J
a
b
f. J f (5) bl (s ) cis
1 a ')2

We can now solve thIs system of equations (providing It is a nonslngular system having a nonzero determinant) for the numbers ci and then substitute these values into equation (2) to find the solutIon ~(x) • That Is,

n

~ (x) = f (x) + 1: j=1

c. a. (x) J J

Remember that a. (x) and b. (xl a re as sume d known [they can be

j J

computed from the kerne J J • ! n other words, if the kerne J is

degenerate and if there is a solution, then it has the form

<jJ(xl

f(x) +

c. a. (x) J J

where the c.s are found by solving a system of algebraic J

equatIons. We will now give an example.

Solution to an integral equation

Suppose we have the equation

I

¢(x) - f XI ~(5) ds o

o < x < I

We know from equation (5) that the solutIon has the form

n

¢(x) m f(x) + 1: c. a.(x) j=l J J

(c an unknown constant)

and 50, substituting it in the integral equation, we get

I

)(2 + ex - f xs (s' + csl ds )(2

o

EvaluatIng this integral, we arrive at c = 3/8, and, hence, the so l ut l on is

93

cp(x) x~ + ~ X

We can easIly check this answer. There are more methods for solving other types of integral equations, and a few of them can be found in reference 2 of the recommended reading. We will now use ou r techn i que too 501 ve a prob 1 em in POE.

Dirichlet Problem

Suppose we wIsh to solve the Dirichlet problem (which we have solved before)

POE

o

(inside the unit circle)

Be

u{I,e) '" He)

o < e <: 211

We know the solution can be written as a double-rayer potential

u(r,O)

where Rl = r~ - lrpcos(e-cp) + pl and ~(e) satIsfies the integral equation

(6)

I 111 () 1

~(6) - - f - InH ~(<fl) dcp

11 0 <lP R

1 --f(6)

11

o <: f) <: lT1

1 f we now compute the kerne J (wh 1 ch w j 11 be eva I us ted on the boundary r ~ p ~ 1), we get

<I 1 I

- In(-)

dp R 1

p=

and so we arrive at the integral equatIon

l'll

~(e) + 1- f ~(CP) dCP '"

211 0

- L f (EI)

11