• Nie Znaleziono Wyników

The Quarterly of Applied Mathematics, Vol. 1, Nr 4

N/A
N/A
Protected

Academic year: 2022

Share "The Quarterly of Applied Mathematics, Vol. 1, Nr 4"

Copied!
84
0
0

Pełen tekst

(1)

Q U A R T E R L Y

OF

A P P L I E D M A T H E M A T I C S

H. L. DRYDEN J. M. LESSELLS

H, BATEMAN J, P. DEN HARTOG j. N. GOODIER R. V. SOUTHWELL

E D IT E D BY

T. C. FRY TH.

W. PRAGER L S.

J. L. SYNGE

w i t h t h e c o l l a b o r a t i o n o f

M, A. BIOT L. I

H. W. EMMONS K. (

F. D. MURNAGHAN W.

G. I. TAYLOR S. P

,-U? k

if ’

v. KARMAN SOKOLNIKOFF

. BRILLOUiN L FRIEDRICHS I, SEARS . TIMOSHENKO

Vo l u m e I J A N U A R Y ; 1944 / Nu m b e r 4

■ J Q 4 3

l

}

(2)

Q U A R T E R L Y O F

A P P L I E D M A T H E M A T I C S

T h is periodical is published u n d er th e sponsorship of B row n U n iv ersity . F o r its su p p o rt, an op eratio n al fu n d is being s e t u p to w hich in d u strial o rg a n iz atio n s m a y co n trib u te. T o d ate, co n trib u tio n s of the following in d u stria l com pan ies are g ratefu lly acknow ledged:

Bel l Te l e p h o n e La b o r a t o r ie s, In c.; New Yo r k, N . Y ., Th e Bristo l Co m p a n y; Wa t e r b u r y, Co n n.,

Ge n e r a l Electric Co m p a n y; Sc h e n e c t a d y, N . Y .

Gu l f Re se a r c h a n d De v e lo pm e n t Co m pa n y, Pit t s b u r g h, Pa. Le e d s & No r th rup Co m p a n y; Ph il a d e l p h ia Pa.

Pr att & Wh it n e y, Div is io n Nil e s- Be m e n t-Po nd Co m p a n y; We st Hart­

f o r d, Co n n.,

R e p u b lic A v ia t io n C o r p o r a t io n ; F a r m in g d a le , L o n g I s l a n d , N . YV, U n it e d A i r c r a f t C o r p o r a t io n ; E a s t H a r t f o r d , C o n n .

T h e Q u a r t e r l y p rin ts original p apers in ap p lied m a th e m a tic s w hich h av e an in tim a te connection w ith application in in d u s try o r p ra c tic a l science. I t is ex­

pected t h a t each p ap e r will be of a high scientific s ta n d a r d ; t h a t th e p re se n ta tio n , will be of such ch a rac te r t h a t th e p ap e r can be easily re ad b y those to w hom it w ould be of in te re st; an d t h a t th e m a th e m a tic a l arg u m e n t, ju d g ed b y th e s ta n d a rd of the field of ap p licatio n , will be of an a d v a n ced ch a rac te r.

Manuscripts submitted for publication in the Qu a r t e r l y o f Ap p l i e d Ma t h e m a t ic s should be sent to the Managing Editor, Professor W, Prager, Quarterly of Applied Mathematics, Brown University, Providence 12, R, I., either directly or through any one of the Editors or Collaborators.

In accordance with their general policy, the Editors welcome particularly contributions which will be of interest both to mathematicians and to engineers. Authors will receive galley proofs only.

Seventy-five reprints without covers will be furnished free; additional reprints and covers will be supplied at cost.

The subscription price for the Quarterly is §6.00 per annual volume, single copies §2.00.

Subscriptions and orders for single copies may be addressed to: Quarterly of Applied Mathematics, 450 Ahnaip St., Menasha, Wisconsin or to Brown University, Providence 12, R. I.

Pr i n t e d b yt h e

Ge o r g e Ba n t a Pu b l is h i n g Co m p a n y Me n a s h a, Wis c o n s in

(3)

Q U A R T E R L Y

OF

A P P L I E D M A T H E M A T I C S

E D I T E D B Y

H. L. DRYDEN J. M. LESSELLS

T. C. FRY W. PRAGER J. L. SYNGE

TH. v. KARMAN I. S. SOKOLNIKOFF

H. BATEMAN J. P. DEN HARTOG J. N. GOODIER R. V. SOUTHWELL

W I T H T H E C O L L A B O R A T I O N O F

M. A. BIOT H. W. EMMONS F. D. MURNAGHAN G. I. TAYLOR

L. N. BRILLOUIN K. O. FRIEDRICHS W. R. SEARS S. P. TIMOSHENKO

V o l u m e

I /V*- ^ • *943

(4)

Printed by the

Ge o r g e Ba n t a Pu b l is h i n g Co m p a n y

Menasha, Wisconsin

(5)

CONTENTS

W . S. A m en t: T h e lines of principle stress in th e plane problem of p la stic ity . 278 H . B a te m a n : T h e tra n sfo rm a tio n of p a rtia l differential e q u a tio n s . . . . 2 8 1 L. B e rs : C oncerning th e acceleration p o t e n t i a l ... 93

an d A. G e lb a rt: On a class of differential eq u a tio n s in m echanics of c o n t i n u a ... 168 L. B rillo u in : T h e a n te n n a p r o b l e m ... 201 C. L. B row n : T h e tre a tm e n t of d isco n tin u ities in beam deflection prob lem s . 349 W . Z. C h ien : T h e in trin sic th e o ry of th in shells a n d p l a t e s ...297

■ an d A. W e in ste in : On th e v ib ra tio n s of a clam ped p la te u n d e r tension 61 L. H . D o n n e ll: A c h a rt for p lo ttin g re la tio n s betw een v aria b les over th e ir e n ­

tire real r a n g e ... 276 H . L. D ry d e n : A review of th e sta tis tic a l th eo ry of t u r b u l e n c e ... 7 W . F e lle r: On A. C. A itk e n ’s m eth o d of i n t e r p o l a t i o n ...86 K. O. F rie d ric h s an d J . J . S to k e r: F orced v ib ra tio n s of sy stem s w ith n on­

lin ear re sto rin g f o r c e ... 97 R . E. G a sk e ll: On m o m en t b alancing in s tru c tu ra l d y n a m i c s ...237 A. G e lb a rt: (See L . B ers)

G. H . H a n d e lm a n : A v a ria tio n a l principle for a s ta te of com bined p lastic stress 351 G. E. H a y an d W . P ra g e r: On plane rigid fram es loaded p erp en d icu larly to

th e ir p la n e ... 49 M . H e rz b e rg e r: A d ire c t im age erro r t h e o r y ...69 N. J . H o ff: A stra in energy d e riv a tio n of th e torsional-flexural bu ckling loads

of s tra ig h t colum ns of th in -w alled open s e c t i o n s ...341 T h . v. K a rm d n : T ooling up m a th e m a tic s for en g in e e rin g ... 2 P . W . K e tc h u m : On th e d isco n tin u o u s flow aro u n d an airfoil w ith flap . . . 149 W . M . K incaid an d V. M o rk o v in : A n ap p lica tio n of o rtho g o n al m o m en ts to

problem s in sta tic a lly in d e te rm in a te s t r u c t u r e s ... 334 Y. H . K u o : On th e force a n d m o m en t a c tin g on a b o d y in sh ea r flow . . . 273 J . L e h n e r an d C. M a r k : A n ap p lica tio n of th e m eth o d of th e acceleration

p o t e n t i a l ... 250 C. C. L in : On th e m otion of a pen dulu m in a tu rb u le n t f l u i d ...43 S. L u b k in an d J . J . S to k e r: S ta b ility of colum ns an d strin g s u n d e r periodically

v a ry in g f o r c e s ...215 C. M a r k : (See J . Lehner)

V. M o rk o v in : On th e deflection of an iso tro p ic th in p l a t e s ...116 ---(See W . M . K i n c a i d )

W . P r a g e r : (See G. E . H a y )

W . C. R a n d e ls : A new d e riv a tio n of M u n k ’s f o r m u l a e ... 88 K . R ie s s : E lec tro m a g n etic w aves in a b e n t pipe of re c ta n g u la r cross section . 328 S. A. S ch elk unoff: T h e im pedance of a tra n sv e rse w ire in a re c ta n g u la r w ave

g u i d e ... 78 On th e a n te n n a p r o b l e m ... 354 H . J . S te w a rt: P eriodic p ro p e rtie s of th e sem i-p erm an en t atm o sp h eric pressure

s y s t e m s ... 262

(6)

J . J. S to k e r: (See K . 0 . Friedrichs)

;----(See S . L u b k i n )

A. G. S tra n d h a g e n : Use of sine tran sfo rm for non-sim ply su p p o rte d beam s . 346 J . L. S y n g e: O n H e rzb erg er’s d irec t m ethod in geom etrical op tics . . . . 268 H . S. T s ie n : S ym m etrical Jouk o w sk y airfoils in sh ear f l o w ...130 A. W e in ste in : (See IV. Z . Chien)

Book re v ie w s ... 189

(7)

281

Q U A R T E R L Y O F A P P L I E D M A T H E M A T I C S

V o l. I J A N U A R Y , 1 9 4 4 N o . 4

THE TRANSFORMATION OF PARTIAL DIFFERENTIAL EQUATIONS*

B Y

H. BATEMAN

California Institute of Technology

1. In tro d u c tio n . In th e early stages of th e use of p a rtia l differential eq u a tio n s for the solution of problem s of m echanics an d physics th e sep a ratio n of v ariab les and constru ctio n of sim ple solutions was the p rim a ry aim . T h e in tro d u ctio n of the idea of an exact differential b y F o n ta in e an d E u ler led to th e idea of associated differential eq u atio n s such as those for the velocity p o ten tial and strea m fun ctio n in h y d ro ­ dynam ics, th e a d jo in t eq u a tio n s of L agrange a n d R iem an n, th e c o n ta c t tra n sfo rm a ­ tions of L egendre an d A m pere, th e tran sfo rm atio n s of E u ler an d L aplace for the solution of differential eq u a tio n s by definite in teg rals an d o th e r tran sfo rm atio n s too num erous to m ention. A n o th er aim w hich led to th e s tu d y of tran sfo rm atio n s was t h a t of reducing an eq u a tio n to a canonical form . L ap la ce’s red u ctio n of a lin ear p a r­

tial differential eq u a tio n of th e second o rd e r to a form in w hich only one p artia l d eriv a tiv e of th e second o rd e r occurs led to th e s tu d y of tran sfo rm atio n s w hich p re­

serve this form an d of q u a n titie s w hich have a p ro p e rty of invariance. C onditions were th en found t h a t an eq u a tio n m ay be reducible b y m eans of a specified ty p e of change of v ariables to some p a rtic u la r eq u a tio n s which h ad been fully stu d ied . T h e conditions found by C am pbell [l ] t (constan cy of his tw o in v a ria n ts) t h a t L a p la c e ’s canonical eq u a tio n m ay be reducible to th e eq u a tio n of E u ler and Poisson, m ay be cited as an exam ple.

A classification of tran sfo rm atio n s m ay be m ade by including in group A all tran sfo rm atio n s w hich arise from th e condition o r conditions t h a t a lin ear differential form m ay be of a specified ty p e (for exam ple an ex a c t differential). T ra n sfo rm a tio n s arising from th e s tu d y of a n u m b e r of lin ear differential form s m ay be included in this group. T ra n sfo rm atio n s associated w ith th e C alculus of V a riatio n s are also included because th e eq u a tio n s of E u ler and L agrange are closely asso ciated w ith th e condi­

tions for an ex a ct differential. T h e extension of L egen d re’s tra n sfo rm a tio n found b y C ara th e o d o ry [2] m ay be m entioned here. In this article a tte n tio n will be d evo ted a l­

m ost en tire ly to tran sfo rm atio n s of group A.

T ra n sfo rm atio n s of g roup B include all those w hich arise from th e con ditio ns t h a t a q u a d ra tic differentia] form m ay be of a specified typ e. T h e tra n sfo rm a tio n of a

* Received March 2, 1943.

f Numbers in square brackets refer to the list of references at the end of the paper.

(8)

282 H . B A T E M A N [V ol. I , N o . 4

linear differential eq u atio n to a form in w hich the v ariables are se p a rated is th u s a B- tran sfo rm atio n . T h e tran sfo rm atio n s of g roup B are n o t necessarily p o in t tra n sfo rm a ­ tions, for instance, if Qia, b, c) is a non-negative q u a d ra tic form in th e real variables a, b, c tran sfo rm atio n s from (x, y , z, t, u, v, w) to (X , Y, Z , T , U, V, W ) m ay be con­

sidered in w hich Q{dx — udt, d y —vdt, dz — wdt) goes over into Q ( d X — Udt, d Y — Vdt, d Z — W d t) th e coefficients of Q in the first case being fun ction s of y , z, t, u, v, w and in the second case functions of X , Y, Z , T , U, V, W . Since the eq u a tio n <3 = 0 im plies th a t d x = udt, d y = v d t , d z = wdt it also im plies t h a t d X = U d T , d Y = V d T , d Z = W d T . In o th e r w ords if u, v, w can be regarded as th e co m p o n en t velocities of a recognizable m oving p article of fluid th en U, V, W can be reg ard ed as co m p o n en t velocities of a recognizable p article of a corresponding fluid. Such a tran sfo rm atio n is of in te re st because the d en sity of each fluid can be defined in such a w ay t h a t th e eq u a tio n of c o n tin u ity is in v a ria n t u n d er th e tran sfo rm atio n .

G roup C m ay be regarded as including all o th e r tran sfo rm atio n s an d som e tra n s ­ form ation of the o th e r g roup w hich arise in th e reduction of an e q u a tio n to a canonical form .

2. A ssociated equations of th e typ es of M onge an d L eg en d re. In his w ork on p a r­

tial differential eq u atio n s of the second o rd e r in tw o variables x , y w hich can be re­

g arded as indep en d en t, M onge [3] used z as d e p e n d e n t variable, p an d q as th e first d eriv a tiv e s zx, z„ respectively an d r, s, t as th e second d eriv ativ es z zx, zxy, z vy. As we shall have app licatio n s to fluid dynam ics in m ind, we shall d ev iate sligh tly from the n o tatio n of M onge an d use u, v in place of p and q so t h a t when z is th e velocity po­

te n tia l u an d v rep resen t th e co m p o n en t velocities as usual. T h is plan also allows us to use th e sym bol p to d en o te th e pressure a n d q to d en o te th e re s u lta n t velocity.

T h e eq u atio n s of ste a d y m otion of a com pressible fluid u n d er no body forces when the flow is irro ta tio n a l an d th e fluid b aro tro p ic (d en sity a function of pressure only) can, as we know, be deriv ed from a v aria tio n a l principle

') d x d y = 0, u — z x, v = s„, (1)

in w hich p is a specified function of q. F o r g re a te r g en e rality a t th e o u tse t we shall suppose, how ever, t h a t p is a specified function of u an d v. L a g ran g e’s p a rtia l differ­

ential eq u atio n for this v a ria tio n a l problem is th en d eriv ab le from H a a r's condi­

tion [4] th a t p v d x — pudy should be an exact differential. W hen th e differen tiatio ns are m ade, th e eq u atio n has the form

puuf + 2 p uvs + p vvt = 0. (2)

W hen L egend re’s tran sfo rm atio n is applied to th is differential eq u a tio n th e new d e­

p e n d e n t v ariable is

w = u x + vy — z (3)

and since d z = u d x + v d y ,

du> = x d u + ydv. (4)

W hen u an d v can be regarded as in d ep en d e n t th is eq u atio n gives th e relation s

x = y = wv, (5)

(9)

an d th e e q u a tio n for w is

p u u ^ w 2.p Ul'Wuv T pvv^uil ~ 0. (6)

W hen, how ever, u an d v are related so th a t th ey can be regarded as fu nctio ns of a single v ariab le r th e eq u atio n (4) indicates t h a t w is then also a fun ctio n of r an d we have th e equ atio n s

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 283

w { t ) = x u(t) + yv(r) — s, w ' ( t ) = x u ' ( t ) + yv'ir), which furnish a solution of (2) if

PuuU,2( t) + 2p Uvu ' ( t ) v ' \ t ) + Pvx-v'-(t) = 0.

L e t us now seek the conditions th a t 3 q u a n titie s

R = R ( u , v), S = S ( u , v), T — T (u , v),

(7)

(8)

(9) m ay be the second d eriv ativ es Z zx, Z xy, Z21 of a single function Z ( x , y ) . T h e re q u ired conditions R V = S X, S y = T x m ay be w ritten in the form

R,,s T R vtS ur + S t.s, S„s + S rt = T ur + T rs. (10) W e now seek the conditions th a t these tw o eq u atio n s arc b o th satisfied in v irtu e of eq u a tio n (2). T h is will be th e case when

R v 1L‘p vr) R u wp ltv ~i~ by Su w p„ u, Svb IVp y, , Tu = w'puu, T t = w ' p uv + li'y S„ = h'w ' p uv, S r = — w ' p vv.

E q u a tin g th e different expressions for R uv, 5 UC, T uv we o b tain the e q u a tio n s llyi — Wyypyyy. lVVp Utly IlVWyiPw >

bu 1C r p uupuvy b v — W i; p u v Wu pvt"

T h e elim ination of h and h ' yields the tw o eq u atio n s IVuuprv 2.wurp ur T w vvp uu — 0, IVuUpW 2 Wyirpny “(~ Wvrpytu — 0

(11)

(12)

(13) which show th a t w and w ' are solutions of eq u a tio n (6). T h e case of chief h y d ro d y - nam ical in te re st is t h a t in which th e second d eriv a tiv e s of w an d w ' are all zero. W e shall, how ever, look first a t a possible a lte rn a tiv e case.

E q u a tin g the different expressions for S u an d S v we o b tain th e eq u a tio n s

H ence

Wpuu = w'pur +

J

[»„' d ( p r)w ld{pu)]y

w'pvv = Wpuv +

J

[wud(pr) — w t d(p„)].

Wv prv + w'purv = 2 w vpuv + Wpuvr ~ Wup,.v, 2wû puv — Wv pu u + w ' p uuv = Wupuu + Wpuuu,

WÛ prv + w'Purv = Wvpuu + Wpuuv

(14)

(15)

(10)

T h ese e q u a tio n s lead to th e relation

w 'D v = w D u, where D = p uup xv — p l v. (16) T his eq u atio n is satisfied iden tically w hen D is c o n sta n t b u t it m ay also be satisfied if w = E„, w ' = E U w here D an d E are related. An im p o rta n t case of this second ty p e occurs when D is a function of q only a n d w = v , w ' = u. In th is case

h — — p u, h' = — p v, R = v p v — p, T = u p u — p, S —u p v — — vp„, (17) an d p is a function of q only. If p u = — up, p v = —vp, w here p is th e d en sity of th e fluid we have

R = — p — pv2, S = puv, T — — p — pu2, (18) and so Z is a kind of stress function w hich satisfies th e eq u atio n

R T - S 2 = p ( p + pu2 + pv2) = p (p + pq2) = F ( R + D (19) since R - \ - T = — 2p —pq2. T h e p a rtia l differential eq u atio n for Z is th u s of L egen d re’s ty p e [5]

3C(R, 5, T ) = 0. (20)

In the special case in which — F ( R + T ) = K Z — |( i ? + 7’) 2 th e eq u atio n reduces to one w hich occurs in S a in t V e n a n t’s th eo ry of plastic bodies. T h is e q u a tio n has been discussed by H en ck y [6], P ra n d tl [7] and C ara th e o d o ry [8]. Oseen [9] uses th e m e th ­ od of L egendre in which th e eq u a tio n is first solved for R , d ifferen tiated w ith resp ect to y an d so reduced to an eq u atio n

(.K2 - V % y \ V XX - Vyy) + 2 V XV Xy = 0 (21) in which

S = V X, T = Vy.

I t should be rem arked t h a t if p x — u = z x, p u = ~ v — —z y,

p + pu2 + pu2 = p, 1 /p — — p, (22) we m ay w rite

R —ppi'2, S = puv, T = — ppii2, (23) an d the eq u atio n s R V = S Z, S y — T x lead to th e p a rtia l differential e q u a tio n for th e strea m -fu n ctio n z.

In th e th eo ry of plane w aves of finite am p litu d e eq u a tio n s of L eg en d re’s ty p e oc­

cu r in a t least tw o w ays one of w hich is discussed b y J. R. W ilton [10]. In th e o th e r w ay use is m ade of th e eq u a tio n s

R = p, S = — pu, T — p + p u 1, (24)

where now y denotes the tim e an d x- a co-o rdinate in th e direction in w hich th e w aves are travelling. T h e q u a n titie s u, v are again th e d eriv a tiv e s of a velo city p o ten tial z, p is th e d en sity , p th e pressure an d u th e velo city of th e fluid. T h e q u a n titie s R , S , T are th e second d eriv a tiv e s of a stress-function Z . T h e ad d itio n al eq u atio n s from w hich the relation betw een R , S an d T m ay be derived, are

v + h i2 = f'(p), P = /(p) — pf'(p)- (25)

2 8 4 H . B A T E M A N [V o l. I , N o . 4

(11)

T h e desired relatio n is th u s

T - R ~ lS2 = /(R) - R f ' ( R ) . (26) T h is eq u atio n , like t h a t considered by W ilton, m ay be solved by th e m eth od of L egendre in w hich th e eq u a tio n is differen tiated w ith re sp ect to one of th e in d ep en d ­ e n t variables (in th is case a;) so as to reduce it to an eq u a tio n of th e M onge-A m pere type. T h e tran sfo rm atio n to the new e q u a tio n can be regarded as a special B acklund tran sfo rm atio n [ l l ] as Oseen [9] observes. If U = Z X, th e new eq u a tio n is

U yy - 2{ U y/ U z) U zy + ( U l / U l ) U zz = - R f " { R ) U zx = c2U zz • (27) or

I I U ZZ + 2 K U zy + L U yy = 0,

w here I l = { U2v/ Z f x) - c 2, K = - ( U y/ U z), Z = 1. T h e in v a ria n t G is

G = K - — H L — c~ (28)

an d th e condition Gt^O is satisfied so long as c - ^ O . In the p re sen t case

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 285

(29) R u = — p u / c2, R v = — p / c2, S u = p(u2/ c2 — 1)S„ = pii/c2,

r „ = Pu( 1 - u2/ c2), T v = - p( 1 + m2/ c 2) . T h e tw o eq u a tio n s (10) are b o th e q u iv a le n t to

(u2 — c2)r + 2 us + 1 = 0, (30)

an d it is readily seen th a t w — — 1, w ' =u, g = k <l — hl = c!-7i §.

I t should be noticed t h a t if we solve eq u atio n s (9) for u an d v in the form

u = F ( R , S ) , v = G(S, T ), (31)

the eq u a tio n u y = vz is satisfied on ac co u n t of R y = S x, S y = T x if

Fr = Gsi F s — Gt. (32)

T hese tw o eq u atio n s th en are consequences of th e single e q u a tio n 3C(R, S, T ) = 0 . T h e expression of such an e q u a tio n in th e tw o form s (32) m ay be reg ard ed as a p ro blem of som e in te re st.

In th e case w hen D is a c o n s ta n t an d p is a function of q only

D = p qp j q = P2[l ~ (? A )2]- (33)

an d th e flow is e ith e r en tire ly subsonic ( D > 0 ) o r en tire ly supersonic (D <0). In m an y cases in w hich p is a function of q only, D can have e ith e r sign an d so th e flow is p a rtly subsonic an d p a rtly supersonic. I t is then of som e in te re st to seek th e co n d i­

tion satisfied by th e function 3C(R , S, T) w hen D> 0 . F o r this purpose we w rite the e q u a tio n in th e form

0 = 3C{R, S , T) = [(£ - T )2 + 4 5 2] I/2 - J { R + T ) , (34) w here / is a function w hich is such th a t

-pf'(p ) — ~ J [ ~ 2/(p)]. (35)

(12)

H ere p = f ( p ) —pf'(p) is the relation betw een th e pressure p an d d en sity p. Now we find by differentiation th a t

3C„ + X T - 2 = - 2 - 2J ’(R + T ) = - 4 - 2 p /"(p )//'(p ) = 4(cV ?2 ~ D- (36) H ence 3C/i + 3Cj- — 2 > 0 when c2> g 2and 3C« + 3Cr — 2 < 0 when c2< q 2. In th e case of th e p lastic e q u a tio n J is a c o n s ta n t an d so

3C« 3Cr = 0. (37)

T h e corresponding flow is ch aracterized b y the relation q- = 2c- an d is co n seq u en tly supersonic.

A sim ple case in w hich D is c o n sta n t is o b tain ed b y w riting

p = a n2 + 2 cuv + fa.’2, (38)

w here a, b and c are c o n stan ts. T h e functions iv, w ' b oth satisfy

bw„u + a w vv2cwUv = 0, (39)

and we m ay w rite w — b V u, w ’ = a V v, where V is a solution of this eq u a tio n . If a fu nc­

tion W is defined b y th e eq u atio n s

W u = c V u -

avv,

W v = b V u - c V „ (40) we m ay w rite h — 2 b W u an d it is read ily found th a t we can w rite

i? = 2b(lV + cV), S = - 2abV, T = 2 a ( W + cV), (41) w here V an d W are connected b y th e foregoing eq uatio ns. In th is case the relatio n betw een R , S an d T is sim ply

a3C = a R - b T = 0. (42)

T h e q u a n tity 3Cb + 3Ct —2 is now sim ply — (a+Z>)/a, a c o n sta n t. T h ere is no change in sign of th e expression. I t will be noticed t h a t th e eq u atio n s a R —b T = 0 an d

ar + 2 cs + bt = 0 (43)

sa tisfy the cond ition of a p o la rity

A b + B a - 2Cc = 0, (44)

w hen the first eq u a tio n is w ritten in th e form A R - \ - 2 C S - \ ~ B T = 0.

3. T h e tran sfo rm atio n of th e M onge-A m pere equation. If for th e e q u a tio n hr + 2 k s;'+ It -T m + n{rt — s2) = 0, (45) th e expression

g = k2III -f- m n (46)

is n o t zero a n d so th e tw o system s in the m eth o d s of M onge an d Boole are d istin c t, th e eq u atio n is tran sfo rm ed b y a c o n ta c t tran sfo rm atio n

2 8 6 . H . B A T E M A N [V o l. I , N o . 4

X = X ( x , y, z, u, v), Y = Y ( x , y, z, u, v), Z - Z { x , y, z, u, v), \ U = U{x, y, z, 11, v), V = V ( x , y, z, it, v), dZ = U d XV d Y = a{dzu d xvdy) j (47) in to an eq u atio n

H R + 2 K S + L T + M + X ( R T - S 2) = 0 (48)

(13)

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 287

for w hich th e q u a n tity G = K i - I i L - { - M N = 0.

In a p ap e r published in 1904 S ophus L ic -[l2 ] rem ark ed t h a t it would be d esirab le to have a d irec t proof of this th eorem an d K iirsch ak [13] gave one based upon a re p resen tatio n of the eq u atio n in th e form of a Jacob ian

d{a, b )/d { x , y) (49)

w here a an d b are functions of x , y , z, u, v a n d d / d x = d / d x - \ - u { d / d z ) -\-r{d /d u ) -\-s{d/dv), d / d y = d / d y + v ( d / d z ) -\-s{d /d u ) -)rt{d/dv). W hen th is re p re se n ta tio n is n o t used th e proof is algebraically m ore difficult b u t th e an aly sis is w o rth giving on ac­

co u n t of th e n um erous re la tio n s to w hich it leads. R eference for th is ty p e of proof m ay be m ade to a p ap e r b y R. G arn ier, S u r la transform ation des dérivées secondes d a n s les transform ations de contact et les transform ations ponctuelles, B ull, des Sci. M a th . (2),

64, 13-32 (1940).

W e shall suppose th a t dz = udx-\-vd y an d th a t co n seq u en tly dZ = U d X - \- V d Y . T o m ake d U = R d X + S d Y , d V = S d X + T d Y consequences of d u = r d x + s d y , d v = s d x f - t d y we shall require th a t

d U - R d X - S d Y = ( U u - R X U - S Y f ) { d u - r d x - sdy) + ( U vR X V — S Y u)(dvs d xtdy), d V - S d X - T d Y = (Vu - S X U - T Y u){dn - r d x - sd y)

+ ( F , - SX„ - T Y v)(dv - s d x - tdy).

W ith the n o ta tio n

(50)

Z y + v Z z, etc., {uv) = U v

(vv) = V v

R X V s x v -

■ S Y V, T Y V,

r{uv) + s{vv) + Vi - S Xi - T Y i = 0, s{vu) + t(vv) + F 2 — S X2 — T Y 2= 0.

Z i — Z i + uZz,

{uu) = Uu - R X U - S Y U {vu) = V u - S X u - T Y u th e eq u atio n s to be satisfied are

r{uu) + s(uv) + U i - R X i - S Y i = 0,- s{uu) + t{uv) + Ui - R X2 ~ T Y i = 0,

H ence

rA = U vV x- U1V v+ R { X1V v- FiA%) + 5 ( 7 ^ , + U xX v- X xU v- F i F ,) + T{ U \Yv - IT U v) + ( R T - S > ) { Y xX v - ATK.),

tA = U 2Vu - V i U u+ R { X UV2 - X2V U) + S { X2U u + Y u V2- X v U2- Y 2V u) + T { Y2U u- Y u Ut) + { R T - S * ) { X2Yu - X u Y 2) ,

s A ^ U xV u - V1U u + R { X u V1- X 1V u ) + S { X1U u + Y u V1- X u U1- Y 1Vu) + T { U u Y1- U 1Y u) + { R T - S * ) { X1Y u- X u Y 1),

sA = V2U v- U2V v+ R { X2Vv- X z V 2) + S { U2X V+ Y i V , - X t U , - V t Y r) + T{ U2Y v - Y2 U v) + { R T —S2) { Y2X V - X 2Y v) ,

A { r t - s 2) = U xV 2- ¿72F i + R ( F 1A 2- F 2AT) + T { U2Y y- U xY 2)

+ 5 (U2X x- U xX2+ V xY2- V2Y x) + (R T - S2)(A iK *- X2Y x) , A - U uV v- U v V u + R { X vV u - X u V v) + T { U vY u - U u Y v)

+ S { Y vV u- Y u V v+ U vX u - U u X l) + { R T - S ' - ) { X u Y v- X vYu).

(5 1)

(52)

(53)

(14)

T h e two expressions for s are eq u iv ale n t on ac co u n t of th e relatio ns

[.U V] = [ X V ] = [ Y U ] = [X V] = 0, [AT£/] = [ Y V ] = a (54) w hich, in ad d itio n to th e relation s [F Z ] = [ Z X ] = 0, \ U Z ] =crU‘[ V Z ] = a V are s a tis ­ fied because th e tran sfo rm atio n is a c o n ta c t tran sfo rm atio n . In these re la tio n s [.A B] is th e Poisson b ra c k e t

[A B ] = A uBx - A t . B u + A vB2 - A2B V. (55) T h e relations arc derived b y Lie [12] b y a clever device. In th e book of C erf [14] th e relations are derived from the eq u atio n

<t\ A B ] = [a i] (56)

w here A , B are the expressions for a (x, y , z, p, q), b(x, y , z, p, q) in th e new co-ordinates [AT, Y, Z , P , Q \, while F. Engel [15] o b tain ed th em w ith th e aid of th e bilin ear co­

v a ria n t b y a dev elo p m en t of a m etho d used by G. D arboux.

I t is read ily seen t h a t th e e q u a tio n h r ~ Y ^k s-\-lt-\-n i-\-n (rt—s 2) = 0 becom es H R + 2 K S + L T + M + N ( R T - S ) = 0 , w here

H = H P, + k ( P b + P q) + IP a + m P r + n P c,

2 K = h ( R p + Cp) + k ( C b + R b + Cq + R q) + l { R a + C.)

+ m ( R r + Cr) + n ( R c + Cc), L = hA p -f- k ( A b + A q) -j- IA a -f- tnA r T u A c,

M = hQp -f- k{Qb + Q,,) + IQ a + tnQr + nQc, A' = hBp + k ( B b + B q) -T l B a T- m B r T n B c, w here

2 8 8 H . B A T E M A N [V o l. I , N o . 4

A a = U UY2 - u2y u, A i U u Y i - U \ Y U, A c = U2Yi - UxIT, A p = UyYv - U VY x, A q = U2Y v - u q y2, A r = U VY U- E T F „ Ba = X2Y U - x uy2, B b = X iT « - X PY h B c = X i Y2- AT2Fx, B P = X v Y i - AT ,F„ Bq = X vY2 - x 2y v, B r = X UY V- ATIT, C a = X2U U - AT U 2, c b = X i U u - X u U u Cc = X1U2- x 2u h C p = X vU i - X i U v, C, = x vu2 - X2U V, Cr = X UU V- x vu u, Pa = XuVa - X 2V u, P b = x uv l - x 1v u, Pc = X2Vx- X i V 2, P p = ATF„ - X vV u Pq = X2V v - X vV 2, Pr = X vVu - X uV v,

Qa = U2V U - U UV 2, <36 = UiV„ - U uV u Qc = U2V2- u2v h Qp = U vv 1 - U i 7 „ Qq = U vV2 - UiVv, Qr = U uV v - UvVu, Ra = Y uV2 - y2 i t , Rb = Y uv1 - Y iV u , Rc = Y2V1- F xF 2, R P = Y i V v - Y vV i, Rq = Y2V v - YvV 2, Rr = Y vV u- Y uV v.

T hese eq u atio n s give th e relation

(15)

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 289

' K2- H L + M N = J ( k2- h l + m n ) , if ï(C& + i2|,-t-Cf! + i?ij)2— {Pb~\~ Pq){A b~\~A q) + (Qî> + <25) (Bb~\~Bq) —JI

\r(Rr+ C r ) ( R c+ C c) + (QrBc+QcBr) ~ (P rA c+ P cAr) = 7 , 1CRp+Cp) (Ra + C a) + ( Q p B a+ Q aBp) - (PpSl a+ P a A p) = - / , ï ( R p + C p )2+ Q PB p - P p A p = 0, l ( R a+ C ay + Q a B a- P „ A „ = 0,

\ ( R r + C ry + Q r B r - P rA r = 0 , KRc + Ccy + Q cB c - P ' A c = 0,

(58) i(-Z?p+C,p)(C j,+i?i,+C a+ i2 s)-f-Q p(5i,-|-3!I)+ 5 p (Q (,-|-0 a)

= P p ( A b+ A q) + A p ( P b+ P q),

■i{Ra-\-Ca)(Cb-\-Rb-\-Cq-\-Rq)-\-Qa(Bb-{-Bq)-\-Ba{Qb-\-Qq)

= P a ( A b+ A q) + A a ( P b + Pq), h(Pr + Cr)(C6 + i?i, + Ca + i?a) + ( ) r (BbA-Bq)-\-Br (06 + (?a)

= P r ( A b+ A i ) + A r ( P b + Pq), KR c + C c) (C b + R b + C q + R q) + Q C ( B b + B q ) + B c (Qb+Qq)

= P c ( A b+ A q) + A c ( P b+ P q), K R P+ C p ) ( R r + C r ) + Q p B r + Q r B p - P p A - , - P rA p = 0,

i( R p A - C p ) ( R cA-Cc)ArQpBcA-QcB p — P p / l c— P cA p = Q ,

%(R* + C a) (R r+ C r) + Q aB r+ Q r B a ~ P aA t - i V l „ = 0, f (i?a+ c „ ) (J?c+ C c) + ^ + < 2 A - P o A - P cA . = 0.

T hese relations m ay be established by using a p a ra m e tric re p resen ta tio n of th e q u a n titie s satisfy in g L ie’s conditions for a c o n ta c t tran sfo rm atio n , we th erefo re w rite

X \ — d\e -f- die', X2 = b\e -\~ b2e', X u — c\e T c2e', X v = die -T d2e', U i = a i f + a2f , U2 = b i f + b2f , Up = c i f + a f ,

Y u = a2p -f- dip', — Y v = bzp -f- bip', Y i = Cap -f- Cip',

V u = az q + diq', — V v = b3q + b ^ ' , V i = c3q + Ciq', w here th e q u a n titie s

U v = d i f + d2f , Y2 = dzp + dip', V i = dzq + d iq ',

(59)

d 1 0 2 & 3 &\

b1 #2 ^4

Cl C2 Cl C\

d\ d% dz di

form an orthogonal m atrix an d th e q u a n titie s e, p, p ', q, q ' are such t h a t (ciß2 —xC2ßi T dpbid2b i ) ( e f T e'f) = {d2di — CI4C3 T bzdi — bidz)(pq'p'q) = cr. (60) I t is th en found t h a t

- H = eq{ 1, 3) + e'q(2, 3) + eq'{\, 4) + e'q'{2, 4), - L = f p ( 1, 3) + f p (2, 3) + / / ( ! , 4) + / y ( 2 , 4),

2IT = («/' - < /) ( 1, 2) + G Y - p 'q ) (3, 4), (61) M = / ç ( l , 3) + A ( 1 , 3) 4) + / Y ( 2 , 4),

= ep{ 1, 3) + e'p{2, 3) + 4) + e'ÿ '(2 , 4),

(16)

where

(1, 3) = li(dic3 + dibz)l(bifl3 -|- Cydz) -f- k(d \d3 -f- 6163 — d\d3 — C1C3) + n (a id3bi.Cz) + m ( d ia3 — Cib3),

(2, 3) = h{dzCz "I- <22^3) — l(bzd3 4* c3dz) -|- k (d3d3 4“ ¿263 — <2203 — C2C3) 4~ ti(d3d3bzCz) 4- m(dzdzc3bz),

(1, 4) = h(diCi 4- 01^4) — l(bidi 4* Cidi) 4- k ( d id tbibi — 0104 — C1C4) 4- n (d \d \ — J1C4) 4- m (d \d \C\bP),

(62) (2, 4) = h(d3d.\ 4- 02^4) — lifizdi 4- c->di) 4- k(dzdi 4~ b3b\ — <220-4 — C2C4)

4- n (d3d* — ¿2C4) 4" m^dzdi — C2J4),

(1,2) = h(did3dzd\) 4- l(b\Cz — ¿2^1) 4- k(diCz — O2C1 4- d\bzdzbi) 4~ u(d\bz + <22^1) 4- m{c\dzdiCo),

(3, 4) = h(bzCib.iCz) 4” l(d.\d3 — <23^4) 4" k(c3d3 — C4O3 4" b3dibid3) 4- n (d 3Ci — dAc3) 4- m{b3di — b4O3).

I t is readily seen th a t

M N - I I L = [(1, 3)(2, 4) - (2, 3)(1, 4)](e’f - e f ) ( p ' q - pq') (63) an d th a t, on ac co u n t of the p ro p erties of an o rthog on al m atrix

( e 'f - e /') 2 = {p'q - Pq'Y: (64)

T h e expression for K also sim plifies considerably an d th e proof m ay be re ad ily com ­ pleted. T h e q u a n tity J as in K u rs c h a k ’s analysis, is equal to a2 a n d so is n o t zero.

C o n ta c t tran sfo rm atio n s are n o t th e only ones in w hich th e con ditio n gp^O is in v a ria n t. In th e th eo ry of th e ste a d y tw o-dim ensional m otion of an inviscid elastic fluid th e eq u atio n s satisfied b y th e velo city p o te n tia l z an d strea m -fu n ctio n z are respectively

puur 4- 2p uvs 4- p w t = 0, pxivX 4" 2p uvs -f- p vvi = 0. (65) In this case d z = udx-{-vdy, d z = ii dx-\-v d y = p td x — p ud y an d so

u = p v, 5 = — p u , g = g = p l vp Uupw - c2(u2 - f i>2 c2). (66) T h u s g = 0 e ith e r w hen c = 0 or w hen q = c. T h e supersonic region is ch a rac te rized by th e condition g > 0 an d th e subsonic region b y the condition g < 0 . T h e curve for w hich c = 0 is a b o u n d a ry for th e flow ju s t as in th e case of th e P ra n d tl-M e y e r flow ro u n d a corner. T h e tran sfo rm atio n u n d er consideratio n is a special B acklund tra n sfo rm a tio n an d is included in th e g roup of B ack lu n d tran sfo rm atio n s

x = X ( x , y, u, v), y = Y ( x , y, u, v), u = U (x, y, u, v), v = V ( x , y, u, v), (67) for which th e Jac o b ia n d ( X , Y, U, V ) / d ( x , y , u, v) is n o t zero. T h ese tran sfo rm atio n s have been stu d ied carefully by G o u rsa t [16]. T h e re q u irem en t th a t

u d x -f- v d y

should be ex a ct leads to an eq u a tio n of th e M onge-A m pere ty p e in w hich z does n o t

2 9 0 H . B A T E M A N [V o l. I , N o . 4

(17)

occur explicitly. I t is show n, how ever, t h a t th e general M onge-A m pere e q u a tio n of th is ty p e c a n n o t be o b tain ed in this w ay an d a sim ilar re su lt has been found b y J.

C lairin [ l 7 ] in his stu d ies of m ore general B iicklund tran sfo rm atio n s. C lairin has stu d ied in p a rtic u la r tran sfo rm atio n s of ty p e

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 291

x ' = f i ( x , y, z, it, v, z'), y ' = / 2(x, y, z, u, v, z'),

u ' = f3(x, y, z, u, v, z'), v' = / ¿ x , y, z, u, v, s'). (68) Som e of C lairin ’s w ork is su m m arized in th e book of F o rsy th [18] an d illu stra te d b y m eans of exam ples.

A n o th er tran sfo rm atio n of ty p e (67) w hich preserves th e condition 0 is o b ­ tain ed b y w ritin g

d U = R d x + S d y , d V = S d x + T d y ,

w here R , S , T are th e functions of u an d v used in section 2. M ak in g use of th e e q u a ­ tions .

(■u p u + v p v) d x = p udz + vdz, (iipu + v p v) d y = p vdzudz, (69) we find t h a t

( T n - S v ) d U + (Rv - S t f i d V = ( R T - S2)dz, ( S p u + T p v) d U - ( R p u + S p v) d V = ( R T - S2)dz.

H ence, if

= u , y ' = V , u ' = d z / d U , v' = d z / d V , ii' = d z / d U , v' = d z / d V , (71) we have th e re la tio n s

u ' = ( T u - S v ) / ( R T - S2), ' / = (Rv - S u ) / ( R T - 5 2), u' = (Spu + T p l) / ( R T - 5 2), v' = - ( R p u + S p t) / ( R T - S 2)

(70)

(72)

w hich, in co nju nction w ith the preceding relations define tw o B acklun d tra n s fo rm a ­ tions. T h e tran sfo rm atio n s considered in m y p ap e r on th e lift a n d d ra g fu n c tio n s are of this ty p e [19] an d are n o t generally c o n ta c t tran sfo rm atio n s as is a p p a re n tly im ­ plied b y a s ta te m e n t re la tin g to the correspondence of the supersonic regions in th e tw o associated ty p es of flow.

In th e case in w hich p is a fun ction of q only th e relatio ns betw een u ' , v' , u, v are u ' = - u / p , v' = — v /p , q' = q / p

an d , if p ' - — 1 I p , p ' + p ' q ' 2= — 1 / ( P + p q 2) we have

, pP

p = --- >

P + p q 2

d p ’/ d q ' = — qp/('p + pq2) = — p'q',

1 — q'2/ c' 2 = (1 — q2/ c 2) [ p / ( p + pq2) ] 2, where c '2 = d p '/d p '.

T h is tran sfo rm atio n m ay be com pared w ith th a t o b tain ed b y m eans of H a a r ’s a d jo in t v a ria tio n problem s [20]. In this case

u* = p j ( p - Upu ~ v p v) , v* = p j ( p - upu ~ v p v), p* = 1 / ( p - Upu - v p v),

(18)

and, w hen p depends only on q,

u* = - p u / ( p + pq2), v* = - p v /(p + pq2), P* = + I / O + q2), q* = ' q / ( p + ?2)- D efining p* b y the eq u a tio n d p * = —q*dq*, we have

p* = (p + pq2) / p p , p* + q* 2 = 1 / p ,

c * 2 = p tp2q*(q2c2) / ( p + pq2)2[c2(p + pq2) 2 + p2(q2 - c2)], c*2 _ q*2 = _ p Y c2/ [ c 2(p + pq2) 2 + p2(q2 - c2)].

H ence c*2 = q* 2 w hen c2 = 0 and c*2 = 0 w hen q2 = c2. I t should be n oticed, how ever, th a t

c*2{c* 2 - q2) = c2(c2 - q2) p2Py / ( p + pq2)2[c2(p + pq2) 2 + p2(q2 ~ C2) ] 2.

T h is m ay be com pared w ith H a a r’s general relation*

(p*.v.p * .v. - p t l v')(puupvv - p lv) = p ^ p * - * .

T ra n sfo rm atio n s m ore general th a n those of B ack lund have been considered by G au [21] b u t so fa r no hydro d y n am ical app licatio n s h ave been found for these so fa r as I know . M en tio n should be m ade, how ever, of th e equ iareal tra n sfo rm a tio n s from th e E ulerian to L ag ran g ian co-o rdinates in th e tw o-dim ensional flow of an inco m pres­

sible fluid. T hese tran sfo rm atio n s have been m uch used in m ap p in g b u t th e hy dro- dynam ical app licatio n s are b eset w ith form idable difficulties.

N o m ention has been m ade of th e use of tran sfo rm atio n s in th e th eo ry of surfaces, congruences, etc. T h is is a su b je c t w hich is well tre a te d in th e books of D a rb o u x [22], F o rsy th [ l 8], G o u rsa t [ l 6], B ianchi [23] an d E ise n h a rt [24].

4. T ra n sfo rm atio n of th e lin e a r equation. In th e special case in which h, k a n d I are fu nctio ns of x an d y only, n — 0 an d m is a lin ear hom ogeneous fu nction of u, v a n d z w ith coefficients d epending o nly on x an d y , th e M onge-A m pere e q u a tio n reduces to a linear eq u atio n . T h e behav io r of th is e q u a tio n in tran sfo rm atio n s of ty p e

X = X ( x , y ) , Y = Y ( x , y ) , Z = zF (x, y) (73) has been stu d ied b y D arb o u x [22], C o tto n [25], R iv ereau [26], J. E. C am pbell a fte r th e case F = 1 h ad been discussed b y L aplace [27], C hini [28] an d o th ers [ l 6].

C am pbell uses th e eq u a tio n in a form in w hich g = 1, a form to w hich th e general e q u a ­ tion can be reduced b y m u ltip ly in g it b y a su itab le factor. H e th en shows t h a t th ere are tw o in v a ria n ts I , / an d an ab so lu te in v a ria n t J / I w here if suffixes d en o te p artia l d eriv a tiv e s

I = h a x T k ( a y T b x) T lbv T (lix T k y) a T (k x -j- l v)b + Ita2.-f- 2kab + lb2m x, J — a yb x, 2a = l ( h x + k ym u)k ( k x + l y — w„)>

2b = h ( k x + /„ — m v) — k ( h x + k ym u).

* This is a consequence of the relations

= P P ' ^ P - - P-), ^ = P ' P K P i ^ - P i -0.

In the correspondence between the two hodograph planes complications arise on account of the relation of PuuPui Puvto these Jacobians.

292 H . B A T E M A N [V ol. I , N o . 4

(74)

(19)

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 293

L ap lace's in v a ria n ts are 1(1 — 7 ), ^ ( I + T ) . T hese are used w ith a d ifferent n o ta tio n , in D a rb o u x ’s T heorie des Surfaces, t.2. C am p b ellsh o w s th a t in th e case of th e e q u a ­ tion of E uler an d Poisson th e in v a ria n ts I an d 7 are c o n sta n t. T h is m ay be co m pared w ith C o tto n ’s result. T h e h arm o nic eq u atio n s belong to th e g ro up ch a rac te rized b y th e relation 7 = 0 . T h e eq u a tio n s considered by B u rg a tti [29] are such t h a t 7 = 0.

In m a th em atica l physics th e sim ple solutions of lin ear eq u atio n s p lay an im po r­

ta n t p a r t an d th e p rim ary problem is t h a t of sep a rab ility . E ven in th e case of th e eq u a tio n w ith tw o in d e p e n d e n t v ariables th ere are som e unsolved problem s. A good idea of the progress w hich has been m ade m ay be derived from D a rb o u x ’s boo k [22].

T h e m eth o d of L aplace provides an im p o rta n t w ay of reducing eq u a tio n s by a c a s­

cade process w hich is p a rtic u la rly useful in th e tre a tm e n t of eq u a tio n s arising in th e th eo ry of plane w aves of finite am p litu d e . R eference m ay be m ade to a p a p e r of Love and P id d u c k [30], an article b y P la trie r [31 ], som e p apers b y B ech e rt [32] an d to tw o p apers b y Oseen [9] in w hich th e tran sfo rm atio n a n d redu ction is given for e q u atio n s occurring in th e th e o ry of e a rth pressure an d in th e th eo ry of p lasticity .

In th e th eo ry of th e ste a d y m otion of an inviscid com pressible fluid th e eq u a tio n s in th e hodograph plane are linear. T hese equ atio n s are

PvvlVuU 2p uvWuc ~\~ puu^vv — 0, pvvWHU 2puv^uv “1“ Puu'&VV b) 2 = 1115, + VWv — W = qivq — W, Z = UWu + V~d>vw = q w q — w.

W hen p is a function of q only th e eq u atio n s u = p v, v — —p u ta k e th e form

u = — pv =q sin r, i' = pu = q cos r, q = pq, (76) a n d th e eq u atio n s becom e

Wr, + q(qwQ)q = 0, w TT + q(qWq)q = 0 (q a function of q). (77) T hese are consequences of sim ple relations betw een w an d w

w T + qw^ = 0, w r — qw^ = 0. (78)

T h e corresponding relations betw een z an d s are found to be

q % = qzr, q \ = - qzT (79)

an d so th e equatio ns for z an d z are

(? 7 S )](2 2/?)Z5L + Srr = 0, 1 (q.*/q){(q2/ q ) z q h + ZrT = o. j

T hese are e q u iv a le n t to th e eq u a tio n s o b tain ed b y M olenbroek [33] an d T schap lyg in [34] for th e case in w hich th e relation betw een p an d p is of th e p o ly top ic o r a d ia b a tic ty p e. T h e sym m etrical form s of th e eq u atio n s are easy to rem em ber.

I t is som etim es useful to in tro d u ce o th e r q u a n titie s w hich sa tisfy lin ear relation s.

T h u s we m ay o b tain th e desired relations betw een z, z, w, w b y w riting w = eT = qeq, w = eT = — qeq,

z = — eT — (q /q )eTT, z = {q/q)e TTeT, w here

cTt “h ^tt ~h q

(20)

294 H . B A T E M A N [V ol. I , N o . 4

T h e lite ra tu re dealing w ith th e tran sfo rm atio n of linear eq u a tio n s in several v a ri­

ables is v ery extensive an d only a brief su m m ary can be a tte m p te d here. B e ltra m i’s w ork on differential p aram eters [35] was ex ten ded by Ricci a n d L e v i-C iv ita [36], C o tto n [25], L ev i-C iv ita [37] an d m an y o th e r w riters. T h e d ev elo p m en t of general re la tiv ity , electrodynam ics a n d the th eo ry of elasticity has m ade this w ork m ore or less know n. T h e w ork of L am é on sim ple solutions of the p o te n tia l eq u a tio n [38] was m uch developed b y la te r w riters an d a good su m m a ry of resu lts up to 1893 is given in th e book of B ôcher [39]. T h e use of a v aria tio n a l principle for o b ta in in g th e tra n s ­ form atio n of th e eq u a tio n w as recom m ended b y L arm o r [40], V o lterra an d o th ers [41]. Since th e a d v e n t of th e new q u a n tu m th eo ry the in te re st in sep arab le equ atio n s an d separable system s has m uch increased. M en tio n m ay be m ad e of th e w o rk of S taeckel [42], E ise n h a rt [43] an d R ob ertso n [44].

In ad d itio n to th e sim ple solutio ns of p a rtia l differential eq u atio n s th ere are solu ­ tions having th e form of pro d u cts in w hich one or m ore o r th e facto rs satisfies a p a r­

tial differential eq u atio n in stead of an o rd in ary differential eq u a tio n . C o m p arativ ely little w ork has been done on th is problem . In th e case of L ap lace’s eq u a tio n V xx+ V vu

+ F*2 = 0, th e aim is to find a solution of form [45]

V = Z F ( X , Y ) , (generalized binary potential) where F satisfies a p artia l differential eq u atio n of the second order in th e variab les X and Y. T h e problem seems to depend on the form ation of a relation of type

(p2 + q2 + r2) ( d x - + d y2 + dz2) - (p d x + qdy + rdz) 2 = a d X2 + I h d X d Y + b d Y2 in which a , b and h are functions of X an d Y only. T h ere is a sim ilar relation for the corresponding problem in a n y n u m b e r of variables.

5. T h e tran sfo rm atio n of L e g e n d re ’s equation. L eg en d re’s eq u atio n 3C(R, S , T ) = 0

is u n altered in form by a L egendre c o n ta c t tran sfo rm atio n

X ' = U , Y ' = V , U ' = X , V = Y \ Z ' = U X + V Y - Z , w hich m akes

R ' T / ( R T — S 2), S ' = - S / ( R T - S 2), T ' = R / { R T - S 2).

In p a rtic u la r, th e eq uation

R ' T ' - S '2 = F { R ' + T ' ) becom es

f R + T \

an eq u a tio n of th e sam e general type. A gain, if a , b and h are c o n sta n ts th e c o n ta c t tran sfo rm atio n

Z ' = U a X 2 + 2 h X Y + 5 F 2) + Z , X ' = X , Y ' = Y , U ' = a X + h Y + U , V = h X + b Y + V

m akes R ' = a + R , S ' = h + S , T ' = b + T and so tran sfo rm s an eq u atio n of L egendre’s ty p e in to a n o th e r e q u a tio n of th e sam e ty p e. E q u a tio n s of th e preceding ty p e usually go in to L egendre eq u a tio n s of a slightly d ifferen t typ e.

(21)

O th e r tran sfo rm atio n s m ay be found b y first tran sfo rm in g th e eq u a tio n to the M onge-A m père form by L egen d re’s device. If, for instan ce, th e eq u a tio n is

T = F ( R , S )

an d we d ifferen tiate w ith respect to x using th en th e new n o ta tio n z = U , R = U x = u, S = U y = v, T x = S y = t , R x = r, S x = s, the new eq u a tio n is

i = F u(u, v)r + F v(u, v)s.

C om parin g this w ith th e eq u atio n p u u r + 2 p uvs - \ - p vvt = 0 we find th a t F u Puu/Pwt F v ~ 2p u t ' / p w

E lim in a tin g F we find th a t the eq u a tio n for z is n o t a general eq u atio n of th e ty p e considered in §2 because th e fu nctio n p ( u , v) satisfies th e condition

d D /d v = 0 where D — p uup vvp uv.2

A n eq u a tio n for w hich D is c o n s ta n t satisfies th is condition an d th e e q u a tio n of ty p e

X ( R , S , T ) = 0

associated w ith it b y th e analysis of section 2 m ay be regarded as a tran sfo rm of th e original eq u a tio n 3C(2?, S, T ) = 0 . In th e case w hen D = 1, we m ay w rite

puu = ea sec b, p vv = e~a sec b, p uv = tan b, w here a an d b are functions of u an d v w hich m u st be chosen so th a t

sec2 bbu = e2 sec b(a t) + e° sec b tan b(bv), sec2 bbv = — e~a sec b(au) + e~a sec b tan b(b„).

Also, since F u = —e2°, F v— — 2ea sin v, we m u st have th e ad d itio n al eq u ation 2e“ cos b(bu) — 2eu { a l) + 2e“ sin v(au) = 0

which is seen, how ever, to be a consequence of th e o th e r two. E lim in atio n of the d eriv a tiv e s of a gives th e eq u atio n

6 bw I e buu 2 sin bbuv — H or puubw *E pwbuu 2puvbuv — d

an d it is readily seen th a t a satisfies the sam e eq uation . T h e eq u atio n 19 = 1 is given as an exam ple in F o rs y th ’s book, p. 220, Ex. 11.

Re fe r e n c e s

[1] J. E. Ca m p b e l l, Proc. London Math. Soc. (2) 10, 406-422 (1910).

[2] C. Ca r a t h é o d o r y, Bull. Greek Math. Soc. 3, 16-24 (1922); Math. Ann. 86, 272-275 (1922);

85, 78-88 (1922).

[3] G. Mo n g e, Application de l’analyse d la géométrie, Bernard, Paris, 1809.

[4] A. Ha a r, J. f. Math. 149, 1-18 (1919); see also T. Rad6, Hungarian Univ. Acta, Szeged, 2, 147-156 (1925).

[5] A. M. Le g e n d r e, Histoire de l’Académie des Sciences 1787, 309.

[6] H. H e n c k y , Zeit. f. angew. Math. u. Mech. 3, 241-251 (1923).

[7] L. Pr a n d t l, idem., 1, 15-20 (1921); 3, 401-408 (1923).

[8] C. Ca r a t h é o d o r y a n dE. Sc h m id t, idem. 3, 468-475 (1923).

[9] C. W. Os e e n, Arkiv f. Math., Ast., o Fysik 20A, Nos. 24, 25 (1928).

1944] T R A N S F O R M A T I O N O F P A R T I A L D I F F E R E N T I A L E Q U A T I O N S 295

Cytaty

Powiązane dokumenty

It is shown in the present paper that the w ave fronts associated with those parts of a disturbance which are derivable from a potential propagate in a rotational

a) Stability of plane Poiseuille flow. T he stab ility of plane Poiseuille flow has been attem pted by m any authors. T he papers of Goldstein and Synge and one of

tics. Liénard, Élude des oscillations entretenues, R ev. K irschstein, Uber ein Verfahren zu r graphischen Beliandlung eleklrischer Schwingungsvorgange, Arch.. M ean

versity for valuable criticisms and suggestions.. H istorical survey of existing theories.. As is now well-known, this m ethod can only give sufficient conditions for

T he keynote of this book is th a t familiarity with the general principles of mechanics is indispensable to a thorough understanding of the analysis of stresses in trusses

T he advantage of formulating this particular class of boundary value problems as Wiener-Hopf integral equations is th at such equations are susceptible to a

teristic for a given object point thus assumes a two-dimensional manifold of values. In this case, all the diapoints fall together, and, according to F erm at’s law,

m ents in cosine series (for the even solutions) and in sine series (for the odd solutions). T he points betw een tw o successive curves for which the periods of the