VOL. 17, 1931 MA THEMA TICS: L.

LA PAZ 459
then we find readily that
p
i1(f) . [l(fi) (f)] P+1, 1 f
p< 00,
I () _< I(,,) (fO.
* NATIONAL RESEARCH FELLOW.
1 M. Frechet, "Sur divers modes de convergence d'une suite de fonctions d'une
variable," Bull. Calcuttk Math. Soc., 11, 187-206 (1921); "Les espaces abstraits,"
Paris, Gauthier-Villars, p. 91 (1928).
2 The proof here is essentially the standard proof in the theory of convergence in
measure. See, for instance, F. Riesz, Compt. Rend., 144, 615-619 (1907); E. Fischer,
Compt. Rend., 144, 1022-1024 (1907); H. Weyl, Math. Annal., 67, 225-245 (1909).

THE EULER EQUATIONS OF PROBLEMS OF THE CALCULUS

OF VARIATIONS WITH PRESCRIBED TRANSVERSALITY
CONDITIONS
By LINCOLN LA PAZ
DEPARTMENT OF MATHEMATICS, OMO STATE UNIVERSITY
Communicated June 30, 1931
Yi' . 1..
If n functions Ti(x, yi, ..., Y I' y,) are of class C1 in a region
S of (x, yi, .. ., Yi ... y')-values and in S satisfy the inequalities
D=I68Ti/ly|IO (i,j=1, ...,n), 1+y T,,O (1)
and the n(n -1)/2 relations
rmk=o (m,k= 1, ...,n,m <k), (2)
where
rmk - YM,(Tm8Tk/6Y,s - Tk?Tml/ bY) + (8Tk/by' - .Tm/)y*) (3)
then there always exists a non-singular problem of minimizing an integral
I= f(x ls
Y* * 1n
1
I
sy)dx (4)
with an integrand function f of class C"' which has the functions T, as
its transversality coefficients.2 In (1) and (3) as elsewhere in this paper a
repeated Greek letter is an umbral index indicating a summation with
the range 1 to n.
The most general such problem (4) has an integrand function
f = geh (5)
where g(x, yi, ..., yn) is a function different from zero and of class C"'
in the (x, yi, ..., y.)-projection of S but otherwise arbitrary, and where
h = T,dy,,/(l + y,T,) (6)
is a line integral which in a space of suitably simple connectivity properties
460 MA THEMA TICS: L. LA PAZ PROC. N. A. S.

can be shown to be independent of the path of integration in view of rela-

tions (2).3 In (6) the symbols 0 and y' used as limits represent sets of
n quantities (0, ... 0) and (y'l ..., Y.)
It is clearly no restriction to assume g > 0, since a problem for which
g < 0 can always be transformed into a problem for which g > 0 by re-
quiring that the integral I in (4) be maximized instead of minimized, Set
I = log g, H= h+1. (7)
Then the Euler equations of (4) formed with the integrand function (5) are
d /eH be
_ ay ) =°0
dx-tay (i = 1, ..., n). (8)
In view of relations (6) and (7), the system (8) reduces at once to
+ y,(TOTiv, - TiTY)-]y = Mi (i = 1, .. ., n)
[Tiy,I/o'Y (9)
Y
where
Mi = [(1 + y,T,)Hy - T -yT,Hy,TiHx](1 + yT,,)
+ [Tiy, Tfyy- (1 + y,T,)Tiy)]y (10)
+ Ty T-O(1+ )Tix
To obtain a system of differential equations in normal form from these
equations the determinant of the coefficients of the second derivatives
must be evaluated. This determinant
d = Tiy, + y,(TTTTiyj'-TiT,ff') | (i,j = 1, .. ., n), (11)
is recognized as the product of the determinants
A = bI8 (1 + y,,T..) - y,.Ti| (12)
and D = .| bTilby'l. To evaluate A we note that
Xi + ',T A bji (1 + y',&T;,) |=(1 + y,,T,,) . (13)
It is not difficult to show that 6 + y'Tj = (1 + y,,,T). Hence (13)
gives A = (1 +yATO)
Consequently d = D (1 + 'T,)" and therefore in the region S the
second derivatives y" can be obtained in the form
ff = dk/D .(1 +
y'T;)"1 (14)
where dk is the determinant d with the element dik replaced by the quantity
Mi in (10) for i = 1, . . ., n. The determinant which arises when the
element in the ith row and kth column of d is replaced by Mi is found
to have the value
dk = Dk* (1 + y,TM)"1, (15)
where Dk is the determinant D with the element bTi/by' replaced for
i = 1, ..., n by the quantity
Ni = (Hy, - TiHx)(1 + 'T,) - (Tix + Tiy (16)
VOL. 17, 1931 MA THEMA TICS: L. LA PAZ 461

Hence the system (14) reduces to the compact system of equations in

normal form described in the following theorem:
THEOREM 1. For problems (4) with a prescribed set of transversality
coefficients Ti, the Euler equations may be put in the normal form
Yk = D/D (k = 1, ...,n). (17)
In (17) isD the Jacobian of the Ti with respect to the y, and Dk is the de-
terminant obtained from by D replacing the element bTi/by' in D by the
quantity Ni in (16) for i = 1, .. ., n.
For the special case in which transversality, in the sense of the calculus
of variations, is to be equivalent to orthogonality the transversality coeffi-
cients T, must have the values
Yn I1
I I
Ti(xf Ylf I I I
wY)=y (i =1 ** n). (18)
When these values for Ti are employed the system (17) reduces to
yk = (al/ayk - ykl/tx)(1 + y1y) .(k = 1, ..., n). (19)
An obvious necessary and sufficient condition for (19) to reduce to the
system
Yk = (k = 1, ...,n) (20)
is that I be a constant. Hence, since two integrals (4) for which the inte-
grand functions differ only by a constant factor are not regarded as dis-
tinct, we reach the following theorem:
THEOREM 2. The only problem of the calculus of variations in space of
(n + 1)-dimensions for which transversality is orthogonality and for which
the extremals are the straight lines satisfying the system (20), is the problem
of minimizing the generalized length integral
rX2
I1= f/ y +... + y')2 dx.
(1y2+ (21)
In case n = 2 and we set y = yl, z = y2 the system (19) reduces to the
pair of equations
yI = (?l/Jy - y'bl/X)(j + y/2 + z'2), (22)
z" = (61/6Z - Z'Zl/ZX)(j + y'2 + z'2).
The distinctive form of -the equations (22) has been employed4 as a
criterion for a natural family, that is to say, a four-parameter family of
curves in three-space which are.identifiable with the extremal system of
a problem of minimizing an integral of the special form
f'x2 1
I F(x, y, z)(1 + y'2 + Z'2)2dx. (23)
In this same sense the system (17) may be regarded as a characteristic
normal form of the Euler equations of a problem of minimizing the general
integral (4) in space of (n + l)-dimensions. For, as has been shown, in
462 MATHEMATICS: L. LA PAZ PROC. N. A. S.

terms of the transversality coefficients (bf/ayj)/(f - y,f/ay,") (i = 1,

... n), of a problem (4) the Euler system can always be reduced to the
normal form (17); and, conversely, if a system of differential equations
Yk = Fk(x, yi ....,
nY} , *SY)( yn),, .)( (24)
permits of representation in the special form (17) then it is a normal Euler
system of a problem (4) for which the transversality coefficients are the
quantities Ti and the integrand function is f = e'f. Thus a necessary
and sufficient condition for the differential equations (24) to be a system
of normal Euler equations of a problem (4) in the calculus of variations
is that there exist n functions Ti(x, yi, . y. Yi ,Y , Y) (i = 1y **n,
possessing the properties of transversality coefficients, and a function
H(x, Yi, ..., y I,, . . . X, Y) which together satisfy the system of partial
differential equations
bH/byk = Tk/(1 + y,T,), Fk = DA/D (k = 1, ..., n). (25)
The integrability conditions for the sub-system (251) are the n(n - 1)/2
relations
Rij =0 (i, j = 1, ...,n, i < j), (26)
where
Rik = y,[TmJTo/ly - Tk?)T./1yY, + T;,(bTkI/y, (27)
- Tml/ayk)I + (Tk /IyY' - JTml/ Yky)
If we note that in terms of the R,> the quantities rmk in (3) are expressible
in the form
= (Rmnk - Y,TkR,m - Y,sTmRk,,)/(1 + y,T,), (28)
and that in view of (252) the quantities Fj are the unique solutions obtained
by Cramer's Rule from the system of equations
(bTj/1yO,)F, = Nj (j = 1, .. ., n), (29)
the statements of the last paragraph may be formulated in the following
theorem:
THEOREM 3. A necessary and sufficient condition for the equations
(24) to be a system of normal Euler equations of a non-singular problem (4)
of the calculus of variations with integrand function f of class C"' is that
there exist n functions Ti(x, yi, ..*. yn, Yi' . . . Yn) of class C" satisfying
the inequalities (1) and a function H(x, yi, .... Yn,y1I.I.. y') such that the
following 2n equations
H/Jby' = Tk/(l + y'T,), (Tkl/by')Fa = Nk (k = 1, ..., n), (30)
where Nk is the function defined in (16), are satisfied identically in x, yi, ..
Yn, Y1l ** Yn-
VOL. 17, 1931 MA THEMA TICS: G. A. MILLER 463
Since
Rmk = (1 + y,"Tp)rmk + y'sTk"rm + yiTmr, (31)
the compatibility of the sub-system (301) is insured by the relations (2)
which n functions T, must satisfy if they are to be transversality coeffi-
cients, but the determination of the integrability conditions for the entire
system (30) appears to be difficult and will be considered in a later paper.
As an immediate corollary to Theorem 3 we have
THEOREM 4. A necessary and sufficient condition for the integral curves
of the system (24) to be the extremals of a problem (4) for which transversality
is orthogonality is that there exists a single function l(x, yi, ...yYn) satisfying
the system of partial differential equations
Fk = (l/ by, - yIl/ x)(I + y,y,) (k = 1, ..., n). (32)
1 The term class is used in the sense of Bolza. Compare Vorlesungen uber Varia-
tionsrechnung, p. 13.
2 La Paz, L., Bull. Amer. Math. Soc., 36, p. 674, 1930.
3 In this connection a function f of the form (5) was first obtained by Rawles, Trans.
Amer. Math. Soc., 30, p. 778, 1928. A proof that the formula (5) actually furnishes
the most general integrand function of a non-singular problem with a specified trans-
versality was first given by the writer, loc. cit., p. 680.
4Kasner, B., "The Theorem of Thompson and Tait and Natural Families of Tra-
jectories," Trans. Amer. Math. Soc., 11, pp. 121-140, 1910; "Differential-Geometric
Aspects of Dynamics," Princeton Colloquium Lectures, § 38, 1912.

THEOREMIS RELATING TO THE HISTORY OF MA THEMA TICS

By G. A. MILLER
DEPARTMENT OF MATHEMATICS, UNIVERSITY OF ILLINOIS
Communicated June 22, 1931
Attention has frequently been directed to evidences of the dominance
of the group concept in the development of mathematics, and in the second
edition (1927) of his well-known Theorie der Gruppen von endlicher Ordnung
A. Speiser emphasized symmetry in ancient ornaments as an evidence of a
prehistoric group theory extending back to about 1500 B. C. It may be
desirable to note here a few evidences of a lack of dominance of the group
concept in developments closely related thereto in order to obtain a truer
picture of the actual influence of this concept. The very late explicit
formulation of the group concept is a striking instance in the history of
mathematics of extreme slowness in coordinating related developments and
it deserves emphasis on this account.
Special arithmetic and geometric series appear in the writings of the