DYNAMICAL TRAJECTORIES AND GEODESICS.*
By Luraer Pranter Ersennanr.
According to the general theory of relativity the motion of a material
particle under the action only of inertia and gravitation is deséribed by
a geodesic of the four-dimensional space-time continuum of the gravitational
field. Thus there is a correspondence between the spacial paths of particles
and the geodesics of the space-time continuum. In this paper we show
that the trajectories of a general holonomic conservative system in classical
dynamics can be put into correspondence with the geodesics of a suitable
Riemannian manifold. In § 2 we consider the general case when the time
enters in the constraints and in the potential function, and in §§ 3, 4 the
relation between the theory of geodesics and the Hamilton-Jacobi theory
of dynamical trajectories. The relation between contact transformations
of the Riemannian manifold and transformations of the Hamiltonian equations
of the dynamical system is set forth in § 5; the result brings out the
significance of the form of the latter transformations. In § 6 the restricted
case when ¢ enters neither in the constraints nor potential function is
treated not as a special case of the more general theory, but on an in-
dependent basis.
1. Equations of dynamics. Consider a conservative holonomic
dynamical system of n degrees of freedom such that the restraints as well
as the potential function V involve the time. If we denote by g'(i = 1, ---, n)
the independent variables determining’ the position of the system and put
t=", the kinetic energy is of the formt }gagg%4q?, where « and &
run from 1 to n-+1 and the dot indicates differentiation with respect to
the time as usual. Then the Lagrangian function Z is given by
(a) L = damp di? —¥.
When this expression is substituted in the equations of Lagrange, namely
(1.2)
a (:)- aL _
dt \agt age
we have as the differential equations of the trajectories of the motion
* Received April, 1929.
} Throughout this paper we use the convention that a repeated index indicates sum-
mation, unless stated otherwise; also latin indices will be supposed to take the values 1 to n.
591592 L. P. EISENHART.
gui + ik, BY B+ (OA + genes — On
«54 Ogintt
4 54 gins
aq agi )i +58
1 4 -
—y jgf Uatints—2V) = 0,
where | jk, é] is the Christoffel symbol of the first kind, namely
jhe, i) = 2 (Ome 4 Bg —1, the functions F not involving ¢’, u and uw’, In this case the
expressions (5.22) for »; and Pj are the same as for the general case,
but in place of the other two we have
ar Fe) __ (@
Pin = — (GE —Paa ta at ot
Hence the corresponding function K is obtained by substituting in the
quantity in parenthesis the expressions for gi, p: and 2° obtained from the
equations F; =... = F, 0 and the first equations of (5.22). Then the
transforms of equations (1.13) are*
dq’ aK aP _ OK
dt OP’ dt ag
6. Natural dynamical systems. When the constraints of the dyna-
mical system and the potential energy V do not involve t, we have that
Jinti = Jntints = 0 and that gy are independent of . Then equations
(1.8) are
ptt aspen OV
(6.1) GE +UK AY GT Se =0.
These equations admit the first integral
1 a.
(6.2) gut tv = B,
where E is the energy constant.
The results of §§ 2 and 3 apply to this case directly, but in place of
using these we consider a Riemannian space of n+1 dimensions, Si,
for which the fundamental form is
(6.3) ds? = gy dq dgit+ Adu,
where it is assumed that A does not involve w. The equations of the
non-minimal geodesics of this space are readily obtained by replacing ¢ by u
in (2.4), (2.5) and (2.6), and putting w= B=0 in the latter equations;
this gives
a agi 1 4a) dg 1 0A my
(6.4) Giger Tk, AG ae 2 og (ads ’
du
(6.5) en
* Cf. Whittaker, 1 c., p. 309, 310.604 L. P. EISENHART.
where @ is a constant. For every finite value of a different from zero
a parameter ¢ for each non-minimal geodesic is defined by
(6.6) t=as.
In terms of ¢ as independent variable equations (6.4) and (6.5) become
(62) wi +k, de
du 1
(68) “wnt
If the parameter ¢ is identified with the time and A is defined by
1
(6.9) sa = Vth
where } is a constant, equations (6.7) are the same as (6.1). From (6.3),
(6.6) and (6.8) we have
jd
(6.10) 4 = gti +z = (E+).
When | E| has an upper-bound, 6 can be chosen so that equation (6.10)
is consistent; in this case it gives the relation between the constant a
along a geodesic and the energy constant Z for the corresponding trajectory.
Then the non-minimal geodesics of Sy4: are defined by
(6.11) g=f', w= of vattone,
the first » of these equations being the solutions of (6.1). In consequence
of (6.2) the last of equations (6.11) may be written in the form
(6.12) = —2frateerye,
which reveals the relation between the codrdinate w and the action.
The determinant of the form (6.3) is Ag, where g is the determinant
of gy. Making use of the notation (3.1) as applied to (6.3), where now
@ and & take the values 1 to n+1, and g"+1= u, we have in place
of (3.2)
(6.13) 7
gy, 750, gone + = 2(V45).
In this case equation (3.3) becomes
ag by | ay \?DYNAMICAL TRAJECTORIES AND GEODESICS. 605
For equations of the form (3.4) for « =n+1 we have
= 2+yee
Comparing this result with (6.5), we have that the geodesics for a given
value of a are the orthogonal trajectories of the hypersurfaces » = const.,
when
(6.15) y= aut),
where wy is independent of u. On substituting this expression in (6.14), we
find that y must be a solution of
aw oy a
(6.16) pa qi +20 +0) = Ze.
Hence if we have a solution of this equation involving n —1 constants a’,
other than a, none of which is additive, then (6.15) defines a complete
integral of (6.14) and the geodesics of Sy: are given by
(6.17) utwpatt ae, Shae,
where the c’s are arbitrary constants.
In consequence of (6.10) equation (6.16) may be written
ow ow _
(6.18) ou Gqi = BEY).
Hence if we have a solution of this equation involving »—1 arbitrary
constants @%, none of which is additive, the last set of equations (6.17)
give in finite form the trajectories for the given value of the energy 2.
For an Sy41 with the fundamental form (6.3) we have in place of (6.13),
in consequence of (6.6) and (6.5),
ay du
B= Ii Gs = Ws Pap = Aa =a,
and in place of (6.16), because of (6.9) and (1.11)
(6.19) C= WRB + TPL, = 2TH.
In view of these results equations (5.6) for « = 1, ---, reduce to (1.13)
and for «= n-+1 to (6.8).
If we take any homogeneous contact transformation such that the equation
of the first set of (6.8) for @ = n-+1 is of the form (5.18), where f and606 L. P, EISENHART.
the other functions of both sets do not involve w, it follows from the first
of equations (6.7) for +1 that Posi = Pai = a. If then we
replace Pj by api, equation (5.9) becomes
(6.20) pidge = pidg'+af.
In this case the corresponding equations (5.10) are of the form (5.21),
in which the F’s do not involve u, w, ¢ and ¢’, and in place of (5.22) we
have
ok oF 1 ar, oFo
= as See | = — Pee Ee
(6.21) Ps age tage? Pt aq" ag"
Equations (6.12) reduce to
dq’ _ oH’ dp! aH dw’
(622) “ar = ay cae wg? ae
where H’ and 4’ are the transforms of H and A. Conversely, if we have
a set of equations (6.21) of the type under consideration defining a contact
transformation, then equations (1.13) are transformed into the first two sets
of (6.22).*
* Cf. Whittaker, I c., p. 305.
PrinceTon UNIVERSITY.