higherdstraightforward. We are calculating the first-order change inSdue to
an infinitesimal changeδγ= (δq(t),δq ̇(t))
δS[γ] =
∫t 2
t 1
δL(q(t),q ̇(t))dt
=
∫t 2
t 1
(
∂L
∂q
(q(t),q ̇(t))δq(t) +
∂L
∂q ̇
(q(t),q ̇(t))δq ̇(t)
)
dt
But
δq ̇(t) =
d
dt
δq(t)
and, using integration by parts
∂L
∂q ̇
δq ̇(t) =
d
dt
(
∂L
∂q ̇
δq
)
−
(
d
dt
∂L
∂q ̇
)
δq
so
δS[γ] =
∫t 2
t 1
((
∂L
∂q
−
d
dt
∂L
∂q ̇
)
δq−
d
dt
(
∂L
∂q ̇
δq
))
dt
=
∫t 2
t 1
(
∂L
∂q
−
d
dt
∂L
∂q ̇
)
δqdt−
(
∂L
∂q ̇
δq
)
(t 2 ) +
(
∂L
∂q ̇
δq
)
(t 1 ) (35.1)
If we keep the endpoints fixed soδq(t 1 ) =δq(t 2 ) = 0, then for solutions to
∂L
∂q
(q(t),q ̇(t))−
d
dt
(
∂L
∂q ̇
(q(t),q ̇(t))
)
= 0
the integral will be zero for arbitrary variationsδq.
As an example, a particle moving in a potentialV(q) will be described by a
Lagrangian
L(q,q ̇) =
1
2
m|q ̇|^2 −V(q)
for which the Euler-Lagrange equations will be
−
∂V
∂qj
=
d
dt
(mq ̇j) =mq ̈j
This is just Newton’s second law, which says that the force due to a potential
is equal to the mass times the acceleration of the particle.
Given a Lagrangian classical mechanical system, one would like to be able
to find a corresponding Hamiltonian system that will give the same equations of
motion. To do this, we proceed by defining (for each configuration coordinate
qj) a corresponding momentum coordinatepjby
pj=
∂L
∂q ̇j