The Principle of Least Action
Throw a ball across the room. Out of the infinitely many curves it could in principle
trace between your hand and its landing point, it picks exactly one — the graceful parabola. Newton
explains this locally: at every instant gravity tugs, and F = ma nudges
the ball's velocity a little. But there is a second, breathtaking way to say the same thing, and it
is global: assign a single number to each whole path, and the path nature actually follows
is the one that makes that number stationary. One number, computed over the entire
journey, silently selects the trajectory. That number is the action, and the
statement is the principle of least action — arguably the most beautiful idea in
all of physics.
It sounds almost mystical, as if the ball "tries out" every route and shops for the cheapest. It is
not mystical — it is exactly equivalent to F = ma, as the next pages
prove. But packaging all of mechanics into a single principle is a staggering compression, and it
is the form of physics that survived intact into relativity and quantum theory, where Newton's
forces do not.
Building the action
The recipe has two ingredients you already know: the kinetic energy
T (energy of motion, \tfrac12 m v^2) and the
potential energy V (energy of position, like
mgy for height). Their difference is a quantity called the
Lagrangian:
L = T - V.
Note the minus sign — this is not the total energy T + V. It is
a strange, almost arbitrary-looking combination, and its deep justification is simply that it
works: it is the thing whose accumulated total nature makes stationary. The
action S is that accumulated total — the Lagrangian added
up along the whole path, from the start time t_1 to the end time
t_2:
-
The action of a path q(t) running from
q(t_1) to q(t_2) is
S[q] = \displaystyle\int_{t_1}^{t_2} L\,dt = \int_{t_1}^{t_2} (T - V)\,dt.
-
The principle. The path actually taken between two fixed configurations, at
two fixed times, is the one for which the action S is
stationary — unchanged, to first order, by any small wiggle of the path.
The square brackets in S[q] are a deliberate flag: S
is a functional.
Feed it an entire trial path and it returns one number. Feed it a different path and it returns a
different number. Finding the path that makes a functional stationary is precisely the
subject of the calculus
of variations — so the machinery to turn this principle into equations of motion is
already sitting on the shelf, waiting for us.
See it minimise: the free particle
Take the simplest case — a free particle (no forces, so V = 0) that must
travel from x = 0 at t = 0 to
x = 1 at t = 1. Common sense (and Newton) says
it should coast at constant speed: the straight world-line x(t) = t.
Let us test that against the action.
Compare a whole family of trial paths that hit the same two endpoints but bulge in the middle:
x_a(t) = t + a\,\sin(\pi t),
where the number a sets how much the path bows away from the straight
line (and a = 0 is the straight line itself). With V = 0
and mass m = 1, the action is
S = \int_0^1 \tfrac12 \dot{x}_a^2\,dt, and a short calculation gives a
clean parabola in a:
S(a) = \tfrac12 + \frac{\pi^2}{4}\,a^2.
The graph below plots it. Every wiggle — bulging up (a>0) or down
(a<0) — raises the action above its floor. The minimum sits
exactly at a = 0: the constant-speed straight line, precisely the path
Newton predicts. Nature took the route of least action, and it turned out to be the honest lazy one.
Why the minus sign? A tug-of-war
The combination T - V looks perverse until you read the principle as a
negotiation. To keep the action small over the trip, a path wants to keep T - V
small, which means keeping T low (move gently, don't thrash about) while
keeping V high (loiter where the potential energy is large). Those two
desires pull in opposite directions, and the real trajectory is the compromise between them,
integrated over the whole journey.
For the thrown ball: it "wants" to dawdle high up where gravitational V = mgy
is large, but climbing and descending costs kinetic energy, and it only has a fixed time to get
from start to finish. The stationary-action compromise between spending time high and not moving too
fast is — exactly — the parabola. The minus sign is what makes the trade-off come out right.
The traditional name promises a minimum, and for short enough paths the action
genuinely is minimised. But the honest, general statement is that the action is
stationary — its first-order change under a small wiggle is zero. That is a weaker
condition: a stationary point can be a minimum, but it can also be a saddle, where
some deformations raise the action and others lower it. Over long trajectories (for instance, orbits
that pass through a focus) the true path is a saddle, not a true minimum, and calling it "least"
is then wrong.
So the safe way to remember it is Hamilton's principle: the action is stationary,
\delta S = 0. "Least action" is a fond nickname that happens to be
literally true only sometimes. It is exactly the same subtlety as ordinary calculus, where
f'(x) = 0 flags maxima, minima and inflections alike — stationary does
not always mean smallest.
The idea has a long, slightly mystical pedigree. In the 1740s Pierre de Maupertuis proposed a
"principle of least action" and declared it evidence of divine economy — proof that God arranges
the universe to waste nothing. Euler put it on firmer mathematical footing the same decade, and
Lagrange turned it into a systematic method. The clean modern form — action as
\int (T-V)\,dt, made stationary — is due to William Rowan Hamilton in the
1830s, which is why it is properly called Hamilton's principle.
There is an even older cousin: Fermat's principle in optics (1600s), that light travels the path of
least time. That is why a straw looks bent in a glass of water — light takes the
time-cheapest route through the two media, not the straight one. Least-action thinking runs like a
thread through all of physics, and in the twentieth century Feynman revealed its deepest meaning:
in quantum mechanics a particle really does explore every path, and the classical
stationary-action trajectory is simply where all those quantum contributions reinforce instead of
cancelling.