Skip to content

Hamilton's Principle and the Principle of Least Action

CSIR NET classical mechanics: action, Hamilton's principle, Lagrange's equations by variation, when the action is a minimum, Maupertuis and Jacobi forms.

  • 19 min read
  • Updated
On this page · 13 sections
  1. Prerequisites
  2. The Hook: A Ball That Takes the Cheapest Route
  3. The Action and Hamilton’s Principle
  4. Stationary, Not Always Minimum
  5. Hamilton’s Principle in Phase Space
  6. The Principle of Least Action (Maupertuis, Euler, Jacobi)
  7. Memory Tricks
  8. Don’t Confuse
  9. Common Mistakes
  10. Practice
  11. Quick Revision Table
  12. Key Takeaways
  13. What’s Next

Prerequisites

The Hook: A Ball That Takes the Cheapest Route

The Action and Hamilton’s Principle

For a system with Lagrangian \( L(q, \dot q, t) \) and a path \( q(t) \), \( t_1 \le t \le t_2 \), in configuration space, the action is the functional

$$S[q] = \int_{t_1}^{t_2} L(q(t), \dot q(t), t)\,dt.$$

It has the dimensions of energy \( \times \) time (J s in SI units).

From \( \delta S = 0 \) to Lagrange’s equations

Vary the path: \( q_j(t) \to q_j(t) + \varepsilon\eta_j(t) \) with \( \eta_j(t_1) = \eta_j(t_2) = 0 \). Then

$$\frac{dS}{d\varepsilon}\bigg\rvert_{\varepsilon = 0} = \int_{t_1}^{t_2}\sum_j\left(\frac{\partial L}{\partial q_j}\eta_j + \frac{\partial L}{\partial \dot q_j}\dot\eta_j\right)dt.$$

Integrate the second term by parts. The boundary term \( \left[\frac{\partial L}{\partial \dot q_j}\eta_j\right]_{t_1}^{t_2} \) vanishes because \( \eta_j \) vanishes at both ends:

$$\delta S = \int_{t_1}^{t_2}\sum_j\left(\frac{\partial L}{\partial q_j} - \frac{d}{dt}\frac{\partial L}{\partial \dot q_j}\right)\eta_j\,dt.$$

For holonomic constraints the \( \eta_j \) are independent and arbitrary, so the fundamental lemma of the calculus of variations gives, for each \( j \),

$$\frac{d}{dt}\left(\frac{\partial L}{\partial \dot q_j}\right) - \frac{\partial L}{\partial q_j} = 0.$$

These are Lagrange’s equations, the Euler–Lagrange equations of the functional \( S \). Hamilton’s principle and D’Alembert’s principle are therefore equivalent for holonomic systems with potentials, and the variational form has three practical advantages:

  1. It is coordinate-free. \( S \) is a number attached to a path; it does not care which coordinates you use. Lagrange’s equations hold in every coordinate system for free.
  2. It makes symmetries visible. If \( L \) is unchanged by a transformation, so is \( S \), and Noether’s theorem extracts a conserved quantity.
  3. It extends. Fields, relativity and the path-integral formulation of quantum mechanics all start from an action.

Worked actions

Motion (end data)Classical pathClassical action \( S_{\text{cl}} \)
free particle, \( x(0) = x_1 \), \( x(T) = x_2 \)\( x_1 + (x_2 - x_1)\frac{t}{T} \)\( \dfrac{m(x_2 - x_1)^2}{2T} \)
uniform gravity, \( L = \frac{1}{2}m\dot x^2 - mgx \), \( x(0) = x(T) = 0 \)\( \frac{g}{2}t(T - t) \)\( -\dfrac{mg^2T^3}{24} \)
harmonic oscillator \( L = \frac{1}{2}m\dot x^2 - \frac{1}{2}m\omega^2x^2 \), \( x(0) = x_1 \), \( x(T) = x_2 \), \( \sin\omega T \ne 0 \)\( \frac{x_1\sin\omega(T - t) + x_2\sin\omega t}{\sin\omega T} \)\( \dfrac{m\omega}{2\sin\omega T}\left[(x_1^2 + x_2^2)\cos\omega T - 2x_1x_2\right] \)

Stationary, Not Always Minimum

Write \( S[q + \eta] = S[q] + \delta S + \delta^2 S + \dots \). On the classical path \( \delta S = 0 \); whether the path is a minimum depends on the second variation. For one degree of freedom with \( L = \frac{1}{2}m\dot x^2 - V(x) \),

$$\delta^2 S = \frac{1}{2}\int_{t_1}^{t_2}\left(m\dot\eta^2 - V''(x_{\text{cl}})\,\eta^2\right)dt.$$

For the oscillator, \( V'' = m\omega^2 \). Try \( \eta = \sin\frac{\pi(t - t_1)}{T} \) with \( T = t_2 - t_1 \):

$$\delta^2 S = \frac{m}{2}\cdot\frac{T}{2}\left(\frac{\pi^2}{T^2} - \omega^2\right).$$
Length of intervalSecond variationClassical path
\( T < \frac{\pi}{\omega} \) (less than half a period)positive for every \( \eta \ne 0 \)minimum of \( S \)
\( T = \frac{\pi}{\omega} \)vanishes for \( \eta = \sin\omega(t - t_1) \)conjugate (kinetic focal) point; with \( x_1 = x_2 = 0 \) every \( A\sin\omega t \) is a solution
\( T > \frac{\pi}{\omega} \)negative for the trial \( \eta \) above, positive for rapidly oscillating \( \eta \)saddle: stationary but neither minimum nor maximum

The general fact (Jacobi): the action is a minimum along a classical path as long as the path contains no point conjugate to its starting point. For short enough time intervals there is none, which is why “least action” is a safe slogan locally. The action of a mechanical system with \( L = T - V \) is never a maximum on a non-trivial interval: a rapid, small wiggle \( \eta \) makes \( \int m\dot\eta^2 \) dominate.

Hamilton’s Principle in Phase Space

Treat \( q \) and \( p \) as independent and vary

$$S[q, p] = \int_{t_1}^{t_2}\left(\sum_j p_j\dot q_j - H(q, p, t)\right)dt$$

with \( \delta q_j = 0 \) at the ends (\( \delta p_j \) free). The coefficient of \( \delta p_j \) gives \( \dot q_j = \frac{\partial H}{\partial p_j} \), and the coefficient of \( \delta q_j \), after integration by parts, gives \( \dot p_j = -\frac{\partial H}{\partial q_j} \). This modified Hamilton’s principle is the bridge to Hamilton’s Canonical Equations and to generating functions of canonical transformations.

The Principle of Least Action (Maupertuis, Euler, Jacobi)

Hamilton’s principle fixes the end times and lets the energy vary from path to path. The older principle fixes the energy and lets the times vary.

Define the abbreviated action (reduced action)

$$W = \int \sum_j p_j\,dq_j = \int_{t_1}^{t_2}\sum_j p_j\dot q_j\,dt.$$

For a scleronomic system (\( T = T_2 \)), \( \sum_j p_j\dot q_j = 2T \) by Euler’s theorem, so \( W = \int 2T\,dt \).

Jacobi’s form: trajectories as geodesics

Write the kinetic energy as \( T = \frac{1}{2}\left(\frac{d\rho}{dt}\right)^2 \), where \( d\rho^2 = \sum_{j,k} m_{jk}\,dq_j\,dq_k \) is the line element of the kinetic-energy metric. Then \( T = E - V \) gives \( dt = \frac{d\rho}{\sqrt{2(E - V)}} \), and \( 2T\,dt = \sqrt{2(E - V)}\,d\rho \). Hence

$$\boxed{\delta\int\sqrt{2\,(E - V(q))}\;d\rho = 0.}$$

Time has disappeared. Jacobi’s principle determines the shape of the trajectory in configuration space, not how it is traversed; the timing comes back from \( dt = d\rho/\sqrt{2(E - V)} \).

ConsequenceStatement
Free motion (\( V = 0 \))\( \delta\int d\rho = 0 \): the system moves along geodesics of the kinetic metric at constant speed. A free particle moves in a straight line; a particle confined to a smooth surface with no other force moves along a geodesic of the surface.
Optical analogy\( \delta\int n\,ds = 0 \) (Fermat) with refractive index \( n \propto \sqrt{E - V} \) for a single particle
“Refraction” at a potential step \( V_1 \to V_2 \)the momentum component along the step is conserved: \( \sqrt{E - V_1}\sin\theta_1 = \sqrt{E - V_2}\sin\theta_2 \), angles from the normal
Single particle, \( d\rho = \sqrt m\,ds \)\( \delta\int\sqrt{2m(E - V)}\,ds = \delta\int \lvert\mathbf{p}\rvert\,ds = 0 \)

Comparing the two principles

Hamilton’s principlePrinciple of least action (Maupertuis–Jacobi)
Integral\( \int L\,dt \)\( \int\sum p_j\,dq_j = \int 2T\,dt \)
Fixedend configurations and end timesend configurations and the energy \( E \)
Variesenergy may differ between pathstravel time differs between paths
Applies toany holonomic system with a (generalized) potential, \( L \) may depend on \( t \)conservative systems, \( H \) free of \( t \)
Outputthe full motion \( q(t) \)the path (orbit) in configuration space; timing from energy

Memory Tricks

Don’t Confuse

ThisvsThat
\( \delta S = 0 \): end points and end times fixed\( \Delta W = 0 \): end points and energy fixed, times free
action \( S = \int L\,dt \)abbreviated action \( W = \int p\,dq \); \( S = W - E(t_2 - t_1) \) for conservative systems
stationary actionminimum action (true only for short intervals)
varied paths need not satisfy the equations of motionthe actual path does
boundary term \( [p\,\delta q] \) vanishes because \( \delta q = 0 \) at the endswith free end points it gives natural boundary conditions \( p = 0 \)
\( L + \frac{dF(q, t)}{dt} \): same stationary path\( L + F(q, \dot q) \), not a total derivative: different path

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper. In Level 3, as in Part C of the Mathematical Sciences paper, one or more options may be correct.

Level 1: Foundation

  1. What are the SI units of the action \( S \)?
  2. Write the Euler–Lagrange equation for \( S = \int\left(\frac{1}{2}\dot q^2 - \frac{1}{2}q^2\right)dt \).
  3. In Hamilton’s principle, which quantities are the same for all competing paths?
Answers
  1. Joule-seconds (J s), energy times time, the same units as angular momentum and Planck’s constant.
  2. \( \ddot q + q = 0 \). \( \frac{\partial L}{\partial \dot q} = \dot q \), \( \frac{\partial L}{\partial q} = -q \).
  3. The end configurations and the end times. The energy may differ from path to path.

Level 2: Part B Style

Practice Q1. A free particle of mass \( m \) moves from \( x = 0 \) at \( t = 0 \) to \( x = a \) at \( t = T \). The action along the actual path is:

(a) \( \frac{ma^2}{T} \) (b) \( \frac{ma^2}{2T} \) (c) \( \frac{ma^2}{2T^2} \) (d) 0

Answer: (b)

The path is \( x = \frac{at}{T} \), \( \dot x = \frac{a}{T} \), and \( S = \int_0^T\frac{1}{2}m\frac{a^2}{T^2}\,dt = \frac{ma^2}{2T} \). Option (c) has the wrong dimensions (energy, not action).

Practice Q2. Which Lagrangian does not give the equation of motion \( \ddot q + \omega^2q = 0 \)?

(a) \( \frac{1}{2}\dot q^2 - \frac{1}{2}\omega^2q^2 \) (b) \( \dot q^2 - \omega^2q^2 \) (c) \( \frac{1}{2}\dot q^2 - \frac{1}{2}\omega^2q^2 + 3q\dot q \) (d) \( \frac{1}{2}\dot q^2 + \frac{1}{2}\omega^2q^2 \)

Answer: (d)

(b) is twice (a); (c) adds \( 3q\dot q = \frac{d}{dt}\left(\frac{3}{2}q^2\right) \). Both give the same equation as (a). (d) gives \( \ddot q - \omega^2q = 0 \), exponential rather than oscillatory motion.

Practice Q3. A particle of energy \( E = 2V_0 \) crosses from a region where \( V = 0 \) into a region where \( V = V_0 \), the boundary being a straight line. If it meets the boundary at \( 30^\circ \) to the normal, it continues at an angle to the normal of:

(a) \( 15^\circ \) (b) \( 30^\circ \) (c) \( 45^\circ \) (d) \( 60^\circ \)

Answer: (c)

\( \sqrt{E}\sin 30^\circ = \sqrt{E - V_0}\sin\theta_2 \) gives \( \sin\theta_2 = \frac{1}{2}\sqrt 2 = \frac{1}{\sqrt 2} \), so \( \theta_2 = 45^\circ \). This is the tangential momentum conservation behind Jacobi’s optical analogy.

Practice Q4. For \( L = \frac{1}{2}m\dot x^2 - mgx \), the action of the classical path from \( x(0) = 0 \) to \( x(T) = 0 \) is:

(a) \( -\frac{mg^2T^3}{24} \) (b) \( -\frac{mg^2T^3}{12} \) (c) \( \frac{mg^2T^3}{24} \) (d) 0

Answer: (a)

Kinetic part \( \frac{mg^2T^3}{24} \), potential part \( \frac{mg^2T^3}{12} \), so \( S = -\frac{mg^2T^3}{24} \). It is smaller than the action 0 of the trivial path \( x \equiv 0 \), as a minimum must be.

Level 3: Part C Style

Practice Q5. Consider \( L = \frac{1}{2}\dot x^2 - \frac{1}{2}x^2 \) and paths on \( [0, T] \) with fixed values \( x(0) \) and \( x(T) \). Which statements are true?

(a) For \( T = 1 \), the classical path gives the smallest action among all such paths. (b) For \( T = 4 \), there are admissible paths with smaller action than the classical path. (c) The first variation of the action vanishes on the classical path whenever such a path exists. (d) For \( T = \pi \), the boundary conditions \( x(0) = 0 \), \( x(\pi) = 0 \) determine a unique classical path.

Answer: (a), (b) and (c)

Here \( \omega = 1 \), so the critical length is \( \pi \).

  • (a) True. \( T = 1 < \pi \): the second variation \( \frac{1}{2}\int(\dot\eta^2 - \eta^2) \) is positive for \( \eta \ne 0 \) (Wirtinger’s inequality \( \int_0^T\dot\eta^2 \ge \frac{\pi^2}{T^2}\int_0^T\eta^2 \)), and since \( L \) is quadratic the expansion is exact, so the minimum is global.
  • (b) True. \( T = 4 > \pi \): with \( \eta = \sin\frac{\pi t}{4} \), \( \delta^2 S = \frac{1}{2}\cdot 2\left(\frac{\pi^2}{16} - 1\right) < 0 \).
  • (c) True. That is Hamilton’s principle.
  • (d) False. \( x = A\sin t \) satisfies both conditions for every \( A \).

Practice Q6. Which statements are true?

(a) In Hamilton’s principle, the competing paths join the same end configurations in the same time interval. (b) In the principle of least action of Maupertuis, the competing paths have the same energy but may take different times. (c) Replacing \( L \) by \( L + \frac{d}{dt}F(q, t) \) changes the actual path. (d) For a particle moving on a smooth surface under no force other than the normal reaction, the path is a geodesic of the surface.

Answer: (a), (b) and (d)
  • (a) True and (b) True: these are the defining conditions.
  • (c) False. The extra term adds \( F(q^{(2)}, t_2) - F(q^{(1)}, t_1) \) to the action of every path, the same constant for all, so the stationary path is unchanged.
  • (d) True. With \( V = 0 \), Jacobi’s principle is \( \delta\int d\rho = 0 \), and \( d\rho = \sqrt m\,ds \) for a particle, so the path is a geodesic. The speed is constant because \( T = E \).

Practice Q7. For \( L = \frac{1}{2}\dot x^2 - \frac{1}{2}x^2 \), consider paths with \( x(0) = 0 \) and \( x\left(\frac{\pi}{2}\right) = 1 \). Which statements are true?

(a) The classical path is \( x = \sin t \). (b) The action of the classical path is 0. (c) The straight-line path \( x = \frac{2t}{\pi} \) has a larger action than the classical path. (d) The classical path is a maximum of the action.

Answer: (a), (b) and (c)
  • (a) True. \( x = A\cos t + B\sin t \), \( x(0) = 0 \Rightarrow A = 0 \), \( x(\pi/2) = 1 \Rightarrow B = 1 \).
  • (b) True. \( S = \frac{1}{2}\int_0^{\pi/2}(\cos^2 t - \sin^2 t)\,dt = \frac{1}{2}\int_0^{\pi/2}\cos 2t\,dt = 0 \). The formula \( \frac{m\omega}{2\sin\omega T}[(x_1^2 + x_2^2)\cos\omega T - 2x_1x_2] \) with \( \cos\frac{\pi}{2} = 0 \), \( x_1 = 0 \) also gives 0.
  • (c) True. \( S = \frac{1}{2}\int_0^{\pi/2}\left(\frac{4}{\pi^2} - \frac{4t^2}{\pi^2}\right)dt = \frac{1}{\pi} - \frac{\pi}{12} \approx 0.0565 > 0 \).
  • (d) False. \( T = \frac{\pi}{2} < \pi \), so it is a minimum; (c) is consistent with that.

Quick Revision Table

IdeaFact
Action\( S = \int_{t_1}^{t_2}L\,dt \), units J s
Hamilton’s principle\( \delta S = 0 \), ends and end times fixed \( \Rightarrow \) Lagrange’s equations
Gauge\( L + \frac{dF(q,t)}{dt} \) shifts every \( S \) by the same constant
Free particle\( S_{\text{cl}} = \frac{m(x_2 - x_1)^2}{2T} \)
Thrown ball, back in time \( T \)\( S_{\text{cl}} = -\frac{mg^2T^3}{24} \), a global minimum
Oscillator\( S_{\text{cl}} = \frac{m\omega}{2\sin\omega T}[(x_1^2 + x_2^2)\cos\omega T - 2x_1x_2] \)
Minimum?yes for \( T < \pi/\omega \) (no conjugate point); saddle beyond; never a maximum
Phase-space form\( \delta\int(p\dot q - H)\,dt = 0 \Rightarrow \) Hamilton’s equations
Least action\( \Delta\int p\,dq = \Delta\int 2T\,dt = 0 \) at fixed \( E \), times free
Jacobi\( \delta\int\sqrt{2(E - V)}\,d\rho = 0 \); \( V = 0 \Rightarrow \) geodesics
Optical analogy\( n \propto \sqrt{E - V} \); \( \sqrt{E - V_1}\sin\theta_1 = \sqrt{E - V_2}\sin\theta_2 \)

Key Takeaways

What’s Next