Hamilton’s Principal and Characteristic Functions: What’s the Difference?

Hamilton’s principal and characteristic functions are two related ways of expressing the action in Hamiltonian mechanics. Their distinction is especially useful when solving the Hamilton–Jacobi equation and deciding whether a system’s motion separates into simpler coordinate equations.

The principal function depends on position and time. For an autonomous system, its time dependence can be separated, leaving the characteristic function, which depends only on position and constants such as the energy.

Hamilton’s principal function

For generalized coordinates $q=(q_1,\ldots,q_n)$, Hamilton’s principal function $S(q,t)$ is the action evaluated along a classical path from a fixed initial event to the endpoint $(q,t)$. As a function of that endpoint, it satisfies the Hamilton–Jacobi equation:

$$\frac{\partial S}{\partial t}+H\left(q_1,\ldots,q_n,\frac{\partial S}{\partial q_1},\ldots,\frac{\partial S}{\partial q_n},t\right)=0.$$

The derivatives of $S$ give the canonical momenta along the corresponding path:

$$p_i=\frac{\partial S}{\partial q_i}.$$

Thus, solving for $S$ is a way to encode the dynamics in a single function. A sufficiently general solution, called a complete integral, includes enough independent constants to specify a family of trajectories. Those constants can be related to initial conditions.

Hamilton’s characteristic function

If the Hamiltonian has no explicit time dependence, the energy is conserved and it is natural to try a separated form:

$$S(q,t)=W(q)-Et.$$

Here $E$ is the constant energy, and $W(q)$ is Hamilton’s characteristic function, also called the abbreviated action. Substituting this form into the Hamilton–Jacobi equation gives the time-independent equation

$$H\left(q_1,\ldots,q_n,\frac{\partial W}{\partial q_1},\ldots,\frac{\partial W}{\partial q_n}\right)=E.$$

For a particular path, $W$ can be expressed as the integral of momentum along the path, $W=\int \sum_i p_i\,dq_i$. Unlike the principal function, it contains no explicit time term. The simple form $S=W-Et$ relies on a time-independent Hamiltonian; it generally does not apply when the Hamiltonian depends explicitly on time.

What separability means

The Hamilton–Jacobi equation is a partial differential equation. It is separable in a chosen coordinate system when a solution can be written as a sum of single-coordinate functions, for example

$$W(q_1,\ldots,q_n)=\sum_{i=1}^{n}W_i(q_i).$$

Then each momentum is $p_i=dW_i/dq_i$. Substituting the sum into the fixed-energy equation may allow the equation to split into separate ordinary differential equations, one for each coordinate. The split introduces separation constants; their values, together with the energy, label the solutions. Whether this works depends on both the Hamiltonian and the coordinates: not every system is separable in every coordinate system.

Worked example: two uncoupled oscillators

Consider two coordinates with masses $m$ and independent oscillator frequencies $\omega_x$ and $\omega_y$:

$$H=\frac{p_x^2+p_y^2}{2m}+\frac{m}{2}\left(\omega_x^2x^2+\omega_y^2y^2\right).$$

Because the Hamiltonian is time independent, use $S=W-Et$. Try $W(x,y)=W_x(x)+W_y(y)$. The fixed-energy equation becomes

$$\frac{1}{2m}\left[\left(\frac{dW_x}{dx}\right)^2+\left(\frac{dW_y}{dy}\right)^2\right]+\frac{m}{2}\left(\omega_x^2x^2+\omega_y^2y^2\right)=E.$$

Terms depending only on $x$ can be separated from terms depending only on $y$ by setting each coordinate’s contribution equal to a constant energy:

$$\frac{1}{2m}\left(\frac{dW_x}{dx}\right)^2+\frac{m\omega_x^2x^2}{2}=E_x, \qquad \frac{1}{2m}\left(\frac{dW_y}{dy}\right)^2+\frac{m\omega_y^2y^2}{2}=E_y, \qquad E_x+E_y=E.$$

For example, the $x$ equation gives

$$\frac{dW_x}{dx}=\pm\sqrt{2mE_x-m^2\omega_x^2x^2}.$$

For $E_x>0$ and $\omega_x>0$, choosing a momentum branch and integrating gives

$$W_x(x)=\pm\left[\frac{x}{2}\sqrt{2mE_x-m^2\omega_x^2x^2}+\frac{E_x}{\omega_x}\arcsin\left(\frac{m\omega_xx}{\sqrt{2mE_x}}\right)\right]+C,$$

on a classically allowed interval, where the square root is real. The $y$ part has the same form with $y$, $E_y$, and $\omega_y$. The characteristic function is $W=W_x+W_y$, and the principal function is $S=W-Et$. The separation constants $E_x$ and $E_y$ express how the total energy is divided between the two independent oscillators.

Common mistakes

• Treating $S$ and $W$ as interchangeable. $S$ includes time dependence; $W$ is the spatial part for a time-independent Hamiltonian at fixed energy.
• Assuming every system separates. Additive separation is a property of a particular Hamiltonian in a particular coordinate system, not a universal feature.
• Forgetting the separation constants. Splitting the equation introduces constants such as $E_x$ and $E_y$; they must satisfy the original total-energy equation.
• Ignoring momentum branches. The square root in a separated momentum equation can have either sign. The sign distinguishes directions of motion and may change at a turning point.