Statistical mechanics tries to predict macroscopic behavior of a thermodynamic system from microscopic considerations. Suppose we are dealing with system with phase space $\Gamma = (\Lambda \times \mathbb{R}^{d})^{N}$ where $N \ge 1$ is large and $\Lambda \subset \mathbb{R}^{d}$. Let me write $x = (q,p) \in \Gamma$ for a point in the phase space.

Boltzmann argued that when we measure an observable $f$ (continuous function on $\Gamma$), the measurament takes some time. Hence, it is natural do expect that the equilibrium value of $f$ is defined by the limit: $$\bar{f}(x) := \lim_{T\to \infty}\frac{1}{T}\int_{0}^{T}f(x_{t})dt \tag{1}\label{1}$$ where $x$ is the state of the system at an initial time $t=0$ and and $x_{t}$ is the state of the system at a posterior time $t$. Physically, we would expect that this measurement does not depend on the initial state of the system, so it becomes a constant, say, $\bar{f}$. This is the ergodic hypothesis and it implies that this time average $\bar{f}$ coincides with the ensemble expectation: $$\langle f\rangle = \int f(x)\rho(x)dx \tag{2}\label{2}$$ where $\rho$ is the probability density (of states) on $\Gamma$.

I am happy with the above definitions. However, my professor took a slightly different route in his classes. He starts with a probability density $\rho_{0}$ at time $t=0$ and $\rho_{t}$ is the probability density at time $t > 0$. By Liouville Theorem, one can show that $\rho_{0}(x) = \rho_{t}(x_{t})$. He then wrote that, at time $t$, the ensemble average of $f$ is given by: $$\langle f\rangle_{t} = \int f(x)\rho_{t}(x)dx, \tag{3}\label{3}$$ to which I agree. This is the tricky part: he defines the equilibrium value of $f$ to be: $$\langle f\rangle_{\text{eq}} = \lim_{T\to \infty}\frac{1}{T}\int_{0}^{T}\langle f\rangle_{t}dt \tag{4}\label{4}$$

Even when (\ref{1}) does not depend on the initial state $x$, it seems that the previous time average $\bar{f}=\langle f\rangle$ and the equilibrium value $\langle f\rangle_{\text{eq}}$ are not equivalent definitions. So, I wonder: what is the correct definition of the equlibrium state of $f$, and how to reconcile these two approaches?

  • $\begingroup$ If I were you I would ask the professor first, then I would argue for or against the answer I'd receive (I am assuming this is not a recorded class). $\endgroup$
    – Themis
    Jun 21, 2023 at 23:25

1 Answer 1


Your professor’s approach is a generalized version of yours. You can recover your definition when $\rho_0$ is a Dirac delta centered on your initial condition.

The corresponding $\rho_{eq}=\lim_{t\to\infty}\rho_t$, defined by the weak limit, will thus correspond to your ensemble in the case of an ergodic system: $$ \rho_{eq}=\rho $$ If your system is not ergodic, then your $\rho$ will depend on the initial condition and similarly $\rho_{eq}$ will depend on $\rho_0$.

For a non ergodic case, take for example the harmonic oscillator: $$ H=\frac{p^2+q^2}{2} $$ Going to action angle coordinates: $$ q=\sqrt{2I}\cos\phi\\ p=\sqrt{2I}\sin\phi $$ This gives: $$ \rho_{eq}(I,\phi)=\int \rho_0(I,\phi’)d\phi’ $$ ie the density is averaged over the energy isolines.

Similarly, your $\rho$ will now depend on the initial energy, i.e. initial $I$. Your $\rho$ will be supported on the initial energy isoline: $$ \rho(I,\phi)=\delta(I-I_0) $$

Hope this helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.