# Gibbs' inequality

$$-\sum\limits_{i=1}^n p_{i} \cdot \log{p_{i}} \le -\sum\limits_{i=1}^n p_{i} \cdot \log{q_{i}}$$ is such a popular thing that I cannot find where it was introduced.

# My findings

I looked at "Elementary principles in statistical mechanics" (1902), and the most similar thing I've found is the theorem about the maximum entropy of uniform distributions:

Theorem IX. A uniform distribution of a given number of systems within given limits of phase gives a less average index of probability of phase than any other distribution.

meaning (index of probability is the natural logarithm of the probability) $$\int\limits_{x \in X} \left(\Delta\eta\left( x \right) + \eta \right) \cdot e^{\Delta\eta\left( x \right) + \eta} dx > \int\limits_{x \in X} \eta \cdot e^\eta dx,$$ where $$\int\limits_{x \in X} e^{\Delta\eta\left( x \right) + \eta} dx = \int\limits_{x \in X} e^\eta dx = 1.$$

In "On the equilibrium of heterogeneous substances" (1874) I see

In the first place, if the system is in a state in which its entropy is greater than in any other state of the same energy, it is evidently in equilibrium, as any change of state must involve either a decrease of entropy or an increase of energy, which are alike impossible for an isolated system.

but I can't find formulae looking like the Gibbs' inequality.

# Question

Had Gibbs actually shown the inequality or it's just called after his name as a sign of respect?

If he had, in which work can it be found and how is it formulated?