Introduction to Irreversible Thermodynaics
2026-01-07
Key ideas from last lecture:
\[\tau_{\text{observation}} \gg \tau_{\text{relaxation}}\]
\[S = k_B \ln\!\Omega\]
After this lecture, you will be able to:
Part of the introductions are adapted from MAT E 374 by Dr. Hao Zhang
Dynamics from Newton & Laplace point of view:
Newtonian dynamics in a nutshell
Why can’t we determine the system state using Newtonian Dynamics (yet)?
What is the alternative?

Let’s reintroduce the microscopic picture by Boltzmann
Central principle linking macrostate and microstates:
The observed macroscopic thermodynamic property \(A_{obs}\) is an ensemble average of all possible microstates \(i\)
\[ A_{obs} = \langle A \rangle = \sum_i A_i \, p_i \]
Consider a cannonical ensemble (constant \(N, V, U\)) 1
I and R can only exchange internal energy, not particles or volume.
Q: The total internal energy of the universe \(U_T\) is fixed. How do the inner and reservoir systems balance their energies?
Let’s first see the equilibrium state.
\[P(U_I) \propto \Omega(U_T - U_I^0)\]
\[ \ln\!\Omega(U_T - U_I^0) = \ln\!\Omega(U_T) - \frac{\partial \ln\!\Omega}{\partial U} \bigg\vert_{U=U_T} \cdot U_I^0 + \dots \]
\[ \ln P(U_I^0) \propto \ln\!\Omega(U_T) - \frac{\partial S_T}{k_B \partial U} \bigg\vert_{U=U_T} \cdot U_I^0 \]
Recall the classical thermodynamic relation:
\[ dU = TdS - pdV \]
We have \(\dfrac{\partial S}{\partial U}\bigg\vert_{V} = \dfrac{1}{T}\), therefore
\[ P(U_I^0) \propto \Omega(U_T) \exp(-\dfrac{U_I^0}{k_B T}) \]
This leads the famous Boltzmann distribution
What if we’re not at equilibrium?
\(T_I\) and \(T_R\) may be two distinct values.
We will apply local equilibrium assumption here
One postulate for stat. mech. is that the entropy of the universe always increases.
We go further by say the equilibrium maximizes the universe entropy
\[ \left( \dfrac{ \partial \ln \Omega(U_T) }{\partial U_I} \right)_{N, V, U} = 0 \]
\[ \left( \dfrac{ \partial \ln \Omega(U_I) }{\partial U_I} \right)_{N, V, U} + \left( \dfrac{ \partial \ln \Omega(U_R) }{\partial U_I} \right)_{N, V, U} = 0 \]
Use the relation \(\partial U_I = -\partial U_R\)
\[ \left( \dfrac{ \partial \ln \Omega(U_I) }{\partial U_I} \right)_{N, V, U} = \left( \dfrac{ \partial \ln \Omega(U_R) }{\partial U_R} \right)_{N, V, U} \]
Apply Boltzmann equation
\[ T_I^{-1} = T_R^{-1} \]
Yay! This is something we know, temperature at equilibrium.
In Onsager’s Irreversible Thermodynamics, for any thermodynamic parameter \(X\), we define
Affinity \(F\): the influence of \(X_i\) to the universe entropy \[ F_i = \frac{\partial S}{\partial X_i} \]
Flux \(J\): the change of parameter \(X_i\) (or density of it) per unit time \[ J_i = \frac{d X_i}{d t} \]
Kinetic Coefficient \(L_{ij}\): influence of flux type \(i\) by affinity type \(j\) (linear term). \[ L_{ij} = \dfrac{\partial J_i}{\partial F_j} \]
It is generally possible to use principles of the continuum limit to define meaningful, useful, local values of various thermodynamics quantities, e.g., chemical potential \(\mu_i\).
Useful kinetic theories can be developed by assuming a functional relationship between the rate of a process and the local departure from equilibrium (“driving force”).
Irreversible Process is An Increase of Entropy in the Universe
\[ \frac{dS}{dt} = \sum_i \frac{\partial S}{\partial X_i} \frac{\partial X_i}{\partial t} = \sum_i F_i J_i \]
Or in vector term \[ \mathbf{J} = \mathbf{L}\cdot\mathbf{F} \]
Onsager propose that
\[ L_{ij} = L_{ji} \]
This is a key component in material kinetics!