Non-Equilibrium Processes: An Introduction
CSIR NET Part C: transport laws, entropy production, Onsager relations, master equation, Boltzmann equation, Fokker–Planck, fluctuation–dissipation.
- 18 min
On this page · 16 sections
- Prerequisites
- The Hook: A Wire That Hums With Temperature
- Equilibrium, Steady State and Relaxation
- Linear Transport Laws
- Entropy Production and Onsager Reciprocity
- The Master Equation and Detailed Balance
- The Boltzmann Transport Equation
- The Fokker–Planck (Smoluchowski) Equation
- Linear Response and Fluctuation–Dissipation
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Diffusion Equation, Random Walk and Brownian Motion: the Langevin equation and Einstein relation, which this page generalises.
- Laws of Thermodynamics, Entropy and Heat Engines: the Clausius inequality, which entropy production makes local.
- Phase Space, Microstates and the Microcanonical Ensemble: Liouville’s theorem and why entropy increase is statistical.
The Hook: A Wire That Hums With Temperature
Equilibrium, Steady State and Relaxation
| State | Time dependence | Currents | Entropy production |
|---|---|---|---|
| Equilibrium | none | none (detailed balance) | zero |
| Non-equilibrium steady state (NESS) | none | non-zero, steady | positive, constant |
| Transient relaxation | decays towards equilibrium or NESS | changing | positive |
A rod between a hot and a cold reservoir reaches a steady state, not equilibrium: its temperature profile is time-independent but heat flows through it and entropy is generated continuously.
Local equilibrium. If gradients are gentle compared with microscopic scales (mean free path, collision time), each small region is close to equilibrium with its own \( T(\mathbf r) \), \( \mu(\mathbf r) \), \( \mathbf u(\mathbf r) \). This is the assumption under which thermodynamic quantities can be defined locally and linear transport laws hold.
Linear Transport Laws
Close to equilibrium, fluxes are proportional to gradients:
| Law | Flux | Driving gradient | Coefficient |
|---|---|---|---|
| Fourier | heat \( \mathbf J_q = -\kappa\nabla T \) | temperature | thermal conductivity \( \kappa \) |
| Fick | particles \( \mathbf J_n = -D\nabla n \) | concentration | diffusion coefficient \( D \) |
| Ohm | charge \( \mathbf J = \sigma\mathbf E \) | electric potential | conductivity \( \sigma \) |
| Newton | momentum (shear stress) \( \tau_{xy} = \eta\,\partial u_x/\partial y \) | velocity | viscosity \( \eta \) |
Kinetic-theory estimates for a dilute gas
A molecule carries a quantity across a distance of about one mean free path \( \lambda = 1/(\sqrt2n\sigma_c) \) at mean speed \( \bar v \). With \( c_v \) the heat capacity per molecule:
$$\boxed{D \approx \tfrac{1}{3}\bar v\lambda, \qquad \eta \approx \tfrac{1}{3}nm\bar v\lambda, \qquad \kappa \approx \tfrac{1}{3}nc_v\bar v\lambda}$$The numerical factors are those of the simple model; rigorous theory changes them but not the scaling. Since \( n\lambda \) is independent of density, \( \eta \) and \( \kappa \) of a dilute gas are independent of pressure at fixed \( T \) and grow as \( \sqrt T \). Also \( \kappa/\eta = c_v/m \) in this model. The Kinetic Theory of Gases page works these numbers for real gases.
Entropy Production and Onsager Reciprocity
Fluxes and forces
For a system in local equilibrium the rate of entropy production per unit volume can be written as
$$\boxed{\sigma_s = \sum_iJ_iX_i \geq 0}$$where \( J_i \) are fluxes and \( X_i \) the conjugate thermodynamic forces. For heat conduction, \( J = \mathbf J_q \) and \( X = \nabla(1/T) \), so
$$\sigma_s = \mathbf J_q\cdot\nabla\frac{1}{T} = \frac{\kappa(\nabla T)^2}{T^2} \geq 0$$which requires \( \kappa \geq 0 \): heat flows from hot to cold. Globally, a heat current \( \dot Q \) from a reservoir at \( T_h \) to one at \( T_c \) produces entropy at the rate
$$\boxed{\dot S = \dot Q\left(\frac{1}{T_c} - \frac{1}{T_h}\right) > 0}$$Linear laws and Onsager relations
Near equilibrium each flux depends linearly on all forces:
$$J_i = \sum_jL_{ij}X_j$$The Onsager reciprocal relations, derived from microscopic time-reversal symmetry, state
$$\boxed{L_{ij} = L_{ji}}$$(in a magnetic field, \( L_{ij}(\mathbf B) = L_{ji}(-\mathbf B) \)). Positive entropy production requires the matrix \( L \) to be positive semi-definite: \( L_{ii} \geq 0 \) and \( L_{11}L_{22} \geq L_{12}^2 \).
Thermoelectric example. In a conductor, charge and heat currents are both driven by gradients of \( T \) and electrochemical potential. The cross-coefficients describe the Seebeck effect (a temperature gradient produces a voltage, coefficient \( S \)) and the Peltier effect (a current carries heat, coefficient \( \Pi \)). Onsager symmetry gives the second Kelvin relation:
$$\boxed{\Pi = ST}$$so measuring one coefficient gives the other.
Minimum entropy production. In the linear regime, with some forces held fixed by the boundaries, the steady state is the one that minimises the total entropy production compatible with those constraints (Prigogine’s theorem). It holds only in the linear regime.
The Master Equation and Detailed Balance
For a system that hops between discrete states \( n \) with transition rates \( W_{m\to n} \), the probabilities obey
$$\boxed{\frac{dP_n}{dt} = \sum_{m\neq n}\left[W_{m\to n}P_m - W_{n\to m}P_n\right]}$$gain from other states minus loss to them.
Detailed balance. The equilibrium (Boltzmann) distribution is reached if every pair of states balances separately:
$$\boxed{W_{m\to n}P_m^{\text{eq}} = W_{n\to m}P_n^{\text{eq}} \ \Leftrightarrow\ \frac{W_{m\to n}}{W_{n\to m}} = e^{-\beta(E_n - E_m)}}$$Detailed balance is stronger than stationarity: a stationary distribution can have net probability currents circulating around loops. Those are the non-equilibrium steady states. A quick test: detailed balance holds if and only if, for every closed cycle of states, the product of rates one way round equals the product the other way (Kolmogorov’s criterion).
The Metropolis rule used in Monte Carlo simulation of the Ising model, \( W = \min(1, e^{-\beta\Delta E}) \), is built to satisfy detailed balance.
Two-level relaxation
For states \( \downarrow \) (energy 0) and \( \uparrow \) (energy \( \Delta \)) with rates \( W_\uparrow \) (up) and \( W_\downarrow \) (down):
$$\frac{dP_\uparrow}{dt} = W_\uparrow P_\downarrow - W_\downarrow P_\uparrow = W_\uparrow - (W_\uparrow + W_\downarrow)P_\uparrow$$$$\boxed{P_\uparrow(t) = P_\uparrow^{\text{ss}} + \left[P_\uparrow(0) - P_\uparrow^{\text{ss}}\right]e^{-t/T_1}, \qquad \frac{1}{T_1} = W_\uparrow + W_\downarrow, \qquad P_\uparrow^{\text{ss}} = \frac{W_\uparrow}{W_\uparrow + W_\downarrow}}$$The relaxation rate is the sum of the two rates, not either one. In contact with a bath, detailed balance makes \( W_\uparrow/W_\downarrow = e^{-\beta\Delta} \). This is the spin–lattice relaxation (\( T_1 \)) of magnetic resonance.
The Boltzmann Transport Equation
For a dilute gas, the one-particle distribution \( f(\mathbf r, \mathbf v, t) \) evolves as
$$\boxed{\frac{\partial f}{\partial t} + \mathbf v\cdot\nabla_{\mathbf r}f + \frac{\mathbf F}{m}\cdot\nabla_{\mathbf v}f = \left(\frac{\partial f}{\partial t}\right)_{\text{coll}}}$$The left side is free streaming plus the effect of external forces (Liouville flow); the right side, collisions.
H-theorem. For Boltzmann’s two-body collision integral, \( H(t) = \int f\ln f\,d^3v \) never increases, and stays constant only for the Maxwell–Boltzmann distribution. With \( S = -k_BH \) (per volume, up to constants), this is a kinetic derivation of the second law, resting on the assumption that colliding molecules are uncorrelated before they collide (“molecular chaos”).
Relaxation-time approximation. Replace the collision term by
$$\left(\frac{\partial f}{\partial t}\right)_{\text{coll}} \approx -\frac{f - f_0}{\tau}$$Drude conductivity. For electrons (charge \( -e \)) in a uniform static field \( \mathbf E = E\hat x \), steady state and small deviation give \( f \approx f_0 + \dfrac{e E\tau}{m}\dfrac{\partial f_0}{\partial v_x} \). Integrating \( \mathbf J = -e\int\mathbf vf \):
$$\boxed{\sigma = \frac{ne^2\tau}{m}}$$the same result as the elementary drift picture \( m\dot{\mathbf v} = -e\mathbf E - m\mathbf v/\tau \).
Wiedemann–Franz law. For a degenerate electron gas with the same \( \tau \) for charge and heat transport (Sommerfeld theory),
$$\boxed{\frac{\kappa}{\sigma T} = L_0 = \frac{\pi^2}{3}\left(\frac{k_B}{e}\right)^2 = 2.44\times10^{-8}\ \text{W}\,\Omega\,\text{K}^{-2}}$$independent of \( \tau \), \( n \) and \( m \). Good metals obey it near room temperature.
The Fokker–Planck (Smoluchowski) Equation
The Langevin equation follows one Brownian particle; the Fokker–Planck equation evolves the probability density of many. For an overdamped particle in a potential \( U(x) \) with friction \( \gamma \) and diffusion coefficient \( D \):
$$\boxed{\frac{\partial P}{\partial t} = \frac{\partial}{\partial x}\left[\frac{U'(x)}{\gamma}P\right] + D\frac{\partial^2P}{\partial x^2}}$$The first term is drift in the force \( -U' \); the second is diffusion. The stationary, zero-current solution is
$$P_{\text{st}}(x) \propto \exp\left(-\frac{U(x)}{\gamma D}\right)$$which is the Boltzmann distribution only if \( \gamma D = k_BT \): the Einstein relation once again.
Harmonic trap. For \( U = \frac{1}{2}k_sx^2 \) (an optical tweezer, say), \( \langle x\rangle \) relaxes as \( e^{-t/\tau_r} \) with \( \tau_r = \gamma/k_s \), and the stationary variance is \( \langle x^2\rangle = k_BT/k_s \) (equipartition), independent of \( \gamma \).
Barrier crossing. For a particle escaping over a barrier of height \( \Delta U \gg k_BT \), the rate has the Arrhenius form \( \propto e^{-\Delta U/k_BT} \), with a prefactor set by the curvatures of the well and barrier and by the friction (Kramers’ theory). This is the stochastic basis of chemical reaction rates.
Linear Response and Fluctuation–Dissipation
When a weak force \( F(t) \) couples to a variable \( A \) (\( H = H_0 - AF(t) \)), the response is linear:
$$\langle\delta A(t)\rangle = \int_{-\infty}^{t}\chi(t - t')F(t')\,dt'$$Causality (\( \chi(t) = 0 \) for \( t < 0 \)) makes the real and imaginary parts of \( \chi(\omega) \) a Kramers–Kronig pair. The fluctuation–dissipation theorem relates the dissipative part \( \text{Im}\,\chi(\omega) \) to the equilibrium spectrum of fluctuations of \( A \). Classically, the response function is proportional to the time derivative of the equilibrium autocorrelation function of \( A \). Onsager’s regression hypothesis says the same thing in words: a small spontaneous fluctuation decays, on average, the way a small imposed disturbance does.
Examples you already know:
| Fluctuation | Dissipation | Relation |
|---|---|---|
| Brownian velocity \( \langle v(0)v(t)\rangle \) | friction \( \gamma \) (mobility) | \( D = k_BT/\gamma = \int_0^\infty\langle v(0)v(t)\rangle dt \) |
| Voltage noise across a resistor | resistance \( R \) | Johnson–Nyquist \( \langle V^2\rangle = 4k_BTR\,\Delta f \) |
| Current fluctuations | conductivity | Green–Kubo formula for \( \sigma \) |
| Energy fluctuations | heat capacity | \( \langle\Delta E^2\rangle = k_BT^2C_V \) (equilibrium version) |
The Johnson–Nyquist formula is the classical limit (\( hf \ll k_BT \)), valid across the bandwidth \( \Delta f \).
Beyond linear response: work fluctuation relations
Far from equilibrium the linear theory fails, but some exact results survive. The Jarzynski equality states that for a system started in equilibrium and driven by any protocol,
$$\langle e^{-\beta W}\rangle = e^{-\beta\Delta F}$$where \( W \) is the work done in each realisation and \( \Delta F \) the equilibrium free-energy difference. Jensen’s inequality then gives \( \langle W\rangle \geq \Delta F \), the second law on average, while allowing individual realisations with \( W < \Delta F \) in small systems.
graph TD
A["Microscopic dynamics"] --> B["Master equation: discrete states, rates W"]
A --> C["Boltzmann equation: dilute gas, collisions"]
A --> D["Langevin / Fokker-Planck: Brownian particle"]
B --> E["Detailed balance gives Boltzmann equilibrium"]
C --> F["Relaxation time: Drude, Wiedemann-Franz, H-theorem"]
D --> G["Einstein relation D = kT/gamma"]
E --> H["Near equilibrium: linear laws, Onsager L_ij = L_ji, entropy production >= 0"]
F --> H
G --> H
H --> I["Fluctuation-dissipation: response fixed by equilibrium fluctuations"]Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Equilibrium: no currents, detailed balance | Steady state: time-independent but with currents | |
| Detailed balance (every pair) | Global balance (stationarity only) | |
| Onsager \( L_{ij} = L_{ji} \) (cross-coefficients symmetric) | Positivity \( L_{ii} \geq 0 \) (from the second law) | |
| Seebeck: \( \nabla T \) gives a voltage | Peltier: a current carries heat; \( \Pi = ST \) | |
| Relaxation-time approximation: \( -(f - f_0)/\tau \) | Full Boltzmann collision integral (needed for the H-theorem) | |
| Langevin: stochastic trajectory | Fokker–Planck: deterministic equation for the probability density | |
| Fluctuation–dissipation: near equilibrium | Jarzynski: any protocol, from an equilibrium start |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Use \( k_B = 1.381\times10^{-23} \) J K\( ^{-1} \), \( e = 1.602\times10^{-19} \) C, \( m_e = 9.109\times10^{-31} \) kg.
Level 1: Foundation
- State the Onsager reciprocal relations (no magnetic field).
- What condition on transition rates guarantees that a master equation relaxes to the Boltzmann distribution?
- Can the local entropy production rate be negative?
Answers
- \( L_{ij} = L_{ji} \): the coefficient linking flux \( i \) to force \( j \) equals that linking flux \( j \) to force \( i \).
- Detailed balance: \( W_{m\to n}/W_{n\to m} = e^{-\beta(E_n - E_m)} \).
- No: \( \sigma_s = \sum J_iX_i \geq 0 \), zero only in equilibrium.
Level 2: Part B Style
Practice Q1. A two-level system has an upward transition rate 1.0 s\( ^{-1} \) and a downward rate 3.0 s\( ^{-1} \). Its relaxation time and steady-state upper-level population are:
(a) 0.25 s and 0.25 (b) 0.33 s and 0.25 (c) 1.0 s and 0.33 (d) 0.25 s and 0.75
Answer: (a) 0.25 s and 0.25
\( T_1 = 1/(W_\uparrow + W_\downarrow) = 1/4 \) s. \( P_\uparrow^{\text{ss}} = W_\uparrow/(W_\uparrow + W_\downarrow) = 1/4 \). (b) uses only the downward rate; (d) gives the lower-level population.
Practice Q2. A steady heat current of 100 W flows through a rod from a reservoir at 400 K to one at 300 K. The rate of entropy production is closest to:
(a) 0.25 W K\( ^{-1} \) (b) 0.083 W K\( ^{-1} \) (c) 0.33 W K\( ^{-1} \) (d) zero, because the rod is in a steady state
Answer: (b) 0.083 W/K
\( \dot S = 100\left(\dfrac{1}{300} - \dfrac{1}{400}\right) = 100\times8.33\times10^{-4} = 0.0833 \) W K\( ^{-1} \). The rod’s own entropy is constant, but the reservoirs’ total increases; a steady state still produces entropy.
Practice Q3. A metal has \( 8.5\times10^{28} \) conduction electrons per m\( ^3 \) and a relaxation time \( 2.5\times10^{-14} \) s. Its Drude conductivity is closest to:
(a) \( 6.0\times10^{5} \) S m\( ^{-1} \) (b) \( 6.0\times10^{7} \) S m\( ^{-1} \) (c) \( 3.7\times10^{8} \) S m\( ^{-1} \) (d) \( 6.0\times10^{9} \) S m\( ^{-1} \)
Answer: (b) 6.0 × 10⁷ S/m
\( \sigma = \dfrac{ne^2\tau}{m} = \dfrac{8.5\times10^{28}\times(1.602\times10^{-19})^2\times2.5\times10^{-14}}{9.109\times10^{-31}} = 5.99\times10^7 \) S m\( ^{-1} \).
Practice Q4. The rms thermal noise voltage across a 1.0 MΩ resistor at 300 K, measured over a 1.0 kHz bandwidth, is about:
(a) 0.13 μV (b) 1.3 μV (c) 4.1 μV (d) 41 μV
Answer: (c) 4.1 μV
\( V_{\text{rms}} = \sqrt{4k_BTR\Delta f} = \sqrt{4\times1.381\times10^{-23}\times300\times10^6\times10^3} = \sqrt{1.66\times10^{-11}} = 4.07\times10^{-6} \) V.
Level 3: Part C Style
Practice Q5. A metal has electrical conductivity \( 6.0\times10^7 \) S m\( ^{-1} \) at 300 K. Consider:
- The Wiedemann–Franz law predicts an electronic thermal conductivity of about 440 W m\( ^{-1} \) K\( ^{-1} \).
- In the Sommerfeld model with a single relaxation time, the ratio \( \kappa/\sigma T \) does not depend on \( \tau \).
- In a dilute gas, by contrast, the simple kinetic-theory viscosity is independent of pressure at fixed temperature.
- Onsager symmetry requires the thermoelectric cross-coefficients to satisfy \( L_{12} = -L_{21} \) in zero magnetic field.
Which are correct?
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
- \( \kappa = L_0\sigma T = 2.44\times10^{-8}\times6.0\times10^7\times300 = 440 \) W m\( ^{-1} \) K\( ^{-1} \). True.
- Both \( \kappa \) and \( \sigma \) are proportional to \( \tau \), which cancels. True.
- \( \eta \approx \frac{1}{3}nm\bar v\lambda \) with \( n\lambda \) density independent. True.
- Onsager gives \( L_{12} = +L_{21} \) in zero field. False.
Practice Q6. A system has three states 1, 2, 3 of equal energy arranged in a cycle. Transitions \( 1\to2 \), \( 2\to3 \), \( 3\to1 \) occur at rate 2 s\( ^{-1} \) and the reverse transitions at rate 1 s\( ^{-1} \). Consider:
- The stationary distribution is \( P_1 = P_2 = P_3 = \frac{1}{3} \).
- The stationary state satisfies detailed balance.
- In the stationary state a net probability current of \( \frac{1}{3} \) s\( ^{-1} \) circulates around the cycle.
- The stationary entropy production rate is \( k_B\ln2 \) per second.
Which are correct?
(a) 1 and 2 only (b) 1 and 3 only (c) 1, 3 and 4 only (d) 1, 2, 3 and 4
Answer: (c) 1, 3 and 4 only
- Every state gains \( 2P_{\text{prev}} + 1\cdot P_{\text{next}} \) and loses \( 3P \); the uniform distribution balances this by symmetry. True.
- Detailed balance would need \( 2\times\frac{1}{3} = 1\times\frac{1}{3} \) on each link. Kolmogorov’s test also fails: \( 2^3 \neq 1^3 \). False.
- Net current on each link: \( 2\times\frac{1}{3} - 1\times\frac{1}{3} = \frac{1}{3} \) s\( ^{-1} \). True.
- Entropy production \( = k_B\sum_{\text{links}}J\ln(W_+P/W_-P) = 3\times\frac{1}{3}\times\ln2\;k_B = k_B\ln2 \) per second. True. This is a non-equilibrium steady state.
Practice Q7. A bead of radius 0.5 μm in water (\( \eta = 1.0\times10^{-3} \) Pa s, \( T = 300 \) K) is held in an optical trap of stiffness \( k_s = 1.0\times10^{-6} \) N m\( ^{-1} \). Treat it as an overdamped Brownian particle. Consider:
- The rms displacement of the bead from the trap centre along one axis is about 64 nm.
- The relaxation time of the mean position is about 9.4 ms.
- The stationary position distribution is a Gaussian whose width does not depend on the viscosity.
- Doubling the viscosity doubles the rms displacement.
Which are correct?
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
- \( \langle x^2\rangle = k_BT/k_s = 4.14\times10^{-21}/10^{-6} = 4.14\times10^{-15} \) m\( ^2 \); rms \( = 64 \) nm. True.
- \( \gamma = 6\pi\eta r = 9.42\times10^{-9} \) kg s\( ^{-1} \); \( \tau_r = \gamma/k_s = 9.4\times10^{-3} \) s. True.
- \( P_{\text{st}} \propto e^{-k_sx^2/2k_BT} \): Boltzmann, independent of \( \gamma \). True.
- Viscosity changes only how fast the bead relaxes (\( \tau_r \) doubles), not the stationary width. False.
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| Transport laws | \( \mathbf J_q = -\kappa\nabla T \), \( \mathbf J_n = -D\nabla n \), \( \mathbf J = \sigma\mathbf E \), \( \tau_{xy} = \eta\,\partial_yu_x \) |
| Kinetic estimates | \( D, \eta, \kappa \approx \frac{1}{3}\bar v\lambda\times(1, nm, nc_v) \) |
| Entropy production | \( \sigma_s = \sum J_iX_i \geq 0 \); heat flow \( \dot Q(1/T_c - 1/T_h) \) |
| Onsager | \( L_{ij} = L_{ji} \); with field \( L_{ij}(\mathbf B) = L_{ji}(-\mathbf B) \) |
| Kelvin relation | \( \Pi = ST \) |
| Master equation | \( \dot P_n = \sum[W_{m\to n}P_m - W_{n\to m}P_n] \) |
| Detailed balance | \( W_{m\to n}/W_{n\to m} = e^{-\beta(E_n - E_m)} \) |
| Two-level | \( 1/T_1 = W_\uparrow + W_\downarrow \), \( P_\uparrow^{\text{ss}} = W_\uparrow/(W_\uparrow + W_\downarrow) \) |
| Boltzmann equation | \( \partial_tf + \mathbf v\cdot\nabla f + (\mathbf F/m)\cdot\nabla_vf = \) collisions; H-theorem |
| Drude | \( \sigma = ne^2\tau/m \) |
| Wiedemann–Franz | \( \kappa/\sigma T = \pi^2k_B^2/3e^2 = 2.44\times10^{-8} \) W Ω K\( ^{-2} \) |
| Fokker–Planck | \( \partial_tP = \partial_x(U'P/\gamma) + D\partial_x^2P \); Boltzmann iff \( D = k_BT/\gamma \) |
| Trap | \( \langle x^2\rangle = k_BT/k_s \), \( \tau_r = \gamma/k_s \) |
| Johnson noise | \( \langle V^2\rangle = 4k_BTR\Delta f \) |
| Jarzynski | \( \langle e^{-\beta W}\rangle = e^{-\beta\Delta F} \Rightarrow \langle W\rangle \geq \Delta F \) |
Key Takeaways
What’s Next
- Thermodynamics and Statistical Physics (Advanced): Quick Revision: the whole chapter on one page.
- Diffusion Equation, Random Walk and Brownian Motion: revisit the Langevin equation with the fluctuation–dissipation idea in mind.
- Laws of Thermodynamics, Entropy and Heat Engines: the equilibrium second law that entropy production generalises.
- Rate laws, Arrhenius and transition-state theory from the chemist’s side: Chemical Kinetics.