Statistical mechanics basics

In short

Statistical mechanics derives the bulk thermodynamic behaviour of matter - pressure, entropy, specific heat, magnetisation - from the statistics of the enormous number of microscopic states available to its atoms, molecules or photons. It supplies what classical thermodynamics cannot: a microscopic meaning for entropy and temperature, and quantitative equations of state and specific heats from first principles. It is a core unit of B.Sc thermal physics under UGC CBCS/NEP-aligned syllabi and carries heavy weight in CSIR-UGC NET, GATE Physics (PH), JEST, TIFR and JAM, where the partition function, Fermi-Dirac/Bose-Einstein distributions and the equipartition theorem appear almost every year. Working physicists use it in semiconductor device modelling (carrier statistics at ISRO, SCL Mohali, semiconductor design centres), magnetic and nuclear materials work at BARC, IGCAR and IUAC, and in computational/data-science roles where Monte Carlo and Boltzmann-type methods are the daily tool.

Explained

Microstates, macrostates and thermodynamic probability

• Macrostate: the state specified by bulk variables (N, V, E, or N, V, T). Microstate: one complete specification of every particle's state (position-momentum cell, or quantum state).
• Thermodynamic probability W (or Ω): the number of microstates belonging to a given macrostate. The most probable macrostate - the one with maximum W - is the equilibrium state.
• Fundamental postulate (equal a priori probability): for an isolated system in equilibrium, all accessible microstates of the same energy are equally likely.
• Boltzmann relation: S = k ln W. Entropy is a measure of the number of microscopic arrangements, so the second law becomes a statement that systems drift to the macrostate of overwhelmingly largest W.
• Counting example: 4 distinguishable coins. Macrostate "2 heads" has W = 6 microstates against W = 1 for "4 heads"; for N ~ 10²³ the peak is so sharp that fluctuations are unobservable (relative fluctuation ~ 1/√N).
• Stirling's approximation ln N! ≈ N ln N − N (for large N) is the standard tool for evaluating ln W.

Phase space and ensembles

• Phase space: a 6N-dimensional space of all coordinates and momenta (μ-space is the 6-dimensional single-particle space; Γ-space the 6N-dimensional system space).
• Classical phase space is continuous; quantum mechanics fixes the cell volume at h³ per particle per degree-of-freedom triplet, so the number of states in volume V with momentum p to p+dp is g(p)dp = (4πV p²dp)/h³ × g_s (g_s = spin/polarisation degeneracy).
• Ensemble: a large collection of imagined copies of the system, identical in macrostate and differing in microstate. Averages over the ensemble replace time averages (ergodic hypothesis).
• Microcanonical ensemble (N, V, E fixed): isolated system; every accessible microstate equally probable; entropy from S = k ln Ω.
• Canonical ensemble (N, V, T fixed): system in contact with a heat bath; probability of a microstate of energy Eᵢ is pᵢ = e^−Eᵢ/kT/Z. Energy fluctuates.
• Grand canonical ensemble (μ, V, T fixed): exchanges both energy and particles; used for quantum gases, adsorption and chemical equilibrium. Probability ∝ e^{−(Eᵢ−μNᵢ)/kT}.
• All three give the same thermodynamics in the thermodynamic limit (N → ∞ at fixed N/V); the choice is one of convenience.

Partition function - the bridge to thermodynamics

• Single-particle partition function z = Σ gᵢ e^−εᵢ/kT; for N indistinguishable non-interacting particles Z = z^N/N! (the N! removes the Gibbs paradox).
• Everything follows by differentiation: F = −kT ln Z, U = kT²(∂ln Z/∂T)_V, S = (U − F)/T, p = kT(∂ln Z/∂V)_T, C_V = (∂U/∂T)_V.
• Translational partition function z_tr = V(2πmkT/h²)^3/2 = V/λ³, where λ = h/√(2πmkT) is the thermal de Broglie wavelength.
• Separability: z = z_tr · z_rot · z_vib · z_el, because the energies add. Rotational and vibrational contributions explain the temperature-dependence of molecular specific heats.
• Classical limit holds when nλ³ ≪ 1 (dilute, hot, heavy particles); then quantum statistics reduce to Maxwell-Boltzmann.

The three statistics

• Maxwell-Boltzmann (classical): distinguishable particles, any number per state. nᵢ = gᵢ e^{−(εᵢ−μ)/kT} = gᵢ e^α e^−εᵢ/kT. Applies to dilute gases, where quantum effects are negligible.
• Bose-Einstein: identical indistinguishable particles of integer spin (bosons - photons, phonons, ⁴He atoms), symmetric wavefunction, no occupancy limit. nᵢ = gᵢ/(e^{(εᵢ−μ)/kT} − 1).
• Fermi-Dirac: identical indistinguishable particles of half-integer spin (fermions - electrons, protons, neutrons, ³He), antisymmetric wavefunction, Pauli exclusion limits one particle per quantum state. nᵢ = gᵢ/(e^{(εᵢ−μ)/kT} + 1).
• Derivation outline (examined): write W for the chosen statistics → take ln W with Stirling → maximise subject to ΣNᵢ = N and ΣNᵢεᵢ = E using Lagrange multipliers α and β → identify β = 1/kT by comparing with thermodynamics. The three cases differ only in the combinatorial expression for W.
• Counting forms: MB W = N! Π gᵢ^nᵢ/nᵢ!; BE W = Π (nᵢ + gᵢ − 1)!/[nᵢ!(gᵢ − 1)!]; FD W = Π gᵢ!/[nᵢ!(gᵢ − nᵢ)!].
• Both quantum distributions tend to MB when e^(ε−μ)/kT ≫ 1, i.e. at high temperature and low density.

Applications of classical statistics

• Maxwell speed distribution: follows from the Boltzmann factor plus the p² density of states; gives most probable speed v_p = √(2kT/m), mean speed v̄ = √(8kT/πm) and rms speed v_rms = √(3kT/m), in the ratio 1 : 1.128 : 1.225.
• Equipartition theorem: each quadratic degree of freedom in the Hamiltonian contributes ½kT to the average energy per particle (½RT per mole).
  • Monatomic gas: 3 translational → U = (3/2)RT, C_V = (3/2)R ≈ 12.47 J/mol·K, γ = 5/3.
  • Diatomic (room temperature): 3 translational + 2 rotational → C_V = (5/2)R ≈ 20.8 J/mol·K, γ = 7/5; vibration (2 quadratic terms, kinetic + potential) adds R only at high temperature.
• Dulong-Petit law: a classical solid has 6 quadratic terms per atom → C_V = 3R ≈ 24.9 J/mol·K, matching experiment at room temperature but failing as T → 0.
• Sackur-Tetrode equation gives the absolute entropy of a monatomic ideal gas and agrees with calorimetric entropies - direct experimental support for S = k ln W and for the N! factor.
• Gibbs paradox: without the 1/N!, mixing two samples of the same gas appears to raise entropy. Indistinguishability (correct Boltzmann counting) removes it.

Applications of quantum statistics

• Photon gas (Bose, μ = 0, g_s = 2): the BE distribution with the photon density of states gives Planck's law; integrating gives the Stefan-Boltzmann law u = aT⁴ and, by limits, Wien's law (hν ≫ kT) and the Rayleigh-Jeans law (hν ≪ kT). This resolved the ultraviolet catastrophe.
• Phonons and specific heat of solids: Einstein's model (all oscillators at one frequency) explains the fall of C_V at low T but decays too fast; the Debye model, with a frequency spectrum up to ν_D, gives the correct C_V ∝ T³ law and C_V → 3R at high T.
• Bose-Einstein condensation: below T_c a macroscopic fraction of bosons occupies the ground state; seen in liquid ⁴He (superfluid below 2.17 K) and in laser-cooled dilute atomic gases.
• Free-electron gas in metals (Fermi): at T = 0 all states fill up to the Fermi energy E_F = (h²/8m)(3n/π)^2/3; E_F is a few eV, so T_F = E_F/k ~ 10⁴-10⁵ K and metal electrons are strongly degenerate at room temperature.
  • Average energy at 0 K = (3/5)E_F; electronic specific heat C_el = γT ∝ T, small compared with the lattice 3R - this explains why classical theory wrongly predicted an extra (3/2)R from electrons.
  • Only electrons within ~kT of E_F participate in conduction and heat capacity; this also underlies thermionic emission, contact potential and the Pauli paramagnetism of metals.
• Degenerate matter: electron degeneracy pressure supports white dwarfs (Chandrasekhar limit); neutron degeneracy supports neutron stars.
• Semiconductors: carrier concentration from Fermi-Dirac statistics with the Fermi level in the gap; the Boltzmann approximation f ≈ e^{−(E−E_F)/kT} is valid for non-degenerate doping - the standard basis of diode and MOSFET modelling.

Comparisons, scope and limitations

• Thermodynamics vs statistical mechanics: thermodynamics is empirical, general and silent about mechanism; statistical mechanics predicts the constants (C_V, S, equations of state) but needs a microscopic model.
• Validity conditions: statistical results hold for large N; small systems show measurable fluctuations. Interactions (real gases, phase transitions, critical phenomena) make Z hard to evaluate exactly, hence mean-field models, Ising-model treatments and Monte Carlo simulation.
• Classical statistical mechanics fails at low T and high density (negative entropies, Gibbs paradox, wrong specific heats) - exactly where quantum statistics must be used.
• Indian standards and practice: all quantities use SI units with the 2019 SI definitions, in which the Boltzmann constant k is fixed exactly at 1.380649 × 10⁻²³ J/K and the kelvin is defined through it; legal metrology in India follows SI under the Legal Metrology Act, 2009 with CSIR-NPL, New Delhi as the national metrology institute maintaining the temperature scale (ITS-90).

Key points

• Microstate - one full microscopic specification; macrostate - the bulk description (N, V, E or N, V, T).
• Equal a priori probability - all accessible microstates of an isolated system are equally likely.
• S = k ln W - Boltzmann's entropy; W is the number of microstates of the macrostate.
• Stirling's approximation - ln N! ≈ N ln N − N, used to maximise ln W.
• Phase-space cell - volume h³ per particle, the quantum limit on classical counting.
• Ensembles - microcanonical (N,V,E), canonical (N,V,T), grand canonical (μ,V,T); equivalent as N → ∞.
• Boltzmann factor - probability of a state ∝ e^−ε/kT; ratio of populations = (g₂/g₁)e^−Δε/kT.
• Partition function Z - the generating function of thermodynamics: F = −kT ln Z gives U, S, p, C_V.
• Correct Boltzmann counting - Z = z^N/N! for indistinguishable particles; resolves the Gibbs paradox.
• Three statistics - MB (distinguishable, unlimited), BE (bosons, symmetric, unlimited), FD (fermions, antisymmetric, one per state).
• Classical limit - both quantum statistics → MB when nλ³ ≪ 1 (high T, low density).
• Equipartition - ½kT per quadratic degree of freedom; gives C_V = (3/2)R, (5/2)R, 3R for monatomic gas, diatomic gas and solid.
• Fermi energy - the T = 0 cut-off energy; mean energy (3/5)E_F; T_F ~ 10⁴-10⁵ K in metals.
• Electronic specific heat - linear in T and small, because only electrons within kT of E_F respond.
• Planck's law - photon gas (μ = 0) under BE statistics; yields Stefan-Boltzmann T⁴ and Wien's displacement law.
• Debye T³ law - low-temperature lattice specific heat; Einstein model falls off too steeply.
• BEC - macroscopic ground-state occupation below T_c; superfluid ⁴He below 2.17 K.
• Fluctuations - relative size ~ 1/√N, so negligible for macroscopic samples, dominant in nanoscale systems.

Formulas and examples

Example 1 - Boltzmann counting and entropy. Four distinguishable molecules are distributed between two equal halves of a box. Find W for the 2-2 split and the entropy difference from the 4-0 split.

• W(2,2) = 4!/(2!2!) = 6; W(4,0) = 4!/(4!0!) = 1.
• ΔS = k ln 6 − k ln 1 = 1.380649 × 10⁻²³ × 1.7918 = 2.47 × 10⁻²³ J/K.
• For one mole splitting into two halves, ΔS = Nk ln 2 = 8.314 × 0.6931 = 5.76 J/mol·K - the standard free-expansion result, recovered purely by counting.

Example 2 - Boltzmann factor, two-level system. A two-level atom has levels separated by Δε = 0.05 eV, both non-degenerate. Find the population ratio at T = 300 K and the mean energy.

• kT = 8.617 × 10⁻⁵ × 300 = 0.02585 eV; Δε/kT = 0.05/0.02585 = 1.934.
• n₂/n₁ = e^−1.934 = 0.145, so about 12.7% of the atoms are in the upper level (0.145/1.145).
• Z = 1 + 0.145 = 1.145; ⟨ε⟩ = Δε × 0.145/1.145 = 0.05 × 0.1266 = 6.33 × 10⁻³ eV above the ground level.
• At T = 1,200 K, Δε/kT = 0.4836, n₂/n₁ = e^−0.4836 = 0.617 - heating drives the levels towards equal population.

Example 3 - Fermi energy of copper. Copper has n = 8.5 × 10²⁸ free electrons/m³. Find E_F, T_F and the ratio of electronic to lattice specific heat at 300 K.

• 3n/π = 3 × 8.5 × 10²⁸/3.1416 = 8.117 × 10²⁸ m⁻³; its 2/3 power = (8.117 × 10²⁸)^2/3 ≈ 1.876 × 10¹⁹ m⁻².
• h²/8m = (6.626 × 10⁻³⁴)²/(8 × 9.109 × 10⁻³¹) = 4.390 × 10⁻⁶⁷/7.287 × 10⁻³⁰ = 6.024 × 10⁻³⁸ J·m².
• E_F = 6.024 × 10⁻³⁸ × 1.876 × 10¹⁹ = 1.130 × 10⁻¹⁸ J = 7.06 eV (measured value ≈ 7.0 eV).
• T_F = 1.130 × 10⁻¹⁸/1.3806 × 10⁻²³ ≈ 8.2 × 10⁴ K - far above room temperature, so the gas is degenerate.
• C_el/C_lattice ~ (π²/2)(T/T_F)/3 ≈ 4.935 × (300/82,000)/3 ≈ 0.006, i.e. the electronic term is about 0.6% of 3R at 300 K.

Example 4 - Molecular speeds and the classical limit. Nitrogen gas (M = 28 g/mol) at 300 K and 1 atm.

• m = 0.028/6.022 × 10²³ = 4.65 × 10⁻²⁶ kg; kT = 4.142 × 10⁻²¹ J.
• v_rms = √(3 × 4.142 × 10⁻²¹/4.65 × 10⁻²⁶) = √(2.672 × 10⁵) = 517 m/s; v̄ = √(8/3π) × 517 = 0.9213 × 517 = 476 m/s; v_p = √(2/3) × 517 = 422 m/s.
• λ = h/√(2πmkT): 2πmkT = 6.2832 × 4.65 × 10⁻²⁶ × 4.142 × 10⁻²¹ = 1.210 × 10⁻⁴⁵; √ = 3.479 × 10⁻²³; λ = 6.626 × 10⁻³⁴/3.479 × 10⁻²³ = 1.90 × 10⁻¹¹ m (0.019 nm).
• n = p/kT = 101,325/4.142 × 10⁻²¹ = 2.446 × 10²⁵ m⁻³ → nλ³ = 2.446 × 10²⁵ × 6.91 × 10⁻³³ ≈ 1.7 × 10⁻⁷ ≪ 1: classical statistics are fully justified for air at room temperature.

Example 5 - Specific heats from equipartition. Predict C_V, C_p and γ for helium, nitrogen at 300 K, and nitrogen at a temperature where vibration is fully active.

• Measured γ for He is 1.66 and for N₂ at 300 K is 1.40 - equipartition with rotation active and vibration frozen out matches experiment, which is itself evidence for energy quantisation.

Where it is used

• Semiconductor and VLSI design: carrier statistics, Fermi-level position and the Boltzmann approximation underpin device models at SCL Mohali, ISRO's semiconductor labs and fabless design centres (Intel, Qualcomm, Texas Instruments India).
• Nuclear and materials research: BARC, IGCAR Kalpakkam and IUAC New Delhi use Fermi-gas models for nuclear level densities, reactor physics and radiation-damage statistics; magnetisation and phase-transition models in materials groups.
• Low-temperature and quantum-technology labs: superconductivity, BEC and cryogenics work at TIFR, RRI Bengaluru, IISc and the National Quantum Mission's platforms uses BE/FD statistics directly.
• Astrophysics: degenerate electron and neutron gases, stellar interiors and CMB blackbody physics - IUCAA Pune, IIA Bengaluru, ARIES Nainital.
• Chemical, process and energy industries: statistical thermodynamics supplies equilibrium constants, heat capacities and adsorption isotherms used in catalyst and fuel-cell work (CSIR-NCL, IOCL R&D).
• Computational and data roles: Monte Carlo and Markov-chain sampling, Boltzmann machines, simulated annealing and lattice-Boltzmann CFD - widely used in quantitative finance, machine learning and simulation jobs for physics graduates.

Interview questions

• Q: What is the difference between a microstate and a macrostate? A: A macrostate is specified by a few bulk variables such as N, V and E; a microstate is one complete microscopic configuration consistent with it. A macrostate normally corresponds to an enormous number of microstates.
• Q: Why is the equilibrium state the one of maximum thermodynamic probability? A: Because all accessible microstates are equally likely, the macrostate with the most microstates is overwhelmingly the most probable, and S = k ln W makes it the state of maximum entropy.
• Q: What does the partition function physically represent? A: It is the weighted count of thermally accessible states; it fixes the probability of each state and all thermodynamic functions follow from F = −kT ln Z.
• Q: Why is the N! divided out in Z = z^N/N!? A: Identical particles are indistinguishable, so permuting them does not give a new microstate; omitting the N! makes entropy non-extensive and produces the Gibbs paradox.
• Q: How do Bose-Einstein and Fermi-Dirac statistics differ physically? A: Bosons have symmetric wavefunctions and any number can share a state; fermions have antisymmetric wavefunctions, so the Pauli principle allows at most one per quantum state.
• Q: When do quantum statistics reduce to Maxwell-Boltzmann? A: When occupancies are small, that is e^(ε−μ)/kT ≫ 1 or equivalently nλ³ ≪ 1 - high temperature, low density and large mass.
• Q: Why is the electronic contribution to a metal's specific heat so small? A: Only electrons within about kT of the Fermi energy can be excited, a fraction ~T/T_F of all electrons, so C_el ∝ T and is a tiny fraction of 3R at room temperature.
• Q: State the equipartition theorem and one place it fails. A: Each quadratic degree of freedom carries ½kT on average. It fails whenever the energy spacing exceeds kT - vibrational modes of diatomic gases at room temperature, and all solids as T → 0.
• Q: How does statistical mechanics give Planck's law? A: Apply Bose-Einstein statistics with μ = 0 to a photon gas with the (8πν²/c³) density of modes; the resulting spectrum removes the ultraviolet catastrophe of classical Rayleigh-Jeans counting.
• Q: Which ensemble would you use for a gas of electrons in a metal, and why? A: The grand canonical ensemble, because the chemical potential (Fermi level) is the natural control variable and particle number in any region fluctuates.
• Q: Does the second law ever get violated microscopically? A: Fluctuations can momentarily lower entropy in small systems, but the probability falls off as about e^−ΔS/k, so for macroscopic N the violation is unobservable - the law is statistical, not absolute.
• Q: What is Bose-Einstein condensation and where is it seen? A: Below a critical temperature a finite fraction of bosons occupies the single ground state; it is seen in superfluid ⁴He below 2.17 K and in laser-cooled dilute alkali-atom gases.

Common mistakes

• Treating W as a probability rather than a count of microstates, and then writing S = k ln (probability).
• Dropping the 1/N! in Z for identical particles, giving a non-extensive entropy and the Gibbs paradox.
• Confusing degeneracy g with occupancy n, or using the single-state distribution where the density of states g(ε) is needed in an integral.
• Applying equipartition to every degree of freedom regardless of temperature - claiming γ = 1.29 for nitrogen at 300 K, or 3R for a solid near 0 K.
• Assuming the chemical potential is constant: μ equals E_F only at T = 0 and shifts with temperature, and μ = 0 only for photons and phonons.
• Mixing units - using eV in e^−ε/kT with k in J/K; remember kT = 0.02585 eV at 300 K.
• Saying Fermi-Dirac statistics apply to electrons "because they are charged"; the reason is half-integer spin and antisymmetry, so neutrons and ³He atoms obey them too.

Related topics

Thermal physics and thermodynamics · Kinetic theory of gases · Quantum mechanics: identical particles and spin · Solid state physics: free-electron theory and band theory · Blackbody radiation and quantum theory origins · Semiconductor physics: carrier statistics.