The Thermodynamic Limit and Ensemble Equivalence

Introduction

An isolated solid has fixed energy, while the same solid in a thermostat has fluctuating energy. Why do both descriptions normally predict the same bulk temperature, pressure and heat capacity? In a very large system, the absolute energy variation can be substantial but remains tiny relative to its total energy scale. This concentration of probability underlies ensemble equivalence and explains why macroscopic thermodynamics works without tracking every molecule.

Core explanation

The thermodynamic limit takes N and V to infinity while keeping density N/V finite. For many stable systems with short-range interactions away from singular behaviour, mean energy U scales with N and canonical variance Var(E) = kT²C V also scales with N because C V is extensive. Standard deviation σ E then scales as √N. Relative to a characteristic energy of order N, σ E/N scales as 1/√N and tends to zero. The canonical distribution becomes sharply concentrated in relative terms around one energy density.

A corresponding grand canonical system has fluctuating N, but under ordinary noncritical conditions Var(N) also tends to scale with N, giving relative number fluctuations of order 1/√N. The microcanonical, canonical and grand canonical ensembles can then give the same limiting bulk equations of state when T, density and μ are chosen to represent matching equilibrium states. They still assign different exact fluctuations because their boundary constraints differ. “Equivalent” means agreement on suitable bulk observables in the limit, not identical probability distributions for every finite system.

This reasoning needs conditions. Small clusters can have noticeable relative fluctuations. Near a critical point, correlations extend over many particles, and response functions such as compressibility or heat capacity can become anomalously large; finite-size scaling becomes important. Systems dominated by long-range interactions can have more complicated or even inequivalent ensembles. First-order phase transitions may also make probability distributions bimodal in finite canonical descriptions, complicating a naïve single-peak argument.

The distinction matters in simulations. A molecular-dynamics run with fixed E,V,N represents a microcanonical setup in an idealisation; adding a thermostat approximates canonical sampling. A grand canonical Monte Carlo simulation proposes particle insertions and deletions at fixed μ,T,V. The desired measured quantity and physical contact conditions guide the choice, even if bulk averages agree in a large ordinary system.

Step-by-step reasoning

Identify the extensive variable and a characteristic mean scale proportional to N. Use a fluctuation relation, such as Var(E) = kT²C V, and assume ordinary extensive C V ∝ N. Take its square root to get σ E ∝ √N, divide by an order-N scale, and interpret the 1/√N decrease. State whether small size, criticality or long-range effects could invalidate the simple scaling.

Visual explanation

Draw two probability distributions of energy per particle for systems with N = 100 and N = 10⁶. The large-system curve is much narrower around the same mean energy density. Add a bracket showing absolute energy width grows as √N but width divided by total energy scale shrinks as 1/√N.

Real-world analogy

When independent random contributions are added, their total uncertainty often grows like the square root of their number while the total amount grows linearly. A large class average is therefore steadier relative to its size than a single student's score. Molecular interactions and critical correlations limit this analogy, but it captures ordinary fluctuation scaling.

Real-world example

An ordinary beaker of liquid at room temperature has a stable observed temperature despite continuous microscopic energy exchanges with its surroundings. A nanoscale cluster can show much larger relative variations. Simulation studies of tiny systems must specify ensemble and sample fluctuations carefully; experimental macroscopic thermodynamics usually sees only sharply concentrated averages.

Why?

Many weakly correlated contributions add to an extensive mean, but their random deviations partly cancel. Statistical weights overwhelmingly favour macrostates near the entropy maximum. The fraction of states with a substantially different bulk energy density becomes negligible as system size grows, allowing different reservoir descriptions to agree on the dominant macrostate.

Common misconception

The thermodynamic limit does not make all fluctuations zero in absolute units; σ E can grow as √N. It makes relative fluctuations small under ordinary scaling. Another mistake is to assume equivalence for any tiny system or at every critical point. The conditions on interactions and phase behaviour are part of the claim.

Worked example

Suppose a model has U = Nu and Var(E) = Nv with positive constants u and v at one temperature. Then σ E/U = √(Nv)/(Nu) = (√v/u)/√N. If N rises from 100 to 10,000 with u and v unchanged, this relative ratio falls by a factor √(10,000/100) = 10. The absolute standard deviation rises by a factor 10, but the mean energy rises by a factor 100.

Quick check

1. If heat capacity doubles when N doubles at fixed density and temperature, how does canonical energy standard deviation change? Answer: Var(E) = kT²C V doubles, so σ E = √Var(E) increases by √2. Relative to an extensive energy scale that doubles, its relative size falls by 1/√2.

Exam focus

Distinguish absolute from relative fluctuation and state what is held fixed in the thermodynamic limit. Show the square root between variance and standard deviation. Qualify ensemble equivalence for short-range, stable bulk systems away from critical anomalies. Do not infer that finite-size distributions are identical merely because their bulk means agree.

Advanced insight

Near a continuous critical point, a growing correlation length makes distant regions fluctuate together, weakening the assumption of many nearly independent contributions. Finite-size scaling studies how response peaks depend on system size and can reveal critical exponents. The thermodynamic limit then remains central, but its approach is not the simple 1/√N story.

Summary

For an ordinary large system, means and variances of extensive observables often scale as N, so standard deviations scale as √N and relative fluctuations shrink as N⁻¹ᐟ². This concentration lets different ensembles agree on bulk equilibrium properties when states are matched. Small systems, critical regions and long-range interactions require more careful treatment.

Practice questions

1. If N increases by a factor of 25, by what factor does a typical 1/√N relative fluctuation change? Answer: It becomes one fifth as large, because √25 = 5. 2. Do canonical and microcanonical ensembles have identical energy variance at finite N? Answer: No. Microcanonical energy is fixed by construction, while canonical energy fluctuates. Their suitable bulk averages can agree in the large-system limit despite this difference. 3. Why are critical points a warning for simple fluctuation scaling? Answer: Correlations can extend over large distances and response functions become enhanced. Variance may no longer follow ordinary independent-region N scaling near criticality.