The Canonical Ensemble and Energy Fluctuations

Introduction

A system in contact with a heat bath has a controlled temperature, but its energy is not fixed. It occasionally gains and loses energy as it exchanges heat with the reservoir. The canonical partition function predicts both the average energy and the size of these fluctuations. Their relationship to heat capacity turns thermal noise into a measurable response property.

Core explanation

Let β = 1/(kT) and Q = Σ j exp(−βE j), summing over all microstates with fixed V and N. The probability of microstate j is p j = exp(−βE j)/Q. Differentiate ln Q with respect to β: ∂ln Q/∂β = (1/Q)Σ j(−E j)exp(−βE j) = −⟨E⟩. Therefore U = ⟨E⟩ = −∂ln Q/∂β. If energy levels have degeneracy g i, one may sum g i exp(−βε i) over levels instead of listing every microstate separately.

A second derivative gives ∂²ln Q/∂β² = ⟨E²⟩ − ⟨E⟩² = Var(E) ≥ 0. Equivalently, ∂⟨E⟩/∂β = −Var(E). Since dβ/dT = −1/(kT²), the constant-volume heat capacity is C V = (∂U/∂T) V,N = Var(E)/(kT²), assuming the Hamiltonian's energy levels have no explicit temperature dependence. Thus a larger energy variance corresponds to a larger heat capacity at the same temperature.

For many ordinary short-range interacting particles away from criticality, U grows proportional to N and Var(E) also grows roughly proportional to N. Standard deviation then grows like √N, while the ratio of standard deviation to mean energy typically falls like 1/√N when the mean has a nonzero extensive scale. This is why a macroscopic thermometer reports a stable value even though energy exchange never stops. Near a critical point or in unusual systems, fluctuations can be enhanced and simple scaling needs care.

The canonical ensemble's Helmholtz energy A = −kT ln Q generates the same mean-state thermodynamics as the microcanonical description for a large ordinary system, while making fluctuations explicit. Note that Q is dimensionless and the energy zero can shift: adding a constant E ref to every state multiplies Q by exp(−βE ref) and shifts U by E ref, but leaves variance and heat capacity unchanged.

Step-by-step reasoning

Write β and a properly degeneracy-weighted Q. Differentiate ln Q once for U and twice for variance, treating the fixed V,N energy levels as constants with respect to β. Use dβ/dT to convert the variance relation to C V. Check that variance is nonnegative and that C V has energy-per-temperature units.

Visual explanation

Draw a probability histogram over energy. At low temperature it is concentrated near the ground energy; at a higher temperature it spreads among more levels. Mark its mean and width. Next to it draw the connection Var(E) = kT²C V, with arrows between histogram width and the heat-capacity response.

Real-world analogy

A household with flexible spending can have an average weekly expense and week-to-week variation. A larger response to changing conditions often accompanies more variation. The analogy helps separate mean from fluctuation, but canonical energy variations are equilibrium heat exchanges governed by precise Boltzmann weights, not discretionary decisions.

Real-world example

Small molecules or nanoscale particles in a thermal environment can show noticeable energy fluctuations relative to their size. A macroscopic sample has far more particles, so its relative fluctuation is normally tiny. Heat-capacity measurements reveal how many additional states become populated as temperature changes, while microscopic simulations can directly calculate an energy variance to estimate C V.

Why?

Differentiating Q pulls down an energy factor from each Boltzmann exponential. The first derivative gives a weighted average; the second measures how much individual energies differ from that average. The same spread controls sensitivity of mean energy to temperature, making heat capacity a fluctuation-response relation.

Common misconception

Fixed T does not mean fixed E in a canonical system. Another mistake is to read Q as a simple unweighted count of states: its states carry Boltzmann factors. Finally, energy variance is not the square of the mean; it is ⟨E²⟩ − ⟨E⟩² and cannot be negative.

Worked example

Consider a two-level system with energies 0 and ε, each nondegenerate. Its partition function is Q = 1 + e^(−βε). The excited-state probability is p = e^(−βε)/Q, so U = pε. Because E is either 0 or ε, ⟨E²⟩ = pε² and Var(E) = ε²p(1 − p). Therefore C V = ε²p(1 − p)/(kT²). The heat capacity tends to zero at very low T because excitation is rare and at very high T because populations approach saturation while ε is fixed.

Quick check

1. What is Var(E) if every accessible canonical microstate has exactly the same energy? Answer: It is zero because ⟨E²⟩ = ⟨E⟩². With no energy difference to populate, the corresponding canonical C V contribution is zero under the stated model.

Exam focus

Distinguish derivatives with respect to β from derivatives with respect to T and retain the minus signs. Include degeneracy in Q. Quote C V = Var(E)/(kT²) only for fixed V,N and temperature-independent energy levels. Check limiting behaviour for two-level examples to catch algebra errors.

Advanced insight

Fluctuation-response relations extend beyond energy: particle-number variance in a grand canonical system is linked to response to chemical potential, and volume fluctuations in pressure-controlled ensembles relate to compressibility. Near continuous phase transitions, large correlated fluctuations can produce pronounced response-function anomalies.

Summary

In the canonical ensemble p j = e^(−βE j)/Q. The derivatives U = −∂ln Q/∂β and Var(E) = ∂²ln Q/∂β² yield C V = Var(E)/(kT²). Macroscopic relative energy fluctuations are usually small, while heat capacity records how strongly accessible-state populations respond to temperature.

Practice questions

1. A canonical system has Var(E) = 4.0 × 10⁻⁴² J² at T = 300 K. Express its C V using k = 1.381 × 10⁻²³ J K⁻¹. Answer: C V = Var/(kT²) = 4.0 × 10⁻⁴²/[(1.381 × 10⁻²³)(300²)] ≈ 3.22 × 10⁻²⁴ J K⁻¹ for this small system. 2. Why does adding the same constant to every energy level leave C V unchanged? Answer: It shifts both E and its mean by the same constant, so deviations E − ⟨E⟩ and their variance are unchanged. C V depends on that variance, not the arbitrary energy zero. 3. In a two-level system with excited probability p = 0.20 and gap ε, what is its mean energy and variance? Answer: U = 0.20ε and Var(E) = ε²p(1 − p) = 0.16ε². Both follow from probabilities of 0 and ε.