Fluctuations

Links energy fluctuations with derivatives of the partition function. Theory, formulas, worked examples and interactive graphs.

Conceptual context

In a thermal microscopic system, energy is not nailed to a single value even when the macroscopic state looks steady. The system explores microstates with different probabilities, so $E_i$ values form a distribution around a $U$. The $Z$ stores more than the average; it also stores the width of that distribution. That is why fluctuations link microscopic statistical motion with macroscopic signals such as $C_V$.

🟢 Essential level

Energy fluctuations describe how far the instantaneous system energy typically wanders from its $U$. In a canonical equilibrium picture, the system can exchange energy with a thermal reservoir, so different microstates do not appear with equal importance. $T$ controls that selection: at low $T$, low $E_i$ values dominate; at higher $T$, more energetic microstates can contribute. The $Z$ acts as the statistical ledger of these contributions.

If the occupied energies are tightly clustered, $\Delta E$ is small. If many energy states compete with noticeable probability, $\sigma_{E2}$ grows. Physically, a fluctuation is not a failure of equilibrium; it is the allowed spread of energy inside that equilibrium description.

This link between microscopic fluctuations and macroscopic thermal response is a special case of the fluctuation-dissipation theorem, one of the deepest ideas in statistical physics, with applications ranging from electronic noise to polymer physics. The same logic connects a system's willingness to fluctuate with its willingness to respond to an external push.

🔵 Formal level

Basic formulas

The formal description begins with $\beta$, which rewrites the thermal scale using $T$ and $k_B$.

Beta as inverse thermal scale

With that variable, the $Z$ collects canonical weights for each $E_i$. The first derivative of its logarithm with respect to $\beta$, the $dlnZ_{d\beta}$, gives $U$, with the required sign convention.

Mean energy from the partition function

Real student questions

Does a fluctuation mean energy is not conserved? No. In the canonical ensemble, the system exchanges energy with the reservoir; conservation applies to system plus environment.

Why does a second derivative appear? The first derivative gives the average, while the second describes how that average changes when thermal selection changes.

Does $\Delta E$ always increase with $T$? Not necessarily. It often grows when more states become accessible, but it depends on the $E_i$ spectrum and on model constraints.

Is $C_V$ itself a fluctuation? No. It is a response coefficient, but in the canonical model it is tied to $\sigma_{E2}$.

Can $\Delta E$ exceed $U$? In small systems or near excited levels, yes: the relative spread is a physical signature that the average is not a reliable single-value description. In that regime the full distribution must be reported, not only the mean.

Final synthesis

The $Z$ does not only compute averages. Its curvature with respect to $\beta$ reveals $\sigma_{E2}$, and that microscopic spread explains part of the system thermal response.