3.6 Fluctuations in ensembles

Concept of fluctuations

Fluctuations are random deviations from average (mean) values in physical systems. They serve as the bridge between the microscopic world of individual particles and the macroscopic thermodynamic quantities you measure in the lab. By studying fluctuations, you gain direct insight into system stability, phase transitions, and how systems behave away from equilibrium.

Microscopic vs macroscopic fluctuations

At the atomic and molecular level, thermal motion causes constant microscopic fluctuations in quantities like particle velocity and local energy. These fluctuations propagate upward to affect observable macroscopic properties such as temperature and pressure, but they average out over large scales. That averaging is precisely why macroscopic properties appear stable: for a system of $N$ particles, relative fluctuations typically scale as $\frac{1}{\sqrt{N}}$, so for $N \sim 10^{23}$ they become negligibly small.

Importance in statistical mechanics

• Fluctuations reveal a system's internal structure and dynamics
• They connect microscopic interactions to macroscopic thermodynamic properties (specific heat, compressibility, susceptibility)
• Response functions and susceptibilities can be calculated directly from fluctuation magnitudes
• Near phase transitions, fluctuations grow dramatically and signal the breakdown of mean-field descriptions
• They also underpin the study of non-equilibrium processes and irreversibility

Time scales of fluctuations

Not all fluctuations happen at the same speed:

• Rapid (femtoseconds to picoseconds): individual molecular vibrations and collisions
• Intermediate (nanoseconds to microseconds): collective motions involving many particles
• Slow (seconds to hours): macroscopic relaxation processes, such as phase separation

Time correlation functions $\langle A(t) A(0) \rangle$ characterize how a fluctuating quantity $A$ at one time relates to itself at a later time. The relaxation time $\tau$ extracted from these functions tells you how quickly the system returns to equilibrium after a perturbation.

Fluctuations in the microcanonical ensemble

The microcanonical ensemble describes an isolated system with fixed total energy $E$, volume $V$, and particle number $N$. Because these global quantities are strictly constant, fluctuations here refer to how energy, particles, or volume distribute among subsystems within the whole.

Energy fluctuations

Total energy is exactly fixed, so there are zero fluctuations in the system's total energy. However, if you mentally divide the system into subsystems, energy fluctuates between them. For a subsystem containing $N_{\text{sub}}$ particles:

• The relative energy fluctuation scales as $\frac{\delta E}{E} \sim \frac{1}{\sqrt{N_{\text{sub}}}}$
• These internal fluctuations are connected to the subsystem's heat capacity: larger $C_V$ means the subsystem can absorb more energy without large temperature swings

Particle number fluctuations

The total particle number is fixed, but local regions of the system see particle number fluctuations as molecules move around.

• For an ideal gas subsystem, particle number fluctuations follow a Poisson distribution, with variance equal to the mean: $\langle (\Delta N)^2 \rangle = \langle N \rangle$
• The magnitude of fluctuations scales as $\sqrt{N}$, so relative fluctuations go as $\frac{1}{\sqrt{N}}$
• These fluctuations connect to the subsystem's compressibility and chemical potential

Volume fluctuations

Total volume is constant, but local volume fluctuations occur in subsystems or around individual particles.

• Related to the system's compressibility and pressure
• In solids, volume fluctuations connect to vibrational modes (phonons)
• In soft matter and biophysics, fluctuations in molecular volumes are relevant for understanding protein conformational dynamics

Fluctuations in the canonical ensemble

The canonical ensemble describes a system in thermal contact with a heat bath at temperature $T$. The system can exchange energy with the bath, so $T$, $V$, and $N$ are fixed but total energy $E$ fluctuates.

Energy fluctuations

This is one of the most important results in ensemble theory. Because the system exchanges energy with the reservoir, its energy samples the Boltzmann distribution $P(E) \propto \Omega(E) e^{-\beta E}$, where $\beta = 1/(k_B T)$.

The variance of energy fluctuations is:

$\langle (\Delta E)^2 \rangle = k_B T^2 C_V$

This result is powerful: it directly ties a measurable thermodynamic quantity (heat capacity) to the statistical spread in energy. Relative fluctuations scale as:

$\frac{\sqrt{\langle (\Delta E)^2 \rangle}}{\langle E \rangle} \sim \frac{1}{\sqrt{N}}$

For macroscopic systems ($N \sim 10^{23}$), this ratio is negligibly small, which is why the canonical and microcanonical ensembles give equivalent thermodynamic predictions in the thermodynamic limit.

Specific heat and fluctuations

The connection between specific heat and energy fluctuations is a direct application of the fluctuation-dissipation theorem:

$C_V = \frac{\langle (\Delta E)^2 \rangle}{k_B T^2}$

This tells you several things:

• A large $C_V$ means the system tolerates large energy fluctuations
• Near a phase transition, $C_V$ can diverge, reflecting the enormous (critical) fluctuations in energy
• The temperature dependence of $C_V$ encodes information about the spacing and structure of the system's energy levels

Particle number fluctuations

Total particle number is fixed in the canonical ensemble, so there are no fluctuations in $N$ for the system as a whole. However, for subsystems within the canonical ensemble, local particle number fluctuations do occur and are related to the isothermal compressibility. To study fluctuations in total particle number, you need the grand canonical ensemble.

Fluctuations in the grand canonical ensemble

The grand canonical ensemble describes an open system that exchanges both energy and particles with a reservoir. Temperature $T$ and chemical potential $\mu$ are fixed, while both $E$ and $N$ fluctuate.

Energy fluctuations

Energy fluctuations here have contributions from both thermal exchange and particle exchange. The full expression is:

$\langle (\Delta E)^2 \rangle = k_B T^2 C_V + \left(\frac{\partial \langle E \rangle}{\partial \langle N \rangle}\right)^2_{T,V} \langle (\Delta N)^2 \rangle$

Note that the second term couples energy fluctuations to particle number fluctuations. The relative energy fluctuations still decrease as $\frac{1}{\sqrt{N}}$ in the thermodynamic limit, so ensemble equivalence holds for large systems.

Particle number fluctuations

This is the defining feature of the grand canonical ensemble. The variance in particle number is:

$\langle (\Delta N)^2 \rangle = k_B T \left(\frac{\partial \langle N \rangle}{\partial \mu}\right)_{T,V}$

This connects directly to the isothermal compressibility $\kappa_T$:

$\langle (\Delta N)^2 \rangle = \frac{\langle N \rangle^2 k_B T \kappa_T}{V}$

• For large systems, particle number fluctuations follow a Gaussian distribution
• Near critical points, $\kappa_T$ diverges, causing particle number fluctuations to become very large
• For an ideal gas, $\langle (\Delta N)^2 \rangle = \langle N \rangle$, recovering the Poisson result

Chemical potential fluctuations

The chemical potential $\mu$ is fixed by the reservoir, so it does not fluctuate for the system as a whole. Locally within the system, however, effective chemical potential fluctuations can occur and are related to particle number fluctuations through thermodynamic identities. These local fluctuations drive diffusion and particle exchange processes.

Thermodynamic fluctuation theory

This framework provides a unified way to connect equilibrium fluctuations to measurable response functions. It also forms the foundation for understanding irreversible processes.

Einstein's fluctuation theory

Einstein's approach starts from entropy. The probability of observing a fluctuation away from equilibrium is:

$P \propto e^{\Delta S / k_B}$

where $\Delta S$ is the entropy change associated with the fluctuation (which is negative, since equilibrium maximizes entropy). This single formula:

• Provides the foundation for understanding Brownian motion and diffusion
• Explains the origin of thermal noise (Johnson-Nyquist noise) in electrical circuits
• Leads naturally to the fluctuation-dissipation theorem

Fluctuation-dissipation theorem

The fluctuation-dissipation theorem (FDT) is one of the deepest results in statistical mechanics. It states that the spontaneous fluctuations a system exhibits in equilibrium are quantitatively related to how the system responds (dissipates) when driven by an external perturbation.

Formally, it connects correlation functions (describing fluctuations) to response functions or susceptibilities (describing how the system reacts to a force). Examples:

• Johnson-Nyquist noise: voltage fluctuations across a resistor are proportional to its resistance and temperature
• Brownian motion: the diffusion coefficient of a particle is proportional to the mobility, linked by $D = k_B T \mu_{\text{mob}}$ (the Einstein relation)

The FDT is the backbone of linear response theory and has been generalized to non-equilibrium settings through fluctuation theorems.

Onsager reciprocal relations

When multiple irreversible processes occur simultaneously (e.g., heat flow and particle diffusion), the transport coefficients obey symmetry relations:

$L_{ij} = L_{ji}$

These Onsager relations follow from microscopic reversibility (time-reversal symmetry of the underlying dynamics). A classic example is thermoelectricity: the Seebeck coefficient (voltage from a temperature gradient) and the Peltier coefficient (heat flow from a current) are related by these symmetries. The Onsager relations constrain which couplings between transport processes are physically allowed.

Statistical properties of fluctuations

Gaussian distribution of fluctuations

For large systems, fluctuations in extensive variables (energy, particle number, magnetization) typically follow a Gaussian distribution:

$P(x) = \frac{1}{\sqrt{2\pi\sigma^2}} \exp\left(-\frac{(x - \mu)^2}{2\sigma^2}\right)$

This arises because these quantities are sums of many weakly correlated microscopic contributions. The distribution is fully characterized by its mean $\mu$ and variance $\sigma^2$. Deviations from Gaussian behavior are a signal of strong correlations or nonlinear effects, which become prominent near critical points.

Central limit theorem

The central limit theorem (CLT) explains why Gaussian distributions are so common. It states that the sum of many independent (or weakly correlated) random variables converges to a Gaussian distribution, regardless of the distribution of the individual variables.

• Applies to fluctuations in extensive thermodynamic variables
• Breaks down near critical points, where correlations extend over the entire system
• Also breaks down for heavy-tailed distributions where the variance is not finite

Correlation functions

Correlation functions quantify how fluctuations at different points in space or time are statistically related.

• Time correlation functions $C(t) = \langle A(t) A(0) \rangle - \langle A \rangle^2$ characterize the dynamics and relaxation of fluctuations
• Spatial correlation functions $G(r) = \langle A(\mathbf{r}) A(\mathbf{0}) \rangle - \langle A \rangle^2$ reveal structural order (e.g., the pair correlation function $g(r)$ in liquids)
• The FDT connects these correlation functions to experimentally measurable response functions
• Near critical points, spatial correlations become long-ranged, with a diverging correlation length $\xi$

Measurement of fluctuations

Experimental techniques

• Light scattering (e.g., dynamic light scattering): measures density and concentration fluctuations in fluids and polymers
• Neutron scattering: probes atomic-scale structural and magnetic fluctuations in solids
• Electrical noise measurements: reveal charge and current fluctuations in conductors
• Single-molecule techniques (optical tweezers, fluorescence): observe fluctuations in individual biomolecules
• Atomic force microscopy: detects surface fluctuations and measures forces at the nanoscale

Noise in measurements

Every measurement encounters noise, and understanding its origin matters:

• Thermal noise (Johnson-Nyquist): voltage fluctuations in resistors, proportional to $k_B T R$
• Shot noise: arises from the discrete nature of charge carriers or photons, proportional to $\sqrt{N}$
• 1/f noise (flicker noise): power spectrum scales as $1/f$; origin is often complex and system-dependent
• Environmental noise: mechanical vibrations, electromagnetic interference
• Quantum noise: sets fundamental limits in high-precision measurements (e.g., gravitational wave detectors)

Signal-to-noise ratio

The signal-to-noise ratio (SNR) quantifies how well you can distinguish a real signal from background noise:

$\text{SNR} = \frac{P_{\text{signal}}}{P_{\text{noise}}}$

Averaging over $N$ independent measurements improves the SNR by a factor of $\sqrt{N}$, since random noise partially cancels while the signal adds coherently. Techniques like lock-in amplification exploit this principle by correlating the measured signal with a known reference frequency.

Applications of fluctuation theory

Critical phenomena

Near a critical point (e.g., the liquid-gas critical point), fluctuations become large and long-ranged. The correlation length $\xi$ diverges, and physical quantities exhibit power-law behavior characterized by critical exponents. These exponents are universal: they depend only on the dimensionality and symmetry of the system, not on microscopic details.

• Critical opalescence: density fluctuations near the critical point scatter light strongly, making the fluid appear cloudy
• Critical slowing down: relaxation times diverge as the system approaches the critical point
• Renormalization group methods provide the theoretical framework for analyzing critical fluctuations
• Mean-field theories fail near critical points precisely because they neglect these large fluctuations

Phase transitions

Fluctuations play distinct roles in different types of phase transitions:

• First-order transitions: proceed through nucleation and growth; fluctuations must overcome a free energy barrier to form a nucleus of the new phase
• Second-order (continuous) transitions: the order parameter fluctuations diverge, and there is no latent heat
• Quantum phase transitions: occur at zero temperature, driven by quantum (rather than thermal) fluctuations as a control parameter is tuned

Fluctuations are also central to understanding metastable states, hysteresis, and spinodal decomposition.

Brownian motion

Brownian motion is the random motion of a mesoscopic particle (e.g., a pollen grain) suspended in a fluid, caused by collisions with surrounding molecules. Einstein showed that the mean-square displacement grows linearly in time:

$\langle x^2 \rangle = 2Dt$

where $D$ is the diffusion coefficient (in one dimension). This result directly connects microscopic fluctuations to a macroscopic transport property. Brownian motion is foundational for understanding colloidal suspensions, polymer dynamics, and transport in biological cells.

Fluctuations in non-equilibrium systems

Equilibrium fluctuation theory assumes detailed balance and time-reversal symmetry. When systems are driven away from equilibrium by external forces or gradients, new theoretical tools are needed.

Fluctuation theorems

Fluctuation theorems generalize the fluctuation-dissipation relation to systems arbitrarily far from equilibrium. They describe symmetries in the probability distributions of quantities like entropy production or work:

• The transient fluctuation theorem applies to systems evolving from an initial equilibrium state
• The steady-state fluctuation theorem applies to systems maintained in a non-equilibrium steady state
• These theorems quantify how likely it is to observe "second-law-violating" trajectories (entropy decreasing over short times), showing that such events become exponentially rare for large systems or long times

Jarzynski equality

The Jarzynski equality connects non-equilibrium work measurements to equilibrium free energy differences:

$\langle e^{-\beta W} \rangle = e^{-\beta \Delta F}$

Here $W$ is the work done on the system during a non-equilibrium process, and $\Delta F$ is the free energy difference between the initial and final equilibrium states. This holds regardless of how far from equilibrium the process drives the system. It has practical applications in single-molecule pulling experiments, where you can extract equilibrium free energies from repeated non-equilibrium measurements.

Crooks fluctuation theorem

The Crooks theorem relates the probability of observing work $W$ in a forward process to the probability of observing work $-W$ in the time-reversed process:

$\frac{P_F(W)}{P_R(-W)} = e^{\beta(W - \Delta F)}$

This is more detailed than the Jarzynski equality (which can be derived from it). The Crooks theorem provides a practical method for extracting $\Delta F$ from the crossing point of the forward and reverse work distributions, and it has been verified in single-molecule RNA unfolding experiments and other nanoscale systems.