Opening the book…
Einstein was unaware of Gibbs's work when he published "On the General Molecular Theory of Heat"[2]. He followed a distinct but logically parallel path to arrive at the fluctuation-dissipation relation.
2.2.1 Step-by-Step Statistical Logic
Einstein utilized the canonical distribution \(d W=C e^{-\beta E} \omega(E) d E\). To find the variance, the following steps are employed:
- Identity: Express the average energy as \(\langle E\rangle=\dfrac{\int E e^{-\beta E} \omega(E) d E}{\int e^{-\beta E} \omega(E) d E}\).
- Differentiation: Calculate the derivative with respect to \(\beta\), yielding \(\dfrac{d\langle E\rangle}{d \beta}=-\left(\left\langle E^{2}\right\rangle-\right.\) \(\left.\langle E\rangle^{2}\right)\).
- Conversion: Use the chain rule where \(\dfrac{d}{d \beta}=\dfrac{d T}{d \beta} \dfrac{d}{d T}=-k_{B} T^{2} \dfrac{d}{d T}\).
Substituting the conversion into the derivative leads back to the fundamental fluctuation result:
\[ \left\langle\Delta E^{2}\right\rangle=k_{B} T^{2} \dfrac{d\langle E\rangle}{d T} \tag{2.5} \label{2.5} \]
This leads back to the same fundamental result as Gibbs's. Einstein applied this to blackbody radiation to suggest the dual nature of light. While the 1904 paper didn't explicitly claim that light is made of particles (photons), it proved that if you treat radiation using the laws of statistical mechanics, it behaves in a way that is strikingly inconsistent with pure wave theory and strikingly consistent with a "molecular" theory of heat. He noted that the fluctuations in radiation followed the same mathematical form as fluctuations in a gas composed of discrete particles. This suggested that radiation might not be a continuous wave but could have a "granular" or discrete structure.
Derivation of Particle Number Fluctuations
We will derive density fluctuations assuming discrete states. For continuous spectrum, the derivation will be similar as above with sums replaced by integrals over energy with proper density of states. To understand density fluctuations, we consider a subsystem of volume \(V\) in contact with a reservoir at temperature \(T\) and chemical potential \(\mu\). The number of particles \(N\) in this volume is not fixed. In the grand canonical ensemble, the probability of a system being in a state with \(N\) particles and energy \(E\) is given by:
\[ P(E, N)=\dfrac{1}{\mathcal{Z}} e^{-\beta(E-\mu N)} \tag{2.6} \label{2.6} \]
where \(\mathcal{Z}\) is the grand partition function:
\[ \mathcal{Z}=\sum_{N} \sum_{i} e^{-\beta\left(E_{i}-\mu N\right)}=\sum_{N} e^{\beta \mu N} Z_{N}(V, T) \tag{2.7} \label{2.7} \]
Here \(\beta=1 / k_{B} T\). The average number of particles is:
\[ \langle N\rangle=\dfrac{1}{\mathcal{Z}} \sum_{N} N e^{\beta \mu N} Z_{N}=\dfrac{1}{\beta} \dfrac{\partial \ln \mathcal{Z}}{\partial \mu} \tag{2.8} \label{2.8} \]
To find the fluctuation, we take the second derivative:
\[ \dfrac{\partial^{2} \ln \mathcal{Z}}{\partial \mu^{2}}=\beta \dfrac{\partial\langle N\rangle}{\partial \mu}=\beta^{2}\left(\left\langle N^{2}\right\rangle-\langle N\rangle^{2}\right) \tag{2.9} \label{2.9} \]
Thus, the variance in particle number is:
\[ \left\langle(\Delta N)^{2}\right\rangle=\dfrac{1}{\beta^{2}} \dfrac{\partial^{2} \ln \mathcal{Z}}{\partial \mu^{2}}=\dfrac{1}{\beta}\left(\dfrac{\partial\langle N\rangle}{\partial \mu}\right)_{V, T} \tag{2.10} \label{2.10} \]
Using the Gibbs-Duhem relation at constant \(T(d P=\rho d \mu\), where \(\rho=\langle N\rangle / V)\), we can transform the derivative:
\[ \left(\dfrac{\partial\langle N\rangle}{\partial \mu}\right)_{V, T}=\left(\dfrac{\partial\langle N\rangle}{\partial P}\right)_{V, T}\left(\dfrac{\partial P}{\partial \mu}\right)_{V, T}=\left(\dfrac{\partial\langle N\rangle}{\partial P}\right)_{V, T} \rho \tag{2.11} \label{2.11} \]
Since \(\langle N\rangle=V \rho\), and the isothermal compressibility is \(\kappa_{T}=\dfrac{1}{\rho}\left(\dfrac{\partial \rho}{\partial P}\right)_{T}\), we find:
\[ \left\langle(\Delta N)^{2}\right\rangle=\dfrac{\rho V}{\beta}\left(\dfrac{\partial \rho}{\partial P}\right)_{T}=\langle N\rangle \rho k_{B} T \kappa_{T} \tag{2.12} \label{2.12} \]
Dividing by \(\langle N\rangle^{2}\) and noting that relative density fluctuations \(\dfrac{\Delta \rho}{\rho}\) are equivalent to \(\dfrac{\Delta N}{\langle N\rangle}\) for a fixed volume, we arrive at the Einstein fluctuation formula [6]:
\[ \dfrac{\left\langle(\Delta \rho)^{2}\right\rangle}{\rho^{2}}=\dfrac{k_{B} T \kappa_{T}}{V} \tag{2.13} \label{2.13} \]
Thus, the relative root-mean-square fluctuation in the particle density of the given system is ordinarily \(1 \sqrt{N}\), and, therefore, negligible. However, there are exceptions, like the in phase transitions where at a critical point the compressibility diverges, so it is no longer intensive and a finite size scaling is required[3]. Thus, at the critical points, we encounter unusually large fluctuations in the particle density of the system. Such fluctuations indeed exist and account for phenomena like critical opalescence. We will describe this later in the chapter on Einstein's theory of critical opalescence.
This page titled 2.2: Einstein's Independent Derivation (1904) was last modified on Sat, 07 Nov 2026 23:18:19 GMT and is shared under a CC BY-NC-SA 4.0 license and was authored, remixed, and/or curated by Pabitra N. Sen.
The book could not be loaded just now.