Abstract
In this chapter we motivate the study of disordered media through the example of a porous system. The basic terms in percolation theory are introduced, and you learn how to generate, visualize and measure on percolation systems in Python. We demonstrate how to find exact results for small systems in two dimensions by addressing all configurations of the system, and show that this approach becomes unfeasible for large system sizes.
Introduction
Percolation theory provides a mathematical framework for understanding connectivity in disordered media. Whether considering the flow of water through porous rock, the spread of a forest fire, or the robustness of a social network, the fundamental question remains: at what point does a system transition from a collection of isolated components to a globally connected structure? This transition occurs at a critical threshold, $p_c$, which is the central focus of this study.
Modeling Disordered Media
To study these complex physical systems, we simplify them into discrete "binary" models. In a site percolation model, we utilize square (2D) or cubic (3D) lattices where each site is independently occupied with a probability $p$ or left empty with probability $1-p$.
As shown above, increasing the occupation probability $p$ leads to a higher density of occupied sites and the eventual formation of clusters. Computationally, this is implemented using Python's NumPy for matrix generation and scipy.ndimage for cluster labeling. A "spanning cluster" is defined as a connected path of occupied sites that extends from one side of the lattice to the opposite side.
Dimensionality and Connectivity
The geometry of these clusters changes significantly with the system's dimensionality. In 2D, a spanning cluster of occupied sites effectively blocks connectivity for the empty sites (pores) in the transverse direction. However, in 3D systems, it is possible to have "dual connectivity," where both the solid phase and the pore phase form spanning clusters simultaneously.
The visualization above illustrates the intricate nature of these clusters in a 3D volume. Near the percolation threshold $p_c$, these structures exhibit fractal properties, appearing self-affine across different scales.
Results and Analytical Solutions
For an infinite 2D square lattice, the percolation threshold is approximately $p_c \simeq 0.5927$. As the system size $L$ increases toward the thermodynamic limit, the probability of finding a spanning cluster, $\Pi(p, L)$, transitions from a smooth S-curve to a sharp step function at $p_c$.
For very small systems, we can derive exact analytical solutions by summing the probabilities of all possible configurations. For a $2 \times 2$ system ($L=2$), the percolation probability is derived as: $$\Pi(p, L=2) = 4p^2(1-p)^2 + 4p^3(1-p)^1 + p^4 = 4p^2 - 4p^3 + p^4$$ This yields an estimated $p_c \simeq 0.4588$ for the $2 \times 2$ case.
Conclusion and Limitations
While exact solutions are possible for small grids, they quickly become computationally unfeasible due to the "curse of dimensionality." An $L \times L$ system has $2^{L^2}$ configurations; a $10 \times 10$ grid already exceeds $10^{30}$ possibilities. Furthermore, real-world materials often exhibit correlations between sites, whereas this introductory model assumes statistical independence. Future work will address these complexities through finite-size scaling theories and more advanced Monte Carlo simulations.
Summary updated