Abstract
In this chapter, we introduce the powerful theoretical methods of renormalization. The fundamental idea is that at the critical point $p = p_c$, a rescaling of the system does not change its most important features. By rescaling, we typically mean a coarse-graining of the system, such as merging $2 \times 2$ cells into a single cell. The rule used to determine the occupation probability of the new, coarse-grained cell, $p'$, is a function of the probability $p$ of the original lattice: $p' = R(p)$. Renormalization theory utilizes the properties of this mapping, $R(p)$, to deduce critical properties of the system, such as scaling exponents. This framework is explored in the context of percolation systems, where the geometric nature of the remapping provides intuitive insights into renormalization as a concept across various lattice structures and dimensions.
Introduction
In percolation theory, the "critical point" ($p_c$) represents a phase transition where a system shifts from a collection of isolated clusters to a state containing a "spanning cluster" that connects opposite boundaries. At this threshold, the system exhibits self-similar fractal properties, and the correlation length ($\xi$)—the characteristic size of finite clusters—diverges to infinity. Calculating $p_c$ and the associated universal exponents, such as $\nu$ (which describes the growth of the correlation length), is analytically challenging for most complex lattices. Renormalization Group (RG) theory addresses this by simplifying large-scale systems into manageable mathematical mappings.
The Renormalization Group Framework
The core of the renormalization method involves Coarse-Graining and Rescaling. The process begins by partitioning a lattice into cells of size $b \times b$, where $b$ is the rescaling factor. A specific "renormalization rule" is then applied to determine if the new, larger cell is considered "occupied" based on the configuration of the original sites.
This transformation defines a mapping function $p' = R(p)$. By iterating this mapping, we can observe how the effective occupation probability changes as we look at the system at increasingly larger scales.
Fixed Point Analysis and Scaling
The behavior of the system under renormalization is governed by its fixed points, where $p^* = R(p^*)$.
- Trivial Fixed Points: $p=0$ and $p=1$ are stable fixed points. A system that is mostly empty will, upon coarse-graining, appear even emptier ($p \to 0$). Conversely, a mostly full system will appear completely full ($p \to 1$).
- Non-trivial Fixed Point: The value $p^*$ where the mapping crosses the identity line ($R(p) = p$) between 0 and 1 represents the critical percolation threshold $p_c$. This is an unstable fixed point.
By linearizing the mapping near the unstable fixed point, we can derive the critical exponent $\nu$. Taking the derivative $\Lambda = R'(p^*)$, the exponent is found via the relation: $$\nu = \frac{\ln b}{\ln \Lambda}$$ This illustrates the principle of universality: the exponent $\nu$ depends on the properties of the fixed point itself rather than the microscopic details of the lattice.
Applications and Key Results
The effectiveness of renormalization is demonstrated across various geometries:
- 1D Percolation: Correctly identifies $\nu = 1$.
- 2D Triangular Lattice: Using a "majority rule" (where a cell is occupied if the majority of its sub-sites are occupied) yields the exact solution $p_c = 0.5$ and a highly accurate $\nu \approx 1.355$ (compared to the exact $4/3$).
- 2D Square Site Lattice: Provides a reasonable estimate of $p_c \approx 0.62$ (true value $\approx 0.592$).
Limitations
While powerful, the accuracy of renormalization depends on the choice of the rescaling factor $b$ and the specific "spanning rule" used to define occupation in the coarse-grained cell. Small cells (e.g., $b=2$) are computationally efficient but may fail to capture long-range correlations accurately. Furthermore, the choice of rule must attempt to preserve the topology of the original clusters to minimize errors in the estimated critical values.
This summary is based on Chapter 7 of "Percolation Theory Using Python" by A. Malthe-Sørenssen.
Summary updated