The Extinction of Family Names
In 1874, Francis Galton posed a question to readers of the Educational Times: what is the probability that a family surname goes extinct? Henry Watson provided a mathematical framework using probability generating functions, founding the theory of branching processes. Their model — where each individual independently produces a random number of offspring — revealed a fundamental threshold: if the average number of offspring is at most one, extinction is certain; above one, there is hope for survival.
Criticality and Phase Transitions
The branching process exhibits a sharp phase transition at μ = 1. In the subcritical regime (μ < 1), the population decays geometrically and extinction occurs quickly. At criticality (μ = 1), the population performs an unbiased random walk toward zero — extinction is still certain but takes much longer. In the supercritical regime (μ > 1), the population either goes extinct early or grows exponentially without bound, with nothing in between.
Stochastic Genealogies
Each realization of a branching process creates a random genealogical tree. This simulation visualizes these trees, showing how different parameter choices produce dramatically different structures — sparse, quickly-dying lineages when subcritical versus bushy, exponentially expanding trees when supercritical. The randomness means that even supercritical populations can go extinct if the first few generations are unlucky.
Modern Applications
Branching processes are ubiquitous in modern science. In cancer biology, they model tumor initiation — a single mutant cell must survive stochastic fluctuations to establish a tumor. In nuclear physics, they describe neutron chain reactions (the basis of criticality in reactors). In epidemiology, the early phase of an outbreak is a branching process where R₀ is the mean offspring number. In computer science, they analyze recursive algorithms and random tree data structures.