⚛️ biophysics

A Geometrical Class of Population Growth Models, across Scales and Population Structures

This paper proposes a unified geometrical framework that distinguishes between replicative and direct growth laws by modeling how the fraction of active components and the size of the accessible component space interact, thereby successfully describing diverse phenomena ranging from bacterial and human population dynamics to innovation metrics like patents and arXiv submissions.

Original authors: Tunstall, T., Zaccone, A.

Published 2026-10-10

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Tunstall, T., Zaccone, A.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Growth is a universal story, told in the language of numbers whether we are watching a colony of bacteria multiply in a petri dish, a human civilization expand across continents, or the number of new scientific ideas appearing each year. For centuries, scientists have tried to write the rules for this story. Some have treated growth as a simple, steady climb, like a line drawn straight up a graph. Others have seen it as a curve that slows down as resources run out, like a car running out of gas. But a new perspective suggests that the shape of this growth curve depends on something deeper: the size of the playground available to the growing things. In this view, growth is not just about how many individuals are present, but about how the very space of possibilities for those individuals changes as they multiply.

Two researchers, Thomas Tunstall and Alessio Zaccone, have developed a new way to look at these patterns. They propose that the speed at which a population grows depends on two things: how many "helpers" are active at any given moment, and how large the total pool of potential helpers is. Imagine a construction crew building a house. The speed of the work depends on how many workers are actually picking up tools right now. But it also depends on the size of the warehouse where the tools are kept. If the warehouse is fixed in size, the crew can only work so fast, no matter how many workers show up. But if the warehouse itself expands every time a new worker arrives, creating more tools and more space for new workers, the speed of construction can accelerate in a way that feels almost explosive. This distinction between a fixed pool of resources and an expanding one is the key to understanding why different systems grow in such different ways.

The researchers tested this idea across three very different worlds: single-celled organisms, human societies, and the generation of new ideas. They found that while the underlying math is the same, the behavior changes dramatically depending on whether the "pool" of growth-supporting components stays the same size or gets bigger.

First, they looked at bacteria. In a closed environment, a bacterial cell has a set number of internal processes—like nutrient intake or energy production—that help it grow. These processes can be turned on or off depending on conditions, but the total number of processes available does not change just because the bacteria multiply. The researchers found that bacterial growth fits a model where the pool of helpers is fixed. The speed of growth is determined by how many of these helpers are active at any moment. When food is plentiful, most helpers are active, and the colony grows fast. When food runs out, helpers turn off, and growth slows. This is a system where the rules of the game do not change, only the players' energy levels do.

Next, they turned to human populations. Here, the story is different. Humans do not just reproduce; they build networks. Every new person adds to the web of trade, knowledge, infrastructure, and cooperation that supports everyone else. In this case, the pool of "helpers" is not fixed; it expands as the population grows. The researchers found that human history follows a pattern where the growth rate accelerates because the network of support grows faster than the population itself. They analyzed data from three distinct eras. Between 1790 and 1948, the data suggested that the support network was expanding very rapidly, leading to a super-fast growth rate. However, between 1970 and 2023, the pattern shifted. The support network was still expanding, but the effect of adding more people began to slow down the growth rate per person, suggesting that the system was hitting new limits or that the nature of the network had changed. The model showed that human history cannot be described by a single, unchanging rule; the way people connect and support each other has changed over time, altering the growth curve.

The researchers also looked at the birth of new species, a process called speciation. Just like humans, species create opportunities for other species to exist. A new species might open up a new niche in an ecosystem, allowing others to evolve. The data on the history of eukaryotic life (which includes plants, animals, and fungi) fit the same expanding-network model. The rate at which new species appear depends on the number of existing species and the complex web of relationships they create. The model suggested that the diversity of life has been driven by this expanding web of interactions, where the presence of more species makes it easier for even more species to arise.

Finally, the team examined innovation, looking at the number of new scientific papers and patent applications. This is a case where the "helpers" do not reproduce. A patent does not give birth to another patent in the same way a bacterium divides. Instead, new ideas are built on the existing stock of knowledge. The researchers found that the rate of innovation follows a different rule: it depends on the size of the existing pool of ideas, but it does not multiply itself. The growth of new ideas is driven directly by the expanding space of possibilities created by past ideas, without the extra factor of self-replication. This distinction allowed them to separate the growth of living populations, which reproduce, from the growth of ideas, which accumulate.

The most important finding of this work is that a single, unified framework can explain all these different types of growth. The difference between a bacterial colony, a human society, and a library of patents is not in the basic math of how they grow, but in the nature of the space they grow in. If the space is fixed, growth is limited by how many parts are active. If the space expands as the population grows, growth can accelerate. If the space expands but the items do not reproduce, growth follows a different, direct path.

This approach changes how we think about limits. In the past, scientists often assumed that all growth must eventually slow down because of a lack of resources. This new model suggests that the slowdown or acceleration depends on how the system itself evolves. For human populations, the model indicates that the rules of growth have changed over history, moving from periods of rapid acceleration to periods where the feedback from the network becomes less positive. It does not predict a single, inevitable end to growth, but rather suggests that the future depends on how the network of human interaction continues to evolve.

The researchers did not claim to have solved the mystery of growth forever. Instead, they offered a new language to describe it. They showed that by looking at the size and activity of the underlying support system, we can understand why some things grow slowly and steadily, while others explode in complexity. The model suggests that the future of any growing system—whether it is a city, a species, or a scientific field—depends on whether that system is merely using a fixed set of tools, or whether it is constantly building new tools as it grows. This distinction helps us see that the story of growth is not just about numbers, but about the changing landscape of possibilities that those numbers create.

Technical Summary: A Geometrical Class of Population Growth Models, across Scales and Population Structures

Problem Statement
Population growth models traditionally range from phenomenological fitting laws (e.g., logistic functions) to mechanistic descriptions linking macroscopic growth to microscopic processes (e.g., bacterial nutrient allocation). A gap exists in unifying these descriptions across vastly different scales and population structures. Specifically, there is a need to distinguish between systems where the "growth-supporting component space" is fixed (controlled by activation/deactivation dynamics) and systems where the component space itself expands as the population grows (driven by interactions, innovation, or speciation). While recent work introduced a geometrical model for bacterial growth based on the fraction of active components, it remains to be seen if this construction generalizes to expanding component spaces and non-replicative systems like technological innovation.

Methodology
The authors develop a general geometrical framework where the per-capita growth rate, $r(t) = \frac{1}{X}\frac{dX}{dt}$ , depends exponentially on the number of active growth-supporting components, $M(t)A(t)$ . Here, $M(t)$ represents the size of the accessible component space, and $A(t)$ is the fraction of those components that are active.

The framework distinguishes three regimes based on the dynamics of $M(t)$ and $A(t)$ :

Fixed-Size Component Space: $M(t) \approx M_0$ is constant. Growth dynamics arise solely from changes in the active fraction $A(t)$ . This recovers the previously proposed bacterial growth model.
Replicative Geometric Growth: The component space expands with population size, scaling as $M_{eff}(X) \propto X^\gamma$ (where $\gamma > 0$ ). In this regime, existing entities reproduce, and the geometrical factor modulates the per-capita rate. This yields the law:
$\frac{dX}{dt} = \alpha X \exp(\beta X^\gamma)$
This is termed the "Replicative Geometric" model.
Direct Geometric Growth: Applicable to innovation indicators (e.g., patents, manuscripts) where existing items do not directly reproduce. Here, the growth rate lacks the linear prefactor $X$ , resulting in:
$\frac{dX}{dt} = \alpha \exp(\beta X^\gamma)$
This is termed the "Direct Geometric" model.

The authors test these models against historical data using empirical fitting and model selection criteria (AIC and BIC). The datasets include:

Bacterial populations (closed systems).
Historical human population data (three distinct time windows: 1790–1948, 1930–1978, 1970–2023).
Eukaryotic lineage-through-time data (speciation).
Innovation indicators: arXiv manuscript submissions and U.S. patent applications.

Key Results

Bacterial Growth: The model successfully recovers bacterial growth dynamics as a fixed-component-space realization where growth is governed by the temporal evolution of the active fraction $A(t)$ , consistent with physiological constraints like nutrient availability and (p)ppGpp regulation.
Human Population: The Replicative Geometric model fits historical human population data better than simple exponential or power-law models, but the parameters vary significantly by regime:
- 1790–1948: Best fit with a superlinear exponent $\gamma \approx 2.2$ , indicating a rapidly expanding effective component space.
- 1970–2023: The model independently converges to $\gamma \approx 1$ (recovering the Trachenko-Zaccone limit) but with a negative feedback coefficient ( $\beta < 0$ ), suggesting the expanding component space has become suppressive to per-capita growth.
- 1930–1978: The data is better described by a power law, which the authors identify as the asymptotic boundary of the Replicative Geometric family when $\gamma \to 0$ .
Speciation: Applied to eukaryotic speciation data, the Replicative Geometric model outperforms both exponential and power-law models (lower AIC/BIC), supporting the hypothesis that biotic niches (interactions between species) expand the configuration space for diversification.
Innovation: For arXiv submissions and patent applications, the Direct Geometric model (lacking the $X$ prefactor) provides the superior fit compared to Replicative Geometric and power-law models. Notably, the Replicative fit for patents converges to $\gamma \to 0$ and $\beta < 0$ , effectively collapsing to a power law, whereas the Direct Geometric model maintains a distinct mechanistic interpretation appropriate for non-replicative production.

Significance and Claims
The paper claims to provide a "common description" for growth processes across biological, demographic, and technological scales by separating the fraction of active components from the size of the accessible component space. The primary contribution is the demonstration that qualitatively different growth laws (logistic-like, exponential, power-law, and super-exponential) can arise from the same underlying geometrical construction depending on how $M(t)$ and $A(t)$ evolve.

The framework distinguishes between Replicative growth (where entities reproduce, modulating a per-capita rate) and Direct growth (where production is not per-capita). The authors argue that the "Trachenko-Zaccone" nonlinear rate-feedback equation is a specific case ( $\gamma=1$ ) of a broader family of Replicative Geometric laws.

Crucially, the paper posits that the geometrical framework serves as a "common language" to ask whether a system grows within a fixed set of possibilities or within a configuration space generated by the growth process itself. The authors emphasize that the fitted parameters (specifically $\gamma$ and $\beta$ ) should be interpreted as effective parameters for specific historical regimes rather than universal constants, noting that physical systems may transition between regimes due to resource limitations or changes in network structure. The work does not prescribe a universal saturation mechanism but rather focuses on identifying the geometrical dependence within specific growth regimes.

More like this