A multiscale theory for network advection-reaction-diffusion
Hadrien Oliveri, Emilia Cozzolino, Alain Goriely · 2025
Read A multiscale theory for network advection-reaction-diffusion by Hadrien Oliveri, Emilia Cozzolino, Alain Goriely (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-06546.txt.
Introduction
Movement between regions in a network is rebuilt from basic physics by Hadrien Oliveri and two co-authors in a 2025 applied mathematics paper. Their cases range from infections between countries to toxic proteins between brain areas. Standard network models move quantities between nodes using a graph Laplacian, a rule based on connections alone. That shortcut ignores what happens inside each link.
Taking flow, reaction, and spreading along each link as the true mechanism, the authors derive the network-level transport rule from first principles. The resulting effective graph Laplacian is fully fixed by the small-scale physics inside the edges. They show the operator captures transport correctly and study how it scales when edges grow longer. The work gives network modelers a transport rule with a clear physical basis instead of a fitted guess.
Catalog record
| CREATOR | Hadrien Oliveri, Emilia Cozzolino, Alain Goriely |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-06546
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A multiscale theory for network advection-reaction-diffusion Hadrien Oliveri, Emilia Cozzolino, Alain Goriely Abstract Mathematical network models are extremely useful to capture complex propagation processes between different regions (nodes), e.g. the spread of an infectious agent between different countries, or the transport and replication of toxic proteins across different brain regions in neurodegenerative diseases. In these models, transport is modelled at the macroscale through an operator, the so-called graph Laplacian based on the edge properties and topology, capturing the fluxes between different nodes of the network. However, this phenomenological approach fails to take into account the physical processes taking place at the microscale within the edge. A fundamental problem is then to obtain a transport operator from mechanistic principles based on the underlying transport process. Using advection-reaction-diffusion as a generic mechanism for inter-nodal exchanges, we derive a multiscale network transport model and derive the corresponding linear transport operator at the macroscale from first principles. This effective graph Laplacian is fully determined by the transport mechanisms along the edges at the microscale. We show that this operator correctly captures the transport, and we study its scaling properties with respect to edge length. 1 Introduction Systems where mass or information move between well-defined discrete regions connected by physical pathways can often be modelled as dynamical systems on a network. In these systems, a quantity is defined at a node , where it follows some local dynamics, and is carried through other nodes by edges through some form of physical transport. Applications range, among others, from mobility-driven epidemiology ( Kuhl 2021 ) to protein trafficking across the brain connectome ( Fornari et al. 2019 ; Fornari et al. 2020 ) . While modelling the dynamics at the level of nodes is relatively straightforward, a central mathematical question is to specify the transport operator governing the nodal exchanges. In the simplest case, by analogy with the continuum case of diffusion, one can postulate that transport is modelled by a graph Laplacian built from the network topology and weighted by its physical properties such as edges length and diameter ( Gautreau et al. 2007 ; Linka et al. 2020 ; Raj et al. 2012 ) . However, there is no first principle dictating the particular form of the transport operator, or how it is weighted apart from satisfying global physical constraints such as mass conservation, or the Fickian property that mass is transported from high to low concentrations. Here, we look at this issue as a multiscale problem. Positing an actual physical mechanism of transport along the network edges at the microscale, we determine an effective graph Laplacian at the macroscale. At the macroscale, we wish to obtain a generic network advection-reaction-diffusion system by combining i) transport between node through the edges; and ii) local reaction in the nodes (e.g. autocatalysis). The generic form of such network model is a system of ordinary differential equations (w.r.t. time t ) for the densities at the nodes of the form ( Kuhl 2021 ) : \frac{\mathrm{d}{}}{\mathrm{d}{t}}\left({\bm{\mathcal{V}}\bm{\rho}}\right)=\bm{\mathcal{V}}\mathbf{G}(\bm{\rho})+\mathbf{L}\bm{\rho}. (1) Here \bm{\rho} is the vector of densities (with unit mass per volume), expressing the density of the considered species across all nodes; and \bm{\mathcal{V}} the diagonal matrix of nodal volumes so that \bm{\mathcal{V}}\bm{\rho} has the dimension of a mass. The function \mathbf{G}(\bm{\rho}) captures the local nodal reaction which may encompass various chemical reactions, as well as natural decay or clearance ( Brennan et al. 2024 ; Ahern et al. 2025 ) . Transport is assumed to be linear and encompasses various contributions, such as passive Fickian diffusion and active transport. Typically, this contribution is modelled through the so-called graph Laplacian \mathbf{L} built from the weighted adjacency matrix of the network ( Putra et al. 2021 ) . This type of system is straightforward to analyse and solve numerically, and has proven highly effective in capturing key propagation mechanisms, as it can be readily validated and parameterized against data ( Chaggar et al. 2025 ) . However, from a physical perspective, a key issue is to obtain the appropriate form of both the reaction term \mathbf{G} and the transport operator \mathbf{L} . Indeed, various authors have used different ad hoc conventions ( Abdelnour et al. 2014 ; Raj et al. 2015 ; Pandya et al. 2017 ; Pandya et al. 2019 ; Thompson et al. 2020 ) — some using Laplacian matrices that violate mass balance, as pointed out by Putra et al. 2021 . This inconsistency is rather unsettling, considering each model aspires to describe the same underlying physical processes of diffusion and transport. The fundamental question being addressed here is then: given well-defined physical processes specifying reaction between species and transport along the edges (the microscale), can we define a model — defined by both a reaction term and a transport operator in ( Eq. 1 ) — at the network scale (the macroscale), from first principles? To obtain the transport operator \mathbf{L} and the correct form of the reaction term, we must therefore start by considering the actual physical transport within each edge at the microscale. Here, we will assume a generic advection-reaction-diffusion transport on each edge. We will show that under suitable simplifying assumptions, we can solve the microscale problem on each edge for multiple species to obtain the change of density along the edges and match the fluxes at the end of the edges to those at each node. If the density profile along an edge depends on the nodal densities in a linear fashion, one can, in principle, obtain a linear operator, which we call the effective graph Laplacian , hence connecting the physics at the microscale to the macroscale behaviour ( Fig. 1 ). Figure 1: Schematic of the model’s structure. At the microscale, the system is a physical network where nodes are compartments with volumes \mathcal{V}_{i} and the edges are (one-dimensional) tubes defining a continuum domain for the advection-reaction diffusion process (with orientation of the advection shown by the arrows). At the macroscale, the system is viewed as a mathematical network of nodes and edges where the nodal densities \rho_{i} evolve according to the macroscopic governing equation ( Eq. 21 ). 2 General setup A network of size N is defined as a pair \mathcal{G}=(\mathcal{V},\mathcal{E}) , where \mathcal{V}=\{1,2,\dots,N\} is the set of nodes and \mathcal{E}\subseteq\mathcal{V}\times\mathcal{V} is the set of edges connecting them. Since active transport between nodes is generally not symmetric, the network is directed, with edge (i,j) distinct from edge (j,i) . An edge (i,j) has arc length \ell_{ij} and is assumed to have uniform cross-sectional area w_{ij} so that, the edge volume is given by v_{ij}\approx\ell_{ij}w_{ij} (without summation on repeated indices). Note that, in the case where no physical edge (i,j) actually exists in the graph, we simply take w_{ij}=0 . We first consider the advection-diffusion-reaction of n species on a single edge before connecting the different edges to nodes. 3 Advection-reaction-diffusion along an edge At the microscale, we view a single edge as a slender one-dimensional tube with arclength coordinate s\in\left[0,\ell\right] in which n species with density \mathbf{x}=\mathbf{x}(s,t)\in\mathbb{R}^{n} interact through \frac{\partial{\mathbf{x}}}{\partial{t}}+\frac{\partial{\mathbf{J}}}{\partial{s}}=\mathbf{F}(\mathbf{x}). (2) Here, \mathbf{F}(\mathbf{x})\in\mathbb{R}^{n} captures any source or sink, or chemical reaction between the species. The flux \mathbf{J} through the domain is given by \mathbf{J}=\mathbf{V}\mathbf{x}-\mathbf{D}\frac{\partial{\mathbf{x}}}{\partial{s}}, (3) with \mathbf{V} the diagonal transport