Physarum Lagrangian: Adaptive Transport Networks

• Physarum Lagrangian is a variational framework that models adaptive transport networks in slime mold by minimizing energy dissipation and metabolic costs.
• The approach integrates fast flow relaxation with slow conductance adaptation, explaining shortest-path selection and network pruning through local feedback.
• It applies to predefined graphs with fixed source-sink boundary conditions, providing insights into biological computation and network optimization.

The Physarum Lagrangian is a physics-based variational framework capturing the adaptive dynamics of transport networks formed by the slime mold Physarum polycephalum. This mathematical formulation describes how the organism’s protoplasmic network organization arises as the constrained minimization of a least-action functional that balances viscous dissipation with metabolic maintenance, providing a quantitative explanation for the observed dynamics in shortest-path selection and pruning of transport tubes. Under this framework, network adaptation is interpreted as the self-organized solution of energy minimization, operating across predefined graphs under fixed source–sink boundary conditions, with steady states corresponding to extrema of a global action functional (Solé et al., 11 Nov 2025).

1. Formulation of the Physarum Lagrangian

Let $G = (V, E)$ denote a connected undirected graph representing the possible transport network. Each edge $(i,j) \in E$ has:

• $L_{ij} > 0$: geometric length of the tube between nodes $i$ and $j$,
• $D_{ij}(t) > 0$: hydraulic conductance, proportional to tube cross-section and inversely proportional to viscosity, and
• $Q_{ij}(t) = -Q_{ji}(t)$: volumetric flux from $i$ to $j$.

Each node $i$ is assigned an external source/sink $(i,j) \in E$0 ($(i,j) \in E$1) and a Lagrange multiplier $(i,j) \in E$2, identified as the pressure $(i,j) \in E$3. The instantaneous Lagrangian functional is given by:

$(i,j) \in E$4

The first term quantifies the total viscous power dissipation (with per-tube resistance $(i,j) \in E$5). The second term encodes Kirchhoff’s law for node mass balance, enforcing $(i,j) \in E$6 with $(i,j) \in E$7.

2. Stationarity Conditions and Physical Principles

The extremization of $(i,j) \in E$8 with respect to $(i,j) \in E$9 and $L_{ij} > 0$0 yields the Euler–Lagrange equations:

• Varying $L_{ij} > 0$1 gives:

$L_{ij} > 0$2

This equates to Poiseuille’s law: $L_{ij} > 0$3.

• Varying $L_{ij} > 0$4:

$L_{ij} > 0$5

This enforces Kirchhoff’s node-balance law. The system rapidly relaxes to instantaneous flow configurations minimizing dissipation at fixed conductance.

3. Trade-Off: Metabolic Dissipation versus Transport Efficiency

Transport network adaptation in Physarum involves two coupled components:

• Instantaneous dissipation: Minimized at fixed $L_{ij} > 0$6,

$L_{ij} > 0$7

• Slow morphological (maintenance) cost: Captured by a reduced free energy,

$L_{ij} > 0$8

with $L_{ij} > 0$9 tuning the metabolic-transport balance, and $i$0 representing per-length maintenance costs (e.g., $i$1 for linear costs).

At steady state, minimizing $i$2 enforces:

$i$3

A plausible implication is that unused or inefficient links shrink to zero ($i$4), thereby pruning the network.

4. Boundary Conditions and Problem Encoding

Sources and sinks are specified through the $i$5 terms. Standard configurations set $i$6 at a source node $i$7 and $i$8 at a sink node $i$9, with $j$0 elsewhere. These fixed-flux boundary conditions produce uniquely defined flow regimes, determining the transport problem Physarum must adaptively solve.

5. Case Studies: Ring, Tree, and Lattice Geometries

The Physarum Lagrangian quantitatively recovers experimentally observed network morphologies in canonical graph topologies:

Geometry	Lagrangian Structure	Emergent Network Behavior
Ring (parallel paths)	$j$1	Prunes longer branch; all flow via shortest arc
Binary tree	Local node Lagrangians representing inflow/outflows, e.g., $j$2	Only branches to sinks remain (minimal tree); others collapse
Square lattice	Discrete sum plus node balance: $j$3	In the continuum, flows concentrated along geodesics (diagonal)—shortest path

In all cases, conductance and flux configurations produced by the variational framework correspond to shortest-path or minimal-resistance configurations characteristic of Physarum’s experimental behavior.

6. Variational Principle and Shortest-Path Selection

Network adaptation is governed by a two-stage process:

1. Fast relaxation: For fixed $j$4, flows $j$5 and pressures $j$6 instantly minimize dissipation, corresponding to a Dirichlet (least-dissipation) problem.
2. Slow conductance adaptation: Conductances $j$7 evolve along a gradient flow, minimizing the free energy $j$8. Unused links are pruned ($j$9).

Because minimal-dissipation backbones coincide with shortest (or shortest-effective) paths in typical graph domains, the framework yields network configurations concentrated on these geodesics. The apparent “problem-solving” of Physarum—e.g., tracing shortest paths—arises from local mechanochemical feedback implementing a least-action principle, without explicit symbolic computation.

7. Synthesis: Biological Computation as Least-Action Dynamics

The Physarum Lagrangian establishes a unified account of adaptive network organization, situating Physarum’s computation as energy-minimizing action in a spatially extended active matter system. The succession of fast flow relaxation and slow conductance adaptation self-consistently produces shortest-path architectures, with direct correspondence to observed tube morphologies and laboratory network formation experiments. This formalism elucidates the fundamental mechanism behind Physarum’s celebrated problem-solving: an emergent variational process minimizing global energy dissipation and maintenance cost under network and boundary constraints (Solé et al., 11 Nov 2025).