Module 3: Fractals & Branching
Life has solved a universal engineering problem: how to transport resources efficiently to every cell in an organism using the minimum amount of material. The answer, discovered independently by Leonardo da Vinci, Cecil Murray, Benoît Mandelbrot, and the WBE group, is a fractal branching network. From the bronchial tree with D \(\approx 2.97\) to the cardiovascular system obeying Murray's law, nature builds self-similar, space-filling hierarchies that minimise dissipation. This module derives the mathematics of fractal dimension, Murray's cubic law, West-Brown-Enquist scaling, and shows how L-systems encode the rules of botanical growth.
1. Fractal Dimension
A fractal is a set whose intricate, self-similar structure persists across scales. Unlike smooth manifolds, fractals fail to have a well-defined tangent space and their “size” depends on the resolution of measurement. Mandelbrot (1967, 1982) formalised this with the concept of fractal dimension.
1.1 Box-Counting Dimension
Cover a set \(S \subset \mathbb{R}^n\) with cubes of side \(\varepsilon\). Let\(N(\varepsilon)\) be the minimum number of such cubes needed. Define:
\[ D_{\text{box}} = \lim_{\varepsilon \to 0} \frac{\log N(\varepsilon)}{\log(1/\varepsilon)} \]
For a smooth curve \(D=1\), a smooth surface \(D=2\); fractals give fractional values.
1.2 Self-Similarity Dimension
If \(N\) copies of the set, each scaled by factor \(s<1\), reproduce the whole, the similarity dimension is:
\[ D_s = \frac{\log N}{\log(1/s)} \]
Cantor set: \(N=2, s=1/3 \Rightarrow D = \log 2/\log 3 \approx 0.631\); Koch curve:\(D = \log 4/\log 3 \approx 1.262\).
1.3 Hausdorff Dimension
The most rigorous construction. For any \(d \geq 0\) and \(\delta > 0\) define the\(d\)-dimensional Hausdorff \(\delta\)-content:
\[ \mathcal{H}^d_\delta(S) = \inf\left\{ \sum_{i=1}^{\infty}(\mathrm{diam}\,U_i)^d : S \subset \bigcup U_i,\ \mathrm{diam}\,U_i \leq \delta \right\} \]
Then \(\mathcal{H}^d(S) = \lim_{\delta \to 0} \mathcal{H}^d_\delta(S)\), and the Hausdorff dimension is the unique critical value:
\[ \dim_H(S) = \inf\{d : \mathcal{H}^d(S) = 0\} = \sup\{d : \mathcal{H}^d(S) = \infty\} \]
For “nice” self-similar sets satisfying the open set condition, all three dimensions coincide:\(D_{\text{box}} = D_s = \dim_H\).
1.4 The Mandelbrot Set and the Coastline Paradox
Richardson (1961) observed that the measured length of a coastline depends on the ruler size:\(L(\varepsilon) \sim \varepsilon^{1-D}\). Britain's west coast gives \(D \approx 1.25\); Norway's fjord coast \(D \approx 1.52\). No “true” length exists — only a dimensional scaling.
The Mandelbrot set \(M = \{c \in \mathbb{C} : z_{n+1}=z_n^2+c \text{ stays bounded}\}\) has a boundary of Hausdorff dimension exactly 2 (Shishikura 1998), yet zero area. Its fractal boundary is a prototype for biological roughness: lung alveolar surfaces, leaf margins, neuronal dendrites.
2. The Lung Bronchial Tree
The human lung is the archetypal biological fractal. Starting from the trachea (generation 0, radius 9 mm, length 120 mm), it bifurcates 23 times before reaching the terminal alveoli (generation 23, radius 0.15 mm). At each branching, radii and lengths scale by a factor\(\beta \approx 2^{-1/3} \approx 0.794\) — exactly as Murray's law predicts.
2.1 Weibel's Regular Dichotomy
Weibel (1963) proposed the first quantitative model: a symmetric, regular bifurcation with radius ratio \(\beta\). After \(n\) generations, there are\(2^n\) branches each of length \(L_0 \beta^n\) and radius \(r_0 \beta^n\). Total cross-sectional area at generation \(n\):
\[ A_n = 2^n \pi r_0^2 \beta^{2n} = \pi r_0^2 (2\beta^2)^n \]
With \(\beta = 2^{-1/3}\) we have \(2\beta^2 = 2 \cdot 2^{-2/3} = 2^{1/3}\), so \(A_n\) grows geometrically — area expands by \(2^{1/3} \approx 1.26\) per generation. At generation 23 the total alveolar area is ~ 100 m², enabling oxygen diffusion to match metabolic demand.
2.2 Fractal Dimension of the Bronchi
The 3D fractal dimension of the bronchial tree is:
\[ D = \frac{\log N}{\log(1/s)} = \frac{\log 2}{\log(1/\beta)} = \frac{\log 2}{\log(2^{1/3})} = 3 \]
Space-filling! But real lungs have a slight asymmetry, yielding \(D \approx 2.97\) (Mandelbrot 1982; West 2017).
2.3 Asymmetric Branching (Horsfield)
Horsfield (1971) showed real lungs branch asymmetrically: one “major” daughter of diameter\(d_M = 0.86 d_p\) and a “minor” daughter of diameter \(d_m = 0.68 d_p\). Murray's cubic law \(d_M^3 + d_m^3 = d_p^3\) is approximately obeyed:\(0.86^3 + 0.68^3 = 0.636 + 0.314 = 0.950 \approx 1\).
3. Murray's Law
Cecil Murray (1926) asked: what vessel radius minimises the combined cost of pumping blood against viscous resistance plus the metabolic cost of maintaining a blood volume? The answer reshaped vascular biology.
3.1 Hagen-Poiseuille Power Dissipation
For laminar Newtonian flow of volumetric rate \(Q\) through a pipe of radius\(r\), length \(L\), and viscosity \(\mu\), the pressure drop is\(\Delta P = 8\mu L Q/(\pi r^4)\) and the power dissipated is:
\[ P_{\text{visc}} = Q \Delta P = \frac{8\mu L Q^2}{\pi r^4} \]
3.2 Metabolic Cost of Blood Volume
Maintaining blood (oxygenating, osmoregulating, transporting metabolites) costs energy per unit volume. Modelled as a constant \(b\) per unit volume:
\[ P_{\text{met}} = b \cdot \pi r^2 L \]
3.3 Minimise Total Cost
Define total cost \(E(r) = 8\mu L Q^2/(\pi r^4) + b\pi r^2 L\). Set \(dE/dr = 0\):
\[ -\frac{32\mu L Q^2}{\pi r^5} + 2b\pi r L = 0 \;\Rightarrow\; Q^2 = \frac{b\pi^2 r^6}{16\mu} \]
Thus, Q scales as \(r^3\):
\[ Q = k\, r^3 \quad \text{with}\quad k = \sqrt{\frac{b\pi^2}{16\mu}} \]
3.4 Branching Law
At a bifurcation, conservation of mass gives \(Q_p = Q_{d_1} + Q_{d_2}\). Applying Murray:
\[ \boxed{\; r_p^3 = r_{d_1}^3 + r_{d_2}^3 \;} \]
Murray's cubic law — verified for cardiovascular, plant xylem and pulmonary networks.
3.5 Constancy of Wall Shear Stress
The wall shear stress in Poiseuille flow is \(\tau = 4\mu Q/(\pi r^3)\). Under Murray,\(Q \propto r^3\), so \(\tau\) is constantthroughout the network. This matches experiment: arterioles and capillaries have nearly identical shear stresses (Kamiya-Togawa 1980), driving endothelial mechanobiology.
4. West-Brown-Enquist (WBE) Theory
West, Brown, and Enquist (1997) generalised Murray's law to derive quarter-power scaling laws from three postulates applied to a branching network:
- Networks are space-filling: they reach every cell.
- Terminal units are invariant: capillaries (or petioles) have a fixed radius across species.
- Total resource transport is minimised (Murray-like variational principle).
4.1 Derivation Sketch
Let \(n\) be a bifurcation ratio per level. After \(N\) levels, the terminal number is \(N_{\text{cap}} = n^N\). Space-filling requires the volume per branch to scale as \(L_k^3\) so that \(L_k \propto n^{-k/3}\); minimising dissipation forces \(r_k \propto n^{-k/2}\) (quite different from Murray's pulsatile optimum which gives \(r_k \propto n^{-k/3}\)).
Total network volume (whole-body blood mass) scales as:
\[ V \propto \sum_{k=0}^{N} n^k \pi r_k^2 L_k \propto N_{\text{cap}}^{4/3} \]
Since \(V \propto M\) (body mass) and metabolic rate \(B \propto N_{\text{cap}}\)(one capillary per metabolising cell):
\[ \boxed{\; B \propto M^{3/4} \;} \]
Kleiber's 3/4-power law, derived from geometry. See Module 4 for details.
4.2 Fractal Dimension of Networks
The WBE network has effective fractal dimension \(D = 3\) (space-filling). The bronchial tree is slightly sub-3 because of asymmetric branching. Tumour vasculature, by contrast, shows\(D \approx 1.8\text{--}2.3\) — a geometric signature of malignancy (Baish & Jain 2000). This connects directly to Module 8.
5. Plant Branching & Leonardo's Rule
Leonardo da Vinci observed (Notebooks, c. 1500) that trees obey a striking rule: the sum of the cross-sectional areas of all branches at one height equals the trunk area.
\[ \sum_i r_i^2 = r_{\text{trunk}}^2 \]
Leonardo's quadratic rule — conservation of xylem area, ensuring constant sap flow.
This is the tree-species analogue of Murray's cubic law. The difference arises because xylem tracheids transport water by capillary tension (driven by transpiration), not pressurised pumping; the dominant constraint is mechanical strength and hydraulic safetyrather than viscous dissipation.
5.1 Mechanical Derivation
Consider a branch subject to bending stress from its own weight (density \(\rho\)) and wind drag (velocity \(U\)). The tip deflection for a cantilever of length\(L\), radius \(r\), Young's modulus \(E\) is\(\delta = FL^3/(3EI)\) with \(I = \pi r^4/4\). For a constant safety factor,\(\delta/L\) must be constant, giving \(r \propto L^{3/2}\) — McMahon's elastic similarity scaling (1973).
5.2 Romanesco Broccoli
Brassica oleracea var. botrytis (Romanesco) displays striking self-similar fractal geometry: the entire head is a logarithmic spiral of miniature spiral florets, each itself a logarithmic spiral. The branching dimension is approximately \(D \approx 2.66\), intermediate between surface (\(D=2\)) and volume (\(D=3\)) — a space-partial-filling form driven by the same meristematic rules (phyllotaxis, Module 1) at every level.
6. Visualisations
A symmetric recursive tree with Murray scaling.
7. Lung Bronchial Tree (Weibel Model)
8. Mandelbrot Set Detail
The boundary of \(M\) has Hausdorff dimension 2 (Shishikura 1998). Zooming reveals infinite self-similar detail, a canonical illustration of fractal complexity.
9. Simulation 1 — Box-Counting Dimension
A synthetic 12-generation bronchial tree is generated using Murray scaling with a little noise. The box-counting dimension is then estimated from the log-log slope of \(N(\varepsilon)\). A projection of a true 3D space-filling network gives \(D \approx 2\) (filling the 2D plane) while a truly fractal projection remains sub-2.
Click Run to execute the Python code
Code will be executed with Python 3 on the server
10. Simulation 2 — Murray's Law
We minimise the combined viscous + metabolic energy cost as a function of vessel radius, recover the optimal radius, verify the \(r^3\) scaling of daughter-parent radii, and demonstrate the invariance of wall shear stress.
Click Run to execute the Python code
Code will be executed with Python 3 on the server
11. Simulation 3 — L-system Fractal Trees
Lindenmayer systems (L-systems) are string-rewriting grammars that generate realistic-looking plants from a few simple rules. Three production rules produce dramatically different morphologies.
Click Run to execute the Python code
Code will be executed with Python 3 on the server
Module Summary
Fractal dimension
D = log N / log(1/s); box, similarity, and Hausdorff agree for self-similar sets.
Lung D ~ 2.97
Space-filling bronchial network; 23 generations; alveolar area ~100 m^2.
Murray's law
r_p^3 = r_d1^3 + r_d2^3 minimises viscous + metabolic cost.
Shear stress constancy
Under Murray, tau is invariant across levels — drives endothelial biology.
WBE theory
Space-filling + invariant terminal + minimum dissipation -> B ~ M^(3/4).
Leonardo's rule
Plant xylem obeys quadratic area conservation, not cubic.
Tumour vasculature
D ~ 1.8-2.3 (abnormal); fractal dimension as a cancer biomarker.
L-systems
Symbolic grammars generate realistic plants — botanical form as computation.
References
- Mandelbrot, B. B. (1982). The Fractal Geometry of Nature. W.H. Freeman.
- Murray, C. D. (1926). The physiological principle of minimum work. I. The vascular system and the cost of blood volume. PNAS, 12(3), 207–214.
- West, G. B., Brown, J. H., & Enquist, B. J. (1997). A general model for the origin of allometric scaling laws in biology. Science, 276, 122–126.
- Weibel, E. R. (1963). Morphometry of the Human Lung. Springer.
- Horsfield, K., Dart, G., Olson, D. E., et al. (1971). Models of the human bronchial tree. J. Appl. Physiol., 31(2), 207–217.
- Sherman, T. F. (1981). On connecting large vessels to small. The meaning of Murray's law. J. Gen. Physiol., 78, 431–453.
- Shishikura, M. (1998). The Hausdorff dimension of the boundary of the Mandelbrot set. Ann. Math., 147, 225–267.
- McMahon, T. A. (1973). Size and shape in biology. Science, 179, 1201–1204.
- Prusinkiewicz, P. & Lindenmayer, A. (1990). The Algorithmic Beauty of Plants. Springer.
- Baish, J. W. & Jain, R. K. (2000). Fractals and cancer. Cancer Research, 60, 3683–3688.
- Kamiya, A. & Togawa, T. (1980). Adaptive regulation of wall shear stress to flow change in the canine carotid artery. Am. J. Physiol., 239, H14–H21.
- West, G. B. (2017). Scale. Penguin Press.