Murray's Law as an Entropy-per-Information-Cost Extremum

Abstract

Transport networks must balance viscous pumping losses with the energetic cost of maintaining an operative architecture. This paper formulates that trade-off as an entropy-per-information-cost (EPIC) extremum that prices structural upkeep in calibrated units (joules per bit). An upkeep law r^m distinguishes volume-priced (m = 2) from surface-priced (m = 1) maintenance. In laminar Poiseuille flow, stationarity yields (i) a generalized Murray scaling Q proportional to r^alpha with alpha = (m + 4)/2; (ii) a tariff-weighted vector balance that fixes bifurcation geometry and predicts near-symmetric daughter openings of about 75 degrees for m = 2 and about 97 degrees for m = 1; and (iii) a universal partition of power between pumping and upkeep. Eliminating radii gives a strictly concave flux cost proportional to Q^gamma with gamma = 2m/(m + 4), favoring mergers and deep tree hierarchies, and defines a routing index that induces Snell-like refraction of optimal paths across spatial tariff contrasts. A preregistered, held-out test on retinal bifurcations from the High-Resolution Fundus dataset (N = 19,126) shows sharp vector closure: the median residual is R = 0.232 with a nonparametric 95 percent bootstrap interval [0.229, 0.236], 91 percent of junctions fall under the pre-specified strict threshold, and structure-preserving nulls shift decisively to larger residuals. These results render classical branching relations explicitly unit-bearing (J/bit) and provide falsifiable geometric targets and quantitative design rules for transport networks.