Butcher's theory of numerical integration uses rooted trees to encode the Taylor expansion of differential equations. Each tree represents a term in the expansion; composition of integration methods corresponds to operations on these trees. The algebraic structure — a Hopf algebra — organizes how methods combine, invert, and approximate solutions. For general ODEs, this framework is classical and complete.
Volume-preserving flows break it. When the vector field is divergence-free, the relevant trees acquire an additional structure called “aromaticity” — a decoration that tracks whether terms respect the volume-preservation constraint. The standard Hopf algebra of Butcher trees doesn't distinguish these; a richer algebraic framework is needed.
Zhu and Laurent show the right framework is a Hopf algebroid, not a Hopf algebra. The distinction is categorical: a Hopf algebroid lives over a non-commutative base ring, while a Hopf algebra lives over a field. The aromatic multi-indices — a simplified representation of the Taylor expansion for divergence-free flows — embed naturally into the BCK Hopf algebra within this richer structure. The embedding is clean, preserving the algebraic operations that matter for numerical analysis.
The structural point: the category-level upgrade from Hopf algebra to Hopf algebroid is not a technicality. It records the fact that volume preservation introduces a non-commutativity that the original framework cannot absorb. The constraint (divergence-free) doesn't just restrict the solutions — it upgrades the algebraic structure needed to describe them. The richer the physics, the richer the algebra, and the correspondence is not metaphorical: it is a functor.