Conserved Quantity Audit¶
Diagnostic audit — instantiates Hamiltonian Mechanics and Canonical Transformations
Enumerates the quantities a system should keep constant and checks — before and after a transformation — that each one actually stays put, flagging any invariant the reformulation quietly broke.
A transformation is only allowed to change how a system looks, never what it conserves. Conserved Quantity Audit is the ledger that enforces that. It first writes down, explicitly and in advance, the quantities the system is supposed to hold constant — total energy, a momentum, an angular momentum, a Casimir, a population invariant — and why each one is conserved (which symmetry or law underwrites it). Then it measures each quantity in the original frame and in the transformed frame and confirms the two agree along the motion. Its distinguishing idea is that it audits specific, physical constants of the motion — a named list with values — rather than the structural machinery of the map; it answers "does energy still not drift?" not "is this coordinate change formally canonical?".
Example¶
A population ecologist has a predator–prey model in the Lotka–Volterra form and wants to rewrite it in variables that make its cycles easier to reason about. The model has a known conserved quantity — a specific combination of the two population densities that stays fixed along every closed orbit, which is exactly why the populations cycle rather than spiral in or blow up. Before trusting the rewritten model, the ecologist runs a Conserved Quantity Audit.
The audit lists the invariant, states its origin (the divergence-free structure of the flow), and evaluates it in both frames at a spread of points along a trajectory. In the original variables it holds to machine precision. In the reformulated variables it drifts by a few percent per cycle — a quiet signal that the "equivalent" rewriting introduced a subtle error in one term. The audit doesn't fix the transformation; it catches that something is wrong before a season of predictions is built on a model that slowly loses the very invariant that made its cycles real.
How it works¶
- Enumerate the invariants. List every quantity that should be constant, with its physical or mathematical justification — ideally traced to a symmetry via Noether's theorem[1] rather than assumed.
- Set tolerances. Decide how much drift counts as "conserved" versus "broken," separately for exactly-conserved and adiabatically-conserved quantities.
- Measure in both frames. Evaluate each invariant along representative trajectories in the original and transformed coordinates.
- Compare and flag. Any quantity that holds in one frame but drifts in the other indicts the transformation; a quantity that drifts in both indicts the model itself, not the map.
The move that makes it an audit rather than a check-of-structure is that it works from a pre-declared list of named quantities and their values — the contract is written before the answer is seen, so success can't be redefined around whatever the transformed model happens to preserve.
Tuning parameters¶
- Invariant coverage — how many conserved quantities to track. Auditing only energy is cheap but blind to broken momenta; tracking the full set is thorough but costly.
- Drift tolerance — the threshold that separates "conserved" from "broken." Tight tolerances catch subtle errors but flag benign numerical noise; loose ones do the reverse.
- Sampling depth — how many trajectories and how far along each. Long, varied sampling exposes slow drift that a single short run would miss.
- Exact vs. adiabatic stance — whether a quantity is required to be strictly constant or only slowly varying within a stated bound.
When it helps, and when it misleads¶
Its strength is that it converts a vague "the reformulation should be equivalent" into a checkable, itemized verdict, and it localizes failure: a broken invariant points at which relationship the transformation damaged. Writing the contract first is also what protects a team from the seductive drift of redefining success around whatever survived.
Its failure mode is the illusion of completeness. An audit only catches violations of the invariants on its list; a transformation can preserve every quantity you thought to track while breaking one you didn't, and the clean report reads as a full endorsement. The classic misuse is treating a passed audit as proof of equivalence rather than as failure to find a counterexample — absence of a caught violation is not presence of correctness. The guarding discipline is to derive the invariant list from the system's symmetries rather than intuition, keep the list open to additions, and pair a clean audit with an independent structural check rather than resting on it alone.
How it implements the components¶
invariant_preservation_contract— it is the contract: the explicit, pre-declared list of what must stay constant and why, against which the transformation is held.equivalence_validation_evidence— the before/after measurements of each invariant are the concrete evidence that the transformed model still means the same thing.
It does not verify the map's formal structure through bracket algebra (poisson_bracket_or_symplectic_check — that is Poisson-Bracket Identity Test and Symplectic Form Preservation Check), construct the canonical_transformation_rule it audits, or compare trajectory shapes visually (phase_space_state_representation — that is Phase Portrait Comparison). Where the Poisson-Bracket Identity Test asks whether the new coordinates are canonical, this audit asks whether named physical constants of the motion still hold their values.
Related¶
- Instantiates: Hamiltonian Mechanics and Canonical Transformations — supplies the invariant-preservation evidence the archetype requires before a reformulation is accepted.
- Sibling mechanisms: Phase Portrait Comparison · Poisson-Bracket Identity Test · Symplectic Form Preservation Check · Structure-Preserving Numerical Integration · Action-Angle Variable Substitution · Canonical Pair Normalization · Generating Function Derivation · Perturbative Canonical Transformation · Inverse Transform Backtranslation
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Enumerates the quantities a system should keep constant and checks — before and after a transformation — that each one actually stays put, flagging any invariant the reformulation quietly broke, making its operative form a bounded evaluation of existing evidence or work that produces a finding or disposition.
Independent corroboration: The frozen evidence defines Conserved Quantity Audit as 'Enumerates the quantities a system should keep constant and checks — before and after a transformation — that each one actually stays put, flagging any invariant the reformulation quietly broke', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Physics
Origin pattern: Single lineage
Present-day reach: Specialized
Rationale: Analytical mechanics cohered checking energy, momentum, and other constants of motion before and after a coordinate or model transformation.
Related originating lineages:
- Mathematics — Symmetry, invariant, and canonical-transformation theory supplies the formal reason each audited quantity must remain fixed.
Review resolution: Both reviewers agree on physics as primary. The source audits a physically conserved quantity across a boundary; mathematics materially supplies invariant and balance formalism, while the generalized audit expression is an encyclopedia synthesis for physical systems.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Noether, E. "Invariante Variationsprobleme". Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Mathematisch-Physikalische Klasse, 235–257 (1918). Connects continuous variational symmetries with corresponding conservation laws. registry ↩