The Poisson conjecture asserts that every endomorphism of the canonical polynomial Poisson
algebra over a field of characteristic
0 is an automorphism. Here the endomorphisms
must preserve the canonical Poisson bracket. On
, this Poisson
bracket is determined by
and zero Poisson
brackets between the
variables or between the
variables.
The integer
counts canonical pairs, so the algebra has
polynomial generators.
The conjecture is closely related to the Dixmier conjecture. Preserving the canonical Poisson bracket
forces the associated polynomial map to have constant
Jacobian determinant 1. The Jacobian
conjecture in variables therefore implies the Poisson conjecture for
canonical pairs over the same field of characteristic
0 (Adjamagbo and van den Essen 2007).
Long (2026) gave an explicit noninvertible polynomial map preserving the Poisson bracket with two
canonical pairs, hence four variables. One coordinate
is , and a fiber of the
map has three points. This refutes the conjecture
for
and, by adjoining unchanged canonical pairs, for
; the case
remains open. The paper also gives a counterexample
to the Dixmier conjecture for the fourth Weyl algebra. ChatGPT supplied the construction with
additional AI auditing, and Long checked its proof.
Blore (2026a) formalized Long's counterexample and its extension to every over any field of characteristic
0 in Lean, with Claude Code assistance. The Palomar Registry (2026a) checking record
reports acceptance by the Lean, nanoda, and con-ron kernels. The paper's exact three-point
fiber count and its separate claim about the fourth Weyl algebra lie outside the checked statements.
Blore (2026b) separately formalized the implication from the Jacobian conjecture in Lean with Claude Code assistance, using the condition that the Jacobian determinant is a unit. The Palomar Registry (2026b) checking record for this conditional statement reports acceptance by Lean, nanoda, and con-ron.