← All problems

Gaussian Moments Conjecture in dimension two

11 years old · traced to

Let X,YX,Y be independent standard real Gaussian random variables. Suppose that P∈C[X,Y]P\in\mathbb{C}[X,Y] satisfies

E(P(X,Y)m)=0\mathbb{E}\left(P(X,Y)^m\right)=0

for every integer m≥1m\geq 1. Must it follow that, for every Q∈C[X,Y]Q\in\mathbb{C}[X,Y],

E(Q(X,Y)P(X,Y)m)=0\mathbb{E}\left(Q(X,Y)P(X,Y)^m\right)=0

for all sufficiently large integers mm?

Equivalently, does the Gaussian Moments Conjecture GMC(2)(2) hold?

References

Progress summary

Refreshed
Open · solution reported

A July 2026 preprint claims to prove the conjecture in two dimensions, completing the known classification, but the claim has not yet been independently verified.

For independent standard real Gaussians X,YX,Y and a complex polynomial PP, the conjecture asks whether vanishing of every positive moment E(Pm)\mathbb{E}(P^m) forces eventual vanishing of E(QPm)\mathbb{E}(QP^m) for every polynomial QQ. Dimension two was the last unresolved case after the one-dimensional result and counterexamples in dimensions n≥3n\geq 3.

Known results

  • The conjecture holds for n=1n=1.
  • The homogeneous two-variable case follows from the theorem of Duistermaat and van der Kallen.
  • Long gave explicit counterexamples for every n≥3n\geq 3: a cubic for n≥4n\geq 4 and a quartic for n≥3n\geq 3.

July 2026 claimed proof

A preprint claims that a prime-isolation argument on a lower Newton-polygon face forces all monomials to have weights of one strict sign, yielding the required eventual vanishing and hence truth of GMC(n)\mathrm{GMC}(n) for n≤2n\leq 2.

Current status (as of July 2026): The one-dimensional case and all counterexamples for n≥3n\geq 3 are settled; dimension two is claimed solved by a preprint, pending independent verification.

Sources

Solutions 0

No solutions have been posted yet.