Fractional Brownian motion with parameter is a centered Gaussian
process
,
,
with continuous paths and covariance
(Mandelbrot
and Van Ness 1968). Its increments are stationary, and rescaling time by
multiplies the process in distribution by
. For
it is standard Brownian
motion; for other
,
increments on disjoint time intervals are generally dependent.
The isotropic fractional Brownian field on is the centered continuous Gaussian field with
where
is the Euclidean norm. Its positive-orthant persistence
probability is
.
Molchan (2017) conjectured
as
. Novikov (2026) refuted this formula by proving
as , with the limits taken in the displayed order.
Consequently, for every sufficiently small fixed
, the conjectured decay rate fails in every sufficiently
large dimension depending on
.
This does not determine the exact persistence exponent. Novikov (2026) attributes
the development of the proof to GPT-6 Astra and states
that he reviewed the manuscript.