The Pólya-Szegő buckling conjecture (Pólya and Szegő 1951) asserts that a disk minimizes the first clamped buckling eigenvalue among smooth bounded planar domains of prescribed area. The problem is distinct from minimizing the fundamental vibration frequency of a clamped plate (Ashbaugh and Laugesen 1996, pp. 383-386).
The first buckling eigenvalue of an open planar set
is the infimum of the Rayleigh
quotient
|
(1)
|
where
is the topological closure in the Sobolev
space
of the real-valued smooth
functions with compact support contained in
. Here
is the gradient and
is the Laplacian. On a sufficiently smooth bounded
domain,
is the least eigenvalue of
with clamped boundary
conditions
,
where
is the outward unit normal vector. The vibrating
clamped-plate equation instead has the form
.
|
(2)
|
where
is the first positive zero of the Bessel
function of the first kind
.
The conjectured inequality is therefore
|
(3)
|
with equality only for disks, where denotes area.
Tang and Zhang (2026ab) reported this inequality and its equality characterization for every open planar set of finite positive area, without assumptions of connectedness, boundedness, boundary regularity, or a fixed sign for a first eigenfunction. Their claimed equality condition is exact equality with an open disk, rather than equality up to a set of area zero. They also reported the sharp comparison
|
(4)
|
where
is the first Dirichlet eigenvalue of the Laplacian
and
is the first positive zero of the Bessel
function of the first kind
.
The authors report ChatGPT assistance with their mathematical proof and Aristotle and Codex assistance with the Lean formalization. As of Oct. 25, 2026, the formalization was publicly registered and automated checks were reported, but independent specialist review of the new argument had not been reported.
The conjecture concerns the first buckling eigenvalue at fixed volume. A separate ratio conjecture compares the first two buckling eigenvalues and is not equivalent to this minimum-first-eigenvalue problem.
Dai et al. (2026b) reported an extension to every dimension . The same variational definition applies
to an open set
of finite positive volume.
Their claimed sharp comparison is
|
(5)
|
where
is the first positive zero of the Bessel
function of the first kind
.
For a ball
of the same volume, this implies
|
(6)
|
No boundedness, connectedness, boundary regularity, or sign assumption on a first eigenfunction is imposed. Equality in this volume
inequality holds precisely for a ball with a relatively
closed set removed such that
.
This forces an actual ball in dimensions
2 and 3, while removal of a single point is allowed in
dimensions at least 4. The calibration used in the
argument fails for
,
which does not refute the conjecture in those dimensions.