Researchers analyze a version of the nonlinear stochastic heat equation in one space dimension. They start from an equation where the Laplacian (the diffusion term) is driven by a very rough forcing given by the spatial gradient of space–time white noise. That rough noise is smoothed in space at scale ε > 0, and multiplied by a small factor ε3/4. The main result is that as the smoothing scale ε goes to zero, the random solutions converge in law to the solution of another stochastic heat equation. The limiting equation has ordinary space–time white noise, but with a changed nonlinearity: the noise appears multiplied by c g′(u) g(u) for a positive constant c and the same nonlinear function g that appeared in the original model.

Concretely, the prelimit equation studied is ∂_t u^ε − Δu^ε = ε3/4 g(u^ε) ∇ξ_ε, where ξ_ε is the spatial mollification of space–time white noise ξ. The factor ε3/4 is crucial. The authors explain that the spatial gradient of white noise is “supercritical,” meaning it is too singular for standard methods to work without reducing its strength. A naive scaling argument would not have predicted the 3/4 exponent. The paper works on the one-dimensional torus (periodic space) and starts from initial data ψ that is 1/4‑Hölder continuous.

Their rigorous statement is that the Itô solutions u^ε — that is, solutions interpreted using the Itô stochastic integral — converge weakly (in law) in the space of continuous functions on [0,1]×T as ε→0 to the solution u of a stochastic heat equation driven by white noise ξ with right-hand side c g′(u) g(u) ξ. The result holds under mild smoothness assumptions on g: it must have bounded derivatives up to third order. The convergence is probabilistic (weak convergence of laws), not a pathwise deterministic limit.

The proof uses elementary stochastic analysis tools rather than the heavy machinery sometimes used for very singular stochastic PDEs. Key ingredients are Itô calculus, standard heat kernel bounds, martingale inequalities (Burkholder–Davis–Gundy), and the stochastic sewing lemma. These allow the authors to control the sequence u^ε, show tightness of their laws, and identify any subsequential limit as the solution of the stated limiting equation. When g(u)=u, the result recovers, via the Cole–Hopf transform, a previously known renormalisation result for the Kardar–Parisi–Zhang (KPZ) equation, but here the same approach also covers genuinely nonlinear g.