Schauder estimates and Hölder data

For an exponent strictly between zero and one, the Hölder seminorm is the supremum of the absolute difference of function values divided by the distance to that exponent. A C-two-alpha norm combines suprema of derivatives through order two with the alpha-Hölder seminorm of second derivatives. This is stronger information than merely saying the Hessian is bounded. The exponent and the region must be included in any estimate, since a function can have different regularity on a compact interior set and near the boundary.

Consider the nondivergence equation minus a-ij times the second derivatives of u equals f on the unit ball. Assume the symmetric leading matrix is uniformly elliptic, its entries have bounded C-alpha norms, and f is C-alpha. A standard interior Schauder a priori estimate controls the C-two-alpha norm on the half ball by the sup norm of u and the C-alpha norm of f on the full ball. The constant depends on dimension, exponent, ellipticity, and coefficient norms, not just on a matrix's pointwise positivity.

A useful proof idea freezes the coefficients at one point and compares the equation to a constant-coefficient elliptic operator. On a sufficiently small ball, Hölder continuity makes the coefficient difference small. The difference contributes an error involving the unknown second derivatives, so the estimate is not complete until that contribution is absorbed into the left side. Interpolation inequalities and nested-ball arguments handle lower derivative terms and keep the final estimate expressed in terms of the solution's sup norm.

This paragraph describes a mechanism rather than a full proof. An a priori estimate assumes a solution of sufficient regularity and bounds it; existence requires a separate construction, such as a continuity method with uniform estimates and a known solvable starting operator. Likewise, a divergence-form equation with merely C-alpha A cannot simply be expanded into a nondivergence equation with bounded first-order coefficients. Differentiating A costs regularity. One must choose the divergence-form Schauder theorem appropriate to the actual coefficient class.

A global Dirichlet Schauder estimate for a bounded C-two-alpha domain requires compatible C-two-alpha boundary data, C-alpha coefficients and source, and the usual uniform ellipticity bounds. An estimate can include the sup norm of u; removing it through a maximum principle needs suitable zero-order signs or another uniqueness estimate. A polygon with reentrant corners does not meet the same boundary hypotheses, so smooth forcing alone cannot justify a global C-two-alpha conclusion there.

At alpha zero, continuity is not a substitute for a positive Hölder modulus in this theorem. More refined Dini conditions support other classical results, while general bounded sources belong naturally to an L-p theory rather than a uniform Hessian bound. An equation can smooth a source by two derivatives in one chosen scale without supplying arbitrary extra regularity. The following one-dimensional solution makes that limitation explicit: its second derivative has exactly the same noninteger Hölder behavior as the forcing.