The paper tackles a subtle gap in the mathematical theory of certain partial differential equations called fully nonlinear elliptic equations. These equations appear in areas like optimal control, geometry, and financial mathematics. A classical theorem from the 1980s guarantees that solutions to these equations are "smooth enough" in a precise technical sense, specifically that their gradients satisfy a Holder continuity condition with some exponent strictly less than a critical value. The theory has always left open whether solutions achieve that critical level of smoothness exactly, or whether something slightly worse is the best one can hope for. This open question is what the paper addresses.
The authors show that solutions do essentially reach the critical smoothness threshold, but with a small logarithmic correction. Think of it this way: if you are trying to measure how well a function can be approximated by a linear function near a point, the classical theory gives you a bound that falls just short of optimal. This paper proves that the true behavior at the boundary case involves an extra multiplicative factor that grows only logarithmically as you zoom in, which is an extremely mild defect. The precise form of this logarithmic correction depends on whether you are exactly at the critical threshold or just above it, and the two cases require different technical approaches.
The key technical innovation is a new method for choosing the right sequence of scales at which to analyze the problem, which the authors call a scale-selection mechanism for endpoint Campanato-type recurrences. Classical iterative arguments break down at the critical exponent because the standard geometric progression of scales no longer closes up the estimates. By choosing scales more carefully, the authors manage to extract the sharpest possible information. They also discuss whether the logarithmic corrections they find are unavoidable or just artifacts of the method, suggesting the results may be essentially sharp.