The Liouville theorem under second order differentiability assumption (Q2437410)

Let \(C^+(n)\) denote the space of \(n\times n\) matrices that are nonnegative scalar multiples of the elements of the special orthogonal group. A diffeomorphism \(f\) between two domains \(\Omega\) and \(\Omega'\) in \(\mathbb R^n\) is called conformal if for every \(x\in \Omega\) the Jacobian matrix \(Df(x)\) is in \(C^+(n)\). A classical theorem of Liouville asserts that conformal diffeomorphisms are rigid for \(n\geq 3\), namely they are Möbius transformations. Liouville proved this result under the assumption \(f\in C^3\). \textit{F. W. Gehring} [Trans. Am. Math. Soc. 103, 353--393 (1962; Zbl 0113.05805)] extended it to Sobolev homeomorphisms in the class \(W^{1,n}\). Later, Reshetnyak removed the assumption of being a homeomorphism. In even dimensions \(n\geq 4\), \textit{T. Iwaniec} and \textit{G. Martin} [Acta Math. 170, No. 1, 29--81 (1993; Zbl 0785.30008)] achieved the ultimate form of Liouville's theorem: if a nonconstant Sobolev map \(f\in W_{\text{loc}}^{1,n/2}\) is such that \(Df(x)\in C^+(x)\) for almost every \(x\), then \(f\) is the restriction of a Möbius transformation. They also showed that for \(p<n/2\), the assumption \(f\in W_{\text{loc}}^{1,p}\) is not sufficient for this conclusion, in any dimension \(n\geq 3\). Thus, the remaining problem is whether conformal maps in \(f\in W_{\text{loc}}^{1,n/2}\) are necessarily Möbius in odd dimensions \(n\geq 3\). This is the problem that the article under review addresses. While the Iwaniec-Martin conjecture remains open in full generality, the following theorem establishes the rigidity of conformal Sobolev maps under a second-order differentiability assumption. Theorem. Let \(\Omega\) be a domain in \(\mathbb R^n\), \(n\geq 3\). Suppose that \(f\in W_{\text{loc}}^{1,1}(\Omega;\mathbb R^n)\), \(Df(x)\in C^+(x)\) for almost every \(x\), and \(|Df|^p\in W^{1,2}_{\text{loc}}\) for some \(p\geq (n-2)/4\). Then \(f\) is either constant or the restriction of a Möbius transformation. The exponent \((n-2)/4\) above is shown to be optimal. The author also shows that the Iwaniec-Martin conjecture would follow from a certain Caccioppoli-type estimate.