On superposition operators between higher-order Sobolev spaces and a multivariate Faà di Bruno formula: the subcritical case. (Q364238)

The present article deals with the superposition operators \((N_gu)(x) = g(u(x)): W^{m,p}(\Omega) \to W^{l,q}(\Omega)\) between two Sobolev spaces \(W^{p,m}(\Omega)\) and \(W^{l,q}(\Omega)\) of functions on an open bounded domain \(\Omega \subset \mathbb {R}^n\) with the cone property. It is assumed that the function \(g: {\mathbb R} \to \mathbb {R}\) is of class \(C^{l-1}\) and the derivative \(g^{(l-1)}: \mathbb {R} \to \mathbb {R}\) is locally Lipschitz. In the case \(1 \leq p <\frac {n}{m}\) and \(1 \leq q \leq \frac {np}{l[n - (m - 1)p]}\), the operator \(N_g\) acts between \(W^{p,m}(\Omega)\) and \(W^{l,q}(\Omega)\), provided that the following growth condition \[ | g^{(l)}(t)| \leq a| t| ^{\frac {n(p-lq)+(m-1)lpq}{q(n-mp)}} + b \] holds a.e.\ in \(\mathbb {R}\); moreover, in this case, the following inequality \[ \| N_gu\|_{W^{l,q}(\Omega)} \leq C\bigg [\| u\|_{W^{m,p}(\Omega)}^{\frac {p(n-lq)}{q(n-mp)}} + 1\bigg ], \quad u \in W^{m,p}(\Omega), \] is true; \(C\) is a constant depending only on \(\Omega ,n,m,l,p,q,a,b,g(0),\dots ,g^{(l-1)}(0)\). In the case \(p =\frac {n}{m}\), \(1 \leq q <\frac {n}{l}\), the operator \(N_g\) acts between \(W^{p,m}(\Omega)\) and \(W^{l,q}(\Omega)\), provided that the following growth condition \[ | g^{(l)}(t)| \leq a| t| ^d + b \] holds; in this case, the following inequality \[ \| N_gu\|_{W^{l,q}(\Omega)} \leq C\bigg [\| u\|_{W^{m,p}(\Omega)}^{d+l} + 1\bigg ], \quad u \in W^{m,p}(\Omega), \] is true. Similar statements are formulated in some important partial cases, in particular when \(g\) is a polynomial of degree \(h \leq l\). The sufficient conditions presented in the article guarantee the continuity of the operator \(N_g\), the validity of the higher-order chain rule (the Faá de Bruno formula, cf. [\textit{C. F. Faà di Bruno}, ``Note sur une nouvelle formule du calcul différentiel'', Quart. J. Math. 1, 359--360 (1855)]) and also the important embedding \(N_g(W^{m,p}(\Omega) \cap W_0^{k,p}(\Omega)) \subset W_0^{l,q}(\Omega)\). A comparison of these results with classical theorems by \textit{M. Marcus} and \textit{V. J. Mizel} [Trans. Am. Math. Soc. 251, 187--218 (1979; Zbl 0417.46035)] and \textit{G. Bourdaud} [Invent. Math. 104, No.~2, 435--446 (1991; Zbl 0699.46019)] is given.