\(L^1-L^p\) estimates for radial solutions of the wave equation and application (Q740034)

The paper is devoted to a never ending story, namely, to the semilinear Cauchy problem for the wave equation with power nonlinearity \[ u_{tt} -\Delta u =|u|^\sigma,\quad u(0,x)=0,\quad u_t(0,x)=g(x). \] It is known that the proof of local (in time) existence of large data solutions or global (in time) existence of small data solutions depends heavily on \(L^p-L^q\) estimates away of the conjugate line for solutions to the corresponding linear Cauchy problem with homogeneous right-hand side. It is known that such \(L^p-L^q\) estimates can be only expected for some admissible range of \(\big(\frac{1}{p},\frac{1}{q}\big)\). Inspired by new results for convolution operators acting on spaces \(L^p_{rad}(\mathbb{R}^n)\) of radial symmetric functions from \(L^p(\mathbb{R}^n)\) the authors prove \(L^1_{rad}-L^q_{rad}\) estimates in odd space dimensions \(n \geq 3\) and for \(q \in \big[1,\frac{2n}{n-1}\big)\) for solutions to the corresponding linear Cauchy problem with homogeneous right-hand side. Here they use a special representation of the kernel function \(K=K(t,|x|,y)\) in the explicit representation of solutions \[ u(t,x)=\int_0^\infty K(t,|x|,y)g(y)dy. \] Interpolating these estimates with known results gives now \(L^p-L^q\) estimates for radial symmetric data \(g\) in a wider set of admissible \(\big(\frac{1}{p},\frac{1}{q}\big)\). This is the main tool to prove the global existence of global (in time) small data Sobolev solutions for \(\sigma \in \big(\sigma_0(n),\frac{n+3}{n-1}\big)\), that is, for powers \(\sigma\) between Strauss exponent \(\sigma_0(n)\) and conformal power \(\frac{n+3}{n-1}\).