Convergence to a single steady state is shown for non-negative and radially symmetric solutions to a diffusive Hamilton-Jacobi equation with homogeneous Dirichlet boundary conditions, the diffusion being the p-Laplacian operator, p 2, and the source term a power of the norm of the gradient of u. As a first step, the radially symmetric and non-increasing stationary solutions are characterized.
G. Barles et al. / Convergence to steady states for radially symmetric solutionsWe further assume the initial condition u 0 ∈ W 1,∞ 0 (B) is radially symmetric and non-negative and u 0 ≡ 0, (1.2) while the parameters p and q satisfy p 2 and 0 < q < p − 1.(1.3)The partial differential equation in (1.1) is a second-order parabolic equation featuring a diffusion term (possibly quasilinear and degenerate if p > 2) and a source term |∇u| q counteracting the effect of diffusion and depending solely on the gradient of the solution. The competition between the diffusion and the source term is already revealed by the structure of steady states to (1.1). Indeed, while it follows from Theorem 1 in [4], that zero is the only steady state in C(B) when p 2 and q p − 1, several steady states may exist when p 2 and q ∈ (0, p − 1) [6,14,20]. Another typical feature of the competition between diffusion and source is the possibility of finite time blow-up in a suitable norm, and this phenomenon has been shown to occur for (1.1) when p = 2 and q > 2, see [16] and the references therein. More precisely, it is established in [18] that, when p = 2 and q > 2, there are classical solutions to (1.1) for which the L ∞ -norm of the gradient blows up in finite time, the L ∞ -norm of the solution remaining bounded. These solutions may actually be extended to all positive times in a unique way within the framework of viscosity solutions [5,21], the boundary condition being also satisfied in the viscosity sense. According to the latter, the homogeneous Dirichlet boundary condition might not always be fulfilled for all times, a property which is likely to be connected with the finite time blow-up of the gradient.Coming back to the case where p and q fulfil (1.3) and several steady states may exist, a complete classification of steady states seems to be out of reach when B is replaced by an arbitrary open set of R N . Nevertheless, there are at least two situations in which the set of stationary solutions can be described, namely, when N = 1 and B = (−1, 1) [14,20] and when N 2 under the additional requirement that the steady states are radially symmetric and non-increasing, the latter being the first result of this paper. More precisely, we show that (1.1) has a one-parameter family of stationary solutions and that each stationary solution is characterized by the value of its maximum. Theorem 1.1. Assume (1.3). Let w ∈ W 1,∞ (B) be a radially symmetric and non-increasing viscosity solution to −Δ p w − |∇w| q = 0 in B satisfying w = 0 on ∂B. Then there is ϑ ∈ [0, 1] such that w = w ϑ , where w ϑ (x) := c 0