I would like to solve the $p$-Laplace dirichlet problem \begin{align*} -\Delta_p u & = 1 \quad\text{in } \Omega:=[0,1]^2 \\ u & = 0 \quad\text{on }\partial\Omega \end{align*} for some $p>1$ with linear finite elements.

This can be done using a Newton method where we iterate over the solution $u^k$ and assemble the stiffness matrix $K = (k_{ij})$ where $$k_{ij} = \int |\nabla u^k|^{p-2}\nabla\varphi_j^T\nabla\varphi_i dx + (p-2)\int |\nabla u^k|^{p-4}(\nabla (u^k)^T\nabla\varphi_j)(\nabla (u^k)^T\nabla\varphi_i)dx$$ in each iteration.

Now my problem is that I do not know how to compute $\nabla u^k$. In each step I know $u^k$, i.e. a vector with the value of $u^k$ at each node, but I don't know how to compute the gradient from this.

Does anyone know any references for this or has experience in how to program this? Thanks!


If $u^k=\sum_{i=1}^NU^k_i\varphi_i$, where $N$ is the number of degrees of freedom, and $\varphi_i,\,\,\,i=1,\ldots,N$ are the basis functions, then $\nabla u^k=\sum_{i=1}^NU^k_i\nabla\varphi_i$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.