1: \begin{abstract}

2: In \cite{Kopt_mc14}

3: a counterexample of an anisotropic triangulation was given on which

4: the exact solution has a second-order error of linear interpolation,

5: while

6: the computed solution obtained using linear finite elements is only first-order pointwise accurate.

7: %

8: This example was given in the context of a singularly perturbed reaction-diffusion equation.

9: %the exact solution of which exhibit sharp boundary layers.

10: %

11: In this paper, we present further examples of unanticipated pointwise convergence behaviour of Lagrange finite elements on anisotropic triangulations.

12: %

13: In particular,

14: we  show that

15: linear finite elements may exhibit lower than expected orders of convergence

16: for the Laplace equation, as well as for certain singular

17: equations, and their accuracy depends not only on the linear interpolation error, but also on the mesh topology.

18: %

19: %It is also shown that  lumped-mass quadrature may in certain situations (although not always) improve the orders of convergence to two.

20: %

21: Furthermore, we  demonstrate that %similar, i.e.

22:  pointwise convergence rates which are worse than one might expect

23: are also observed when higher-order finite elements are employed on anisotropic meshes.

24: %

25: A theoretical justification  will be given for some of the observed numerical phenomena.

26: %which relies on finite difference representations of the considered finite element methods.

27: \end{abstract}

28: