LSFEM of 2-D Div-Curl Systems

Abhijit Tilak

Hi all,

recently I wrote a 2-D code based on Least Squares FEM to solve some simple BVP's. I am following the Jiang's book. Now here's my question:

I tried to solve a simple potential flow problem by solving div.V =0 and curl V =0 by LSFEM. Jiang's book says that you cannot use 3-node linear triangle with LSFEM. 4-node Iso-P quad works great. 6-Node Iso-P and subparametric elements don't work. The code has been validated, so I am sure there are no bugs that could result in failure. I overheard somebody at a conference saying LSFEM is more of a art than a science. I contacted Dr. Jiang, but he informed me that he has never used higher order elements with LSFEM. Also, my advisor does'nt know anything about LSFEM.

Any comments/suggestions/references greatly appreciated.

Thanks in Advance,

Abhijit Tilak.