mothur can make a neighbor-joining tree through the clearcut() command, or a UPGMA tree through the tree.shared() command. Both of these require a distance matrix to compute, which you can generate using dist.seqs (for aligned data) or pairwise.seqs (for unaligned data).

If you need maximum likelihood you'll need to use another program, such as FastTree or MEGA. As long as you get a Newick-formatted tree from your tree-building software you can drop it straight back into mothur once you're done, if you want to continue phylogenetic analysis in mothur.

http://www.mothur.org/wiki/Dist.seqs

http://www.mothur.org/wiki/Pairwise.seqs

http://www.mothur.org/wiki/Clearcut

http://www.mothur.org/wiki/Tree.shared

Many thanks  David, It is very useful and I got the tree finally. Do you know how to make a consensus phylogenetic tree from different treeing algorithms? 

The only software I've ever done this in is ARB, which can also do NJ, MP and ML trees under a range of models and assumptions.

I know that MEGA can't do this. While MEGA does have something called consensus trees, this is actually just the outcome of bootstrapping within a single method. Also, I realised I set you wrong before - MEGA can do maximum parsimony, but not likelihood.

Above link information are not able to open at Mothur server.....?

Dear Manoj Kumar, 

I noticed that a number of mothur's special websites are down two weeks ago. I don't know the reason, maybe they are under maintenance. Hopefully they will be back on very soon. thanks for reading.