Hi Ahmed, Thank you for initiating the discussion. Actually I have not done anything. Things are being planned. Genetics is not my speciality but quite familiar with terms and some of the uses.
I guess the best way would be to use "neutral" markers (such as microsatellites, genes which do not show signal for selection or perhaps mitochondrial genes). That way, mutations probably have accumulated at a steady state during evolution and reflect the time that passed since the species you are studying have diverged.
If you get nucleotide data for those, you can use any kind of models for measuring genetic distance in the software MEGA (the models are well explained in the manual): http://www.megasoftware.net/
I also think that the more genes/ the longer sequences you analyze, the better your estimate will be. You could concatenate multiple genes prior to your analysis in MEGA for example.
Perhaps, others scientists here will have better suggestions which I will also be interested in reading.
I know I am telling you something that is not very informative, but I would answer your question as follows.
"To find the evolutionary history of a number of plant species, you need a sufficient number of evolutionary informative polymorphisms."
The issue is that it is very difficult to decide on the kind of genetic marker that gives you the best possible evolutionary tree within a defined budget (time and money) until you have done analyses of multiple markers and are able to compare their merits. So that does not help you much.
It depends on how closely related the species you are going to investigate are. There are some crude guidelines. For plant species, the chloroplast sequence can be very informative. But which locus do you need to amplify (rbcL, matK, trnL...) is not straightforward. There are many studies that have used chloroplast markers, and the results are not always conclusive for every species in the study. In most studies there are still certain species or groups of species that remain either unresolved because maybe not enough polymorphisms (different species can share same DNA sequence at some chloroplast loci sometimes), or at the other end, difficult to place because too much differences and parallel or consecutive or reverse mutations interfere with deducing a clear bifurcating evolutionary pattern. From the species you are interested in and what is known about them from morphological studies and genetic studies of related species, you can try to estimate what markers would be more informative. If the species are very closely related and have large diversity within the species, or if you expect that there might be related species that have undergone hybridization in the past, or undergone genome duplications (auto- or allopolyploidy etc) then you could even expect sharing of the same chloroplast haplotype among species and you may need to analyse some nuclear markers (maybe in addition to chloroplasts). Your evolutionary relationships may actually be network like rather than a bifurcating tree.
If the species belong to different genera, then most chloroplast markers will do... on condition they are not too different in sequence. You might need to sequence more than one chloroplast marker to find an acceptable solution.
In summary, DNA sequencing is definitely the way to go for meaningful phylogenetic anaysis (stay away from RAPDs or AFLPs if you can), chloroplast is a good candidate to focus on... but be open-minded and look out for potential problems.
I hope this helps you focus your search for a good marker system a bit.
Hugo already gave you a good and complete answer. And I agree in that you should stay away of polymorphic markers like AFLP, SSR, etc. to compare species. I have used matk and rbcl sequences in plants, which are good for bar code of life. There is a lot of literature around, but you may find this useful and enlightening: http://www.pnas.org/content/106/31/12794.short
There's a little bug in detecting evolutionary distances by coding sequences: that's convergent selection. An useful coding sequence could develop in two very different species just because it's the fittest for that environment. In orther to avoid this problem, someone suggests analysing some non coding sequence too (if possibile). Others suggest to consider synonimous/non synonimous substitution rate (synonimous substitution are certainly not subjected to selettive pressure).