Yes, hydration matters. If the lipids aren't sufficiently hydrated (and each lipid has a typical hydration number), then this can happen. If the water layer is not thick enough, you'll also run into periodicity artifacts if the headgroups "see" each other across the boundary.

Thanks Justin. Following your answer, i have water molecules over and above hydration number for my lipid (link attached below).I equilibrated the system for 5 ns and and analyse the thickness after 50ns of production run but still has different thickness values. Could there be any other issue that we can suspect?

http://www.ncbi.nlm.nih.gov/pubmed/15137088

Plenty of things - force field, unconverged simulation, proper run settings, whether or not the difference between the simulations is actually statistically significant, whether or not you're seeing a phase transition (look at order parameters), etc.

Many Thanks for your answer :)

Interesting discussion - that's why I still interrupt you. I have always thought that the system should contain preferably too much water than just a minimum, so that proper equilibration is not prevented at least because of the lack of water. I have often used as much as 36/lipid.   -marja-

Indeed, I would consider the bare minimum very risky.  Whether or not you need a large excess of water, I'm not sure.  I would imagine the results converge after full hydration.

Experimentally some changes are seen already with 20 water per lipid:

http://dx.doi.org/10.1039/B508190D

http://dx.doi.org/10.1016/j.bpj.2010.11.010

You can link the acyl chain order parameters to the bilayer thickness.

Behaviour in different simulations and experiments regarding the choline and glycerol backbone is discussed in the nmrlipids.blogspot.fi project, see

https://github.com/NMRLipids/nmrlipids.blogspot.fi/blob/master/HGmodelMANUSCRIPT/HGmodel_ACStemplate.pdf

There are lot of trajectories available through this project with different hydration levels (see citations in the publication). These may be helpful as well. Also Fig. 1 here may be interesting:

http://nmrlipids.blogspot.fi/2015/03/mapping-scheme-for-lipid-atom-names-for.html

Thanks Samuli for such an informative answer. Your references were detailing about bilayer deformation/phase change under different pressure conditions and hydration number.However the effect of hydration number was not much considerable beyond ~25 waters/lipid as against 5-10 waters/lipid range. Moreover we can intuitively expect the order parameter and bilayer thickness to be related. To summarize i can say from your and previous answers that for a fixed pressure (if we are simulating under NPT ensemble) and byond minimum hydration number of lipid, the bilayer thickness should not be a function of hydration number because byond this number hydration forces should neutralize the repulsive forces between the head groups (water acts as a glue between them) and we should get a constant value of bilayer thickness !