Hi Serdar,

I have slightly different opinion. In fact, it just depends on the level of accuracy you need but, in general, BLYP and B3LYP may not be good enough. At least, they're not designed to treat weak interactions such as H bonding. You should try truhlar's M06L for a meta-GGA or M06-2X for a hybrid. Or course newer versions of those functionals should provide similar accuracy. Alternatively, you can include the Grimme correction to your calculations with Gaussian (GD3BJ is the newest version). In the manual you can find how to do it: http://www.gaussian.com/g_tech/g_ur/k_dft.htm

Regarding the basis sets. Again it depends on the accuracy and on how large is your system. In any case, I'd recommend you to include diffuse functions. 

I'm sorry I can not provide any specific reference. These are things that I've learnt when working with molecules, specially organic molecules. Most of my assumptions are based on preliminar benchmark studies that have not been published as such, so I can not give you further details.

Anyway, I hope it helps.

Cheers

Sergi

Thanks Oleg. I'll check those references you gave.

Thanks Sergi. I'll reconsider my previous calculations by employing the level of theory you mentioned to see how it matters. 

My experience suggest that GGA is OK.

Hi Serdar,

you may also want to look at the works by Steve Scheiner (Utah State) as listed in his publication list attached: he's devoted a lot of his research to the H-bonding.

Hope it will be helpful, too.

Aleksey.

Thank you, Aleksey.