"I used hatree fock method before ..is it really need to used DFT method? and basis set?"

Yes, it's very, very needed. Nowadays it's very hard to justify HF calculations, since we have a huge number of much better methods that are as fast as HF. A good reference, which shows benchmarks for many DFT methods (and comparison woth HF) is here: (http://comp.chem.umn.edu/info/docs/773.pdf

I'd like to use M06 (or M06-2X) functionals, which work well for many systems, but there is a few very good, other functionals - the accuracy may also depend on the system. Also, it is worth adding the D3 dispersion correction of Grimme (http://www.thch.uni-bonn.de/tc/index.php?section=downloads&subsection=DFT-D3&lang=english). Most QM programs have these implemented. As for the basis set, the standard today for 50-100 atoms systems is a double-zeta basis set with polarization and diffuse functions for geometry optimization and triple-zeta basis set for energies.

Dear Thongam,

You have not provided how many atoms your system has. Generally, Hartree-Fock method is not capable of studying intermolecular interaction due to ignorance of coulomb correlation, so you should use post Hartree-Fock method. Also, generally the current fucntionals are having problem with long-rang part of dispersion energy. Also, Grimme correction (+D3) adds a number to interaction energies as dispersion, but physically it is not real dispersion energy, but in practice it can be a recourse if you are using DFT functional, Also, in usage of Gimme correction one should take care of double-counting, especially when one wants to use Minnesota functionals from Truhlar like M06-2X. I prefer to use correlated method, the golden one is CCSD(T), but  in most of cases it is out of reach. The cheapest correlated method is Moller-Plesset truncated at second-order (MP2), but it is not recommended for conjugated systems and pi-pi stacking interactions because it overestimates interactions energy(because MP2 accounts for dispersion part of interaction  in an uncoupled level), so we can use the proposed methods which deal with this deficiency of MP2, like SCS-MP2,MP2C or SOS-MP2. Another key factor to account for dispersion energy correctly is basis set. Dispersion lies in correlation part, so basis set used should be included diffuse functions. Also, dispersion part of interaction has shown slow convergence behaviour with basis set, so you should use in a type of basis set, the basis set with larger cardinal number, at least triple zeta. If you vote for correlated method, it is necessary to choose proper basis, if you apply frozen core approximation and do not include core electrons in correlations, so you can use dunning basis set (aug-cc-pVnZ n=D.T,Q,..), but if you desire to include core electrons, I vote for aug-cc-pCVnZ, or aug-cc-pwCVnZ.

Best wishes,

Sirous 

Dear Thongam,

If you study large molecules than you are limited in the choice by the DFT-D and MP2. However, as Sirous said, DFT-D provide not a real dispersion energy.

If you have a saturated compounds, it is best to use MP2.

If unsaturated, the choice is not good. SCS-MP2 give a better energy to the unsaturated moiety and equally worse for the saturated. Therefore the choice of SCS-MP2 questionable.

If you can confine 4-5 atoms (excluding hydrogen), then you can consider an accurate coupled clusters (or MP4). However, the aug-cc-pVnZ little help. With their to obtain reasonable dispersion energy n must be at least 5, which is unrealistic.

The way out - to use the mid-bond functions.

I work on the personal computer with F12/RI-CCSD(T)/cc-pVTZ-bf

Best wishes,

Alexei