1)what is the use of internal standard in lcms analysis
To correct for fluctuations in sample introduction or detector response. The idea is that these fluctuations will affect the analyte peak and the internal standard peak in the same proportion, and so will cancel out when the ratio of peak areas is calculated.
2)what is the criretia of its selection
It should have similar response tot he analyte and elute at the same time or just after. The "best" internal standards are isotope-labelled analyte.
3)will it do to make standard curve without internal standard.
In many cases, yes. If the precision without using IS is good, then an internal standard can actually make things worse (having to measure two peaks can add error).