Density functional theory (DFT) has gained popularity because it can frequently give accurate energies and geometries. The evaluation of DFT integrals in a fully analytical manner is generally impossible; thus, most implementations use numerical quadrature over grid points. The grid-free approaches were developed as a viable alternative based upon the resolution of the identity (RI). Of particular concern is the convergence of the RI with respect to basis set in the grid-free approach. Conventional atomic basis sets are inadequate for fitting the RI, particularly for gradient corrected functionals [J. Chem. Phys. 108, 9959 (1998)]. The focus of this work is on implementation of and selection of auxiliary basis sets. Auxiliary basis sets of varying sizes are studied and those with sufficient flexibility are found to adequately represent the RI.