We describe the preliminary evaluation of the frequency corrections and their uncertainty in the cesium fountain primary frequency standard (PFS) NMIJ-F2 under development at National Metrology Institute of Japan (NMIJ). In NMIJ-F2, cold atoms generated from a vapor-loaded optical molasses in the (001) configuration are optically pumped to the Zeeman sublevels of m F = 0 to increase the number of atoms involved in the Ramsey interrogation. Moreover, a cryocooled sapphire oscillator with ultralow phase noise is employed as the local oscillator to avoid degradation of the frequency stability due to the Dick effect. As a result, we have obtained a very high fractional frequency stability of 9.7 × 10 −14 τ −1/2 . As for systematic frequency shifts, the fractional correction for the second-order Zeeman shift is experimentally estimated to be (−165.5 ± 0.5) × 10 −15 from the first-order Zeeman shift of atoms in m F = +1 launched to various heights. The fractional frequency correction for cold-atom collisions is estimated to be (+3.3 ± 0.4) × 10 −15 by extrapolating the frequency to zero density from the frequencies measured for various nonzero atom numbers. We will soon be able to make a comparison with other atomic fountain PFSs at the 1 × 10 −15 level.Index Terms-Atomic fountain clock, primary frequency standard (PFS), uncertainty evaluation.