A quantitative estimation of metal-induced oxide charge (Qmi) density is performed on the surface of n-type silicon (Si) wafers rinsed with trivalent aluminum (Al)- and iron (Fe)-contaminated RCA alkaline solution by analyzing the frequency-dependent AC surface photovoltage (SPV). Qmi arises from (AlOSi)- or (FeOSi)- networks in native oxide which are responsible for inducing negative oxide charge. On the basis of Munakata and Nishimatsu's half-sided junction model [C. Munakata and S. Nishimatsu: Jpn. J. Appl. Phys. 25 (1986) 807], the network densities are estimated in depletion and/or weak inversion in which the cutoff frequencies of the frequency-dependent AC SPV curves are defined. It is found that the charge density Qmi increases with the time of exposure to air and it is calculated that about 4% of Al atoms in the native oxide are activated in the form of an (AlOSi)- network for 1 h of exposure. The (FeOSi)- network density is calculated as a function of Fe concentration. As a result, the frequency-dependent AC SPV measurements carried out here enable a successful evaluation of impurity level in a nondestructive and noncontact manner.