ABSTRACT The deep, wide-area (∼800–900 arcmin2) near-infrared/WFC3/IR + Spitzer/IRAC observations over the CANDELS fields have been a remarkable resource for constraining the bright end of high-redshift UV luminosity functions. However, the lack of Hubble Space Telescope (HST) 1.05 μm observations over the CANDELS fields has made it difficult to identify z ∼ 9–10 sources robustly, since such data are needed to confirm the presence of an abrupt Lyman break at 1.2 μm. Here, we report on the successful identification of many such z ∼ 9–10 sources from a new HST program (z9-CANDELS) that targets the highest-probability z ∼ 9–10 galaxy candidates with observations at 1.05 μm, to search for a robust Lyman-break at 1.2 μm. The potential z ∼ 9–10 candidates were preselected from the full HST, Spitzer/IRAC S-CANDELS observations, and the deepest-available ground-based optical+near-infrared observations (CFHTLS-DEEP+HUGS+UltraVISTA+ZFOURGE). We identified 15 credible z ∼ 9–10 galaxies over the CANDELS fields. Nine of these galaxies lie at z ∼ 9 and five are new identifications. Our targeted follow-up strategy has proven to be very efficient in making use of scarce HST time to secure a reliable sample of z ∼ 9–10 galaxies. Through extensive simulations, we replicate the selection process for our sample (both the preselection and follow-up) and use it to improve current estimates for the volume density of bright z ∼ 9 and z ∼ 10 galaxies. The volume densities we find are 5 and lower, respectively, than those found at z ∼ 8. When compared with the best-fit evolution (i.e., ) in the UV luminosity densities from z ∼ 8 to z ∼ 4 integrated to (−20 mag), these luminosity densities are and lower, respectively, than the extrapolated trends. Our new results are broadly consistent with the “accelerated evolution” scenario at z > 8, consistent with that seen in many models.