A series of Ni2P/SiO2 catalyst samples with loading from 6 to 22 wt% Ni2P was prepared by the method of temperature-programmed reduction (TPR). The samples were characterized by BET, CO uptake, X-ray diffraction (XRD), and extended X-ray absorption fine structure (EXAFS) measurements. The activity of these catalysts was measured at 643 K and 3.1 MPa in a three-phase, packed-bed reactor for hydrodesulfurization (HIDS), hydrodenitrogenation (HDN), and hydrodeoxygenation (HDO) using a model liquid feed containing 3000 ppm S as dibenzothiophene, 2000 ppm N as quinoline, and 500 ppm 0 as benzofuran. An optimum Ni2P loading for HDS, HDN, and HDO activity was found around 18 wt% which gave an HDS conversion of 99%, an HDN conversion of 91 %, and an HDO conversion of 80% at a WHSV of 2.1 h(-1). These were much higher than those of a commercial Ni-Mo-S/Al2O3 catalyst which gave an HDS conversion of 76%, and an HDN conversion of 38% based on equal sites (70 mumol) loaded in the reactor. The sites were counted by CO chemisorption for the phosphide and by low-temperature O-2 chemisorption for the sulfide. XRD and EXAFS results confirmed that a Ni2P phase was formed on the support. The crystallite size increased slightly from 7.7 to 9.8 nm when the loading was increased, and there was evidence from CO chemisorption that some crystallite agglomeration occurred in the higher loading samples. It was found that the HDS performance did not change appreciably with Ni2P loading, but that the HDN activity and stability went through a maximum at intermediate loading. It was concluded that the HDS reaction is structure-insensitive, while the HDN reaction depends on the local arrangement of surface Ni, P, and S atoms, and is structure-sensitive. XRD and EXAFS analysis indicated that the active catalyst formed a surface phosphosulfide phase on top of a Ni2P core in the Course of hydrotreating, (C) 2003 Elsevier Inc. All riahts reserved.