Effective reactions have been discovered for ALD of many metal compounds, including oxides, nitrides, sulfides and fluorides. However, ALD of pure metals has proven to be more difficult. Tungsten and ruthenium are the only metals previously deposited from molecular precursors by self-limiting, complementary ALD reactions. ALD of titanium, tantalum, copper and aluminum used atomic hydrogen as a reactant, which limits the step coverage attainable in deep trenches because of the rapid recombination of hydrogen atoms. It is also difficult to design ALD apparatus to distribute hydrogen atoms over large areas, and to avoid plasma damage to substrates. We synthesized many new metal acetamidinates, a class of metal precursors that we designed to have the properties needed for ALD of metals. These compounds have the required high thermal stability because two metal-nitrogen bonds hold each ligand onto the metal (chelate stabilization). Nevertheless, each metal-nitrogen bond is individually weak (formally a bond order of only one half) and highly reactive. The high volatility of these metal compounds (vaporization temperatures less than 100 C) arises from the outer surfaces of the molecules being entirely saturated hydrocarbons. For example, homoleptic N,N'-diisopropylacetamidinato metal compounds and molecular hydrogen gas were used as ALD reactants to form highly uniform and conformal thin films of transition metals including iron, cobalt, nickel and copper. We propose that these ALD layers grow by a catalytic hydrogenation mechanism that should also operate during the ALD of many other metals. The process should allow improved production of many devices, such as copper interconnections in microelectronics, magnetic information storage, micro-electromechanical structures and catalysts. Use of water vapor in place of hydrogen gives highly uniform, conformal films of metal oxides, while ammonia gives metal nitrides.