We present a technique for obtaining atomic resolution ultrahigh vacuum scanning tunneling microscopy images of diamond (100) films, and use this technique to study the temperature dependence of the etching of epitaxial diamond (100) films by atomic hydrogen. We find that etching by atomic hydrogen is highly temperature dependent, resulting in a rough and pitted surface at T ഠ 200 and 500 ± C, respectively. At T ഠ 1000 ± C etching results in a smooth surface and is highly anisotropic, occurring predominantly in the direction of dimer rows. This observation supports recent theoretical models that propose anisotropic etching as the mechanism for the growth of smooth diamond (100) films. DOI: 10.1103/PhysRevLett.86.3368 PACS numbers: 81.65.Cf, 68.37.Ef Diamond films grown using chemical vapor deposition (CVD) [1] have recently attracted considerable interest due to the unique mechanical and electronic properties of this material [2]. CVD growth of diamond films is not as well understood as CVD growth of Si and other semiconductors. Growth of diamond films requires a large ratio of molecular hydrogen ͑H 2 ͒ to hydrocarbon gas. During growth, H 2 is dissociated into atomic hydrogen by a hot-tungsten filament or microwaves. It is known that atomic hydrogen etches away graphite, and promotes diamond growth by terminating the diamond surface in an sp 3 configuration [1][2][3]. It is also known that atomic hydrogen etches diamond [1][2][3]. However, the effects of atomic hydrogen etching of diamond during growth are not well understood. Scanning tunneling microscopy (STM) in ultrahigh vacuum (UHV) has been extensively used to study the etching and growth of Si [4,5] and other semiconductors [5] at the atomic scale. To our knowledge, UHV STM has not been extensively used to study diamond. Most STM studies of diamond have been performed in air where sample preparation is limited to room temperature techniques and usually a hydrogen-terminated surface [6][7][8]. A recent study reported that STM was not possible in UHV because the diamond surface was not conductive enough until it was exposed to air [7]. One would like to see results of UHV STM of diamond comparable in quality to those that have been reported for Si.Epitaxial diamond (100) films are of particular interest because such films grow smooth at substrate temperatures T ഠ 1000 ± C [9], unlike epitaxial diamond (110) and (111) films that grow rough. The growth of smooth diamond (100) films is not well understood. Surface diffusion of adsorbates, which is responsible for the growth of most smooth films, is not widely accepted as responsible for the growth of smooth diamond (100) films because the hydrogen-terminated surface prevents diffusion [3]. Diffusion on the diamond (100) surface during growth has been modeled as occurring along surface sites where hydrogen has been removed by abstraction [10]. However, recent theoretical models have shown that the addition of surface diffusion to models would lead to higher growth rates than are experimentally observed [11...