We present a scanning tunneling microscopy study of the direct comparison between homoepitaxial deposition and surface ion sputtering on the Ag(001) system. At a temperature of 200 K, sputtering results in mound formation similar to the epitaxy case, while at higher temperatures an erosive regime sets in with the appearance of regular square pits. Contrary to the conventional wisdom, which considers ion sputtering as a deposition of vacancies, the analysis of single ion impact events reveals

more »

... t the process produces both adatom and vacancy clusters. The key parameter determining the temperature dependence of surface morphology turns out to be the mobility of the adatom clusters which exceeds that of vacancy clusters. Ion sputtering, i.e., the bombardment of surfaces by means of energetic ionized particles, is used in a large number of experimental techniques for the analysis and preparation of solid interfaces. For example, sputtering is employed for depth profile analysis in secondary ion mass spectroscopy and has been shown to improve the quality of thin films if used in combination with molecular beam epitaxy [1] . Recently, it was demonstrated that ion bombardment at low energies is also a powerful tool for surface nanostructuring: By simply changing a few experimental parameters such as substrate temperature or sputtering geometry, a large variety of self-assembled periodic structures was reported on metals [2], semiconductors [3] , and amorphous materials [4] . According to this, GaSb quantum dots were produced by means of Ar 1 sputtering [5] and showed quantum confinement effects marked by photoluminescence properties. In spite of the wide spreading of surface applications based on ion bombardment, a detailed understanding of the time-extended surface sputtering process is still lacking. In particular, while the effects of ion irradiation of solid materials have been extensively studied at the bulk level, surface modifications received only minor attention, being mainly focused on the damage process caused by the ion-substrate interaction, which ends in the first tens of psec after the impact [6, 7] . On the other hand, especially for what concerns surface nanostructuring, a great importance is held by the later evolution of the surface, since the final morphology depends on how the surface defects created by the ion impact diffuse on the substrate and interact with each other. A continuum theory for surface sputtering has actually been developed [8] but, obviously, it gives only a coarse-grained description of the process, completely neglecting the atomistic aspects which, on the contrary, are fundamental for the full comprehension and control of this phenomenon. A simplifying assumption is often made by considering the sputtering process as equivalent to a "negative deposition" where adatoms are replaced by single atomic voids (monovacancies) (see, e.g., [9, 10] ). Experimental evidences supporting this picture have been reported in the case of low energy ion bombard-ment of Ge(001) [11] ; moreover this statement has also been used in computational models which tried to simulate surface sputtering [12, 13] . In order to achieve a better understanding of the sputtering phenomenon and to directly test the presumed equivalence between ion sputtering and negative deposition, we performed a temperature dependent comparison between the morphology of an ionbombarded Ag(001) surface and the morphology which results from exposing the same surface to a thermal atomic beam. We chose this particular metallic substrate both for its simple crystalline structure and because surface diffusion is isotropic, which makes it an ideal model system. Moreover on this system single adatoms and vacancies have almost the same surface diffusivity [14, 15] , which in a simple guess should lead to specular surface morphologies if the negative deposition assumption were true. The experiments were made by means of a variable temperature scanning tunneling microscope (STM) housed in an ultrahigh vacuum chamber with a base pressure of 1 3 10 210 mbar [16] . Ne 1 ions with an energy of 1 KeV were used for sputtering, while an electron-bombardment Ag source was employed for homoepitaxial growth. The direction of the ions was chosen normal to the surface, so to eliminate from the final surface morphology any effect caused by the curvature-dependent sputtering yield [8] . In order to make the comparison between sputtering and deposition more meaningful, the atom deposition rate was chosen to be equivalent to the rate of total displaced material during sputtering. In particular, the ion flux (F sput 2.5 3 10 12 ions ? cm 22 ? s 21 ) was scaled down in respect to the deposition flux (F dep 8.8 3 10 13 atoms ? cm 22 ? s 21 ) by the total number of atomic defects created by each impact (discussed later). In both experiments the surface was exposed for the same time (t 450 s, corresponding to a nominal deposition of 33 monolayers) to the ionic or atomic beam and, immediately after turning off the particle source, the surface temperature was rapidly quenched to 130 K in order to avoid subsequent surface restructuring. For both types of experiments the final surface morphology is marked by quite regular three-dimensional 838 0031-9007͞01͞86(5)͞838(4)$15.00