surface science reports ELSEVlER Surface Science Reports 28 (1997) 1777245 Elastic and inelastic processes in the interaction of l-10 eV ions with solids: ion transport through surface layers Mustafa Akbulut, Norbert J. Sack, Theodore E. Madey * Department of Physics and Astronomy and Laboratoryfor Surface Modification, Rutgers, The State University of New Jersey, Piscataway, NJ 08855, USA Manuscript received in final form 5 March 1997 Abstract We review the escape.depth of secondary ions (or neutrals) desorbing from solid surfaces under the impact of electrons, photons or ions. We survey ion (or neutral) transport through many materials, but most are wide band gap insulators such as rare-gas solids and molecular solids. We address the issue of low-energy (< 10eV) ion-solid interactions, and review experimental and theoretical studies that provide insight into the physical mechanisms of these interactions, such as elastic scattering,charge transfer and ion-molecule reactions. Although it is usually assumed that most of the secondary ions stem from the top surface layer, we show that this is not necessarily the case: In certain instances, 1-1OeV ions are able to transmit solid films which are several monolayers thick. The transport oflow-energy ions through materials has very broad implications. We point out the importance of these results for electron or photon stimulated desorption (ESD/PSD), secondary ion mass spectrometry (SIMS), and ion-sputtering of surfaces, and discuss their relevance to other fields, such as ion beam deposition (IBD), low-energy ion implantation, and electrochemistry. 1. Introduction Electron, photon or ion bombardment of solid surfaces can lead to the desorption of secondary particles, including ions or neutrals. Analysis of the secondary particles and their kinetic energy and angle of desorption has proven useful in the identification of the structure and elemental and chemical composition of the surface from which they desorb. * Corresponding author. Fax: + 1908-445-4991; e-mail: [email protected]. 0167-5729/97/$32.00 0 1997 Elsevier Science B.V. All rights reserved ‘PII SOl67-5729(97)00007-l 182 M. Akbulut et al. 1 Surface Science Reports 28 (1997) 177-245 Secondary particles ( < 10 eV) sputtered during ion bombardment of solid samples are utilized in methods such as secondary ion mass spectrometry (SIMS) or secondary neutral mass spectrometry (SNMS) [l]. The identification of the mass of secondary ions allows insight into the elemental composition of the surface layers. Electron/photon stimulated desorption (ESD/PSD) are based on the principle that energetic electron/photon beams (a few eV to several thousands of eV) incident on a solid can cause desorption of ions and neutral species (including metastables) from the surface layers of the solid by inducing electronic transitions to dissociative states [2]. (Mechanisms of ESD/PSD are discussed in Appendix C in detail.) These beam damage processes are termed desorption induced by electronic transitions (DIET). The most probable range of kinetic energies of ESD and PSD ions is 1-1OeV. In ESDIAD (ESD ion angular distribution) one not only measures the ion desorption yield, the ion mass and the ion kinetic energy distributions but also the angular distribution of the desorbing ions [3,4]. This provides information about the geometric orientation of the chemical bonds that are broken by the electrons and therefore about the geometric orientation of the atoms and molecules on the surface. It is clear that the secondary ions must come from the surface layers. However, the question of the exact depth of origin of secondary ions has until relatively recently not been addressed in detail. Do most of the ions stem from the top surface layer as it is often assumed [1,2,5], or can some ions which are produced below the surface traverse the surface layers and escape from the surface? The physics behind these questions is the physics of the interaction of ions having kinetic energy in the range 1-1OeV (low energy ions) with solids or thin films. l-1OeV ions generated in a layer beneath the surface can interact with the topmost layers through various elastic and/or inelastic processes. The elastic and inelastic processes that influence ion transport through ultrathin films depend on the nature of the ion and its kinetic energy, and on the structure and electronic properties of the surface materials. Although there exists a wealth of information on ion-solid interactions at collision energies higher than 1 keV [6], the interactions of l-10 eV ions with solids (such as rare gas and molecular solids) are not known in detail. It is commonly assumed that both kinetic energytransfer and charge transfer neutralization are the dominant types of interactions in the low collision energy range. Low-energy ion-solid interactions can be rather complex because the cross-sections of the elastic and inelastic collision processes depend strongly on the details of the interaction potential between the collision partners. Hence generalizations that can be made for higher-energy collisions (e.g. keV-MeV) cannot be applied in this energy regime. The transport of low-energy (l-10eV) ions through atomic and molecular layers is of great importance in many diverse areas such as electrochemistry, radiation chemistry and physics, low-energy ion implantation and ion beam deposition [7]. Irrespective of the source of ions, in all low-energy ion transport processes through ultrathin films, the main physical mechanisms of elastic and inelastic scattering should be basically the same. In this review, we survey data on ion and atom transmission through a variety of insulators, semiconductors and metallic layers. Since most of the available data are for ions passing through rare-gas solids and molecular films, this review deals mainly with the question of the escape depth of secondary ions from wide band gap insulators. In order to address the escape depth of secondary ions from solids, we have developed a novel experimental approach in our laboratory. The essence of our experimental approach is illustrated in Fig. 1. We generate a source of atomic ions (e.g., O+, H+, F+, F-, Cl+, Cl-) with k nown kinetic energy and angular distribution by bombarding an appropriate surface with a focused electron beam, and causing electron stimulated desorption (ESD) (see M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 183 e- beam (-300 eV) l-10 eV ion (Of I F+I F-, H+) substrate as an ion source (e.g., oxidized W(lOO),l-MLPF3/Ru(OOOl)) Fig. 1. Illustration of experimental approach in Section 4.1. An ion is desorbed by ESD from a substrate, and overlayers are adsorbed onto this substrate and the ion yield is monitored as a function of overlayer thickness. Appendix C for more detail). We then cover the surface, cooled to m 20 K, with an ultrathin film of a condensed atomic or molecular solid (Ar, Kr, Xe, NH,, H,O, etc.). During electron bombardment of the layered surface, we measure the changes in the intensity, kinetic energy distributions, and angular distributions of the ion beam as a function of film thickness. ESD-produced ions are detected by means of a high sensitivity digital ESDIAD detector that allows mass, energy and angle-resolved ion detection. Except for metastable species, escaping neutrals are not detected. By measuring the changes in total intensity, angular distribution and kinetic energy distribution of ions that originate at the substrate beneath an overlayer film, as a function of film thickness, we can gather much information. The change in total ion intensity is expected to be due mainly to 184 M. Akbulut et al./ Surface Science Reports 28 (1997) 177-245 backscattering and large angle scattering (> 90” with respect to the surface normal) as well as charge transfer neutralization and ion-molecule reactions leading to formation of new species (for molecular overlayers). The changes in ion angular distribution and energy distribution are expected to arise from elastic and inelastic scattering in a forward direction. In studying ion transport through surfaces, we address the above-mentioned questions concerning 1-1OeV ion-solid interactions. A comparison of the velocity of the outer shell electron of a target atom (u,) with a projectile ion velocity (u,) is very useful in determining the interaction potential during a collision. When u, >>v, (slow collisions) the collision is called adiabatic and the collision can be described using quasimolecular potentials formed by the colliding partners during the collision. On the other hand, when u, >>u, (fast collisions) the electronic distributions of the colliding partners are static, except for abrupt transitions that occur when the particles are at their closest approach. Hereafter, in this review, we define the energy ranges of ions as follows: (a) ultralow energy ions: 0.0251 eV, (b) low-energy ions: 1 eV to the energy at velocity u, > u, (where u, is the velocity of the outer shell electron of a target atom of order lo* cm/s, and u, is the projectile ion energy); and (c) high-energy ions: u, > u,. This review is organized as follows: we discuss ion attenuation mechanisms in ultrathin films in Section 2. In Section 3-5, we discuss a selection of experimental and theoretical results related to the question of the escape depth of neutrals and ions. We present implications for the interpretation of ESD, PSD and SIMS measurements, and an outlook to neighboring fields in Section 6. Conclusions are given in Section 7. We discuss basic physics of elastic and inelastic atomic collisions in Appendix A, and charge transfer processes between ions and solids in Appendix B. We survey ion desorption from surfaces in Appendix C. 2. Low-energy ( < 10 eV) ion attenuation mechanisms in ultrathin films Energetic electrons, photons (few eV to several keV) or ions (usually higher than 1 keV) can penetrate a solid and create secondary particles (such as ions, neutrals) on or below the surface via electronic excitation or momentum transfer processes (see Appendix C for details of desorption processes). For DIET of an adsorbed monolayer on a metal surface, all desorbing species (ions and neutrals) originate from the adsorbed monolayer, because substrate electronic excitations are quenched too rapidly for desorption of bulk atoms. For compound materials such as oxides and fluorides, subsurface layers might contribute to the desorption signal. The desorption signal from an atomic or molecular multilayer film might also include contribution from subsurface layers. This is the main focus of this review: from what depth do desorbing species originate? For ions created below the surface, their desorption probability depends strongly on their interaction with the surface layers on top. As an energetic ion created below the surface enters the surface layer region, its trajectory and charge state are determined primarily by a series of elastic and inelastic collisions between the projectile ion and target atoms in the surface layers, and on the electronic properties of the projectile and the surface layer. The probability that an ion created below an ultrathin overlayer film survives and exits the film as an ion depends on both the energy loss and charge transfer reactions between the desorbing ion and target atoms in the film. Elastic scattering between a projectile ion created below the surface and target atoms in the overlayer leads to changes in the trajectory and kinetic energy of the ion. In an experimental- M. Akbulut et al. /&r&ace Science Reports 28 (1997) 177-245 185 measurement ofion escape probability, those ions that are elastically backscattered or scattered with such a large angle that they cannot escape from the surface are not detected. In the case of scattering by an angle of 90” or more with respect to the surface normal, an ion can undergo reneutralization by resonant electron tunneling or Auger neutralization via coupling to the substrate density of states and finally, either desorb as a neutral or become trapped in the overlayer film. If an ion is scattered in a forward direction (by an angle of < 90” with respect to the surface normal) through the overlayer, it may escape from the surface as an ion. Charge-transfer and ion-molecule reactions are expected to influence low-energy ion transport through ultrathin films by reducing the ion yield. Note that we use the term ion-molecule reaction to refer to a chemical reaction leading to formation of new molecular species. By a charge-transfer reaction we mean a reaction in which an electron is transferred between projectile and film. If projectile ions, in their passage through a molecular overlayer, interact with the overlayer molecules via charge transfer or chemical reactions to form new species [S], this results in a decrease in the projectile ion yield. Elastic scattering of l-10 eV ions by weakly bonded thin films can be explained in terms of a very simple model of binary collisions. In this simple model, each of the atoms or molecules in the films is considered to be free and at rest; the lattice binding energy and vibrational motion of the target atoms or molecules are neglected. Although this is an oversimplified assumption of collision because it neglects the influence of the neighboring atoms, it is a very useful model. Since virtually all measurements described in Section 4 are for condensed films of insulating solids which are weakly bonded, and the collision time (10-‘5-10-‘4s for 1-1OeV for F+, O+ ion) is smaller than the characteristic period for lattice vibrations (- lo- ’ 3 s), the binary collision approximation should be valid, at least for the initial collision(s). If the ion kinetic energy is ultralow (< 1 eV), the interaction of a desorbing ion with the lattice vibrational modes or molecular vibrational modes can become important, because the vibrational period of the atoms or molecules in the films can be comparable with the collision time. The ion can lose a small fraction of its energy due to the lattice vibration. In order to take the lattice vibration into account, the velocity of the target atom is included in the collision kinematics. The influence of the target velocity on the final energy of the projectile ion can be seen as a shift in the measured ion elastic energy as a result of a binary collision and cause energy broadening, depending on the temperature of the solid [9,10]. The effect of lattice vibrations on ion scattering can be estimated from the ratio of the shift in the final elastic scattering energy (6E) to the final ion energy (E,), 6E/E,. For example, for ion impact energies larger than N 1 eV, the lattice vibrations have a negligible effect on the scattering at temperatures between 25 and 300K, i.e. dE/E, z k,T/E, z 2 x 10e3-2 x lop2 for initial ion kinetic energy E, = 1 eV. Therefore, in ion attenuation processes the effect of lattice vibrations on ion scattering is important only if the ion kinetic energy is ultralow (< 1 eV). In a dipolar medium such as water or water ice, it is also possible that a moving ion can lose energy to dipolar relaxation. If the ion loses sufficient energy due to multiple elastic and inelastic processes without charge transfer and chemical reaction in the films, the ion may not escape from the overlayer films; it may become trapped in the overlayer films. This leads to a decrease in the ion yield. Because most of the processes we consider are based on binary collisions between projectile ions and target atoms of the wide band gap insulators, we discuss the basic physics of atomic collisions, binary elastic collisions, one electron charge transfer processes, and ion-molecule reactions elsewhere in this review (see Appendix A). M. Akbulut et aLlSurface Science Reports 28 (1997) 177-245 i86 -.....-.. Poisson layer-by-layer thickness (ML) Fig. 2. Ion attenuation in a film of thickness < 5 ML. Comparison of Poisson statistics (corresponds to a Poisson distribution of the film thickness) to a layer-by-layer film growth model. 2.1. Statistics of ion attenuation in thinjilms In the following we assume that ion transport through a film is determined by a series of binary collisions during which changes in trajectory and charge transfer can occur. In this case, ion transport can be described by Poisson statistics: a@ dx- -No@, (2.1) with CDbeing the ion flux at position x in the film, N the atom/molecule number density (per unit volume) in the film, and crthe attenuation cross-section. This leads to an exponential attenuation of the ions in the film, as depicted in Fig. 2 (solid curve): CD= QOexp( - Nod), (2.2) where d is the film thickness. This result assumes that the attenuation cross-section of the ions is independent of film thickness. The Poisson statistics are based on a random distribution of the atoms/molecules in the film, i.e., the film thickness is not uniform. However, a film may grow in a non-Poisson way, such as a layer-by-layer growth for which the individual layers can be ordered (crystalline) or disordered. For layer-by-layer growth, the attenuation of an ion signal in the film is expected to be linear as a function of atom density within one layer, as shown in Fig. 2, provided that the atomic separation (radius) is larger than the collision radius measured from the experimental attenuation cross-section. The ion transmission signals at each completed layer still lie on an exponential curve, but within one M. Akbulut et aLlSurface Science Reports 28 (1997) 177-245 187 layer the attenuation is linear. Therefore, for very low attenuation, the attenuation cross-section can be derived using the exponential attenuation mode (Eq. (2.2)). When the attenuation cross-section is very large (a >>l/N& atom separation less than collision radius), the effective interaction areas of neighboring overlayer atoms overlap with increasing coverage, even in a single monolayer. This leads to an exponential attenuation of the ion signal within one layer for both statistical and layer-by-layer growth modes. Note that there are similarities between the statistics of ion transmission through thin films and that of Auger electrons emitted from substrates and transmitted through adsorbates [ll]. If the adsorbate growth is statistical, then the Auger electron signal from the substrate is attenuated exponentially by the adsorbate. If the adsorbate grows in a layer-by-layer fashion, then the attenuation curve shows kinks at each completed layer. Attenuation can be caused entirely by elastic scattering effects with little or no charge transfer, or charge transfer can dominate. Consider first what happens in a gas phase charge transfer process. In a gas phase one-electron charge transfer reaction between a positive ion and an atom or molecule, the ion captures an electron from the target atom or molecule, and becomes neutral. In the case of the negative ion-atom/molecule interaction, the negative ion becomes neutral by transferring an electron to the target atom or molecule. The energy defect, which represents the change in the total internal energy in a charge transfer reaction, is an important parameter governing the charge transfer at low energy. Experimental results and theoretical calculations indicate that the smaller the energy defect, the larger the charge transfer probability [12,13]. The physics of charge transfer in ion-atom/molecule collisions is discussed in detail in Appendix A. For ion transmission through overlayer films near a conducting surface, the valence levels of both the desorbing ions and the target atoms or molecules of weakly interacting ultrathin film are shifted by the surface image potential [14]. However, we expect that the energy difSerence between the relevant levels does not change significantly near the surface. Therefore, we believe that an understanding of gas phase charge transfer processes is useful to explain a charge transfer reaction between a desorbing low energy ion and weakly interacting film. In a thick overlayer film (2 3 ML) a band structure can be well developed [l&16], and one might consider that this could make the direct application of inelastic binary ion-atom/molecule collision concepts to the transport problem inappropriate. In the wide. band gap insulating films, the electronic excitations are highly localized, because the electronic excitation energies are much larger than the cohesive energies. Hence, in first approximation, we expect that gas phase ion/atom charge transfer concepts can still be useful to describe a charge transfer reaction during ion passage through a thick insulating film. 3. Depth of origin of desorbing neutrals Although the focus of this paper is the depth of origin of secondary ions from solid surfaces, we address briefly the depth of origin of desorbing neutrals. It can be expected that the depth of origin of a neutral, e.g. 0, differs from that of an ion, e.g. O+, because the interaction potentials are different. This will affect the elastic scattering and energy loss of the different species in the surface layers; furthermore, inelastic scattering cross-sections, such as charge transfer, will be very different for the ion and the neutral. Also, the ion escape depth is a lower limit on escape of M. Akbulut et aLlSurface Science Reports 28 (1997) 177-245 188 0 1 2 3 4 5 6 7 8 9 10 14.5 Ar intemwdiote spacer thickness d I nm Fig. 3. Attenuation of Kr,F fluorescence intensity as a function of Ar intermediate spacer layer. From [17]. all particles, because species which begin as ions may still escape from the surface as neutrals after charge transfer. Very recently, Bressler and Schwentner [17] have reported a long range migration (mean penetration depth of 2.8 nm) of photomobilized - 4.2 eV F atoms in an Ar film. As shown in Fig. 3, they used a sandwich of three films: the top rare gas film Ar was doped with F,, the intermediate film is pure Ar, and the third film contains Kr that was used for detection. They generated F atoms by F, photodissociation in the top layer and monitored the migration of the F atoms across the Ar spacer layer with variable thickness by measuring the intensity of fluorescence from Kr,F formed at an Ar/Kr interface. Although the authors did not discuss F atom transport mechanisms (such as elastic and inelastic collisions) through the Ar film, it is expected that a - 4.2 eV F atom can interact with the Ar layers through various scattering processes. We compare this result with - 4eV F+ and - 1 eV F- ion transmission through ultrathin films of Xe, Kr and H,O in Section 4. In the following we point out some experimental sputtering measurements and theoretical studies which address the depth of origin of secondary neutrals. Dumke et al. [18] studied the sputtering of gallium-indium alloys with 15 and 25 keV Ar+ ions. The principle underlying these experiments is that the surface layer of a gallium-indium alloy (composition: 16.5% indium, 83.5% gallium) is enriched with indium: 94% indium vs. 6% gallium. Hence by measuring the indium to gallium ratio of the sputtered atoms one can derive the surface layer contribution to the total neutral sputter yield (if the sputter yields of pure gallium and indium are known). They conclude that for 15 keV Ar+, 85% of the sputtered neutrals stem from the surface layer, while for 25 keV At-+ it is only 70%. The authors do not comment on the difference. We suggest that this may be due to a difference in beam damage caused by the different energy ions in the surface layers, which may allow a different amount of neutrals from subsurface layers to escape from the surface. Burnett et al. Cl93 determined the depth of origin of neutrals from a Cu overlayer (< 2 ML) on Ru(000 1) by 3.6 keV Arf ions. By measuring the ratio of sputtered Cu to Ru neutrals M. Akbulut et al. /Surface Science Reports 28 (1997) 1777245 189 using the SARISA method (surface analysis by resonance ionization of sputtered atoms) they could derive the amount of Ru atoms that can desorb through the overlayer. They conclude that N 66% of the sputtered particles come from the top surface layer. Their experimental data are in agreement with the results they obtain from the computer simulation program “TRIM” POIHarrison et al. [21] performed classical computer simulations on the ion bombardment of Cu(1 00), Cu(1 lo), and Cu(1 1 1). They find differences in the sputter yield for the three crystallo- graphic orientations. More important for the present paper, they find that the majority of the sputtered atoms originate in the top layer ( > 99% for the (10 0) face, 81% for the (1 10) face, and 96% for the (1 1 1) face). The differences are explained by the different structures of the three faces: The (1 10) direction has the most open structure of the top layer and therefore also the largest percentage of sputtered particles from subsurface layers. In the experimental studies mentioned above, the attenuation of the neutrals is dominated by elastic scattering, or “blocking”. A theoretical study concerning the depth of origin of sputtered atoms has been performed by Vicanek et al. [22]. The authors use a Monte Carlo simulation, and assume an isotropic initial angular distribution and an E- 2 energy distribution without an upper or lower cutoff. They conclude that the most important factor influencing the escape depth is the elastic energy transfer from the low-energy atoms to the surface atoms. The effect of angular scattering on the escape depth of atoms is found to be rather small. A round robin computer simulation published by Sigmund et al. [23] compared the results from various simulations of the ejection of low-energy copper atoms ( < 50 eV) through a planar copper surface. In most cases, a Born-Mayer repulsive interaction potential, called Gibson 2 was used: V(r) = Aexp( - r/a), (3.1) where A = 22.5 keV and a = 0.196 A. This potential was augmented where necessary by an attractive interaction for distances, r, around and above the interatomic distance. Some of these results are shown in Fig. 4. Although it is shown that some energetic atoms can traverse a surface layer several A thick (- 5 A), most of the sputtered atoms are found to originate in the surface layer. As seen in Figs. 4(e)-4(f), the Cu atom ejection probability is very low for Cu atoms having kinetic energy less than lOeV, and it increases slowly with increasing atom energy up to 50 eV. Recently, Lill et al. [24,25] have measured abundance distributions of ionic and neutral clusters sputtered by 4 keV Ar ion bombardment of liquid gallium, liquid gallium-aluminum eutectic alloy and gallium-indium eutectic alloy by time-of-flight mass spectrometry. They have found that the depth of origin of sputtered clusters increases with increasing cluster size. Their data indicate that for small clusters (I containing 3 atoms) the ejection region growth is lateral, whereas for large clusters (2 containing 3 atoms) the ejection region growth is vertical in the surface. The escape depths of neutrals are of the same order of magnitude as those obtained for ions in cases where elastic scattering dominates the attenuation. The sputtering studies described above reveal that roughly 30% of the desorbing neutrals originate below the top monolayer (i.e., 70% of the neutrals originate in the surface layer). Similarly, from the results discussed below in Sections 4.1.2 and 4.1.4 we estimate that for many systems roughly 50% of the ions transmit 1 ML overlayer. M. Akbulut et al./ Surface Science Reports 28 (1997) 177-245 190 i 7 +yIi I 6 x(A) P p I (4 05 06 01 01 01 x(A) Fig. 4. (a) -(d) Depth dependence of ejection probability from a random Cu target (from [23]). The straight line is inserted for orientation (see [23] for detail).(e) -(f) Energy dependence ofejection probability from a random Cu target for(e) x = 0 (surface layer) and(f) x = 1.8 8, (second layer). In (e) the straight line represents P, = 0.5 for E 2 U and 0 otherwise, valid for a spherical barrier in the absence of collisions (see [23]), and the solid curve represents the expression P, = )[ 1- ( U/E)1’2], where U is the binding energy of the target atom, E is the kinetic energy of the ion. 4. Depth of origin of desorbing ions: experimental results There have been a number of experimental studies investigating the depth of origin of secondary ions from solids that are mainly wide band gap insulators. In each of the following sections we first introduce the experimental approach and then discuss selected results. M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 191 4.1. Ion desorption from substrates through overlayers 4.1 .I. Principles The approach used in the authors’ laboratory is illustrated in Fig. 1. The basic idea is to desorb ions from a substrate by ESD, and to condense overlayers (of a different material from the substrate) on top of the substrate which are nearly transparent to the primary electron beam. The yield, energy distribution, and angular distribution of the substrate ions are then monitored as a function of overlayer thickness. There are a few principal assumptions underlying this approach. First, the substrate and the overlayer have to be chosen so that the adsorption of the overlayer does not lead to a mixing of the two components. This is achieved by choosing a “rigid” substrate (such as an oxide) and an overlayer that is only weakly bound to the substrate. To date, most overlayers used have been physisorbed rare gases and condensed molecular layers that are weakly bound to the substrate. Another assumption is that the primary electrons can penetrate the overlayer film without losing a significant fraction of their energy. This is achieved by using electrons with a kinetic energy > 100 eV; in most solids, the energy loss is small enough that electrons can penetrate an overlayer film several monolayers thick without losing a significant portion of their energy [26]. It is also assumed that the primary electrons do not disturb the overlayer significantly in the vicinity where the secondary ion traverses the films, i.e., that the secondary ions do not traverse a strongly perturbed film. In order to keep radiation damage in the overlayer (sometimes referred to as “second order effects” see [l,p. 6721) to a minimum, low electron fluences (- 101”e/cm2) are used. However, this still does not rule out the possibility that the same electron that initiates one ESD process may 1ocalIy disturb the overlayer in the vicinity of the desorbing ion (“first-order effect”). But the total desorption cross-section from a condensed atomic and molecular thin film for Xe), so it is not expected that under > 1OOeV electron bombardment is low ( - 3 x lo-“cm2 disturbance of the condensed overlayer films by primary electrons (or secondary electrons) is a significant problem affecting the measured parameters [27-30-J. A further assumption is that the adsorption of the overlayer does not affect the ESD process on the substrate surface significantly (see Appendix C for mechanisms of ESD). This assumption may not always be valid; however, one can assume that such quenching at the substrate-overlayer interface is limited to the first overlayer monolayer only, so that there is true ion-thin film transmission for the second and following monolayers. When the above conditions are fulfilled, this is a good approach to study the interaction of 1-1OeV ions with thin films, and it can reveal information about the depth of origin of secondary ions. 4.1.2. Experimental procedures The experimental measurements of ion transmission through condensed films are carried out in an ultrahigh vacuum (UHV) chamber equipped with facilities for surface characterization; details of the apparatus and the methods used have been described previously [31]. The following experiments can be performed in this UHV system: (a) thermal desorption spectroscopy (TDS), (b) low-energy electron diffraction (LEED), (c) low-energy ion scattering (LEIS), (d) Auger electron spectroscopy (AES), and (e) electron stimulated desorption ion angular distribution (ESDIAD) of both positive and negative ions. 192 M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 TDS and ESDIAD are the two main techniques used in these studies. TDS is used to determine the overlayer coverage, while ESDIAD is used to measure total ion yield, energy and angular distribution of ESD ions as a function of the overlayer thickness. The sample, either a W( 100) or a Ru(000 1) single crystal in our studies to date, is mounted on an XYZ-rotary sample manipulator that is coupled to a closed cycle helium refrigerator to achieve a sample temperature of N 20 K. The sample can be heated in the range of 20- 1600 K by radiative heating and electron beam heating from a tungsten filament located behind the sample. The temperature of the sample is measured by using a Ni/Cr/Si-Ni/Si/Mg, N-type (generally known as Nicrosil-Nisil) thermocouple. The overlayer gases are deposited onto the substrate at 20K using a leak valve through a directional gas doser capped with a microcapillary array aimed at the surface [29,32]. This procedure allows precise dosing of various coverages of the gas, from fractional monolayers to multilayers. The exposure in units of Langmuir (L) is determined by measuring the dosing pressure with an uncalibrated ionization gauge located in the UHV chamber, and the dosing time. The coverages of the atomic and molecular overlayers are determined using TDS, which allows us to resolve the desorption peaks from the first monolayer and subsequent multilayers [29,30]. Monolayer identification is possible if the desorption peaks from the first monolayer and subsequent layers occur at different temperatures. Since the integrated area of the pressure vs. time trace is directly proportional to the number of gas atoms or molecules which desorb (coverage), the relation between coverage and exposure can be measured and calibrated. The ESDIAD/TOF (time of flight) detector allows mass, energy, and angle-resolved positive or negative ion detection [29,30]. Positive and negative ions can be generated by bombarding a surface with a focused electron beam (typically N 300 eV). The ESDIAD/TOF detector (Fig. 5) includes a set of four high transparency planar grids of which all except the second are grounded, a stack of five microchannel plates, and a position sensitive resistive anode encoder (RAE). The RAE is connected to a position analyzing computer to provide direct digital acquisition of two-dimensional data. By pulsing the primary electron beam, which provides a start pulse coincident with desorption from the surface, and by gating the retarding potential grid G,, we can perform TOF analysis of the desorbing ions. The TOF capability allows us to separate easily lighter ions (shorter flight times) from heavier ions (longer flight times). For appropriate bias and pulse conditions we can detect both positive and negative ions in a mass-resolved mode. Energy analysis of desorbing ions is measured in two ways: by use of a retarding field method in the ESDIAD/TOF detector and by careful analysis of ion signals in TOF measurements. In the retarding field method, the ESDIAD ion intensity is monitored as a function of retardation voltage applied to the grid G, to obtain an ion retardation curve. Upon taking the derivative of the retardation curve with respect to retardation voltage, an ion kinetic energy distribution is obtained for ions of known mass. The polar angle of detection (under field free conditions around the crystal) ranges from 0” to 22”. Application of a positive bias voltage to the substrate compresses the angular distribution of the desorbing positive ions so that even ions which desorb with angles substantially larger than 22” can be detected. For example, when a substrate bias of + 100 V is applied, the polar angle of detection is increased from O”-22” to O”-70” for a positive ion with a kinetic energy N 7 eV [29,30]. Hence, the application of a positive sample bias allows us to collect nearly the total desorption yield, but it makes the quantitative measurement of the angular and energy distribution of desorbing ions more M. Akbulut et al. 1 Surface Science Reports 28 (1997) G2 grids \ 177-245 193 /c--i Microchannel Plates (MCP / primary electrons / Resistive Anode Encoder (RAE) IIII IIII J III I crystal, bias: V, ions III I t +i lo I Fig. 5. ESDIAD >OeV. detector electro;; 1 : : G2 bias voltage and time of flight (TOF) arrangement. The ion trajectories are indicated for a substrate bias difficult. The position sensitive RAE also enables us to measure angle resolved yields over arbitrary regions. Fig. 5 shows the ESDIAD detector and TOF arrangement. Typically, the electron pulse length is 0.1 ps, the average electron current is 1 nA and the total electron fluence for a measurement is - 2 x 1Ol3cmd2 (beam area - 1 mm2) [29,30]. Different samples are used as sources for different ion beams. In our first series of experiments, we study the transmission of low-energy ESD-produced O+ through ultrathin atomic and molecular overlayers. In order to generate a well-defined ESD O+ ion beam, we use a W(100) single crystal; it is oxidized at 860 K in an 0, atmosphere (5 x lop6 Pa) for 10 min to produce a thin oxide film on the crystal [30]. The oxide surface prepared in this way exhibits a (1 x 3) LEED structure. Based on earlier studies, the stoichiometry of the surface is assumed to be WO,_, [33,34]. ESD of this surface produces a well-defined 0 ’ beam having a peak energy of - 7 eV and full width at half maximum (FWHM) of - 15” desorbing normal to the surface [35]. This is our reference O+ ion source. Fig. 6 shows the profile of the ESD O+ beam obtained under field free condition (zero sample bias) from the oxidized W( 100) surface measured with the ESDIAD/TOF detector. The inset of Fig. 6 is the O+ ESDIAD pattern from which the angular profile is extracted. In order to study transmission of F+ and F- ions through condensed layers, a chemisorbed PF, layer on a Ru(0001) surface is used to produce F+ and F- ion beams with well-defined energy and angular distributions [36,37]. The clean Ru(0001) surface exhibits a (1 x 1) LEED pattern. A saturation coverage of PF, (opF3 = 0.33 monolayer (ML)) is dosed onto the Ru(000 1) surface at lOOK; subsequently, the crystal is annealed at 270 K for a few seconds in order to produce a hexagonal array of off-normal ion emission under electron bombardment, for both F- and F+ 194 M. Akbulut et al. / Surface Science Reports 28 (1997) 100 200 300 400 177-245 500 position channel Fig. 6. Profile of the O+ desorption beam from oxidized W(10 0) surface measured with the ESDIAD detector. Primary electron energy: 300 eV. The inset shows the O+ ESDIAD pattern from which the profile is obtained. The periodic structure in the O+ yield is due to ion transmission through a metal grid. ions. The F+ ions desorb with a peak energy of - 4eV, while the F- ion kinetic energy distribution from adsorbed PF, has a peak at - 1 eV [38]. This surface exhibits a (fi x ,_/?)R30” LEED pattern. Afterwards the PF,/Ru(OOO 1) surface is slightly damaged under electron bombardment; ESD of this surface gives rise to a strong F+ emission (due to PF fragments) [36,37] normal to the surface, in addition to the hexagonal off-normal beams. Fig. 7 shows F+ and F- ESDIAD patterns from the slightly damaged PF,/Ru(OOOl) surface obtained with a bias voltage of + 160 and - 120 V, respectively. The F- ESDIAD pattern is unchanged by beam damage [36,37]. After these initial surface treatments, the slightly damaged surface remains stable throughout the measurements. A well-defined ESD H+ (D+) ion beam can be produced either from a bilayer H,O adsorbed on a Ru(000 1) surface [39] or a thick H,O film (> 7 ML) on Ru(000 1) [40]. ESD of the bilayer H,O adsorbed on Ru(000 1) produces an H+ (D+) beam having a peak energy of - 4 eV and a broad angular distribution (FWHM - 40”) centered on the normal to the surface. Surprisingly, H+ (D+) ions generated from the thick H,O film on Ru(0001) desorb in directions closer to the surface normal with a FWHM - 22” under field free conditions [40]. 4.1.3. 0 ‘transmission through rare gas$lms Sack et al. [29,32] have investigated the transmission of 7 eV O+ ions from an oxidized W(100) substrate through ultrathin overlayer films of Ar, Kr and Xe. They find that the ions can penetrate several monolayers of Kr or Xe, while < 2 ML Ar suppress the 0’ signal to -C 1% (Fig. 8). They also M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 195 (4 (b) F’lg. 7. (a) F’ ESDIAD pattern from slightly damaged 1 ML PF,/Ru(O 00 1) beam surface obtained with a sample bias of + 160 V (incident electron energy: 360 eV) and (b) F- ESDIAD pattern from a slightly damaged 1 ML PFJRu(000 1) beam surface obtained with a sample bias of - 120 V (incident electron energy: 200 eV). Note that the differences in the angular distribution of F+ and F- hexagonal beams obtained from the PF,/Ru(OOO 1) surface are due to field distortion and the differences in the sample mounting. find a change in the kinetic energy distribution towards lower energies for Kr or Xe films thicker than 2 ML (Fig. 9), and they present indirect evidence for a slight broadening (change) in the angular distribution of the desorbing Of ions for Xe films thicker than 2 ML. M. Akbulut et al./ Surface Science Reports 28 (1997) 177-245 196 i I I 4 5 rare gas coverage (ML) V m .“” I 0 h\ m 2 1 I 3 6 Fig. 8. Total O+ yield from an oxidized W( 10 0) substrate as a function of rare gas overlayer thickness. The lines are guides to the eye. From [32]. 1 .I ? =r 0.t 4 a z 0.E ‘& g ‘PI c; 0.4 & % 0.2 1, 2 II,,,I,,,,,,,,,I,,,,,,,,,,,,,,,,, 4 kinetic at erg) (eV) I 10 I I 1 1 12 Fig. 9. Energy distribution of the O+ ions from clean oxidized W( 10 0) surface and of O+ after passage through Xe films of various thickness (in monolayers), measured by a retarding-field method. Yields are normalized to unity at their maximum. From [32]. M. Akbulut et al. / Surface Science Reports 28 (1997) / 3. Iaye? 177-245 1. 197 layer 2. layer Fig. 10. Schematic structure of a fcc( 1 1 1) single crystal film. It can be seen that there exist channels perpendicular to the surface which are being closed by the third layer. They attribute the attenuation of the O+ ions in the Kr and Xe films predominantly to elastic scattering (see Appendix A.1 for the atomic collision discussion). The attenuation cross-sections for Kr and Xe (of order lo- ’ 5 cm’) agree well with a molecular dynamics simulation [41] (see Section 5); this simulation assumes only elastic scattering between the 0’ ions and the rare gas atoms, but no charge transfer. However, the simulations do not produce the strong attenuation observed for the Ar films, which is so far unexplained. Furthermore, the change in the attenuation of Of in Kr and Xe around 2 ML is correlated with the suggested rare gas structure: Bulk rare gas solids are known to be fee which means that in the (111) direction they show an A-B-C-A _. . layer structure [42] (Fig. 10). The authors conclude that the existence of channels allows the O+ ions to traverse the rare gas films with relatively small attenuation and energy loss up to 2 ML [29]. The third monolayer closes the channels and hence leads to increased attenuation and energy loss. Channeling leads to a lower attenuation cross-section of an ion in a crystalline solid than in an amorphous solid if the ion traverses the solid along one of the channeling directions. Steering means that an ion changes its trajectory from a non-channeling direction to a channeling direction. Hence, this study shows that the structure of the surface layers of a solid can influence the escape depth of ions from the solid significantly. If the surface layers are well ordered (crystalline) and if there are channels through which ions can escape from the surface, the ion escape depth may be larger than if the surface layers were amorphous. The angular distributions of the desorbing ions are expected to be affected by the structure of the surface layers. An alternative interpretation of Of attenuation by rare gas films has been presented by Ageev [43], who bases the attenuation on elastic backscattering and local Coulomb surface field relaxation. According to this model, an 0 + ion backscattered by an overlayer atom decelerates toward the surface and then accelerates from the surface by means of the same local repulsive Coulomb surface field that produces the ESD of Of from the surface. If the backscattered O+ ion .has sufficient time to attain kinetic energy exceeding the image surface potential it may desorb, M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 198 0.0 0.5 1.0 1.5 Coverage 2.0 2.5 3.0 (ML) Fig. 11. Total O+ yield from an oxidized W( 100) substrate as a function of H,O and NH, overlayer thickness. The lines are exponential curve fits. From [30]. otherwise it readsorbs. This model predicts quite well the experimental attenuation and kinetic energy distributions of Of ions in rare gas films. cross-sections 4.1.4. 0’ transmission through H,O, NH,Jilms Akbulut et al. [30] have studied the transmission of 7 eV O+, generated from an oxidized W(10 0) substrate, through ultrathin films of H,O and NH,. They find strong suppression of O+ by H,O (Fig. 11): Less than 1 ML H,O suppresses the O+ signal to < 0.1%. The attenuation cross-section for Hz’*0 is 9 f 2 x lo- l5 cm2; for NH, the cross-section is 3 k 1 x lo- l5 cm2. The authors suggest that the observed attenuation cross-section cannot be explained primarily by backscattering and large angle elastic scattering alone, for several reasons. First, NH, and H,‘*O have nearly the same mass and similarly large dipole moments, so that very similar Of attenuation cross-sections for both NH, and H,‘*O would be expected due to elastic scattering; this is not observed. Second, the measured cross-sections are too large to be caused only by elastic scattering (see below). They attribute the strong attenuation mainly to charge transfer to the O+ ions in the overlayer: O++H20-,0+H20+, (4.1) O++NH,-tO+NH; (4.2) (see Appendix A.3 for more details about charge transfer reactions in the gas phase). In thermal gas phase reactions [44] the cross-sections of reactions (4.1) and (4.2) differ by a factor 3, similar to the ratio of the cross-sections found in these studies. The absolute cross-sections at thermal energies are significantly larger than those at 7 eV collision energy as expected on the basis of the orbiting theory (see Appendix A.3). M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 199 The exponential attenuation of the ions even in the fractional monolayer range is explained by statistical growth of the overlayer films at 25 K, at which temperature the mobility of the overlayer molecules is strongly suppressed. Because the effective interaction area is much larger than the molecular size (based on intermolecular spacing), with increasing overlayer coverage the effective interaction areas of neighboring molecules overlap. This leads to Poisson statistics of ion attenuation in the overlayer film. Earlier studies by Diebold and Madey [45] revealed an attenuation cross-section for 5eV 0’ transmission through NH, similar to that for 7 eV O+ through NH, [30]. In that experiment [45], the oxygen ions are desorbed from a TiO, substrate. The fact that the O+ attenuation cross-section by NH, is nearly the same in these two experiments suggests that the experimental approach allows insights into ion-thin film transmission, since the results are independent of the substrate used as a source of 0 +. 4.1.5. Of transmission through alkali metal overlayers Very recently, Ageev et al. [46] have investigated ESD produced O+ transmission through fractional monolayer films of Li, Na, K and Cs on an oxidized tungsten substrate. Fig. 12 shows a semilogarithmic plot of the normalized ESD O+ signal as a function of alkali metal concentration. As seen in Fig. 9, the ESD Of signal is attenuated by the alkali overlayer: For alkali metal coverage < 0.3 ML the attenuation rate of the O+ signal is lower than the attenuation rate of O+ for alkali coverage > 0.3 ML. For alkali coverage > 0.3 ML, the attenuation cross-sections of O+ are (18&2)x 10-15cm2 for Cs, (13+1)x 10-15cm2 for K, (5+1)x 10-‘5cm2 for Na and (3.3 + 0.3) x lo- l5 cm2 for Li. In an earlier study, Yu [47] has investigated ESD O+ and 0 - yields from an oxygen chemisorbed Mo(100) surface as a function of Cs overlayer coverage, for thickness less than a monolayer. Yu has found that at low Cs coverages the 0 - yield increases with coverage, whereas the 0 ’ yield decreases slightly, and at high Cs coverages both ESD Of and O- yields are strongly attenuated. He suggested that shielding of the oxygen atoms from incident 200 eV electrons by elastic electron scatte?ing from o- -1 - ;” Z-2J- I 0 4 2 I I e a N.10"an" Fig. 12. Total O+ yield from an oxidized W( 10 0) substrate .C461. as a function of Li, Na, K and Cs overlayer thickness. From 200 M. Akbulut et al. /Surface Science Reports 28 (1997) 1777245 the Cs overlayer may be the reason for the high coverage attenuation. However, as discussed in Section 4.1.1, attenuation of > 1OOeV electrons by elastic backscattering in an overlayer film is expected to be very low [29,30]. Based on the above argument, Ageev et al. [46] do not believe that primary electron backscattering in the alkali overlayer has strong influence on the measured attenuation of O+ in their measurements. These authors also rule out the possibility of elastic scattering of ions as a dominant process in transmission of O+ ions through alkali metal films, because the measured O+ attenuation cross-sections are much too large to be realistic for elastic backscattering and large angle scattering, for which smaller impact parameters are necessary. However, there is a correlation between the ionization potential of the alkali metal and the O+ for attenuation cross-sections for > 0.3 ML, which is consistent with the resonance one-electron charge transfer model proposed by Rapp and Francis [48]. Hence, Ageev et al. attribute the attenuation of O+ in alkali metal films for > 0.3 ML mainly to resonance one-electron charge transfer from adsorbed alkali metal atoms to 0* 3S, and O*?S, excited states of oxygen with excitation energies of - 4.09 and 4.4 eV, respectively. 4.1.6. F+ , F- transmission through rare gas$lms Sack et al. [35,36,49] studied the transmission of F+ and F- ions from a monolayer of PF,/Ru(OOOl) through Kr and Xe overlayers. The kinetic energy of F+ is - 4 eV and that of F- - 1 eV [38]. The F+ ions are attenuated to < 10% by 1 ML of Kr or Xe (Fig. 13). The F+ angular distributions do not change significantly as a function of overlayer coverages. The measured attenuation cross-sections are - 1.4 x lo- l5 cm2 for Kr and - 2.6 x lo-r5 cm2 for Xe. The authors discuss the results in terms of elastic scattering, and perform a molecular-dynamics simulation which indicates that the attenuation may be dominated by elastic scattering. However, since the F+ angular distributions do not change significantly as a function of overlayer coverages, the authors also suggest that charge transfer reactions may contribute to the observed attenuation crosssections. rare gas coverage Fig. 13. Total F+ yield from a monolayer of PF,/Ru(OOO 1) as a function (ML) of rare gas overlayer thickness. From [49]. M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 rare gas coverage Fig. 14. Total F- yield from a monolayer of PFJRu(000 201 (ML) 1) as a function of rare gas overlayer thickness. From [49]. 1.50 f Y Fig. 15. F- ESDIAD surface plots from 1 ML PF,/Ru(OOO 1) and from 0.15-ML-Xc/l-ML-PF,/Ru(OOOl). From [36]. Surprisingly, the total F- yield increases upon adsorption of 1 ML of Kr or Xe (Fig. 14), and decreases exponentially for films thicker than 1.5 ML. This increase is accompanied by a dramatic change in the F- angular distribution (Fig. 15): While F- ions desorb from 1 ML PF,/Ru(OOO 1) on a hexagonal array of trajectories (cf. Fig. 7(b)), the distribution changes to one which is broad and centered around the surface normal upon passage through the Kr film. The authors explain the enhancement in F- by a decrease in the neutralization rate of F- with the substrate during desorption (see Section 4.1.7 for details). The change in the F- angular distribution is caused by elastic scattering: As was concluded in Section 4.1.3 (see also [29]), there are channels in the rare gas films (up to 2 ML) through which ions can escape with little attenuation. These channels are normal to the surface for fcc(ll1) films; hence the F- ions which desorb from the clean PF,/Ru(OOO 1) surface with polar angles of - 60” have their trajectories changed to smaller polar angles (“steering”). It is also suggested that the attenuation of the F- ions in Kr and Xe films thicker than 1 ML is mainly caused by elastic scattering, because inelastic processes such as charge transfer reactions are unlikely to occur. 202 M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 As discussed in Section 3, Bressler and Schwentner [17] have reported that the mean penetration depth (l/e) of - 4.2 eV F atoms in the Ar film is - 2.8 nm; this is an interesting result as compared to the short escape depths (-0.4-0.8 nm) for F+ and F- ions through Kr and Xe layers. Since the diameter of an F atom is smaller than an F- ion, we expect that the mean penetration depth of F atoms in a rare gas film should be greater than that of F- ions. However, considering that the Ar film grows in fee fashion, and the energetic F atoms undergo a series of (mainly) elastic collisions in the Ar film, it is somewhat surprising that a - 4.2eV F atom can penetrate so many Ar layers. 4.1.7. F ‘, F- transmission through H,Ojlms Akbulut et al. [37,50] studied the transmission of Ff and F- from 1 ML PF,/Ru(OOOl) through H,O overlayers. They also compared the behavior of F+ ions which desorb from the surface with different polar angles, 8, - 0” and - 60”. The F+ ESDIAD patterns for various H,O thicknesses are shown in Fig. 16. Fig. 17 shows the total angle integrated F+ ESDIAD intensity from the l-ML PF,/Ru(OOO 1) surface as a function of H,O thickness. As seen in Fig. 17, 1% of the F+ ions survive transmission through several monolayers of H,O. They find that the attenuation of the off-normal (60”) F’ ions is about twice as strong as that of the normal (0’) ions, which is consistent with a simple path length argument (l/cosQ). For F-, Akbulut et al. find a similar increase in F- yield around 1 ML H,O as discussed in Section 4.1.6 for F- through Kr and Xe. Fig. 18 shows the total angle-integrated F- ESDIAD intensity as a function of H,O thickness. It can be seen in Fig. 18 that the F- ESDIAD intensity increases in the H,O coverage range O-l ML, and this increase is accompanied by a strong change in the F- angular distribution [37]. However, it is interesting to note that there is no increase in the F- yield upon adsorption of - 1 ML H,O when the F- ions are desorbed from a multilayer of PF, (> 10 ML) instead of 1 ML, as shown in the inset of Fig. 18. This points out that the increased Fyield is a phenomenon associated with the substrate-overlayer interface, and not a charge transfer process in the film. Based on this observation, the authors suggest that dielectric screening of the H,O overlayer leads to a reduction in the neutralization probability of F- from 1 ML PF,/Ru(OOO 1) with the surface [37,50]. Nordlander and colleagues [SO] have carried out quantum mechanical calculations in order to estimate the influence of the dielectric water overlayer on the resonance electron tunneling rates between an F- ion and the Ru(0001) surface. The water overlayer was modeled as a uniform dielectric film with dielectric constant E and thickness d. They have shown that the presence of a dielectric film introduces a potential barrier between a desorbing F- ion and the metal surface. This potential barrier reduces the electron tunneling rates near the surface, so that the F- survival probability increases. The calculations have also revealed that the increase in survival probability of the F- saturates after a water bilayer is formed, which is in excellent agreement with the experiment. It is noteworthy that condensed H,O attenuates the Ff signal (the attenuation cross-section is - 1 x lo- l5 cm”) much more effectively than the F- signal; - 1% of the F- ions can penetrate 10 ML of H,O (the attenuation cross-section is only 6 x lo- I6 cm”). The long penetration depth for F- is explained by a low probability for electron detachment and dissociative attachment reactions (all of the known reactions between F- and H,O are endothermic), and the open structure of the ice H,O films, through which the ions can channel without significant energy loss and attenuation. In contrast, the strong attenuation of F+ without substantial changes in angular distribution suggests that charge transfer processes are important in this case. Although some ion-molecule reactions M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 203 (a) (b) Fig. 16. F+ ESDIAD pat tern obtai ined wi.th a sample bias of + 160 V (incident electron 1 energy r: 360 eV) from (a) a sligl 7t1y beal m-d amaged 1 ML PF JR u(OO( 3l)sur .face (no H,O present). (b) 0.5 ML of H,O an Id(c) 1.61ML H,O adsorbed on the sligl 1t1y beam-damage :d 1 ML 1 PF,,iRu(O( 3 0 1) surface at 20 K. From [37]. 204 M. Akbulut et aLlSurface Science Reports 28 (1997) Water Coverage 177-245 (ML) Fig. 17. Total F+ yield from a monolayer of PF,/Ru(OOO 1) as a function of H,O overlayer thickness. Shown are the attenuation curves for an F+ beam centered around the surface normal, and for one off-normal beam (polar angle: - 60 “). From [37]. J L, 0.5 :- 8 0 0 3 # Fig. 18. Total F- yield from a monolayer of PF,/Ru(OOO 1) as a function of H,O overlayer thickness. The inset shows semilogarithmic plots of the total angle-integrated F- ESDIAD intensity from 1 ML PFJRu(000 1) (solid circles) and 10 ML PFJRu(000 1) surfaces (open circles) as a function of H,O overlayer thickness. From [37]. (such as F+ + H,O +(HF)+ + OH, F+ + H,O + HF + (OH)+) are energetically possible and may occur in the film, we were not able to detect evidence for ion-molecule reactions. 4.1.8. Cl- from CH,Br/CCI,/Ag(lll) There is another report on the enhancement of an ion yield upon condensation of an overlayer on top of the substrate. Dixon-Warren et al. [Sl] investigated the effect of an overlayer of CH,Br on the M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 205 Cl- yield from a layer of CCl,/Ag(l 11). They observe an increase in the Cl- yield for small CH,Br coverages and an exponential decrease with higher CH,Br coverages. Based on their finding that the photoelectron yield also increases with CH,Br coverage for small coverages, they conclude that the production of Cl- by photoelectrons is increased. The increase in photoelectrons is caused by a decrease in the work function of the surface by the CH,Br overlayer. 4.1.9. Evidence for ion-molecule chemical reactions in overlayers Sanche and Parenteau [52,53] studied the transmission of O- through hydrocarbon overlayers, 1 ML n-C,H,,+, (such as C,H,, and C,H,,) and 1 ML 1 ML n-C,H,, (such as C,H, and C,H,). In these experiments, a Pt substrate was covered by a 4 ML 0, film, on top of which the hydrocarbon layers were deposited. Under bombardment by electrons with energies w 13.5, they find OH- to desorb from the surface. They also observe the desorption of OH- ions from films composed of a mixture of 0, at 25% vol in n-CnH2n+2 and n-C,H,,. Fig. 19 shows the thickness dependence of the O- and OH- yields and O-/OHintensity ratio for a film containing 25% 0, (by vol) in n-C,H,, deposited on the Pt surface. Since n-C,H2n+2 and n-C,H,, contain only hydrogen and carbon, the OH- signal bears the signature of dissociative electron attachment (DEA) from adsorbed 0,. Sanche and Parenteau suggest that the OH- ions are formed via the reaction between ground state molecules (C,H,, +2 and n-C,H,,) and O- produced directly from adsorbed 0, by the I 1 I 2 COVERAGE 1 3 I 4 I 5 (MONOLAYERS) Fig. 19. Thickness dependence of the 0 - and OH - yields and OH -/O - intensity ratio for a film containing in n-C,H,, deposited on Pt. The incident electron energy is 13.5 eV. From [52,53]. 0, at 25% vol M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 206 electron beam: 0- +C,H 2n+2+OH- 0- +C,H,,+OH- + CnH~n+l, +C,H,,_; (4.3) (4.4) This is one of the few direct measurements of ion-molecule reactions in ion transmission experiments in this low-energy regime. In another experiment, Sanche [54] investigated the ESD from thin films of 0, and N,, and observed NO+ as one of the desorption products. By measuring the dependence of the NO+ yield on the primary electron energy, Sanche concluded that the formation of NO’ is initiated by the ionization of O,, and that NO+ must be formed in an ion-molecule reaction O++N,+N,O++NO++N. (4.5) 4.1.10. Cr on SiO, and SiO, on Cr To conclude this section we mention a study of the escape depth of the secondary ions resulting from electronic sputtering by 9 MeV Ar ions in inorganic thin films [55]. Two cases have been investigated: sputtering of Si(OH), from SiO, with Cr as overlayer, and sputtering of CrO,H- from Cr with SiO, as overlayer. A SIMS operating in TOF mode was used in order to determine the depth of secondary ions. Based on the experimental observations, the authors derive escape depths of 0.5 nm for the Cr overlayer and 1 nm for the SiO, overlayer. This result indicates that the depth of origin of the secondary ions resulting for electronic sputtering is larger than that for knock-on (nuclear) sputtering (- 0.5 nm). However, it is possible that beam damage of the overlayers and surface roughness affect the apparent depth of origin of ions in this experiment, because it is expected that the primary ion beam can cause much more damage to the inorganic materials than the metals [6]. This beam damage of surface layers may affect the depth of origin of ions, because the ions traverse strongly disturbed overlayers, instead of nearly undisturbed overlayers. In Table 1, we summarize the available experimental data on low-energy ion and neutral transmission through overlayer films. 4.2. Ion desorption from multilayer adsorbates and compounds 4.2.1. Principles In this final part of Section 4 we summarize a number of experimental results from ESD yields from multilayer adsorbates on metal surfaces, and their dependence on the adsorbate film thickness. We try to extract some information from these data concerning the origin of the detected ions: do they originate from the surface or from subsurface layers? We illustrate this approach in Fig. 20: The total ESD ion yield from a multilayer adsorbate can contain contributions from ions originating on the surface and from ions in subsurface layers. As the first example, we discuss Ccl, multilayers, which are interesting because they produce both positive and negative ESD ions. 4.2.2. Cl+, Cl- desorptionfrom Ccl, The ESD Cl+ and Cl- yields from CCl,/Ru(O 0 0 1) were investigated as a function of Ccl, film thickness from monolayer to multilayer thicknesses by using a - 300eV electron beam (Fig. 21) ESD ESD ESD ESD - 5eVO+ -7eVO+ -7eVO+ -4eVF+ -leVF- MD (theory) -7eVO+ Technique/method W (10 0) Oxidized 1) 1) 1 ML PF,/Ru(OOO tungsten Oxidized 1 ML PF,/Ru(OOO W ( 10 0) Oxidized 10) W ( 10 0) Oxidized TiO,(l or ion source ion and neutral Neutral data on very low-energy ESD (experiment) or ions of the available -7eVO+ Neutral Table 1 Summary overlayer 90 30 33 50 130 180 14 26 10 - 0.5 ML H, I80 (0.18 nm) - 3 ML NH, (0.7 nm) <lML(<O.l6nm) < 1 ML( <0.19nm) < 1 ML( <0.24nm) < 1 ML( <0.27nm) - 1 ML Kr (0.4 nm) - 1 ML Xe (0.44 nm) - 4 ML H,O (1.5 nm) H “0 N;I,e Li’ Na’ Kf cs’ Krp Xeg H,Oh Kra Xeg H,Oh transfer Charge transfer transfer Charge Charge transfer scattering and transfer Elastic and charge Charge transfer (O-l ML) (O-l ML) (0- 1 Charge Charge transfer - 5 ML Ar (1.9nm) -4MLKr(1.6nm) - 6 ML Xe (2.6 nm) Arc Kr’ Xe’ 28 Elastic scattering scattering Elastic 13 (1-5 ML range) 17(224 ML range) ll(226 ML range) - 6 ML Xe( 2.6 nm) Xeb NHxd Elastic Elastic scattering scattering Elastic - 4 ML Kr (1.6 nm) Krb Dominant mechanisms range) range) range) range) range) range) ML ML ML ML ML ML Attenuation cross-section ( x lo-r6 cm’) 5(0-l 60(1-2 5 (O-2 22 (2 4 5 (O-2 15 (226 films - 2 ML Ar (0.75 nm) Escape-depth” through Arb Overlayer transmission Cu(ll1) Cu(ll0) Cu( 100)” -2ML -2ML -3ML Ion-molecule reaction (OH - formation) Enhancement (O-l ML) Elastic scattering Elastic scattering Elastic scattering Dominant mechanisms neutral or ion yield to - 70% of the Ru yield suppressed 11 15 6 - 2 ML Kr (0.8 nm) - 2 ML Xe (0.88 nm) - 10 ML H,O (3.7 nm) -15nm Krg Xeg H,Oh Ar’ n-C, H,,’ n-C, H,, + 2 CH,Brk Cr’ SiO, cum - 0.5 nm - 1.0 nm >lMLCu Attenuation cross-section ( x lo-l6 cm2) Escape-depth” Overlayer a In order to compare the available data from different sources, we define escape-depth as an overlayer thickness that attenuates approximately 1% of the original yield. b Refs. [29,32]. ’ Ref. [41]. d Ref. [45]. e Ref. [30]. ’ Ref. [46]. g Refs. [36,49]. h Refs. [37,50-J. i Ref. [17]. J Refs. [52,53]. k Ref. [Sl]. ’ Ref. [55]. m Ref. [19]. ” Ref. [21]. l-20 eV Cu CCl,/Ag( 11 1) SiO, Cr Ru(000 1) PSD Sputtering clSi(OH) ; CrO,HRu and Cu cu 3.6 keV Ar sputtering and TRIM Theory F,/Ar/Kr WPt Photodissociation ESD - 4.2 eV F - 1.5 eV O- 1 ML PF,/Ru(OOO 1) ESD -leVF- Neutral or ion source Technique/method Neutral or ions Table 1 (continued) M. Akbulut et al. /Surface Science Reports 28 (1997) 209 177-245 eV V ions ‘\ overlayer 4 substrate Fig. 20. Illustration of experimental approach in Section 4.2. An ion is desorbed by ESD from a multilayer on a substrate, and the ion yield is measured as a function of coverage. ! 1.0 - 7 0.8 - 0.6 - 0.4 - O.*0. 3 ( I d 8 1 2 3 CC14 coverage 4 5 6 (ML) Fig. 21. Cl- and Cl+ ion desorption yield as a function of Ccl, coverage for CCl,/Ru( 0 0 0 1). 1 ML corresponds to room temperature saturation coverage. From [57]. 210 M. Akbulut et al./ Surface Science Reports 28 (1997) 177-245 [56,57]. As seen in Fig. 21, the Cl+ yield continues to increase up to 5 ML, while the Cl- signal increases up to 2 ML, and then decreases to a saturation value between 3 and 6 ML. If the Cl+ ESD cross-section were independent of film thickness, and if all the detected ions originated in the surface layer, one would expect a constant ion yield for films thicker than 1 ML. This is obviously not the case. One factor that could lead to an increase in ion yield beyond 1 ML is that quenching of the primary excitation with the surface (which leads to reduced ion emission) is not as strong for the second and subsequent layers as for the first layer which is directly in contact with the metal substrate. This would lead to higher ESD cross-section for the second and subsequent layers than for the first layer. We also cannot exclude the possibility that not all ions detected originate from the surface layer; some ions may originate in subsurface layers and traverse the surface layers and contribute to the ion signal. This may explain why the ion yield increases so strongly for adsorbate thicknesses > 1 ML. As discussed in Section 4.1, it is clear that ions produced in subsurface layers can traverse overlayer films. As seen in Fig. 21, the Cl- yield rises very steeply between 1 and 2 ML. This may be due to a decrease in the neutralization probability of the desorbing Cl- with the surface. This effect is expected to enchance the Cl- yield due to dielectric screening caused by the second Ccl, layer, similar to the observation discussed in Sections 4.1.6 and 4.1.7. There is another factor which can influence the thickness dependence of negative ion yields. Sanche [58] and Akbulut et al. [38] have reported that the ESD O- and F- yields produced via both DEA and dipolar dissocation (DD) processes from 0, and PF, on Pt vary strongly with the thickness of the 0, and PF, layer, respectively. The DEA O- (F-) yield increases to a saturation value with increasing 0, (PF,) thickness, whereas the DD O- (F-) yield first increases as a function of O2 coverage up to l-2 ML and then decreases with increasing 0, coverage. Sambe et al. [59] have also reported a similar observation, when 0, is condensed on the Pt substrate with a rare gas spacer layer between the 0, and Pt. They have argued that the image potential induced at the Pt surface affects the O- yield via DEA and DD processes differently. Since the Cl- yield from CCl,/Ru(O 0 0 1) generated under - 300eV electron bombardment, it is expected that the Cl- ions from CCl,/Ru(O 0 0 1) are produced mainly via DD process. Therefore, it is also possible that the change in the Cl- yield as a function of Ccl, is influenced by the image potential induced in the Pt surface. 4.2.3. H+ desorption from H,O, NH, Madey and Netzer [60] investigated the H+ ESD yield from H,O on Ni( 1 1 1). They observe that the H+ yield increases with H,O film thickness far beyond 1 ML. Benndorf and Madey [61] and Netzer and Madey [62] also studied the H+ ESD yield from NH,/Ru(O 0 0 1) and NH,/Ni(l 1 1) and find that the H+ signal increases strongly with adsorbate film thickness beyond 1 ML. In all these cases, it seems likely that the ion signal contains contributions from ions produced in subsurface layers. Recently, Ma et al. [40] have investigated ESD of H+ ions from amorphous ice films (l-300 ML thick) condensed onto a Ru (0 0 0 1) surface at - 30 K. H + ions generated by using a focused electron beam (- 300 eV) desorb normal to the surface with a peak energy of - 4.5 eV. The FWHMs of H+ become narrow with increasing film thickness, from - 40” at 1 ML to - 20” at 10 ML. Annealing the samples leads to a phase transformation from an amorphous to crystalline ice which decreases the H+ signal and increases the FWHM of H+. This observation indicates that for amorphous ice, the M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 211 H+ are not only created at the top layer; they also originate from deeper layers (3-5 ML), mainly from interstitial water molecules, and contribute significantly to the observed H+ yield. In contrast, the ESD Hf yield from crystalline ice originates mainly from the top layer. Measurements of H+ transmission through ultrathin Xe overlayers condensed onto ice layers support this idea [39,40]. 4.2.4. Oxygen depletion in TiO, by electron bombardment It has been reported [63,64] that electron beam damage can lead to a significant oxygen depletion in the surface layers of a TiO, crystal. McCartney and Smith [63] used a high-resolution electron microscope for both sample analysis and for electron irradiation of rutile TiO,. A layer of rocksalt TiO is produced which is several nanometers thick. Apparently, oxygen species produced by ESD in subsurface layers traverse the surface layers, desorb, and thereby contribute to the reduction of TiO,. 5. Depth of origin of desorbing ions: computer simulations Computer simulations have served as helpful tools in addressing the question of the depth of origin of desorbing ions. In particular, molecular dynamics (MD) methods are suited because they enable the study of trajectories of low-energy (- 7 eV) ions as they penetrate a solid film. One can compare computed parameters such as the total transmission yield, the energy distribution and the angular distribution of ions after they traverse the film to the corresponding experimental values. Moreover, one can study the energy losses and trajectories of the ions in the thin film, which are usually not accessible experimentally. In order for the results of the MD simulation to be accurate and meaningful, one has to know the interaction potentials of the ions and neutrals involved; this is often the main limitation on the accuracy of the MD simulation. In contrast to high-energy collisions, where the exact shape of the interaction potential surface does not matter as much and where good approximations are available, the results of the simulation at these low collision energies depend strongly on the details of the interaction potential. In the following, we discuss one example of a MD simulation: The transmission of 7 eV 0 ’ ions through ultrathin rare gas films of Ar, Kr, and Xe [29,32] by Klein et al. [41]. This MD simulation was performed in connection with the experiments [29] which we describe in Section 4.1.3. The authors use O+-rare gas interaction potentials from the literature [65] and calculate the transmission of O+ ions through fcc(1 1 1) rare gas films. The initial 0 ’ energy and angular distributions are chosen to agree with experiment. No inelastic effects, such as charge transfer, are considered. As can be seen from Fig. 22 the simulation of the total transmission yield agrees very well with the experimental data for Kr and Xe. For Ar, there are deviations for film thicknesses > 1.5 ML. The excellent agreement for Kr and Xe strongly supports the explanations given in Section 4.1.3, and demonstrates that MD can predict the ion transmission yield through a thin film in this low collision energy range. One of the reasons for the excellent agreement in this case is that the interaction potentials for O+-Ar, O+-Kr and O+-Xe are available [65] (based on experimental studies and on calculations). The authors confirmed that the transmission yield depends strongly on the details of the shape of the interaction potentials [41]: If the onset of the repulsive part of the 212 M. Akbulut et al. / Surface Science Reports 28 (1997) f simulation ! IO t . I 0.1 0 177-245 , f i . . 7 i (clAf 12 1 (mo3nolayA) 5 6 Fig. 22. Transmission yield Yas a function of film thickness for three rare gases. From [41], with experimental data taken from [29]. potential energy curve is increased to slightly larger internuclear distances, the ion transmission rate drops significantly. Klein et al. [41] also investigated the sputtering of (the weakly bound) xenon atoms from the overlayer film by the 7 eV oxygen ions (a “kick-off” process, as suggested in [27,28-J).They find that the sputter yields are of order 1, and that the sputter yield exhibits a maximum for Xe films of thickness 3 ML. The decrease in sputter yield for thicker Xe films is attributed to the fact that the O+ ions have lost too much kinetic energy in subsurface collisions to be able to cause Xe desorption; for thinner M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 213 films, the probability for a collision of O+ with Xe becomes small. The sputtering of Xe atoms is correlated with the back scattering of Of ions, i.e., only small impact parameter collisions provide enough momentum transfer from O+ to a Xe atom to sputter the Xe. We discuss in Section 4.1.3 that the rare gases are expected to grow in a layer structure A-B-C-A-B-C-A..., corresponding to a (1 1 1)-oriented fee structure. Klein et al. [41] also investigated the transmission through a hypothetical hcp structure with a layer structure A - B -A B-A... The transmission yield is found to be much higher for thicker rare gas films with the hcp structure than for the fee structure. This is in agreement with a “channeling”mode1: In contrast to the fee case, where the layer C closes the last channels, the channels persist beyond the third layer for the fee structure, and ion transmission yield is therefore higher (Fig. 10). This is a nice example where a MD simulation can provide insights into physics where the experiment cannot. 6. Applications and implications for surface analysis In this section we briefly present some conclusions from the discussions so far concerning the interpretation of ESD, PSD and SIMS measurements. One obvious conclusion is that these techniques are not always sensitive to the top surface layer only, contrary to what is often assumed [1,2,5]. The secondary ion signal from an unknown sample to be analyzed may well contain contributions from subsurface layers. The percentage of subsurface contributions depends on a variety of physical surface properties, such as chemical composition, structure and electronic properties, but total subsurface contributions of order 50% would not be surprising. Since these properties are usually not known for a sample to be analyzed, it may be difficult to attribute a secondary ion signal to a specific surface layer. Besides the total yield, the energy distribution and angular distribution of the secondary ions can contain information about the chemical bonds and the geometry of the bonds on the surface (see Appendix C). However, if ions that originate below the surface layer can escape through the surface layers into vacuum and be detected, their kinetic energy may be reduced upon transmission through the surface layers and their trajectory changed, both due to elastic scattering. Hence in certain cases it can be difficult to draw conclusions from ion energy and angular distributions concerning the properties of an unknown surface. 6.1. Ion beam deposition Low-energy ion beam deposition has been shown to be a promising technique for deposition of smooth, homogeneous and ordered films on substrates. For instance, Lifshitz et al. [7] demonstrated that smooth, diamond-like films can be produced by low-energy C+ deposition on a carbon matrix. Rabalais [66] deposited 15 eV Si+ ions onto a Si substrate and observed homoepitaxial film growth. The physical processes involved in low-energy ion beam deposition are very similar to the ones determining the escape of low-energy ions from surfaces. Instead of the ion being generated in the solid and traversing the surface layers in order to escape from the surface, the solid is bombarded with low-energy ions which may penetrate the solid a few monolayers or be adsorbed on the .surface. The interaction mechanisms of the ions with the surface are reflection of the ions (elastic 214 M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 backscattering), neutralization (charge transfer), or penetration These processes are discussed throughout this paper. of the surface (e.g. channeling). 6.2. Low-energy ion implantation Ion implantation is the most common process to place dopant (impurity) ions (such as B+, I’+, As+) in semiconductor crystals [67,68]. With ion implantation, the penetration depth of the ions and dopant ion concentration are controlled precisely. The ion penetration depth depends mainly on the energy and mass of the ions, and the atomic mass of the solid. For example, the average penetration range of B+ ions incident on Si in the energy range from a few keV to 3 MeV is about 100-4000 nm, while that of 10 keV P+ ions in silicon is 14 nm [67,68]. Continued reduction or scaling of the semiconductor device dimensions is demanding a need for ultrashallow source/drain junctions only 60 nm deep for the 0.25 ,um device generation [67,69]. The ion penetration depth can be, in principle, reduced by decreasing the ion beam energy to form an ultrashallow junction. For p-channel devices where the source/drains are heavily doped with boron, low-energy (< 1 keV) ion implantation is required. This is an extremely challenging problem in the semiconductor industry, because at low energies, it is difficult to produce high beam currents to perform industrial processes [67,69]. (Since the n-type source/drains are heavily doped with phosphorus or arsenic, the formation of ultrashallow junctions with > 1 keV ions is not a major problem for n-channel source/drains.) There has been significant progress to develop low-energyhigh current ion implanters [69,70]. It is expected that continued transistor scaling will require even lower-energy (< 500 eV) ion implantation [69]. However, only little is known about the processes and properties that determine low-energy ion implantation, and hence, there are considerable research efforts to understand l-5 keV or even lower-energy (< 500 eV) processes [69,70]. A future study of low-energy B+ ion transport through ultrathin Si films will be very important to understand the limitation of very low-energy ion implantation processes. 6.3. Electrochemistry Low-energy ion transport through liquids or ultrathin films plays a central role in the field of electrochemistry [71]. In particular, ion transport through liquid water is of fundamental interest in electrochemistry, because water is the most important electrolyte in electrochemistry [72]. Since amorphous ice can serve as a useful model of liquid water [72], ion transport through amorphous ice water layers can have important implications. On the other hand, very slow ions ( < 1 eV) can be solvated in water. For example, as discussed by Akbulut et al., the attenuation of - 1 eV F- ions in amorphous ice films is mainly due to energy loss processes (multiple elastic and inelastic scattering) [37,50]. (Note that - 1 eV F- ions can penetrate - 10 ML of condensed water.) After an F- loses sufficient energy via collisions in ice films, it may become trapped in the ice films as a result of ion-solvent interactions. 6.4. Electronic aging In radiation physics and chemistry, transport and relaxation of excess low-energy charge carriers (such as electrons and ions) is a topic of major importance for engineering materials (such as M. AkbulutlSurface Science Report 234 (1997) 177-245 215 dielectrics) subjected to high electric fields [73-761. Hot electrons produced in dielectric materials subjected to high voltages can induce electronic excitations and ionization events. These can lead to physical and/or chemical changes which contribute to the aging process in the dielectric [76]. As discussed earlier, low-energy ions can interact with target atoms/molecules via various elastic and inelastic processes. If ions, in their passage through an atomic or molecular film, interact with the films via charge transfer and/or ion-molecule reactions these processes can produce reactive chemical species which contribute to the electronic aging. Therefore, it is expected that the ion transmission through ultrathin dielectric films is also relevant to the field of electronic aging. 7. Conclusions In this paper we have addressed the fundamental question of the depth of origin of secondary ions from surfaces. We have provided basic background knowledge and experimental and theoretical results from investigations which can shed light on the interaction of ultralow energy ions with thin solid films. We have correlated the different physical attenuation mechanisms with material properties of solid surfaces. Finally, we pointed to a few neighboring fields of research where similar physical processes occur as in the area discussed here. Contrary to “common belief” secondary ions from solid surfaces desorbing under the impact of electrons, photons or ions can originate in subsurface layers, sometimes several layers below the surface. We have shown that elastic and charge transfer processes are the dominant processes in determining the depth of origin of ions created below the condensed overlayers. Our experimental results on ion transmission through condensed atomic and molecular films have revealed that these processes depend strongly on the overlayer electronic and chemical properties as well as the nature of the ion and its kinetic energy. We believe that ion transmission through ultrathin film measurements can have an impact in areas as diverse as ion transport in electrochemistry, astrophysics, radiation physics and chemistry, ion beam deposition, the physics of atomic and molecular collisions at low energies and surface analysis using electron, photon and ion beams. Acknowledgements The authors acknowledge, with pleasure, valuable discussions with Prof. R.E. Johnson. This work was partially supported by the National Science Foundation, Grant CHE-9408367. Appendix A. Ion-atom/molecule collisions Since most of the processes leading to attenuation of l-10 eV (low energy) ions in condensed films having wide band gaps are believed to involve binary collisions between projectile ions and target atoms in the surface layers, we now focus on the basic physics of atomic collisions, binary elastic collisions, charge transfer processes, and ion-molecule reactions. Collisions between ions and atoms can be classified according to the collision energy. If the relative collision velocity of the two particles is fast relative to the motion of the electrons (u, >>0,; v,: 216 M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 collision velocity; u,: electron velocity), then the electron motion during the collision can be neglected, and the electron distribution is static, except for transitions which occur when the particles are at their closest approach [77,78]. For low-energy collisions (0, <<u,), on the other hand, the electrons adjust constantly to the changing potentials during the collision; this is an adiabatic process. This means that the electrons of the two collision partners may be shared by both centers during the collision. This results in a complex potential diagram with many possible transitions between potential curves, leading to high cross-sections for inelastic processes such as charge transfer. It is useful to remember that the speed of an electron of a ground state H atom is 2.2 x lo6 m/s, which corresponds to a collision energy of about 25 keV/amu. Hence all collisions that we have been discussing in this paper fall in the limit of low-energy collisions. A.1. Binary collisions: collision kinematics and elastic collisions We begin with a two-body binary collision model [79]. If the collision time during the ion-solid collision is much shorter than the period of a lattice vibration, the collision is called fast (or impulsive). The collision time is defined as the time during which the projectile ion experiences the repulsive interaction with a surface atom. (Assuming that the range of a repulsive interaction is d - 1 A, one can estimate roughly the collision time z for a given ion velocity v, as z N d/v.) Assuming that the collision is fast and elastic, any bonding forces acting on the target solid atoms during the collision can be neglected and the target atoms are treated as initially at rest; the projectile has already left the vicinity of the surface before the interaction between the target surface atom and its neighbors becomes significant. In elastic collisions, both momentum and total kinetic energy of the collision partners are conserved during the collision [80-821. Fig. 23 shows a schematic of a two-body collision. Particle A (ion) with mass M, and kinetic energy E,i scatters off particle B (target atom) (mass MB) which is initially at rest. The scattering energy is 0,, and the final energy of A is E,,. Particle B has a final energy of EBP,and the recoil angle is 8,. If there is no change in the internal energy of the system AB, MA Fig. 23. Schematic of an elastic two-body collision. Particle A with mass M, and initial kinetic energy EAi scatters off particle B (mass B) and changes its energy to E,,. Scattering angles: 0, and Oa. Energy of particle B after collisions: E,,. M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 217 which could, for instance, be the result of a charge transfer process, then we can write the equations of conservation of energy and momentum as (A.11 and M,U,i = M,u,, COS 8, + M,u,, COS 8,, 64.2) where uAi and uAf are the initial and final velocities of particle A, respectively, and uBf is the final velocity of particle B. Through elimination of r+$fand 8Bwe can derive the formula for energy loss in an elastic collision in the laboratory system: (A-3) This ratio is referred to as the kinematic factor and gives the energy transfer in a binary collision in terms of the mass ratio p (p = MB/MA) and the scattering angle eA [81,82-J. For MB > M, only the expression with the plus sign in Eq. (A.3) is valid. When the projectile ion is heavier than the target atom (MB < M$ both plus and minus expressions in Eq. (A.3) hold, indicating that two final energies are possible for each scattering angle. For MB < M,, the heavier projectile ion cannot be backscattered by the lighter target atom; there is only a limited regime of scattering angle. The maximum scattering angle is given by sin 8, = MB/M,. Fig. 24 shows energy loss (E,,/E*,) as a function of laboratory scattering angle for various values of p = MB/M, as given by Eq. (A.3). For 8, = 90 ’ and MB > M,, Eq. (A.3) becomes particularly simple: EAr/EAi= (MB - MA)I(MB + MA) = CL- I/P + 1. Note that for MB < M, there is no solution possible for Eq. (A.4). Fig. 24. The ratio E,,/E,, as a function of scattering angles for various mass ratios (MA/MB). From [82]. (A.4) 218 M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 For a scattering angle of QA= 180 ‘, Eq. (A.3) reduces to EAf/EAi = Ct”B - M*)/(“B + M*)12* (A.3 For primary ion energies above lOOeV, it is well established that the scattering of ions from solid surfaces can be easily explained in terms of elastic binary collisions with single surface atoms, because the collision time is very much shorter than the period of a lattice vibration [81,82]. For ion energies below lOOeV, the validity of the binary collision model has been addressed by several authors [83-851. Hulpke [84] has reported that even for low-energy (2-20eV) ion-solid scattering the vibrational period of the surface atoms is larger than the collision time. For example, for a 10 eV Li f -W(1 10) surface collision the vibrational period of the W atoms at 190 K is 40 times larger than the collision time, while at 2 eV the vibrational period of the W atoms is 18 times larger than the collision time. These results indicate that the binary collision approach for ion-solid interactions is still valid even at such low ion energies. However, a comparison of the experimental results with the calculated energy loss from Eq. (A.3) shows that the calculated results deviate slightly from the experimental results. Hulpke [84] has shown that in this low-energy range (2-20 eV) the attractive image potential plays a very important role in determining the energy loss processes; a better agreement between the experiment and the calculation has been found by modifying Eq. (A.3) by including a potential step of height E (E, -+ E, + E,E, + E, + E). Low-energy ion transmission through weakly bonded thin films (such as rare gas solids) can be viewed as proceeding via a single binary collision or a series of binary collisions between a desorbing ion and atoms in the films, because the vibrational period (- l/phonon frequency) of the atoms/molecules in the film is much larger than the collision time of low-energy ions; for example, the vibrational period of atoms in solid rare gas is - 1 x lo- l2 s, the collision time for 7 eV O+ is - 1 x lo-i4 s (time for - 7 eV 0+ ion to travel N 1 A). However, the ion collision time may be comparable with the lattice vibration time after the ion loses sufficient energy due to a series of binary collisions in the film, and ion-phonon scattering in the film becomes important. Ion-phonon scattering may be a very important energy loss process in the transmission of < 1 eV ions through a molecular solid such as ice (sublimation energy of ice is - 0.5 eV/molecule). A.2. Interaction potentials So far we have discussed the situation for a hard sphere elastic collision. Note that Eq. (A.3) does not depend on the actual interaction potential. The details of the collision, such as the exact trajectories of the particles, the differential cross-section, and the possible transitions, depend on the nature of the interaction potential between the two colliding particles [SS]. In principle, the primary force determining this potential is the Coulomb force. For high-energy collisions (u, >>u,), nuclear repulsion dominates the collision, and the interaction potential can usually be approximated by a screened Coulomb potential [20]. However, for the low collision energy regime (a, <<UJ this approximation is not appropriate. In this energy range, both repulsive (short range) and attraction (long range) interactions contribute to the interaction potential; the interaction of the two collision partners at large distance, R (R > rA, rB; the atomic radii of A, B), becomes important. The long-range interaction can sometimes be approximated by a power M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 2 L 6 219 a Fig. 25. Adiabatic potential energy curves vs. internuclear separation, R, for H + -Xe system. From [87]. law potential: V(R) = 1 CJR" (A-6) n with C, being a constant. For ion-dipole interactions the leading term is n = 2; for interactions of an ion with a neutral without a dipole moment, the ion-quadrufiole interaction dominates and the leading term is n = 3. For the collisions between an ion and a polarizable atom (such as Kr, Xe), the ion-induced dipole interaction dominates (n = 4). A reasonable interaction potential describing the interaction of an ion with a polarizable atom/molecule (or a molecule having a permanent dipole moment) can be obtained by superposition of the Lennard-Jones interaction and the ion-induced dipole interaction (or the ion-dipole interaction). At low energies (v, <<u,) an inelastic ion-atom/molecule collision can be best viewed as involving transitions between a set of quasi-molecular electronic states formed in the collision by the colliding patterns. As an example, we depict in Fig. 25 the calculated potential energy curves for the H+-Xe system [87]. It can be seen in Fig. 25 that there are a number of possible transitions at small intermolecular distances. Some of these transitions can lead to charge transfer, which is discussed in the next section. Therefore, for an exact description of scattering a knowledge of the potential energy curves is necessary. These can be obtained from theoretical molecular orbital calculations and from scattering experiments. 220 M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 A.3. Charge transfer reactions in the gas phase A charge transfer reaction between an ion and weakly interacting ultrathin film can be explained by utilizing the knowledge of gas phase charge transfer processes, provided the film is very thin ( < 3 ML) so that the band structure is not well developed. The following is a discussion about charge transfer processes in the gas phase and near solid surfaces. Electron transfer in collisions between ions and atoms or molecules is important from a fundamental as well as a practical standpoint [12,88,89]. During collisions of multiply charged or singly charged ions with atoms or molecules, several reaction pathways are possible. These include electron transfer (single or multiple), electron transfer accompanied by target excitation or ionization, etc. In this review, we are concerned with the one-electron charge transfer processes between a singly charged ion and an atom or molecule, since this case is most appropriate to our experiments. The reader interested in multiply charged ion-neutral charge transfer reactions is referred to reviews by Kamber and Cocke [90], and Appell [91]. In a charge transfer reaction between a singly charged positive ion and atom/molecule, the projectile ion picks up an electron from the target atom/molecule and becomes neutral. This process can be described by the relation: A++B-+A+B++AE, (A.7) where AE is the energy defect of the reaction, and A and B can be atomic or molecular species [77,78,89]. AE is the total change in internal energy; when reactants are in their ground states, AE is just the difference between ionization energies of A and B at infinite separation. For asymmetric systems (A # B) AE can be either positive (exoergic) or negative (endoergic); these processes are known as non-resonant charge transfer [12,88,89]. For exoergic reactions (AE > 0), no external energy is needed for the reaction to proceed. For endoergic reactions (BE < 0), the needed energy usually can be obtained from the kinetic energy of the projectile ion; there is a limitation on the primary ion energy range to which the charge transfer reaction can proceed. The one-electron charge transfer reaction between two identical particles (symmetric system) is described as A++A-,A+A+. (A.8) For symmetric systems, AE = 0 when reactants and products have the same internal energy state. These reactions are classified as symmetric resonant charge transfer reactions [12,88,89]. One-electron charge transfer between an unlike ion and an atom can also be a resonant or nearly resonant process if appropriate electronic states are available so that AE is nearly zero, and these processes are called accidental resonant charge transfer [88,89]. In the following we discuss each charge transfer process separately. We first begin with a symmetric resonant charge transfer process. A.3.1. Symmetric resonant charge transfer The cross-section for a symmetric resonant charge transfer collision increases with decreasing collision velocity. The variation of symmetric resonant charge transfer cross-section, c, with impact velocity, V,is given by a=(a-blnv)‘, (A.9) M. Akbulut et aLlSurface Science Reports 28 (1997) 177-245 221 impact velocity (cm/set) Fig. 26. Resonant charge transfer cross-section for hydrogen (H+ + H -+ H + H+) as a function of impact velocity. From ~921. where a and b are constants [89]. As an example, we depict in Fig. 26 the symmetric resonant charge transfer collision velocity dependence of the process H++H+H+H+. (A.lO) Sakabe and Izawa [92] present calculated and measured cross-section values for symmetric resonant charge transfer for all non-transition elements. The data cover mainly the high collision energy regime, but go down to as far as a few eV; note that for hydrogen a collision energy of 5 eV corresponds to a velocity of 3 x lo6 cm/s. Rapp and Francis [48] have shown that symmetric resonant charge transfer cross-sections depend strongly on the ionization potential. Fig. 27 shows the calculated symmetric resonant charge transfer & as a function of ion velocity for several systems [48]. It is apparent from Fig. 27 that the lower the ionization potential the higher the charge transfer cross-section. A.3.2. Non-resonant charge transfer Experimental and theoretical gas phase studies have shown that cross-sections for non-resonant charge transfer are small at low energies (u, < uJ, rise to a maximum, and fall off in the same way as symmetric resonant processes for u, > u, [48,89,93]. However, at energies from thermal energy to 10 eV, the non-resonant charge transfer cross-sections may increase if the reactions are exoergic. This is mainly due to the long-range polarization potential between the ion and neutral. In the 222 M. Akbulut et al./ Surface Science Reports 28 (1997) 10’5 2 2 s 10' 2 177-245 s v. cm SK-’ - Fig. 27. Variation of symmetrical resonant charge transfer cross-section calculated by Rapp and Francis. From [48]. (X’ + X+X + X’) with impact velocity, following, we first discuss the application of the Massey adiabatic criterion to non-resonant charge transfer in order to explain the energy dependence of the non-resonant charge transfer crosssections. The Massey criterion [13,89] (adiabatic maximum rule) states that if the time of collision is comparable to the time of transition, the charge transfer cross-section c is maximum. The velocity at which c is expected to peak is (A.1 1) Here a is known as the “adiabatic parameter”, the range of the interaction between the colliding partners, u is the relative impact velocity, and AE is the energy defect. The collision time is defined as a/v. If v C-C aAE/h, v is small compared to the electronic velocity of the target (v,). The target atom/molecule can have enough time to adjust the perturbation imposed by the projectile ion without a charge transfer occurring; this type of collision is called adiabatic. Hence, at low energies (E <<u2 1BE1 2m/2h2, but not smaller than 100 eV), the cross-section is small unless AE is very small. Fig. 28 shows clearly that the smaller the energy defect, the larger the charge transfer probability at least for energies down to N 100 eV. At very high velocities (v >>uAE/h), the cross-section decreases with increasing ion energy, because the interaction time becomes too short for the transition to occur; the collision is said to be sudden. Experimental data analysis shows that the Massey criterion is very useful [94]. For example, Hasted [ 13,891 has derived values of a using the experimental cross-section data for non-resonant charge transfer. For a large number of data, Hasted has shown that a is quite large, a -N 7Aln, where n is the number of electrons transferred in the collision. M. Akbulut et aLlSurface Science Reports 28 (1997) 177-245 102 1oJ 10’ 105 Energy (eV) Fig. 28. Charge transfer cross-sections for protons 10’ 223 10’ in rare gases. From [13]. The determination of a and A E can be rather difficult for certain collisions [13,89]. For example, the determination of AE is complicated by the final states of the ion and neutral. Another factor which can influence the energy defect during collision is polarization forces; the energy defect during collision may be rather different from the energy defect at infinite nuclear separation due to polarization forces. Since the determination of a and AE during collision is problematic, by using the Massey criterion one may not accurately estimate the projectile velocity (or energy) for the maximum. However, the Massey criterion can provide qualitative understanding of the energy dependence of the cross-sections for non-resonant charge transfer. Rapp and Francis [48] derived an approximation for non-resonant charge transfer processes at low collision energies ( > 100 eV) which has proved to be very useful. For low energies ( > 100 eV), the Rapp - Francis non-resonant charge transfer cross-section is reduced to the following form: (A.12) with y = JE,/13.6eV, E, is the ionization potential of B, U,is the collision velocity and a is the Bohr radius. Eq. (A.12) indicates that the smaller the energy defect AE and the larger the collision velocity, the larger the cross-section [48,93]. Fig. 29 shows the energy defect (AE) dependence of the non-resonant cross-sections calculated by the approximation of Rapp and Francis. Fig. 30 compares the experimental cross-section curves for charge transfer in collisions of Hf with Xe and Kr atoms with the calculated cross-section curves for A E = O.l,OS, 1,3,5 eV. Since the rare gas ions may be formed in either 2P1,2states or 2P3,2states [48,93], the energy defects for H+-Kr and H f -Xe collisions are: H++K~+H+KI-+(~P,,J, AE= -0.4eV, (A.13) H+ + Kr-+H + Kr+(2P,,J, AE = - l.OeV, (A.14) H++Xe+H+Xe+(2P,,,), AE= +lSeV, (A.15) H+ + Xe--+H + Xe+(2P,,J, AE = +0.2eV. (A.16) 224 M. Akbulut et al. JSurface Science Reports 28 (1997) 177-245 i- . i- /4 / - 1 Impact velocity (cm s-l) Fig. 29. Cross-section for charge transfer calculated by the approximation of Rapp and Francis; (a)-(c) refer to cases in which ionization potentials are chosen to correspond, when AE = 0, to the symmetrical charge transfer reactions He+-He, H+-H, and Cs+-Cs, respectively; (-) symmetrical resonant case (AE = 0); (-----) non-symmetrical resonant case with IAEl as indicated in eV. From [48]. As seen in Fig. 30, for H ‘( D+)-Xe collisions, although the observed cross-sections for both sets of data are different in absolute magnitude, the data show maxima located at nearly the same impact velocity. This corresponds to BE = 1 eV. This is suggestive that Xe+(‘P,,,) is the predominant product. For H+-Kr collisions the observed data disagree both in absolute magnitude and location M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 4 6 810’ Impact vdociry 01 IOQ 24 68 (cm 3-l) I I I f 1 t 2 4 I 10’ 2 + velocity km 9) Impact 225 I tw I 1 2 Fig. 30. Cross-sections for charge transfer collisions between H+ ions and (a) He, Ne and Ar atoms, (b) Kr and Xe atoms. Numbers refer to authors as indicated in the references below; (-) calculated by semi-empirical method by Rapp and Francis for different assumed energy defects as indicated; (-0 -o-) calculated for H+-He by Green [139]. Actual energy defects are 11.0,8.0,2.2 eV for H+ collisions with He, Ne and Ar, respectively (From [48,93]). 1. J.B.H. Stedeford and J.B. Hasted, Proc. R. Sot. A 227 (1955) 466; 2. J.B. Hasted, Proc. R. Sot. A 212 (1952) 235; 3. M. Becher and A.Z. Scharmann, Z. Naturf. 24 (1969) 854. [93]. Therefore, based on these experimental data it is difficult to draw a conclusion about the Kr ion state. Using Eq. (A.1 1) (the Massey criterion) and taking a - 7 A, one can estimate the peak velocity for the H ‘-Xe collision to be - 2.6 x lO’cm/s ( - 360eV) for AE = 1.5 eV and - 3.4 x lo6 cm/s 226 M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 (A+e++B Nuclear Separatron Fig. 31. Effective potential energy curves for the trajectories leading to a transition. The curves cross at an internuclear distance R = R,. From [90]. ( 1~6 eV) for AE = 0.2 eV. A comparison of these results with the experimental result shown in Fig. 30 indicates that (a) reaction (A.15) is dominant and (b) the Massey criterion and the Rapp-Francis model are very useful to predict the final state of the collision products. For low-energy collisions (0, CU,), a charge transfer reaction can be described as a transition between electronic states of a temporary molecule formed as an intermediate state by the colliding partners during the collision. A typical charge transfer process at low energies is A+ +B-QAB)++A+B+. (A.17) A charge transfer is assumed to be induced at curve crossings in the temporary molecular potentials or between temporary states which are energetically close at finite internuclear distance [ 13,88,95]. For example, note that some potential energy curves of the H+-Xe system cross at particular Hf-Xe internuclear distances, as shown in Fig. 25. Hence, the exact value of the cross-section depends strongly on the details of the potential energy curves. One of the most useful formulae for predicting low-energy charge transfer cross-sections was developed independently by Landau, Zener and Stuckelberg [ 13,881. The Landau-ZenerStuckelberg (LZS) model is based on the crossing of two potential-energy curves at a finite internuclear separation R,. According to the LZS model, the charge transfer takes place only at the crossing point. As shown in Fig. 31[903, for impact parameter less than R, the system passes over the crossing point twice, on the incoming and outgoing passages. p is the probability that a transition from one potential curve to another occurs in traversing the crossing point and (1 -p) is the probability that the system remains on the same potential curve. If a transition occurs on the first passage but not the second then the total transition probability is [p(l - p)] and vice versa [( 1 - p)p], this leads to a net change in the state of the system (charge transfer). Therefore, the M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 221 averaged charge transfer transition probability P,,, at the crossing point R, is given by PLZS = 2P(l - PI. (A.18) The crossing probability p is given by -2) p=l-exp( (A.19) with Al/ Of _ dvf dvo I dR I dR R=R,’ (A.20) Here, the quantity H,, is the transition matrix element, u is the relative velocity, V, and I’, are the initial and final potential energies and R is the internuclear separation. Then, using the LZS model, the charge transfer cross-section can be written in the form [95] fsLZS = (A.21) f’,,z(&). Note that LZS model breaks down in the threshold region where the distance of closest approach is approximately equal to R,. Demkov developed a model considering non-crossing interaction potentials [13,39]. For a review of the theoretical low-energy charge transfer models, readers are urged to refer to books by Johnson [95], McDanielet al. [13] and Bransden and McDowell [96], and review papers by Hasted [89] and Johnson [SS]. Although non-resonant charge transfer collisions in the collision energy regime higher than keV have been studied intensively, there exist only limited reports in the low collision energy range l-100eV. Based on the available experimental studies, there does not appear to be a trend in the non-resonant charge transfer cross-sections in the low collision energy range l-100eV [l&13]. At these low energies, charge transfer reactions can take place at relatively large impact parameters and the deflection of the projectile by the target may be negligible (glancing collision). At ultralow energies (0.025-l eV), charge transfer between an ion and a neutral occurs during orbiting due to long-range polarization potential [12,13,95]. The orbiting effect increases the probability for a charge transfer reaction if the reaction is exoergic [88,95]. There are several classical theoretical treatments of ion-neutral atom/molecule collisions. At ion energies much below 1 eV, the charge transfer cross-sections can be estimated according to the Langevin ion-induced dipole theory or according to the average dipole orientation (ADO) theory [97]. The Langevin pure polarization theory considers the interaction between an ion and a non-polar atom/molecule, assuming that both ion and neutral are point particles with no internal energy. If the reaction is exoergic, the Langevin capture cross-section and rate constant are given by CC= (A.22) , (A.23) 228 M. Akbulut et al./Surface Science Reports 28 (1997) 177-245 where aP is the polarizability of the neutral, e is the charge on the ion, E, the relative energy of the system and M, is the reduced mass. The Langevin ion-induced dipole theory predicts that the microscopic capture cross-section is inversely proportional to Ef’* and the rate constant is independent of velocity v. The Langevin theory only agrees well with a few simple low-energy ion-non-polar molecule reaction rate constants but underestimates the rate constants of most of ion-polar molecule collisions. Various theoretical models of ion-molecule collisions in which the neutral molecule has a permanent dipole moment have been developed. For example, the ADO theory, which includes the contribution of the ion-dipole potential to the collision cross-section, predicts quite accurately both the magnitude of the rate constants and their dependence on dipole moment for some selective ultralow energy charge transfer reactions [97]. However, both ADO and Langevin theories estimate cross-sections for ion-molecule reactions very poorly at energies higher than 1 eV. The high-energy limit of the collision cross-section approaches zero, because the collision partners are assumed to be point particles in these theories. Low-energy ion-molecule non-resonant charge transfer collisions are expected to be significantly different from ion-atom charge transfer collisions, provided that the molecule has a permanent dipole moment [12,97]. In addition, in ion-molecule charge transfer reactions, at least one of the collision partners is a molecule, the vibrational and rotational excitations may play an important role during charge transfer reactions. Fig. 32 shows measured cross-sections for 1.0 0.8 0.6 ., _ 4 Impact Fig. 32. Observed cross-section for charge transfer reactions et al. [140], (-) observed by Koopman [141]. From [93]. Ii 8 energy ( I2 IO'eV) ofH+ with O,, N,, CO, and H,O; (---) observed by Stebbings M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 229 H+-0, (AE = + lS3eV), H+-N2 (AE = 1.98eV), H+-CO2 (AE = O.l7eV), and H+-HZ0 (AE = + 0.98 eV) charge transfer reactions [93]. It is clear from the shape of the curves shown that the factors determining the ion-molecule charge transfer cross-sections are considerably more complex than for atoms. A.3.3. Accidental resonant charge transfer Accidental resonant charge transfer processes (charge transfer processes between an unlike ion and an atom/molecule when the energy defect is zero or very small) exhibit characteristics similar to either the symmetrical or the non-symmetrical charge transfer process [93]. The O’(4S) + H(ls) +O(3P,) + H+ collision presents a very good example of an accidental resonance charge transfer process [S9,93]. The energy defect of the 0+-H charge transfer collision depends on the total angular momentum J of the oxygen: 0+(4S)+H(1~)+0(3PJ)+H+ AE= -O.OleV (J=O), A E = 0.00 eV (J = 11, i AE= +O.O2eV (A.24) (J=2). The measured charge transfer cross-sections as a function of ion energy for the 0+-H charge transfer collision and the H+-H symmetric charge transfer collision are shown Fig. 33. A comparison of the Of-H charge transfer cross-section with the H+-H charge transfer cross-section indicates that both exhibit cross-sections which decrease monotonically with increasing impact energy [89,93]. However, note that the 0+-H charge transfer cross-section is smaller than the H +-H charge transfer cross-section. This can be understood by the fact that interaction energies of initial and final states are not identical at a finite nuclear separation. Although the energy defect is zero at infinite separation, it is non-zero as the projectile approaches the target atom; the difference in polarizabilities induces an effective energy defect. In the symmetrical resonant process, the energy defect is always zero, independent of the nuclear separation. A.4. Ion-molecule chemical reactions In this paper, we refer to ion-molecule of an ion-molecule collision, such as A+‘-‘+ BC+AB+‘-‘+ C. chemical reactions when new species are formed as a result (A.25) At low energies (l-lOeV), the ion-molecule chemical reaction cross-section depends on whether a reaction is exoergic or endoergic. For the endoergic reactions, the reaction cross-sections are generally zero up to a threshold energy and increase rapidly as the translational energy increases above the threshold energy (Fig. 33) [8,98-1001. At low energies, the typical ion-molecule reaction cross-section is - lo-i6cm2. Many gas phase ion-molecule reaction constants are listed in the literature (e.g., [44,101]). Several different ion-molecule chemical reaction channels between a projectile ion and a target molecule can be possible [S,SO]. For example, in the case of - 7 eV O+ and N 4 eV F + transmission through H,O overlayers, the ion-molecule reactions: 230 M. Akbulut et al./Surface Science Reports 28 (1997) 177-245 L------ 2.0 cl ICI E/E. n;(v I I 0) l Hr-HeH* 1 l I n I 0.15 - > it 0.10 3 0.05 - 0 0 t Em 2 4 (I 8 Et/N Fig. 33. (a) Translational energy E, dependence of the reaction cross-section for reactions with and without energy threshold (from [98]). (b) Translational energy E, dependence of the reaction cross-section for reaction H: + He -+ HeH+ + H (from [99]). M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 231 0+ + H,O -QOH)+ + OH, (A.26) F+ + H,O +(OH)+ + HF, (A.27) F+ + H,O+(HF)+ (A.28) + OH are energetically possible; reaction (A.26) is endoergic by w 0.4 eV, reactions (A.27) and (A.28) are exoergic by N 1.0 and 4.8 eV, respectively. Appendix B. Charge transfer processes between ions and solids In the following, we present a very brief discussion of ion-solid charge transfer processes. A detailed and complete account of the charge transfer processes near surfaces can be found in papers by Nordlander [14], Yu and Lang [102] and in a forthcoming review paper by Nordlander [103]. Charge transfer is the dominant inelastic process during low-energy ion or atom scattering from surfaces as well as during the desorption of energetic particles from solids [14,104-1071. Charge transfer between an ion and a solid can occur via two mechanisms: (a) Auger neutralization and (b) resonance neutralization. These processes depend strongly on the ionization potential and the electron affinity of the ion (or atom) as well as the work function of the solid. The energy levels of the energetic ion near a surface broaden and shift (the ionization level is shifted up, while the electron affinity level shifted down) due to the screening by the electrons in the metal. In Auger neutralization, the Coulomb field of the ion induces an electronic transition in the solid involving two electrons, if unoccupied electronic states (hole states) of the ion are located well below the conduction band states of the solid (Ei > 2e& Ei is the ionization energy of the ion, and 4 is the work function of the solid). The neutralization of positive ions (except alkali metal ions) by a surface is mainly due to Auger neutralization processes, because unoccupied positive ion states are typically located below the bottom of the metal valence band. Positive alkali metal ions and negative ions near a metal surface usually undergo charge transfer by resonance tunneling from the occupied level of the metal or into the unoccupied level of the metal, respectively. A change in the work function of the substrate could affect the resonant neutralization ion probability significantly. There have been several studies demonstrating the influence of the change in the work function on the ion yields [47,102,108,109]. Since adsorption of alkali metal atoms on a metal surface decreases the work function of the surface significantly, in some of these studies the effect of submonolayers of alkali metal on the secondary ion yields were investigated. For example, Joyce et al. [lOS] studied the coadsorption of potassium, K, with PF,/Ru(O 0 0 1) and its influence on F- and F* (metastable F neutrals) yields. Fig. 34 shows the influence of K on the desorption yields from PF, chemisorbed ( < 10% of saturation coverage) on Ru(0 0 0 1). As seen in Fig. 34 the F + yield decreases with increasing K coverage, while the F- and F* yields increase. The authors argue that an explanation of the decrease in Ff is not straightforward due to a complex chemical interaction of K with PF,, while the increase in the F- and F* yields can be explained due to a decrease in the work function. Since K decreases the work function of the substrate by several eV, this increases the Fand F* desorption yields by decreasing the neutralization probability of ESD-produced F- ions, as illustrated schematically in Fig. 35. 232 Akbulut et aLlSurface Science Reports 28 (1997) 177-245 PF3 Coverage it 0.00 0.02 = 0.02 ML 0.04 Pota88hn 0.06 Coverage 0.08 0.10 0.12 (ML) Fig. 34. Electron stimulated desorption yields of F+, F- and F’ from 0.02 ML of PF, coadsorbed with K on Ru(000 1). From [lOS]. Appendix C. Ion desorption from surfaces Since the main focus of this paper is the depth of origin of secondary ions from the surfaces of solids, we briefly summarize in the following section the principles of ion desorption from surfaces induced by various electronic and momentum transfer processes. We discuss desorption induced by electronic transitions (DIET) (electron and photon stimulated desorption (ESD/PSD)) and ion sputtering. C.1. Electron/photon stimulated desorption In ESD/PSD, beams of energetic electrons or photons (usually I 500 eV) impact on solid surfaces containing either adsorbed atoms or molecules, or terminal bulk atoms. These can induce electronic excitations to states that are repulsive in character, and the subsequent conversion of potential energy into motion of the excited species can result in the emission of energetic particles (e.g., positive and negative ions, and neutral species, including metastables) from the surface layers of the solid. Depending on the quenching rate of the excitations, the desorption of energetic species from the surface may occur (see Appendix C.2) [ 1lo]. Extensive reviews on ESD and PSD phenomena can be found in the literature [2-4,ll l-l 161. Since the penetration depths of electron or photon beams (energies of hundreds to several thousands of eV) incident on solids are of hundreds to thousands of &rgstrbms, electronic excitations can be produced far below the surface. Hence, for DIET of atomic or molecular multilayer films, or compound materials, subsurface layers might contribute to the desorption signal. C.1 .I. Mechanisms of desorption induced by electronic transitions (DIET) Direct momentum transfer between an incident low-energy (I 500 eV) electron and an adsorbed species (even the lightest adsorbed species) is generally insufficient to lead to anything but 233 M. Akbulut et al. J Surface Science Reports 28 (1997) 177-245 b b 4 = 3.5 E 0 = 5.5 4 B E vat eV e* = 4.7 eV % B/B E vat E vat b 2 Fig. 35. Schematic of the energy levels of F relative to the substrate surface region, for three different work functions. From [lOS]. Fermi level and the shifting of these levels in the near vibrational excitation, so that desorption must result from electronic excitation of the adsorbate, the adsorbate-substrate bond, or the bulk atoms. In this energy range, the electronic excitations leading to ESD/PSD ion or neutral desorption can involve either valence or core level excitations. Several models have been developed to explain DIET from surfaces. The Menzel-GomerRedhead (MGR) model and Knotek-Feibelman (KF) models have been very successful in explaining desorption from covalent adsorbates and desorption from ionic adsorbates, respectively. In the following we summarize the fundamentals of ESD/PSD. C.l.l.l. Menzel-Gomer-Redhead (MGR) model. The MGR model provides a good starting point for a discussion of desorption induced by electronic transition processes. The MGR model is 234 M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 Fig. 36. Schematic potential energy diagram illustrating stimulated desorption of surface species for an adsorbate (A)-substrate(M) system. Electronic excitation from the attractive ground-state potential curve (M + A) to a repulsive excited-state potential curve (such as MA’, MA+, MA +*)can lead to desorption of energetic species. very general [2,3], and is based on a description of electronic excitation and dissociation of gaseous molecules. According to this model, the initial excitation is sudden (Franck-Condon-like) occurring on a timescale that is short in comparison to nuclear motion. Fig. 36 shows schematic potential energy curves for an adsorbate(substrate(M) system to illustrate the MGR model. The system is assumed initially in a ground-state configuration (MA), and a sufficiently large excitation energy can make a Franck-Condon (FC) transition from the (MA) state to a repulsive excited-state (such as MA*, MA+, MA+*) possible, as shown in Fig. 36. After transition to a repulsive excited-state potential, the adsorbed species acquires nuclear motion (Fig. 36). If this excited state has a sufficiently long lifetime, the excited species can gain enough kinetic energy to desorb from the surface [4]. (The time required to break the surface bond is typically - lo- r4 s, corresponding to the time necessary for a typical ESD ion to travel - 1 A.) In the gas phase, if a molecule is excited to an antibonding state, it will typically dissociate. However, for a molecule adsorbed on a metal surface, there are final state effects (such as reneutralization and image force) that can influence the desorption probability. At the surface, the electronic excitations can be quenched rapidly; the de-excitation can proceed via reneutralization or by “a bond-healing” transition due to charge transfer to (or from) the desorbing species, e.g., by resonant tunneling or Auger neutralization [ 1lo]. Quenching of the excitation (de-excitation) can lead to recapture of the excited species, unless it has already gained an amount of kinetic energy M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 235 sufficient to result in desorption along the ground-state potential curve. Desorption along the excited-state curve occurs if no quenching transitions occur. Since the quenching processes at surfaces can be very efficient, the overall desorption cross-sections are smaller than those for gas-phase dissociation, even when the primary excitation cross-sections are comparable. The typical cross-sections for gas phase dissociative ionization are N lo-i6 cm2, whereas the maximum crosssections are in the range of 10-20- 10-23cm2 for desorption of ions and in the range of lo-‘*10-20cm2 for desorption of neutrals from surfaces [4]. Thus the ESD cross-section may be expressed in the form 0 = o,P, (C.1) where ce is a primary excitation cross-section and P is an escape probability [3] (see Appendix C.1.3). The MGR model has been successful in estimating the magnitude of P, but it cannot be used to predict specific excitations that will lead to desorption. C.1.1.2. Knotek-Feibelman (KF) model. The model proposed by Knotek and Feibelman describes a mechanism for positive ion desorption from ionic solids [2,3,115]. The KF model is applicable to maximal valency materials, such as TiO, and WO,. In these oxides (maximal valency materials), the cation (metal ion) is ionized down to the noble gas configuration, such as Ti+4[...3p6] in TiO,, and the highest occupied electronic level of the cation is a core level. This model explains both the observed thresholds and the large charge transfer involved in the ESD of positive ions (such as O+ and OH+) from maximal valency oxides. In the KF model, the desorption of ions from surfaces is initiated by the creation of a core-hole on a cation site at the surface [115-J.This core-hole can be filled by an Auger decay process, resulting in a multihole state. Since in the maximal valency configuration there are no valence electrons on the nearest neighbor cations, the core-hole decay can be described as an interatomic Auger process, and two additional electrons are ejected, as shown in Fig. 37 for an Of desorption from a TiO, surface. The interatomic Auger decay of the core-hole creates a two-hole (2h) positive anion. In this way a positive ion is formed at an initially negative ion site leading to a repulsive Coulomb interaction (reversal of Madelung potential), which provides the driving force for expulsion of a positive ion from the surface. C.1 .I .3. Auger stimulated desorption model. An extension of the KF description of ion desorption is the Auger stimulated desorption (ASD) model, which is applicable to both covalent and non-maximal valency materials, such as adsorbed gases (CO, NO) as well as Cr,O, and WO,. The ASD model gives a generalized picture of the desorption of ions from surfaces following creation of a core-hole and Auger decay [117-1221. This model for ion desorption is based on the ionization of a core level as the primary process and the production of an electronic state with multiple holes that are created by an Auger process. If the holes are localized for a sufficiently long time in a bonding orbital, a hole-hole repulsive state can arise resulting in the desorption of a positive ion. The localization of the holes is possible if the effective hole-hole repulsion U’ is greater than the valence bandwidth I/ (U’ > V) [2,123]. Recent experimental and theoretical studies on ESD from adsorbed molecules (such as N,, CO, H,O, C6H,IF, C,H,F, and PF,) have shown that multi-hole final states (such as 2hle and 2h) may initiate the desorption of positive ions from surfaces [3&l 17-1201. The presence of multi-holes in 236 M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 Auger Electrons -_Fermi Level \ IY f Valence Band 02s Ti 3p Ti*+ 02- Fig. 37. Schematic diagram of the Knotek-Feibelman (KF) model. a covalent system can result in a repulsive interaction between the unscreened nuclei and subsequent formation of ion fragments. ESD and PSD measurements from adsorbed CO, NO and N,O at excitation energies near the 0 1s core level have shown dramatic increases of O+ production (by up to a factor of 50-100) [113,122,124]. Although the core excitation cross-sections are much smaller than the valence excitation cross-sections, the ion desorption probability for excitations caused by core ionization is higher than that from valence excitations. An initial core excitation in a covalent system may lead to a two or three valence-hole Auger final state that is highly repulsive, and result in the desorption of ions. C.1 .I .4. Negative ionformation. There are important differences between desorption mechanisms for negative and positive ions from surfaces under photon and electron bombardment. ESD positive ion emission is generally initiated by valence or core ionization processes, as discussed above, and the threshold energies for ESD positive ion emission are typically > 15 eV [4,123]. However, the threshold energies to cause ESD of negative ions can be considerably smaller than for positive ions (O-15eV) [125,126] and the mechanisms of ESD of negative ions are substantially different. Negative ion production by electron impact can proceed via dissociative electron attachment (DA or DEA) and/or DD [125,126]. For electron energies from 0 to N 15 eV, negative ion (anion) formation proceeds via the temporary capture of the incident electron to form a short-lived negative ion resonance, which is dissociative in the FC region. Dissociative electron attachment (DEA) processes usually proceed via one-hole two-electron states. For a diatomic molecule AB this process may be represented by the relation e+AB+[AB]-+A+B-. (C.2) At the surface of a solid, an anion must have sufficient kinetic energy to overcome the induced electronic polarization potential to produce an ESD signal. The most detailed studies of DA processes have been performed by Sanche and co-workers on condensed and physisorbed atomic and molecular layers [75,125-1271. M. Akbulut et al. /Surface Science Reports 28 (1997) 177-245 237 I i- DEA DD .e 4- 5 _ d 2" z a ._ 2- t G l- 0 2 4 6 0 lo 12 14 16 18 20 Electron Energy (eV) Fig. 38. Typical negative fluorine ion yield as a function of electron energy from a 10 ML thick PF, layer on a Pt surface. At electron energies above about 15 eV, negative ion formation may proceed via the formation of an electronically excited neutral intermediate state which decays by DD. For a diatomic molecule this process may be represented by e+AB-+[AB]*+A++B-+e. (C.3) This occurs at energies which lie above the dissociation energy of the fragments (e.g., A+ + B-). The DD process produces a signal, which, beyond threshold, increases monotonically with electron energy. Fig. 38 shows a typical negative ion yield as a function of electron energy from a thick PF, layer on a Pt substrate [38]. C.l.2. ESDIPSD ion angular distribution and kinetic energy distribution In ESD/PSD, ions do not desorb isotropically from a surface; ions are often observed to desorb in well-defined cones determined by the bond direction which is ruptured by electron or photon excitation. Measurements of the ion angular distribution (IAD) can provide direct structural information about adsorbed species on the surface. The ESDIAD experimental technique has been successfully used to obtain direct information about the structure of the molecules oriented on surfaces, and to study dynamical aspects of surface species [3,4,128]. The initial direction of ion desorption in ESD occurs along the chemical bond direction which is ruptured by electron impact excitation, because the excitation to a localized repulsive state is short on a timescale compared to characteristic vibration times of molecules. ESD ions often desorb in M. Akbulut et al/Surface 238 Science Reports 28 (1997) 177-245 electron beam \ F, Fig. 39. Illustration F+ of the relationship between surface bond angle and ion desorption angle in ESDIAD. specific directions with respect to the substrate symmetry axis. That is, ground state bond angles are directly related to ion desorption angles in measurements of ESDIAD. Therefore, the measurement of the angular distribution of the desorbing species can provide information about the bonding geometry of species adsorbed on surfaces. As illustrated in Fig. 39, if the primary excitation leads to breaking of the C-O bond of CO bonded in a “standing up” configuration on a surface, then the oxygen ion (one of the possible desorption products) desorbs on trajectory from the surface that is perpendicular to the terrace plane. ESD of H+ from H,O bonded via the 0 atom of “inclined” OH, and ESD of F+ (6r F-) from PF, bonded through the P atom to the surface are expected to occur away from the surface normal, as shown in Fig. 39. The width of the angular distribution is determined largely by the amplitude of bending vibrations of the surface molecule [3,4]. The energy distributions of ESD/PSD ions generated at the topmost surface layer display peaked structures, and the ion energies usually range from 1-15 eV. The ion energy distribution can provide information about both the chemical state of the surface and the ESD process [129,130]. However, care must also be taken in interpretation of the ion energy distribution; the final state effects (such as image force) can influence the ion energy distribution (see Appendix C.1.3). C.l.3. Final state efects in ion desorption As ions desorb from the topmost layer of a solid surface containing either adsorbed atoms or molecules, or terminal bulk atoms, they interact with the surface in two ways which can change its state of charge and its trajectory: (a) surface neutralization and (b) image force interaction [131,132]. Surface neutralization means that there is an electron transfer from the desorbing ion to the surface o,r from the surface to the ion. This charge transfer takes place at ion-surface distances of up to a few A, and depends strongly on the electronic states of the ion and the surface. Usually, the ion M. Akbulut et al. / Surface Science Reports 28 (1997) 177-245 239 survival probability, P, is modeled as P = exp [ 7R(z(t))dt 1 Z0 with z(t) being the time dependent ion-surface separation, 2, the initial ionsurface R the reneutralization rate: R(z)=Ae-a2 (C.4) separation, and (C.5) (A and a are constants). In Appendix B, we present a more general discussion about charge transfer processes between ions and solids. The image force interaction between a desorbing ion and the surface leads to a change in the trajectory of the ion. Since the interaction is attractive, the image force causes an increase in the polar desorption angle of an ion leaving a planar surface: the ion trajectory is bent towards the surface. Since the larger the desorption angle with respect to the surface, the more time the ion spends near the surface, the image force results in an increase in the ion neutralization probability. Thus, there is preferential reneutralization of ions desorbing off-normal, which can affect the measured ion angular distribution [131,132]. Miskovic et al. [ 1321 have developed a classical model describing the influence of the image force interaction on ion desorption processes. The screened image potential in which the ion moves is approximated by v,(z) = V,l(z + z0) with the image potential at the initial ion-surface V, = - e2/4(s, + k-i). (C.6) separation s,, (C.7) Here k-’ is the Fermi-Thomas screening length, and s0 is the distance of the ion from the surface image plane at the instant the ion is formed with initial kinetic energy E, and z,, = s0 + k-i. The z-axis and the x-axis are chosen along the surface normal and parallel to the surface, respectively. Fig. 40 shows the calculated trajectories of desorbing and trapped ions in the image potentials [132]. These results show clearly that the image potential causes an increase in the polar desorption angle of an ion leaving a planar metal surface. In conclusion, the ion escape probability from a metal surface increases with increasing ion kinetic energy and with decreasing polar angle of desorption (towards the surface normal). The trajectories of the ions are most strongly bent towards the surface by the image force for larger polar angles of desorption and for small ion kinetic energies. C.l.4. Substrates There are three types of substrates which are commonly studied in ESD/PSD experiments. The first type is a metal or semiconductor surface covered by a monolayer of adsorbed gas. Another common substrate for ESD/PSD studies includes homogeneous compounds, such as oxides, fluorides, etc. The stoichiometry in the surface layers may not differ significantly from that in the bulk. Since the primary radiation in ESD or PSD experiments can penetrate well beyond the first monolayer into the solid and cause electronic excitations or ionization in subsurface layers, the desorbing particles could, in principle, originate below the surface. The question (which is addressed M. Akbulut et aLlSurface Science Reports 28 (1997) 177-245 240 I- b 3 . N 4 2 E 20 IO 30 -X* Fig. 40. Trajectories of desorbing and trapped ions in the image potential x* = x/z,y lateral coordinate, z* = z/zO: ion-surface distance (both dimensionless). BO:initial desorption polar angle, BC:critical initial desorption polar angle (for B0> 8,, the image potential will bend the ion back to the surface. From [132]. in Sections 3 to 5) is whether they are able to traverse the surface layers and escape from the surface as charged particles. A third type of substtate is multilayer adsorbates (condensed gases, etc.). Similar to the compound surfaces, a fraction of the ions may be generated below the surface, and their probability for transmission through the surface layers determines whether they can desorb from the surface. C.2. Sputtering Besides electrons and photons, ion bombardment (sputtering) of a surface can also lead to the desorption of ions or neutrals (sputtering) [95,133-1351. An ion with a kinetic energy of a few eV up to several MeV impacts a solid; depending on the energy, the ion can penetrate from a few monolayers up to several pm. On its way into the solid, the ion loses energy by collisions, which leads to motion of atoms in the solid. If the target atoms are close to the surface, they can desorb (either as ions or as neutrals); in some instances, these atoms can collide with and eject other atoms from the surface layer [27,28-J. These processes are indicated in Fig. 41. The two basic physical processes that lead to the sputtering of solids by energetic ion bombardment are knock-on (nuclear elastic) collisions and electronic excitations. When an energetic ion collides with a target solid atom it can lose a substantial fraction of its energy to the target atom; the M. Akbulut et al. 1 Surface Science Reports 28 (1997) 177-245 241 ion - 1_--- secondary secondary _--. ions, neutrals solid Fig. 41. Illustration of sputtering. A primary ion impacts on a surface and penetrates into the solid. Secondary ions and neutrals are emitted from the surface through interaction of the primary ion with the solid and through secondary processes. target atom recoils into the solid. If the energy transferred to the target atom is larger than the local binding energy of atoms within the solid, the target particle overcomes the binding forces. It can either escape from the solid or collide with other target atoms and transfer energy and momentum to the surface, leading to substantial atom displacement in the solid and particle ejection. This process is the so-called “knock-on sputtering” or “nuclear sputtering”. The sputtering yield depends strongly on the amount of energy deposited at the surface and the sublimation energy of the materials. The number density of atoms set in motion is a very important parameter determining the momentum transport in the solid. In electronic sputtering, the incident ion causes excitations of the solid. Secondary ions or neutrals can be created as a result of these processes. Electronic excitations and ionizations in the solid can be produced two ways by ion impact. First, the fast ion can transfer sufficient energy to cause excitation or ionization during a close collision with a target electron in a binary encounter. Second, a fast charged particle produces a time varying field. If the frequency components of this field can be absorbed by a target atom, this can initiate dipole excitations and ionizations. Sputtering in metals is mainly due to the energy deposited in elastic nuclear collisions, whereas sputtering in inorganic insulators (such as rare gas solids, water ice, carbon dioxide ice) is determined by the electronic energy deposition processes. Since in condensed atomic and molecular films (large band gap insulators) the electronic excitations are localized, and the electronic excitation energies are larger than the cohesive energies, the electronic energy deposition by ions can result in the motion of lattice atoms or molecules [136-1381. This is not the case for metals; because the M. Akbulut et al. 1 Surface Science Reports 28 (1997) 177-245 242 electronic excitations in metals diffuse rapidly from their origin around the incident particle track, the energy density is not sufficient to eject an atom or molecule from the solid. The ability of a solid to slow down and stop an incident high-energy ion at an incident energy E, at some characteristic depth is expressed by the stopping power. The total stopping power can be expressed to a first approximation as a sum of the nuclear (elastic) and electronic (inelastic) contributions, dE,/dx = (dE,/dx), + (dE,/dx),. (C.8) The first term is the elastic-nuclear stopping power, while the second term is the electronic stopping. The total stopping power can also be expressed in terms of the nuclear and electronic cross-sections W,) and UE,) as dE,W = N{W,) + W,))> (C-9) where N is the material number density. The reader interested in more details about sputtering and ion stopping is referred to books by Johnson [95] and Nastasi et al. [6]. One of the main differences between sputtering and ESD/PSD processes is that in ESD/PSD only a few bonds are broken in the solid due to the (low-energy) electron/photon impact, while a high-energy ion can lead to significant beam damage (atom displacement) in the vicinity of its impact on the solid. Hence the question of the depth of origin of ions becomes more complicated: Instead of a situation where the ions generated below the surface have to traverse nearly unperturbed surface layers (in ESD/PSD) [29], the surface layers can be disturbed significantly in the case of sputtering. This can affect the depth of origin of sputtered ions. For instance, the atoms/molecules from the surface layers may be displaced by the primary ion so that the secondary ion from a subsurface layer can escape without attenuation through a “crater” in the surface layers. References [l] A. Benninghoven, F.G. Riidenauer and H.W. Werner, Secondary Ion Mass Spectrometry, Eds. P.J. Elving and J.D. Winefordner, Chemical Analysis (Wiley, New York, 1987). [2] M.L. Knotek, Rep. Prog. Phys. 47 (1984) 1499. [3] R.D. Ramsier and J.T. Yates, Jr., Surf. Sci. Rep. 12 (1991) 243. [4] T.E. Madey, Science 234 (1986) 316. [S] J.A. Yarmoff, S.A. Joyce, C.W. Lo and J. Song, in: Desorption Induced by Electronic Transitions DIET IV, Eds. G. Betz and P. Varga (Springer, Berlin, 1990) p. 65. [6] M. Nastasi, J.W. Mayer and J.K. Hirvonen, IonSolid Interactions: Fundamentals and Applications (Cambridge University Press, Cambridge, 1996). [7] Y. Lifshitz, G.D. Lempert and E. Grossman, Phys. Rev. Lett. 72 (1994) 2753. [S] R.D. Levine and R.B. Bernstein, Molecular Reaction Dynamics and Chemical Reactions (Oxford University Press, New York, 1987). [9] D.M. Goodstein, C.A. Dirubio, B.H. Cooper and K. Burke, Surf. Rev. Lett. 1 (1994) 175. [lo] J. Harris, in: Dynamics of Gas-Surface Interactions, Eds. C.T. Rettner and M.N.R. Ashfold (The Royal Society of Chemistry, Cambridge, 199 1) p. 1. [ 1l] H. Liith, Surfaces and Interfaces of Solids (Springer, Berlin, 1993). [12] F. Linder, in: Electronic and Atomic Collisions, Eds. H.B. Gilbody, W.R. Newell, F.H. Read and A.C. Smith (Elsevier, Amsterdam, 1988) p. 287. M. Akbulut et al. 1 Surface Science Reports 28 (1997) 177-245 243 [ 131 E.W. McDaniel, J.B.A. Mitchell and M.E. Rudd, Atomic Collisions: Heavy Particle Projectiles (Wiley, New York, 1993). [ 141 P. Nordlander, Scanning Microsc. Suppl. 4 (1990) 353. [ 151 N. Schwentner, F. Himpsel, V. Saile, M. Skibowski, W. Steinmann and E.E. Koch, Phys. Rev. Lett. 34 (1975) 528. [16] N. Schwentner, E.-E. Koch and J. Jortner, Electronic Excitations in Condensed Rare Gases, Springer Tracts in Modern Physics (Springer, Berlin, 1985). [17] C. Bressler and N. Schwentner, Phys. Rev. Lett. 76 (1996) 648. [lS] M.F. Dumke, T.A. Tombrello, R.A. Weller, R.M. Housley and E.H. Cirlin, Surf. Sci. 124 (1983) 407. [19] J.W. Burnett, J.P. Biersack, D.M. Gruen, B. Jorgensen, A.R. Kraus, M.J. Pellin, E.L. Schweitzer, J.T. Yates, Jr. and C.E. Young, J. Vat. Sci. Technol. A 6 (1988) 2064. [20] J.P. Biersack and J.F. Ziegler, Nucl. Instr. Meth. 194 (1982) 93. [21] D.E. Harrison, P.W. Kelly, B.J. Garrison and N. Winograd, Surf. Sci. 76 (1978) 311. [22] M. Vicanek, J.J.J. Rodriguez and P. Sigmund, Nucl. Instr. Meth. Phys. Res. B 36 (1989) 124. [23] P. Sigmund et al., Nucl. Instr. Meth. Phys. Res. B 36 (1989) 110. [24] T. Lill, W.F. Calaway, Z. Ma and M.J. Pellin, Surf. Sci. 322 (1994) 361. [25] T. Lill, W.F. Calaway and M.J. Pellin, J. Appl. Phys. 78 (1995) 505. [26] L.J. Kieffer, At. Data 1 (1969) 19. [27] Q.J. Zhang and R. Gomer, Surf. Sci. 109 (1981) 567. [28] Q.J. Zhang, R. Gomer and S.R. Bowman, Surf. Sci. 129 (1983) 535. [29] N.J. Sack, M. Akbulut and T.E. Madey, Phys. Rev. B 51(1995) 4585. [30] M. Akbulut, N.J. Sack and T.E. Madey, J. Chem. Phys. 103 (1995) 2202. [31] N.J. Sack, M. Akbulut and T.E. Madey, Nucl. Instr. Meth. Phys. Res. B 90 (1994) 451. [32] N.J. Sack, M. Akbulut and T.E. Madey, Phys. Rev. Lett. 73 (1994) 794. [33] D.A. King T.E. Madey and J.T. Yates, Jr., J. Chem. Phys. 55 (1971) 3236. [34] D.A. King T.E. Madey and J.T. Yates, Jr., J. Chem. Phys. 55 (1971) 3247. [35] T.E. Madey, N.J. Sack and M. Akbulut, Nucl. Instr. Meth. Phys. Res. B 100 (1995) 309. [36] N.J. Sack, M. Akbulut and T.E. Madey, Surf. Sci. 334 (1995) L695. [37] M. Akbulut, N.J. Sack and T.E. Madey, Phys. Rev. Lett. 75 (1995) 3414. [38] M. Akbulut, T.E. Madey, L. Parenteau and L. Sanche, J. Chem. Phys. 105 (1996) 6032,6043. [39] M.B. Raschke and T.E. Madey, in preparation (1997). [40] Z. Ma, M. Akbulut and T.E. Madey, in preparation (1997). [41] P. Klein, M. Vicanek and H. Urbassek, Phys. Rev. B 51 (1995) 4597. [42] J.M. Layet, M. Bienfait, C. Ramseyer, P.N.M. Hoang, C. Girardet and G. Coddens, Phys. Rev. B 48 (1993) 9045. [43] V.N. Ageev, Surf. Sci. 336 (1995) 321. [44] D.L. Albritton, At. Data Nucl. Data Tables 22 (1978) 1. [45] U. Diebold and T.E. Madey, Phys. Rev. Lett. 72 (1994) 1116. [46] V.N. Ageev, N.D. Potekhina, B.V. Yakshinskii, M. Akbulut and T.E. Madey, Phys. Rev. B 54 (1996) R5271. [47] M.L. Yu, Surf. Sci. 84 (1979) L493. [48] D. Rapp and W.E. Francis, J. Chem. Phys. 37 (1962) 2631. [49] N.J. Sack, M. Akbulut, T.E. Madey, P. Klein, H.M. Urbassek and M. Vicanek, Phys. Rev. B 54 (1996) 5130. [SO] M. Akbulut, T.E. Madey and P. Nordlander, J. Chem. Phys. 106 (1997) 2801. [Sl] S.J. Dixon-Warren, E.T. Jensen and J.C. Polanyi, in: Desorption Induced by Electronic Transitions DIET V, Eds. A.R. Burns, E.B. Stechel and D.R. Jennison (Springer, Berlin, 1992) p. 1. [52] L. Sanche and L. Parenteau, J. Chem. Phys. 93 (1990) 7476. [53] L. Sanche, in: Desorption Induced by Electronic Transitions DIET V, Eds. A.R. Burns, E.B. Stechel and D.R. Jennison (Springer, Berlin, 1993) p. 3. [54] L. Sanche, private communication (1995). [55] H. Allali, B. Cabaud, G. Fuchs, A. Hoareau, B. Nsouli, J.-P. Thomas, M. Treilleux and J.-S. Danel, Nucl. Instr. Meth. Phys. Res. B 84 (1994) 303. [56] N.J. Sack, L. Nair and T.E. Madey, Surf. Sci. 310 (1994) 63. [57] L. Nair, N.J. Sack and T.E. Madey, Nucl. Instr. Meth. Phys. Res. B 101 (1995) 79. [58] L. Sanche, Phys. Rev. Lett. 53 (1984) 1638. 244 M. Akbulut et al. 1 Surface Science Reports 28 (1997) 177-245 [59] H. Sambe, D.E. Ramaker, L. Parenteau [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [SO] [81] [82] [83] [84] [SS] [86] [87] [SS] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [loo] [loll [102] [103] and L. Sanche, Phys. Rev. Lett. 59 (1987) 236. T.E. Madey and F.P. Netzer, Surf. Sci. 117 (1982) 549. C. Benndorf and T.E. Madey, Surf. Sci. 135 (1983) 164. F.P. Netzer and T.E. Madey, Surf. Sci. 119 (1982) 422. M.R. McCartney and D.J. Smith, Surf. Sci. 221(1989) 214. T. Ustare, J. Arik and M. Elango, Appl. Phys. Lett. 65 (1994) 2551. M.F. Guest, A. Ding, J. Karlau, J. Weise and I.H. Hillier, Mol. Phys. 38 (1979) 1427. J.W. Rabalais, private communications (1994). E. Rimini, Ion Implantation: Basics to Device Fabrication (Kluwer, Boston, 1995). H. Ryssel and I. Ruge, Ion Implantation (Wiley, Chichester, 1986). P. Singer, in: Semicond. Int. (April 1996) p. 72. D.F. Downey, R.J. Eddy and S. Mehta, Nucl. Instr. Meth. Phys. Res. B 74 (1993) 160. A.J. Bard et al., J. Phys. Chem. 97 (1993) 7147. J.O. Bockris and A.K.N. Reddy, Modern Electrochemistry (Plenum, New York, 1970). D.L. Griscom, D.B. Brown and N.S. Saks, in: Proc. Symp. on the Physics and Chemistry of SiO, and the Si-SiO, Interface, Eds. CR. Helms and B.E. Deal (Plenum, New York, 1988) p. 287. D.J. Dimaria and M.V. Fischetti, in: Excess Electrons in Dielectric Media, Eds. C. Ferradini and J.-P. Jay-Gerin (CRC Press, Boca Raton, 1991) p. 315. L. Sanche, IEEE Trans. Electr. Insul. 28 (1993) 789. R.M. Marsolais, E.A. Cartier and P. Pfluger, in: Excess Electrons in Dielectric Media, Eds. C. Ferradini and J.-P. Jay-Gerin (CRC Press, Boca Raton, 1991) p. 43. D.R. Bates, Atomic and Molecular Processes (Academic Press, New York, 1962). H. Massey and H. Gilbody, Electronic and Ionic Impact Phenomena (Clarendon, Oxford, 1974). P.G. Bertrand and J.W. Rabalais, in: Low Energy Ion-Surface Interactions, Eds. J.W. Rabalais (Wiley, Chichester, 1994) p. 53. G.W.F. Drake, Eds., Atomic, Molecular and Optical Physics Handbook (AIP Press, New York, 1996). H. Niehus, W. Heiland and E. Taglauer, Surf. Sci. Rep. 17 (1993) 213. E. Taglauer, in: Ion Spectroscopies for Surface Analysis, Eds. A.W. Czanderna and D.M. Hercules (Plenum, New York, 1991). W. Heiland and E. Taglauer, Nucl. Instr. Meth. 132 (1976) 535. E. Hulpke, Surf. Sci. 52 (1975) 615. V.I. Veksler, Sov. Phys. JETP 17 (1963) 9. I.M. Torrens, Interaction Potentials (Academic Press, New York, 1972). C. Kubach, C. Benoit, V. Sidis, J. Pommier and M. Barat, J. Phys. B: At. Mol. Phys. 9 (1976) 2073. R.E. Johnson and J.W. Boring, in: Collision Spectroscopy, Ed. R.G. Cooks (Plenum, New York, 1978) p. 91. J.B. Hasted, Physics of Atomic Collisions (Elsevier, New York, 1972). E.Y. Kamber and C.L. Cocke, in: Physics of Ion Impact Phenomena, Ed. D. Mathur (Springer, Berlin, 1991). J. Appell, in: Collision Spectroscopy, Ed. R.G. Cooks (Plenum, New York, 1978) p. 227. S. Sakabe and Y. Izawa, At. Data Nucl. Data Tables 49 (1991) 257. H.S.W. Massey, E.H.S. Burhop and H.B. Gilbody, Electronic and Ionic Impact Phenomena (Oxford University Press, London, 1974). W.W. Duley and D.A. Williams, Interstellar Chemistry (Academic Press, London, 1984). R.E. Johnson, Energetic Charged-Particle Interactions with Atmospheres and Surfaces (Springer, Berlin, 1990). B.H. Bransden and M.R.C. McDowell, Charge Exchange and the Theory of Ion-Atom Collisions (Clarendon, Oxford, 1992). T. Su and M.T. Bowers, in: Gas Phase Ion Chemistry, Ed. M.T. Bowers (Academic Press, New York, 1979) p. 83. R.D. Levine and R.B. Bernstein, J. Chem. Phys. 56 (1972) 2281. T. Turner, 0. Dutuit and Y.T. Lee, J. Chem. Phys. 81(1984) 3475. D.J. Auerbach, M.M. Hubers, A.P. Baede and J. Los, Chem. Phys. 2 (1973) 107. W.W. Duley and D.A. Williams, Interstellar Chemistry (Academic Press, London, 1984). M.L. Yu and N.D. Lang, Nucl. Instr. Meth. Phys. Res. B 14 (1986) 403. P. Nordlander, Surf. Sci. Rep. in preparation (1997). M. Akbulut et al. /Surface Science Reports 28 (1997) [ 1041 [105] [106] [lo73 [lOS] [ 1091 [l lo] [ill] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [ 1351 [136] [137] [138] [139] Cl403 [141] 177-245 245 J.C. Tully, Phys. Rev. B 16 (1977) 4324. P. Nordlander and N.D. Lang, Phys. Rev. B 44 (1991) 13681. B.J. Garrison, A.C. Diebold, J.-H. Lin and Z. Sroubek, Surf. Sci. 124 (1983) 461. J.A. Olson and B.J. Garrison, Nucl. Instr. Meth. Phys. Res. B 14 (1986) 414. S.A. Joyce, C. Clark, V. Chakarian, D.H. Shuh, J.A. Yarmoff, T.E. Madey, P. Nordlander, B. Maschhoff and H.-S. Tao, Phys. Rev. B 45 (1992) 14264. G.A. Kimmel, D.M. Goodstein, Z.H. Levine and B.H. Cooper, Phys. Rev. B 43 (1991) 9403. P. Avouris and R.E. Walkup, Annu. Rev. Phys. Chem. 40 (1989) 173. N.H. Tolk, J.C. Tully and T.E. Madey, Eds., Desorption Induced by Electronic Transitions DIET I (Springer, Berlin, 1983). A.R. Burns, E.B. Stechel and D.R. Jennison, Eds., Desorption Induced by Electronic Transitions DIET-V(Springer, Berlin, 1992). D. Menzel, Nucl. Instr. Meth. Phys. Res. B 13 (1986) 507. D. Menzel, Appl. Phys. A 51 (1990) 163. M.L. Knotek and P.J. Feibelman, Surf. Sci. 90 (1979) 78. M.L. Knotek, Phys. Ser. T6 (1983) 94. D.E. Ramaker, J. Chem. Phys. 78 (1983) 2998. D.E. Ramaker, Chem. Phys. 80 (1983) 183. J.A. Kelber and M.L. Knotek, Surf. Sci. 121 (1982) L499. J.A. Kelber and M.L. Knotek, J. Vat. Sci. Technol. A 1 (1983) 1149. P. Feulner T. Mtiller, A. Puschmann and D. Menzel, Phys. Rev. Lett. 59 (1987) 791. R. Franchy and D. Menzel, Phys. Rev. Lett. 40 (1979) 7427. T.E. Madey, D.E. Ramaker and R. Stockbauer, Annu. Rev. Phys. Chem. 35 (1984) 215. P. Feulner, R. Treichler and D. Menzel, Phys. Rev. B 24 (1981) 7427. L. Sanche, in: Excess Electrons in Dielectric Media, Eds. C. Ferradini and J.-P. Jay-Gerin (CRC Press, Boca Raton, 1991) p. 1. L. Sanche, in: Linking the Gaseous and Condensed Phases of Matter, Eds. L.G. Christophorou, E. Illenberger and W.F. Schmidt (Plenum, New York, 1994) p. 1. L. Sanche, Scanning Microsc. 9 (1995) 619. N.J. Sack and T.E. Madey, Surf. Sci. 347 (1996) 367. S. Prigge, H. Niehus and E. Bauer, Surf. Sci. 75 (1978) 635. D.P. Woodruff, M.M. Traum, H.H. Farrell, N.V. Smith, P.D. Johnson, D.A. King, R.L. Benbow and Z. Hurych, Phys. Rev. B 21 (1980) 5642. Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci. 169 (1986) 405. Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci. 141 (1984) 285. W.L. Brown, Nucl. Instr. Meth. Phys. Res. B 32 (1988) 1. W.L. Brown, W.M. Augustyniak, L.J. Lanzerotti, R.E. Johnson and R. Evatt, Phys. Rev. Lett. 45 (1980) 1632. C.T. Reimann, W.L. Brown and R.E. Johnson, Phys. Rev. B 37 (1988) 1455. R.E. Johnson and J. Schou, in : Fundamental Processes in Sputtering of Atoms and Molecules (SPUT 92), Ed. P. Sigmund (Danish Academy and Science and Letters, Copenhagen, 1992) p. 403. W.L. Brown, L.J. Lanzerotti and R.E. Johnson, Science 218 (1982) 525. W.L. Brown, W.M. Augustyniak, E. Simmons, K.J. Marcantonio, L.J. Lanzerotti, R.E. Johnson, J.W. Boring, C.T. Reimann, G. Foti and V. Pirronello, Nucl. Instr. Meth. Phys. Res. 198 (1982) 1. T.A. Green, H.E. Stanley and Y.-C. Chiang, Helv. Phys. Acta 38 (1965) 109. R.F. Stebbings, A.C.H. Smith and H. Ehrhardt, J. Geophys. Res. 69 (1964) 2349. D.W. Koopman, in: Proc. 5th Conf. on Physics of Electronic and Atomic Collisions (Nauka, Leningrad, 1967).
© Copyright 2026 Paperzz