The in ﬂ uence of oxygen on the neutralization of slow helium ions scattered from transition metals and aluminum surfaces

Low energy ion scattering (LEIS) was employed for the analysis of thin ﬁ lms of Mo, Ru, Hf, Al and their oxides. Measurements with di ﬀ erent He + energies showed that the characteristic velocities for neutralization of the transition metal atoms change when the metal binds with oxygen. However, such behavior was not observed for aluminum. We suggest that the increased neutralization in oxidized Ru, Hf and Mo originates from the presence of the O 2s band. This band is in resonance with the He 1s level, which allows for a quasi-resonant neutralization mechanism (qRN). On the other hand, a decrease of the strong Auger neutralization for metallic Al upon oxidation may compensate for the increase in neutralization by qRN, leading to similar neutralization behavior of Al in both states. We also demonstrate the dependence of characteristic velocity on oxygen content and discuss how this e ﬀ ect can be used to select proper reference samples for quantitative surface analysis by LEIS. measurements and the re-liability of obtained data for surface quanti ﬁ cation of metal oxides. To that end, we investigated the charge exchange of He + on metal and metal oxide samples of transition metals Mo, Hf and Ru, and Al as well-studied reference material. The measurements with di ﬀ erent He + energies show how oxide formation a ﬀ ects ion neutralization and may lead to misinterpretation of data when comparative analyses are performed. We further investigate ways to overcome these drawbacks and obtain a reliable quantitative analysis. This work shows that quantitative surface characterization of transition metal compounds by low


Introduction
The characterization of oxide surfaces has been a subject of considerable research effort for a long time. The control of surface composition is often crucial to achieving the optimal properties of oxide layers in the many areas in which these compounds can be applied, ranging from electrochemistry [1,2] to catalysis [3,4] and microelectronics [5]. In this context, Low Energy Ion Scattering (LEIS) appears as a valuable tool for quantitative analysis, since it provides relatively fast and straightforward way to measure the atomic composition of the topmost layer of materials [6]. This extremely low information depth arises from the strong neutralization of noble gas ions at the applied energy range (typically from 1 to 8 keV), and is unparalleled by other surface analytical techniques, such as X-ray photoelectron spectroscopy (XPS) [7,8] and secondary ion mass spectrometry (SIMS) [9][10][11], which are commonly applied in oxide characterization.
Quantification of surface composition by LEIS is usually done by comparing measured peak intensities of every element against peak intensities of reference samples, which can be either pure elements or their known compounds [6,12]. However, this approach is only valid if the neutralization efficiency of an ion scattered from a specific surface atom does not depend on the surrounding species, that is, no matrix effect due to the different elements is present [13]. In the LEIS regime, many cases have been reported where matrix effects do not play a role in ion-surface interactions [6]. However, several studies have shown that neutralization processes can depend on the chemical environment for a range of compounds [14][15][16][17]. A matrix effect was suspected for NiO [18], but the conclusive proof would require characteristic velocity measurements, which were not performed. Bruckner et al. [19] verified the influence of surface oxygen on the reionization of He + scattered on sub-surface Ta and Zn, showing a significant dependence of the obtained yield on the presence of oxygen. These facts indicate that there is still a lack of knowledge on the role of oxygen on charge exchange in surface processes as studied by LEIS.
Recent characteristic velocity measurements on He + neutralization by elemental Ru and Ru in RuO 2 films were performed in our group [20], and He + neutralization by Ta in Ta 2 O 5 were performed by Bruckner et al. [21]. These measurements provided a direct evidence of a matrix effect caused by oxygen. The present work aims to provide a deeper understanding of the role of oxygen in low energy ion neutralization, verifying the limitations of LEIS measurements and the reliability of obtained data for surface quantification of metal oxides. To that end, we investigated the charge exchange of He + on metal and metal oxide samples of transition metals Mo, Hf and Ru, and Al as wellstudied reference material. The measurements with different He + energies show how oxide formation affects ion neutralization and may lead to misinterpretation of data when comparative analyses are performed. We further investigate ways to overcome these drawbacks and obtain a reliable quantitative analysis. This work shows that quantitative surface characterization of transition metal compounds by low energy ion scattering needs proper choice of reference samples and highlights the importance of investigation of ion-surface charge-exchange mechanisms.

Experimental
The experiments were performed in a home-designed ultrahigh vacuum system at a base pressure of ≤1 × 10 −9 mbar, which allows for in-vacuum transfers between deposition, LEIS and XPS chambers with negligible accumulation of surface contamination. Films of typically 20 nm Mo, Hf, Ru, Al and respective oxides were deposited onto natively oxidized super-polished Si substrates by DC magnetron sputtering at room temperature. The working gas for metal deposition was argon, with an average pressure of 5 × 10 −4 mbar. Oxides were obtained by reactively sputtering the respective metallic target with a mixture of oxygen and argon, for which the proportion between O 2 and Ar flow was adjusted in order to guarantee the formation of stoichiometric oxides.
To verify the chemical composition and measure the valence band states of the deposited films, XPS measurements were performed with a Thermo Theta Probe spectrometer using Al-Kα radiation.
LEIS measurements were performed in an ION-TOF GmbH Qtac 100 high sensitivity LEIS spectrometer. The system is equipped with a double toroidal electrostatic analyzer (DTA) and an electron impact source with ion incidence angle normal to the surface of the sample and scattering angle of 145°. The characteristic velocity for neutralization was measured by the analysis of deposited samples with a He + ion beam, with energies ranging from 1 up to 6 keV. The respective beam current was measured with a Faraday cup before each spectrum was acquired. Whenever sample sputtering was performed, Ar + ions with 500 eV energy and an average 100 nA current were applied at an angle of 59°with respect to the surface normal.

Characteristic velocity calculation
In order to study He + neutralization of metal oxides, we applied an established method for the study of matrix effects in LEIS [14,15,22]. This method is based on the determination of the characteristic velocity of He + neutralization during scattering from surface atoms of given materials (v c ), a measure of the neutralization efficiency. In LEIS, the measured signal S i (in counts per ion dose) from an element i is dependent on the fraction of scattered ions ( + i , commonly referred to as ion fraction), the differential scattering cross section dσ dΩ i of the element at the applied ion energy, the atomic surface concentration of the element N i , ξ an instrumental factor including detector solid angle, detector efficiency and analyzer transmission, and the surface roughness correction factor R, as expressed in Eq. (1): where e is the electron charge. Different models for the ion fraction have been reported in literature over the years [6,[23][24][25]. In this work, we apply the Hagstrum model [25], in which the electron density is considered homogeneous and the neutralization rate is assumed to depend only on the distance between ion and surface. According to this model, the ion fraction + i can be written as: where v 1 represents inverse incident ion velocity, calculated as + with v 0⊥ and v f⊥ the surface normal components of the incoming and scattered ion velocities.
By substituting Eq. (2) in Eq. (1), a linear relationship between the logarithm of LEIS signal divided by the cross section and the characteristic velocity is obtained [6,14,23]: with = C ξR e . In Eq. (3), v c serves as a slope and ln(C × Ni) as a vertical offset of the line. Therefore, v c and N i of an element can be determined by plotting the S i /(dσ i /dΩ), measured at different energies, as a function of the inverse velocity. If no matrix effect is observed, for the same element at surfaces with different concentrations or composition, the plotted lines will present a fixed slope (v c ), but different vertical offsets.  Table 1 shows the characteristic velocities (v c ) of each system, extracted with the help of Eq. (3). The data points presented for each set were obtained from measurements on (at least) three different samples. The energy range corresponds to those energies for which consistent measurements and a reliable fit of the binary collision peak were obtained. For the metals, before each measurement, the surface was sputter cleaned until signal saturation of the metal peak, to exclude a possible influence of trace amounts of oxygen on the measurement results. In this work, the differential cross section was obtained by using the universal ZBL potential [26].

LEIS measurements
For clarification purposes, the characteristic velocity of He +   ("Ox" as a placeholder for the referred oxide). The characteristic velocity of He + neutralization by oxygen in an oxide surfaces will be referred as v c O Ox , . It is clear from Table 1 that Mo, Ru and Hf exhibit differences in He + neutralization efficiency between metal and oxide surfaces. On the other hand, Al shows similar neutralization efficiency for both. On both Fig. 2e and Table 2 it is possible to note the large scatter of data points between different oxide surfaces and, consequently, calculated v c O values. This scatter comes from the lower intensity of the oxygen peaks, a consequence of the lower elemental sensitivity factor of oxygen [27]. However, it is possible to note an oscillatory behavior in Fig. 2e, which is consistent between all four datasets. This behavior strongly resembles the oscillatory behavior of (quasi-)resonant charge transfer (qRCT) [15], a neutralization mechanism further described in session 3.3. This observation directly contradicts the previous work of Tellez et al. [27], where the inverse velocity curves for both oxygen isotopes 16 and 18 are described by linear trends. Since no similarly strong oscillations can be assigned to the data from metals ( Fig. 2a-d), we can exclude the possibility of these oscillations being an artifact of our measurements.
To verify possible effects of sputtering and composition change on the characteristic velocities, characteristic velocity plots were also made based on sputter depth profiles. For each He + energy a sputter depth profile was made on a fresh spot of the sample. The data from those sputter depth profiles were then combined to yield the characteristic velocity plots as a function of sputter ion fluence, and hence as a function of oxygen concentration. This approach was chosen due to the low stability of a surface at an intermediate oxidation state, which might have led to composition variation during the time when sequential measurements with different primary energies were made. This experiment was performed on two metals: Mo and Al, which were chosen as two cases where changes in characteristic velocity are present and absent, correspondingly. Sputter effects on oxide samples were verified by performing the described analysis on 20 nm reactively deposited oxides, with each sputter step removing an average of 0.3 to 0.5 nm of oxide. These samples served as a control group. The effect of composition change of oxygen versus metal was studied by analyzing samples of few nanometer oxide (2-3 nm) deposited on metal. For those samples, sputter depth profile was performed all the way through the oxide layer, with each sputter step removing an average of 0.3 to 0.5 nm of oxide, until saturation of the metal signal was obtained. An example of signal evolution at a specific primary energy and the inverse velocity plot obtained for each sputter step is shown in Fig. 3a-h for all analyzed samples. Lower elemental sensitivity factor of oxygen lowers the quality of the data presented in Fig. 4b. Furthermore, due to preferential sputtering, the oxygen peak rapidly vanishes when approaching the metallic surface. Since each depth profile is performed at a different position on the sample, this fast change serves as additional source of uncertainty when combining measurements with the same fluence together. For these reasons, the interpretation of the data related to neutralization efficiency of O atoms with decrease of its surface concentration is not straightforward. Nevertheless, the obtained data indicates that the neutralization efficiency of oxygen is not systematically different between different thick deposited oxide surfaces. Further comments on the characteristic velocities of oxygen related to the transition from oxide to elemental Al and Mo are provided in Section 3.3.
To recognize the origin of the (not) observed differences between neutralization of He + by metal atoms on oxide and metallic surfaces, charge-transfer processes between the projectile and the sample surface (target) must be explored.

Determination of neutralization mechanisms
In LEIS, the mechanisms of charge transfer between a target and a projectile can be divided into two main groups: resonant processes and Auger processes [28]. Auger processes are two-electron processes in which one electron is transferred to an unoccupied level of the projectile, and the energy of the system is conserved by the creation of surface excitations (electron-hole pair or plasmons) [29,30]. Considering that in the LEIS regime the ion velocity is much smaller than the Fermi velocity of the metal electrons, Auger ionization processes are neglected in this study [30,31]. In Auger neutralization (AN), one electron from the surface is transferred to a bound state (often the ground state) of the projectile, creating surface excitations. AN along the trajectory is possible at any primary projectile energy [32].
Resonant processes are single electron mechanisms in which electron tunneling from a projectile to a target or vice versa takes place when the projectile energy level and a state or continuum of states of the solid are degenerate in energy. However, one must be aware that when a projectile is adjacent to a sample surface, the electronic levels of the projectile are modified with respect to the static levels at infinite distance: the projectile levels shift (and broaden) during the collision, a consequence of the interaction of the projectile states with the valence and core electrons of the target [6,33]. In this context, the promotion of the He levels is important, as this may lead to an energy alignment not observed when the atom and target are in an unperturbed state [34,35]. In the LEIS regime, resonant processes are classified according to the pair of energy levels of target and projectile between which the charge transfer takes place: collision-induced neutralization or reionization between the promoted ground state of the projectile (He 1s, in the present case) and lowest occupied states of target conduction band (metals) or valence band (non-metals) (CIN/CIR, also known as resonant neutralization/reionization in close collision); resonant neutralization from the highest occupied states of conduction band of a target to an excited state of a projectile (C-RN); (quasi-)resonant charge transfer between bound levels of a target and the ground state of the projectile (qRCT); and (quasi-)resonant neutralization from the valence band to the ground state of the projectile (VB-qRN) [6,16,20,[35][36][37][38].
The sum of all mechanisms present in the projectile-target interaction will determine the observed ion fraction related to a specific LEIS signal. For incident ion energy higher than the reionization threshold, we can describe the ion fraction from a single scattering ( + ) by [35]: in RN out in RI out (4) where + P in / + P out is the probability of incident ions surviving AN in the way in/out; P RI the reionization probability by resonant processes: CIR and, when present, reionization by qRCT. P RN is the neutralization probability by resonant processes: CIN and, when present, C-RN, qRCT and VB-qRN. Therefore, the first term describes ions that remained charged along the entire trajectory and the second one describes ions that were initially neutralized, then reionized and survived AN on the way out.
CIN and CIR appear when the projectile energy exceeds a threshold, leading the projectile to be sufficiently close to the target. This proximity will promote the projectile ground state to align with the bottom of the target band (conduction or valence), enabling charge exchange [6,14,35,39]. The presence of this process is clear due to the appearance of an angle in the slope of the energy dependence of the ion yield, as collision-induced neutralization is stronger than reionization [15,40]. The chosen measurement range for the present experiments is above the reionization threshold (E th ) for the studied metals [38]. Therefore, the influence of such mechanism is expected to be constant for each ion-target combination.
As previously stated, C-RN occurs when an excited state of the projectile and conduction band of the target are in resonance. As explained in detail in the work of Cortenraad et al. [14], this phenomenon is expected for materials with work function values below 3.5 eV [14,15]. Considering the work function values of both metallic and oxidized surfaces of the analyzed materials [41], such effect is not expected in this study.
The qRCT process is active for materials with an atomic level nearly resonant with the unperturbed He-1s level [6,42,43]. If the mismatch between these interacting levels is small, the transition rates for neutralization and reionization of qRCT will be similar. The charge state of the scattered He will oscillate between He 0 and He + as a function of interaction time with the target atom, which depends on the incident ion energy. Therefore, qRCT is the only mechanism to have an oscillatory ion velocity dependence [15,44]. However, when He 1s is (quasi-)resonant with a band of energy states, or has a large mismatch with the interacting level, quasi-resonant neutralization (qRN) becomes much stronger than reionization, which leads to damping of the oscillations [43,45]. These neutralization phenomena are then classified as VB-qRN and CIN. As previously mentioned, the latter occurs at very small distances between target nucleus and the projectile, at which the ground level of the projectile is promoted to energy high enough to become resonant with the valence band. On the other hand, VB-qRN occurs when the valence band is wide enough to become resonant with the ground state of the projectile even without strong level promotion, an effect first demonstrated for He neutralization by graphitic carbon [16]. Zameshin et al. [20] observed the presence of VB-qRN for carbides and borides, in particular a continuous change in characteristic velocity of Ru as a function of the amount of B in the film. In the mentioned study, the wide valence band with low lying energy states (as low as 20 eV below Fermi level) for Ru-B and Ru-C films leads to VB-qRN, which was not observed in elemental Ru, as the lowest lying states of the metal are found at 7.5 eV. The difference in energy levels resulted in changes of neutralization efficiency, i.e. matrix effects for the compound surfaces. In the same paper, it was hypothesized that qRN-related matrix effects of a similar mechanism would appear in metal oxides, with a proof-ofprinciple experiment of Ru from an oxidized surface showing  differences in characteristic velocities of Ru from a pristine surface. The resonance was considered to happen between He 1s and the O 2s levels present in the oxide [20].
To verify the energy distribution within the materials, XPS valence band analysis was performed. For this, a thin oxide (2 -4 nm) on top of the corresponding elemental metal was analyzed by XPS. This thin oxide film was obtained by exposing the deposited metal to atomic oxygen at room temperature. With this method, a stoichiometric oxide with a known thickness is grown by controlling the oxygen exposure time [46]. Subsequently, the oxide was completely sputtered and the metallic surface was analyzed under the same conditions. Fig. 5 displays the XPS valence band analysis for metallic and oxidized surfaces of the investigated elements. For HfO 2 , the valence band features are overwhelmed by the Hf 4f core levels that are present in this energy range. It is important to note that values obtained by XPS are relative to the Fermi level, while energy levels for He (ions) are usually referred from vacuum level. Assuming a typical value of work function of 4 eV, He 1 s (with ionization energy of 24.6 eV) will be in resonance with a band as long as it presents energies close to 20.6 eV [20].
In Fig. 5 it is possible to note the contribution of the underlying metal layer to valence band spectra of the oxide thin films. This contribution is noted by the absence of a band gap in the spectra, while Al 2 O 3 and MoO 3 are well-known dielectric materials with band gaps of values close to 6 eV [47,48]. For RuO 2 no clear band gap of the oxide is expected, considering the metallic character of the oxide [49]. A contribution of the metal underlayer to the signal is also expected, as the thicknesses of the top oxide layer (about 1 nm) is smaller than the probing depth of XPS, which typically lies between 3 and 10 nm [50][51][52].
It is noted that elemental Mo and Ru present the lowest lying valence band states between 5 and 7 eV, while for elemental Al electron emission is observed in two energy regions: one up to about 12 eV and another following from near 15 to 25 eV. In this case, the detected electrons of the second region do not correspond to primary electrons, but to electrons that are detected after undergoing inelastic scattering events in the solid, therefore presenting a lower kinetic energy [50]. These features are classified as bulk plasmons, and should not be taken into account as part of the valence band structure of aluminum [50,52]. Therefore, for elemental Al, the lowest lying state is found around 12 eV. For the oxides of Mo, Ru and Al, a wide band in the region between 20 and 25 eV is present, with respective peak position and full width at half maximum of 24.25 and 4 eV for Al 2 O 3 , 21.4 and 3.8 eV for MoO 3 , and 21.2 and 3.5 eV for RuO 2 . This region corresponds to the O 2s band, also named the low valence band (LVB) region of oxides [50,[52][53][54].
As previously mentioned, a band will be resonant with He 1s if it lies in energies around 20 eV below Fermi level [20]. With this, considering the above mentioned observations and description of neutralization mechanisms, AN and CIN are expected to be the responsible for neutralization in all analyzed samples, while oxides should present an extra effect related to the O 2s band. However, it is expected that the LVB region (O 2s) would lead to a neutralization phenomenon with characteristics that lie between the qRCT observed in lanthanides and the VB-qRN observed for graphitic carbon and borides [16,17,20,45]. The contrast with VB-qRN comes from the fact that, in the present case, the resonance does not originate from an continuous valence band, but from an isolated band close in energy to He 1s. On the other hand, oscillations would not be expected as this isolated band relates to slevels, which present wider radial distribution function of electron probability comparatively to the d-levels that lead to qRCT in lanthanides [20]. As reported by Tsuneyuki et al. [45], the width of the surface band strongly influences the final charge-transfer probability, with wide bands being responsible for more effective neutralization. This phenomenon occurs due to the diffusion of the hole (generated by the electron transfer from the band to the projectile) in the target band, impeding the oscillatory charge exchange process to continue. This aspect justifies the absence of oscillations in the inverse velocity plots for the oxides on Fig. 2a-d. With this, we classify the neutralization phenomenon relative to O 2s as a quasi-resonant neutralization (q-RN).
It is known from literature that for transition metals like Pd [55] and Ag [56], AN appears as the dominant neutralization mechanism, with resonant processes being negligible. Therefore, the appearance of the O 2s band leads to an extra (non-local) neutralization process of He + for Ru, Hf and Mo. Considering Eq. (4), the presence of two neutralization mechanisms of significant contribution results in a lower ion fraction and consequently higher characteristic velocity values [6] for the metals when bonded to oxygen. For HfO 2 , this neutralization appears to be even stronger, considering that HfO 2 shows the highest v c values of all oxide surfaces, both for scattering on metal and oxygen atoms (Tables 1and 2). Unfortunately, the overlapping between Hf 4f and O 2s levels does not allow further investigation of this feature. However, the question still remains on why a similar characteristic velocity is observed for Al in both metallic and oxide surfaces, considering a similar O 2s energy level as in the transition metals and the absence of resonant energy levels for Al metal.
Several studies have been developed presenting theoretical and experimental analysis of He + /Al neutralization on both metal and oxide surfaces [34,35,[57][58][59][60][61]. In these studies, it is disclosed that both Auger and collision induced processes are important for the neutralization of ions by Al metal. However, for the scattering on aluminum oxide, a suppression of the Auger neutralization was found [13,29,32]. Therefore, the ion survival probability on the incoming and outgoing trajectories due to this mechanism is close to unity. In this case, if no other neutralization mechanism would be present, the total He + ion fraction scattered from Al 2 O 3 would be higher than the one scattered from Al, which is not observed in neither this study, nor in previous studies presented in literature [13,62]. We put forward a hypothesis that the appearance of neutralization by resonance with O 2s band increases neutralization probability by Al in Al 2 O 3 , compensating the decrease in neutralization by suppression of Auger processes with Al oxidation. This would lead to similarly high neutralization probability and consequent high v c Al in both metal and oxide surfaces. Furthermore, the differences in AN between oxidized and elemental Al may also justify the observed increase in v c O values at the transition from Al 2 O 3 to Al surface (green triangles in Fig. 4b). This transition involves a gradual increase in Al content and consequent appearance of the AN channel. The non-local characteristic of this neutralization mechanism implies that it can contribute to the neutralization for scattering on O atoms, resulting in an increase of v c O . This is not observed for v c O at the transition from MoO 3 to Mo as the AN remains the same in both surfaces.

Conclusions
In this work, a detailed analysis of the He + neutralization efficiency was performed for ion scattering on metal and metal oxide films of Ru, Mo, Hf and Al. The obtained results reveal the presence of a matrix effect for transition metals in oxidized state, which is absent for aluminum. By using XPS valence band analysis, we demonstrate that the increased neutralization in oxidized transition metals may originate from the presence of the wide O 2s band. This band is in resonance with the He 1s level, providing an extra neutralization mechanism: quasiresonant neutralization (qRN). On the other hand, we suggest that the suppression of Auger neutralization and appearance of qRN for aluminum counter-balance each other and result in the same neutralization efficiency of He + for both elemental Al and Al oxide.
We also demonstrate that sputtering of pure metal or pure oxide surfaces does not interfere in the neutralization of He + , even for MoO 3 , which was found to be sensitive for removal of oxygen by preferential sputtering. This points out that by choosing a correct reference sample, quantification by LEIS analysis is not limited by matrix effects in metal oxides observed in this work. As long as the strength of involved neutralization mechanisms does not change, it is possible to choose proper reference samples for LEIS quantification of metals and metal oxides. Based on this work, we propose the following rule of thumb: when the sample of interest contains the atoms of a given metal in its metallic form, a reference sample of a pure elemental metal should be used. On the other hand, if the sample of interest contains an oxide of a given metal, a pure metal oxide should be used as a reference. This way, the presence or absence of additional qRN mechanism associated with metal oxide will not affect the results.
This study indicates that the combined action of different mechanisms might lead to misinterpretation of the existence of matrix effects. This shows that there is still insufficient theoretical knowledge for description of the interaction mechanisms between ions and surfaces at low energies, highlighting the importance of further investigation of the topic.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.