XPS Investigation of Monatomic and Cluster Argon Ion Sputtering of Tantalum Pentoxide

In recent years, gas cluster ion beams (GCIB) have become the cutting edge of ion beam technology to sputter etch organic materials in surface analysis. However, little is currently known on the ability of argon cluster ions (Ar n+ ) to etch metal oxides and other technologically important inorganic compounds and no depth profiles have previously been reported. In this work, XPS depth profiles through a certified (European standard BCR-261T) 30 nm thick Ta 2 O 5 layer grown on Ta foil using monatomic Ar + and Ar 1000+ cluster ions have been performed at different incident energies. The preferential sputtering of oxygen induced using 6 keV Ar 1000+ ions is lower relative to 3 keV and 500 eV Ar + ions. Ar + ions exhibit a steady state O/Ta ratio through the bulk oxide but Ar 1000+ ions show a gradual decrease in the O/Ta ratio as a function of depth. The depth resolution and etch rate is substantially better for the monatomic beam compared to the cluster beam. Higher O concentrations are observed when the underlying Ta bulk metal is sputtered for the Ar 1000+ profiles compared to the Ar + profiles.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Introduction
Tantalum pentoxide, Ta 2 O 5 , is technologically significant due to its dielectric properties [1] and applications in microelectronics [2] and optics [1,3]. In addition, Ta 2 O 5 grown on Ta foil is a well-established standard material for the determination of ion etch rate and depth resolution in compositional depth profiles obtained in electron spectroscopy [4]. Many authors use the European standard (BCR-261T), which has a certified thickness of Ta 2 O 5 grown on Ta foil as a reference to estimate the etch rate when performing XPS/AES depth profiles on other metal oxide thin films [5]. However, it is well known that monatomic argon (Ar + ) sputtering of Ta 2 O 5 leads to the preferential sputtering of O [6] and this is generally considered to result from the difference in the atomic weight between Ta (180.95 u) and O (16.00 u) [7]. XPS studies of the preferential sputtering of oxygen from Ta 2 O 5 have been performed by a number of workers [7][8][9][10][11][12]. Hofmann and Sanz performed the earliest in-depth study and they gave the steady-state TaO x stoichiometry (using 3 keV Ar + ) to be TaO 1.05 [7]. Holloway and Nelson sputtered Ta 2 O 5 at varying incident Ar + energies between 0.5 and 5 keV and reported that greater preferential sputtering of O occurred at 0.5 keV than at higher energies and attributed this to the Ar + ions preferentially transferring their energy to O atoms [8].
It has been shown that there are a number of potential advantages in employing cluster beams (C 60 + , Bi n + , Au n + , Ar n + ) in SIMS depth profiling of organic materials compared to monatomic sources, including reduced damage and roughening, lower penetration depth and higher sputter yield, enhancing the quality of chemical information obtainable, sputter rate and interface resolution [8,9]. As a result, gas clusters are widely accepted as effective sources for the depth profiling of polymer samples without causing chemical damage or crosslinking [10].
Until recently, limited work had been published on the use of Ar n + gas cluster ion beam (GCIB) sources for the XPS analysis and depth profiling of inorganic compounds, in particular metal oxides, despite their importance as functional thin films and corrosion resistant layers. As instruments with Ar n + GCIBs are becoming more widespread, publications are now emerging in the literature. Cumpson and co-workers have investigated Ar n + GCIB analysis of HfO and ZnO [13,14] using Ar gas clusters of 1000 atoms (Ar 1000 + ), corresponding to an average energy/atom (E/n) of 6 eV (the same conditions have also been employed in this work). They found that HfO exhibited no preferential sputtering of O using a 6 keV beam [13], whilst the work on ZnO was focused on optimising analytical conditions of inorganic interfaces and no information was given on the degradation (or not) of ZnO under these conditions [14]. Steinberger et al have examined FeO and Zn 5 (CO 3 ) 2 (OH) 6 (hydrozincite) using a range of different Ar n + CGIB conditions [15]. Preferential sputtering of O was observed both for hydrozincite using Ar 2000 + at incident energies of 4 keV (E/n = 2) and FeO using Ar 2000 + at 6 keV (E/n = 3) [15]. Results of Ar n + GCIB of single crystal SrTiO 3 have recently been reported by Aureau et al [16]. In that work, the Ar n + GCIB experimental conditions are not entirely clear, but would appear to be Ar 3000 + at an incident energy of 4 keV (E/n = 1.33). Under those conditions, the Ti 2p peak showed no evidence of reduced Ti states, but a small amount of reduction is observed in the Sr 3d peaks at longer etching times.
As a precursor to the more extensive work presented in this paper, we reported initial results on the Ar 1000 + GCIB sputtering of the BCR-261T 30 nm Ta 2 O 5 layer at incident energies of 4, 5 and 6 keV (E/n = 4, 5 and 6) [17]. That work showed the preferential sputtering of O for E/n = 5 and 6, but not at E/n = 4. At an E/n = 4, there was no evidence of sputtering occurring, whereas at E/n = 5 and 6, profiles through the 30 nm layer could be recorded. In some of the other work reported above, comparisons have been made between Ar + and Ar n + GCIB sputtering [15][16][17][18]. In all cases, using optimized ion beam conditions, Ar n + GCIB sputtering reduces the extent of damage observed compared to monatomic Ar + bombardment, offering the possibility of performing XPS analysis and depth profiling of metal oxides with lower sputtering induced modification to the metal oxide and hence better quality data for the surface analyst.
The aim of this work is to employ the 30 nm thick Ta 2 O 5 layer (BCR-261T standard) to: (i) investigate changes in chemical state associated with Ar + and Ar n + bombardment; (ii) compare the preferential sputtering, etch rates observed and depth resolution for Ar + and Ar n + depth profiles through the 30 nm thick Ta 2 O 5 layer. Ar + depth profiles were performed at ion beam energies of 500 eV and 3 keV and Ar 1000 + depth profiles were acquired at ion beam energies of 6 keV and 8 keV (with and without sample rotation). The ion beam energies employed for the latter Ar 1000 + depth profiles give rise to E/n values of 6 and 8 eV.

Experimental
The standard 30 nm Ta 2 O 5 layer on Ta foil (BCR® -261T), described in [3] was employed for all XPS analyses undertaken. The XPS work was performed on a Thermo Scientific K-Alpha XPS system, equipped with the Ar + and Ar n + gas cluster ion beam (GCIB) source, MAGCIS. The MAGCIS source is mounted at an angle of 60 o to the sample normal. The ion beam area was rastered over an area of 1 mm 2 and to avoid crater edge effects, an X-ray spot diameter of 200 µm was employed. For monatomic Ar + profiling, energies of 500 eV and 3 keV were employed, operating at beam currents of 1 and 3 µA (measured at the sample holder) respectively. For the Ar 1000 + GCIB at 6 keV and 8 keV, a current of 20 nA was used. Ar 1000 + GCIB profiles were also recorded at 6 keV and 8 keV using sample rotation at a rate of 1 rotation/min. The Ta 2 O 5 /Ta interface was assigned on the depth profiles using the linear drop in the O signal in the interface region. The interface was taken as being the mid-point between the two positions on that O signal line (before and after the interface) where the signal deviates from linearity.
The XPS spectra were acquired employing a monochromated Al kα X-ray source operating at a power of 300 W. The spectrometer was calibrated using the Au 4f 7/2 , Ag 3d 5/2 and Cu 2p 3/2 peaks at 83.98, 368.26 and 932.67 and eV respectively. A pass energy of 50 eV and step size of 0.1 eV were employed. During ion beam bombardment, the X-ray source was blanked.
Quantification was performed after a Shirley background subtraction and used Wagner sensitivity factors, modified to account for the instrument transmission function. The Thermo Scientific Avantage software was employed for peak fitting, using a Gauss/Lorentz mix value of 26%.
As the Ar n + cluster etch rates were much lower than the Ar + etch rates, the XPS data recorded at each level of the GCIB depth profile was performed using the "snapshot analysis" mode. In this mode, the pass energy of the analyser is increased from 50 eV to 150 eV and rather than scanning through the selected energy range, the electron signal from the entire range is recorded simultaneously. This sacrifices some of the energy resolution but greatly reduces the acquisition time. The spectral resolution is then mostly recovered using a deconvolution process during data analysis, see Figure 1. To ensure accurate peak fitting, a 3 keV Ar + depth profile was initially peak fitted to establish the peak fitting parameters with which all subsequent depth profiles were fitted.

XPS spectra and depth profiles
In order to establish the Ta chemical states observed during depth profiling of the Ta 2 O 5 layer, a methodical approach was adopted in fitting the Ta 4f spectra, using the known binding energies for the metallic tantalum doublet and well reported peak positions of the Ta 2 O 5 doublet. These two pairs of peaks represent spectra at the beginning and the end of the depth profile, thus could be easily isolated and accurately fitted. is electrically conductive, both peaks in the doublet show the high binding energy tail associated with shake-up events for electrons close to the Fermi level, yielding an asymmetric peak shape that is unlike the other Ta oxide components fitted in the depth profile. This peak shape was taken into account when fitting the Ta 0 peaks.
The initial Ta 5+ peaks must be fitted correctly as they provide the peak shape for the subsequent sub-oxide peaks that are produced by preferential etching of O from Ta 2 O 5 . Once fitted, the Ta 5+ peaks were used to set the FWHM, the L/G mix ratio as well as the relative intensity and separation ratios between the Ta 4f 7/2 and Ta 4f 5/2 peaks for each sub-oxide state.
In fitting the final Ta peak envelope, all of these variables were fixed for each of the sub-oxide states to ensure that the only variation between them were the binding energies and peak intensities. However, the FWHM of the peaks fitted to the spectra after ion bombardment were allowed to increase to accommodate ion beam induced broadening.
For the Ta 2 O 5 profile, in addition to Ta 0 and Ta 5+ , the other possible sub-oxide reduction states that can be generated during ion etching are Ta 4+ , Ta 3+ , Ta 2+ , and Ta 1+ , representing TaO Table 1. [11][12]. Figure 3 shows that there is a linear progression of the binding energy with oxidation state, as found by Benito and Palacio for Ta 2 O 5 under ion bombardment and the peak binding energies for the different Ta oxidation states are all within ± 0.4 eV of their reported values [12].  The peak positions as shown in Table 1    show that this layer is removed after about 3 nm of material has been sputtered away. As the carbon contamination layer was not of interest in this work, the C 1s data has been excluded from subsequent depth profiles shown in Figure 5 [15][16][17][18].
To provide more information on the change in Ta n+ chemical states during the Ta 2 O 5 depth profiles, Ta 4f spectra for monatomic Ar + sputtering at 3 keV and 500 eV and Ar 1000 + GCIB at 6 keV (not rotated) are given in Figure 7. To make a direct comparison of Ta 4f peakshape changes over the "steady state" region for the depth profiles recorded using 3 keV Ar + and 6 keV Ar 1000 + , it is also useful to plot overlays of the spectra Ta 4f spectra, as shown in Figure 8. The difference between the Ar + and Ar 1000 + is clear, with the variation in Ta 4f envelope for 3 keV Ar + being minimal, whilst for the 6 keV Ar 1000 + , the peakshape progressively changes with depth.  Comparing these depth profiles reveals some interesting information about the existence and depth variation of sub-oxide states in the altered layer for the different etching conditions.
However, due to the complexity of fitting the Ta 4f peak envelope, it is important to be mindful that the peak fitting may not be entirely accurate and truly representative of the behaviour of each oxidation state as a function of depth. Consequently, over interpretation of the data should be avoided. Nevertheless, some important trends can be extracted from the data and it is reasonable to separate the sub-oxide states into 2 groups, (Ta 4+ and Ta 5+ ) representing the higher oxidation states and (Ta 2+ and Ta + ) representing the lower oxidation states. A comparison between the Ar + and Ar 1000 + depth profiles in Figure 9 reveals that when profiling through the bulk oxide, Ar + bombardment gives rise to a clear preference in the formation of the lower Ta n+ oxidation states in the altered layer rather than the higher oxidation states. In contrast, Ar 1000 + bombardment results in a greater concentration of higher oxidation states compared to the lower oxidation states. Furthermore, as expected, for the monatomic Ar + profiles, it can be seen that over the steady state region, the intensities of each of the Ta n+ states remain constant. However, for the 6 keV Ar 1000 + profile, this is not the case, with a gradual decline in the intensity of the higher oxidation states and concomitant gradual increase in the intensity of the low oxidation states as a function of depth. This effect is even more pronounced in the depth profiles for the Ar 1000 GCIB at 6 keV and 8 keV including sample rotation where again the behaviour for each oxidation state is plotted as a function of depth ( Figure 10).

Etch rate
Use of the 30 nm thick BCR® -261T standard enables the etch rate to be accurately determined for all of the Ar + ions and Ar 1000 + ion depth profiles performed. A comparison of the etch rates for the different ion beam conditions, given in Table 2, is of particular interest to the practical analyst. The etch rates for the monatomic Ar + beam (at 3 keV and 500 eV) are approximately 2 orders of magnitude higher than for Ar 1000 + GCIB at 6 keV (rotated and un-rotated) and 1 order of magnitude higher than for Ar 1000 + GCIB at 8 keV (rotated). Rotating the sample for the Ar 1000 + GCIB at 6 keV leads to an increase in the etch rate by approximately 4 times. Taking account of the different ion current densities employed in the different studies, the etch rates for the Ar 1000 + GCIB at 6 keV in this work are consistent with our previous values of 1 x 10 -3 nm/s for the same BCR -261T standard recorded on a Thermo Scientific ThetaProbe instrument [17] and similar to the etch rate of 4.2 x 10 -4 nm/s for HfO reported by Barlow et al [13] recorded on a Thermo Scientific ThetaProbe instrument equipped with the same MAGCIS source as used in this work.

Discussion
This work is the first reported example of an XPS depth profile through an inorganic layer using an Ar n + GCIB. The results have shown that despite the much lower E/n, Ar n + GCIB bombardment using Ar 1000 + GCIB at 6 keV and 8 keV still leads to the preferential sputtering of O for Ta 2 O 5 .
However, the extent of preferential sputtering is lower than that observed using a monatomic Ar + beam at 3 keV or 500 eV. The Ar 1000 + profiles have shown some rather surprising results: (i) the depth resolution is substantially degraded compared to the monatomic Ar + profiles (ii) the absence of a steady state region in the oxide, where the composition of the altered layer is constant. Instead, the O concentration gradually decreases as the profile progresses towards the interface (iii) once the Ta 2 O 5 layer has been removed, the O concentration drops to a value of 10 -20 at.% rather than a much lower value, as seen for the monatomic Ar + depth profiles.
With regard to the penetration depth of the ion beams, using molecular dynamic simulations, Aoki et al [19] reported that the penetration depth of Ar 688 + clusters dropped to below 5 Å when the average E/n was 14 eV. The average E/n of the Ar 1000 + used in this study were lower than this value, being 8 and 6 eV, thus penetrating just 1 -2 monolayers into the sample surface. In contrast, the penetration depths for monatomic Ar + at 3 keV and 500 eV calculated using SRIM calculations [20] is 3.5 nm and 1.3 nm respectively. However, using the experimental set-up described here yields the unexpected result that the lower penetration depth does not lead to improved depth resolution when profiling the Ta 2 O 5 layer on Ta; instead the depth resolution is degraded. For the profiles recorded using Ar 1000 + clusters at 8 keV and 6 keV with rotation compared to that recorded for Ar 1000 + clusters at 6 keV without rotation the depth resolution is further worsened. Surface roughening and nano-topography resulting from Ar + bombardment are well-known phenomena which are influenced by many experimental parameters, including the ion beam incident angle, energy, flux and fluence [21]. Simulations of Ar clusters of various sizes and energies with the incident beam inclined at an angle of 60 0 with respect to the surface normal bombarding different material surfaces have shown the formation of ripple and dot structures (but the extent of the roughening was not stated) [22].
However, considering the known low penetration depth of the Ar clusters, sample rotation causing further interface broadening (when it would be expected to reduce roughening) and the large degree of interface broadening observed for a 30 nm thick film, it seems unlikely that roughening is the main cause of the interface broadening observed for the Ar 1000 + depth profiles reported here for Ta 2 O 5 .
It has been reported that the impact of cluster ions with solids leads to high temperature and pressure transients in the vicinity of the impact which do not occur in an equivalent manner for Ar + bombardment [22][23][24][25]. An increase in surface temperature for samples being bombarded with an argon cluster beam can lead to various effects including an increase in sputtering yield and changes in the interface width [26,27]. Infusion doping of elements into solids through the use of dopants in Ar clusters has been described [28] and in the recently published book by Yamada, it is stated that the intense Ar cluster thermal spike allows the infusion of elements into the solid surface by an atomic mixing process which occurs within the thermal transient region [29].
Preferential sputtering of O from metal oxides can be caused by a number of different processes, where in many cases it is thought that a combination of mechanisms may be influencing the altered layer composition [30,31]. The preferential sputtering of O from Ta 2 O 5 is generally considered to be a ballistic process [7,30].  [33]. For Ar n + cluster depth profiles, in a similar manner to Ar + profiles of different inorganic materials, there are likely to be various different ion beam induced processes which introduce artefacts into the profiles and the susceptibility to these effects is material dependent. However, with Ar cluster depth profiling, there is a large GCIB parameter space to be explored which has the potential to offer greater possibilities for minimising such undesirable effects.

Conclusions
Depth profiles through the certified (European standard BCR-261T) 30 nm thick Ta 2 O 5 layer grown on Ta foil using monatomic Ar + and Ar 1000 + cluster ions have been performed using a GCIB at different incident energies. The preferential sputtering of O, relative intensities of Ta oxidation states, depth resolution and etch rates obtained from the profiles using the different ion beam conditions have been recorded and compared. The following conclusions can be drawn from this investigation: The preferential sputtering of O induced using 6 keV Ar 1000 ions is lower relative to 3 keV and 500 eV Ar + ions. At a point close to the middle of the 30 nm thick oxide, the stoichiometry recorded for the 6 keV Ar 1000 + ion beam was TaO 2.0 , compared to TaO 1.5 and TaO 1.55 for Ar + at 3 keV and 500 eV respectively.
Depth profiles recorded using Ar + ions give rise to a steady state region in the oxide bulk, where the preferential sputtering of O remains constant. The use of 6 keV Ar 1000 + ions shows a gradual decrease in the O concentration over the same region. This progressive loss of O as a function of depth is further enhanced when the experimental conditions are changed through the use of sample rotation and an increase in the Ar 1000 + incident energy to 8 keV. Curve fitting has shown that the concentration of the higher Ta oxidation states (Ta 4+ and Ta 5+ ) is greater than the lower oxidation states (Ta + and Ta 2+ ) in the Ar + profiles and the opposite trend is observed for the Ar 1000 + profiles. Furthermore, for the Ar 1000 + profiles, there is an increased concentration of the higher oxidation states closer to the surface and the concentration of lower oxidation states progressively increases with depth.
The Ar + depth profiles recorded using ion beam energies of 500 eV and 3 keV (without sample rotation) exhibit a better depth resolution than Ar 1000 + profiles at beam energies of 6 keV and 8 keV (with and without sample rotation).
There is a higher O concentration observed when profiling into the underlying Ta for profiles performed using 6 keV and 8 keV Ar 1000 + which is not observed for the 3 keV and 500 eV Ar + profiles.
The etch rate increases with E/n. Using a 3 keV Ar + beam, the etch rate was found to be 4.8 x 10 -3 nm/min. For the 6 keV Ar 1000 + beam, the etch rate decreased to 1.8 x 10 -5 nm/min.
Increasing the Ar 1000 + ion beam energy to 8 keV and rotating the sample during profiling leads to a significant increase in the cluster beam etch rate (3.8 x 10 -4 nm/min).
It is proposed that for the Ar 1000 + depth profiles, the progressive O loss from the oxide and increased interface width may be caused by the high temperature transient resulting from the cluster impact.