Electronegativity-dependent tin etching from thin films

The influence of a thin film substrate material on the etching of a thin layer of deposited tin (Sn) by hydrogen radicals was studied. The amount of remaining Sn was quantified for materials that cover a range of electronegativities. We show that, for metals, etching depends on the relative electronegativity of the surface material and Sn. Tin is chemically etched from surfaces with an electronegativity smaller than Sn, while incomplete Sn etching is observed for materials with an electronegativity larger than Sn. Furthermore, the amount of remaining Sn increases as the electronegativity of the surface material increases. We speculate, that, due to Fermi level differences in the material’s electronic structure, the energy of the two conduction bands shift such that the availability of electrons for binding with hydrogen is significantly reduced.


I. INTRODUCTION
A hot Sn plasma, emitting at a target wavelength of 13.5 nm, is used as an EUV (extreme ultraviolet) source for photolithography applications. 1Tin, in the form of debris resulting from the plasma, can be deposited on mirrors and, consequently, reduce the mirror reflectivity. 2,3It is important to maintain the cleanliness of the collector optical surfaces. 4eactive hydrogen species are able to form the volatile hydride, stannane (Sn + 4H • → SnH 4 ↑), which can, subsequently, be removed from the surface. 4Stannane slowly decomposes (at 35 • C) to metallic Sn and hydrogen, 5,6 which reduces the effective etch-rate due to re-deposition.Atomic hydrogen etching is highly selective in the presence of typical EUV relevant mirror surfaces, such as Mo/Si multilayer mirrors, meaning that, while the Sn is etched, the mirror surface remains intact. 3or an acceptable EUV reflection loss (in the order of few percent), a thickness of Sn on the EUV mirror of less than 0.5 nm is required. 3,7uccessful Sn etching using hydrogen atoms will depend on the surface material. 3,7For example, the cleaning rate of the last few nm of Sn from a Si surface is almost twice as fast as that from a Ru surface. 7This disparity in etching rate increases for Sn layers that become thinner than 2 nm.A possible explanation for this disparity is Sn re-deposition on the surface, due to the decomposition of Sn hydride.Metallic surfaces like Ni, Ru, Rh or Au effectively catalyze the decomposition of SnH 4 . 7During the etching process, the substrate becomes more exposed, which favors Sn decomposition and re-deposition, and may lead to a dynamic equilibrium between SnH 4 formation and decomposition.However, the amount of Sn remaining after hydrogen etching is significantly more than a monolayer, which cannot be fully explained by enhanced re-deposition due to catalytic activity. 7side from environmental factors, such as hydrogen flow rate and substrate temperature, an additional factor that contributes to Sn etching is the surface material's electronic structure, which has remained relatively unstudied as a factor.Experimental studies on the influence of electronic structure are challenging because surface states depend intrinsically on the local geometry of the surface material where the Sn atoms adsorb.As a first step towards quantifying the influence of surface electronic structure, we use a proxy that may allow some general principles to be revealed.
Because electronegativity provides a simple empirical model for binding strength (and, indirectly, the changes to electronic structure), we propose here to use electronegativity as a proxy for electronic structure.
The electron distribution in a bond in a diatomic molecule containing different atoms is not symmetrical between the atoms, 8 the degree to which this redistribution occurs can be described by a single parameter, called the electronegativity. 9The probability density of an electron pair, shared between two elements, has a maximum that is located closer to the more electronegative element.The more electronegative element acquires a partial negative charge, and the molecule may also acquire a dipole moment.
When electrons are shared by two metallic atoms in a bulk material, a metallic bond may be formed.The electrons that participate in metallic bonds are delocalized, thus, metallic bonds have no dipole moment.An interface between two metals acquires a potential difference, however, that depends on the difference between their Fermi levels.At thermodynamic equilibrium, the Fermi level is constant across the interface, meaning that the energy of the two conduction bands must smoothly shift to be coincident at the interface. 10This modifies the electronic structure of the Sn near the interface, and changes the availability of electrons for bond formation with, in our case, hydrogen atoms.Due to this change in charge availability, the relative electronegativities of the surface material and the Sn may be used to predict the degree to which electrons are available for bonding.This allows the success or failure of Sn etching from a thin metallic surface material to be predicted.Specifically, we hypothesize that Sn can be fully chemically etched from surfaces with electronegativity lower than Sn, while incomplete Sn etching is expected for materials with electronegativity higher than Sn.

A. Material selection
From all chemical elements, 15 were chosen for further study (table I).The chosen surface materials belong to different groups and periods in the periodic table.
As stated, electronegativity is a measure of the ability of an atom in a molecule to draw bonding electrons to itself.The most commonly used electronegativity is a relative scale of electronegativity, developed by Linus Pauling in 1932. 11In this scale, electronegativity differences are defined by differences in bond dissociation energies with hydrogen set to an arbitrary reference point value of 2.1.
Elements with electronegativities, χ, higher and lower than Sn (χ=1.8) were chosen.It should be noted that Sn has a rather high electronegativity, limiting the number of appropriate materials with an electronegativity higher than Sn.Finally, to avoid cases dominated by excessively fast oxidation, elements from the first and second groups were excluded.Furthermore, elements that were known to react strongly with hydrogen, such as Pd and Co, were excluded.Within these restrictions, it was desirable to choose elements that covered as broad a range of the periodic table as possible.The aim was to include a large range of possible electronegativities.Elements from groups 3-14, consisting of transition metals, metalloids, other metals and non-metals were chosen for further investigation.

B. Methodology
Thin films of chosen materials, with an average thickness of 5 nm, were deposited onto silicon wafers (10x10x0.5mm) by magnetron sputtering or electron-beam deposition.The film thickness was monitored during deposition using a quartz microbalance.After deposition, the samples were removed from the chamber and exposed to ambient conditions.Grazing incidence x-ray reflection measurements were performed to extract accurate layer thicknesses.The deposition method, and XRR derived layer thicknesses are shown in table I.The uncertainty in the film thickness was estimated to be 10%.
X-ray photoelectron spectroscopy (XPS) (Thermo Scientific Theta Probe) analysis showed that for many materials a surface oxide layer forms upon exposure to ambient conditions.For most of the materials that we chose, the oxide layer is usually self-limiting at atmospheric pressure and room temperature, resulting in a so-called native oxide layer with a thickness of few nanometers at most. 12in deposition and etching experiments were performed in a vacuum chamber (fig.1(a)) with a base background pressure of 10 −9 mbar (10 −7 Pa), which increases to 10 −8 mbar during Sn deposition.The samples described in table I were not subjected to further preparation and were placed directly in the vacuum chamber.The sample was held upside down on a movable sample holder, 4 cm above the exit port of an evaporation source.The temperature of the sample was monitored using a Pt-100 temperature sensor, installed under the sample (Fig. 1(b)).
Tin (Umicore) with a bulk density of 7.3 g/cm 3 was evaporated from an Effusion Cell EF 40C1 (PREVAC) evaporator operated at 980 • C. The Sn deposition rate was calibrated using a quartz microbalance (INFICON XTM/2 Deposition Monitor).It was found that the evaporation rate as measured using the quartz microbalance (QMB) at 980 • C was approximately 0.4 nm/min In a typical experiment, 8 nm of Sn was deposited.During deposition, the sample temperature remains lower than 55 • C, and, after deposition the sample is allowed to cool to room temperature.The influence of surface temperature on mobility of Sn was checked and no mobility of Sn was observed below 150 • C. Tin deposition and etching (see below) on the sample surface was monitored with spectroscopic ellipsometry (J.A. Woollam Co., M-2000, spectral range 245-1689 nm) at an angle of incidence of 76.4 • (with respect to the surface normal).Tin was etched with atomic hydrogen (H • ), which was generated using a hot (2000 • C) tungsten filament (placed 3 cm from the sample holder).In order to achieve sufficient H • flux with a stable flow rate, the hydrogen (purity 99.999%) pressure during exposure was fixed at 8•10 −2 mbar.The hydrogen radical flux at the surface was measured to be 10 17 at/(cm 2 • s). 13 To prevent the sample from heating excessively, the filament was operated at a duty cycle of 0.5 with a period of 10 min, and a shield was placed between the filament and the sample.
Selected samples were removed from the deposition chamber before exposure to hydrogen radicals, allowing Sn layer thickness analysis using X-ray fluorescence X-ray fluorescence (XRF) (PW2400 chemium anode sequential XRF-spectrometer) has a detection limit for Sn of 0.01 µg/cm 2 (0.014 nm).Other samples were analyzed after hydrogen etching.

A. Tin deposition and etching in situ monitored with ellipsometry
Although the absolute thickness of the Sn layer is relevant, the total amount of remaining Sn after etching is the parameter of interest.To avoid uncertainties on the amount of deposited Sn, we use XRF and determine the layer thickness using the bulk density of Sn (7.287 g/cm 3 ).The average thickness of the Sn layer on the samples before etching was measured to be 7.7 ± 0.9 nm (table II).XPS measurements could not detect the underlying material, indicating that the layers were closed to within a few percent.The Sn etching experiments do not include exposure to ambient, and the etching experiments were performed less than 1 hour after deposition, thus we do not expect Sn oxidation.Therefore, for the hydrogen etching measurements, the initial etching is dominated by the chemistry of Sn and hydrogen, while further into the etch process, the substrate starts to play a role.
During Sn deposition and hydrogen radical etching, the surface was monitored in situ with spectroscopic ellipsometry.These measurements were primarily used to distinguish the end point for Sn etching by monitoring the ellipsometric angles Psi (not shown) and Delta (see fig. 2 for Delta).For a sample where complete Sn etching was observed (confirmed afterwards with XRF measurements), e.g., crystalline silicon (fig.3(a)), it can be seen that, after 120 minutes of hydrogen etching, the Delta value for a wavelength of 530 nm, which shows the largest Delta modulation,  nearly returns to the value it had prior to Sn deposition, and is stable thereafter (end point of etching).This is typical for the entire Delta (Fig. 2) and Psi (not shown) spectra for surfaces from which Sn could be etched completely.The similarity of the initial and final Psi and Delta spectra also means that the surface is very close to its original state after etching.It is most likely that the small differences in the Psi and Delta spectra are due to a very slight change in sample alignment during etching.The Ru surface is an example of incomplete Sn etching, which can be clearly seen from the ellipsometry measurements (fig.3(b)).After 100 min of etching, the Delta value is stable, but it has not returned to the value it had prior to Sn deposition.It is clear that Sn remains on the sample surface (confirmed with XRF measurements, see below).
Although the ellipsometry measurements can be used to estimate the Sn layer thickness during deposition and etching, this process proved to be unreliable.The as-deposited Sn surface is rather rough, 3 and the surface roughness is expected to change during etching, which influences the interpretation of the ellipsometry data during etching, but not at the end point of etching.For some samples, the ellipsometry measurements showed evidence that the Sn and hydrogen radicals interacted with the underlying surface, changing its optical properties as well.As a result of these factors, a reliable model fit could not be obtained.Therefore, to quantify the amount of Sn remaining on the surface after etching, XRF measurements were performed.

B. Electronegativity dependent Sn etching
Fig. 4 shows the Sn layer thickness derived from XRF after etching as a function of difference in electronegativity between the surface material and Sn.The surface materials can be divided in two groups in terms of complete and incomplete Sn etching.The first group includes surface materials that have electronegativity values that are lower or equal to Sn ( χ Sn -χ M ≥ 0).For these materials, Sn is completely etched (less than 0.08 nm of Sn on the surface) with two exceptions, where 0.4 nm and 0.14 nm of Sn remain on the Mo and W surfaces.For these materials, it is possible that the re-deposition plays a role, especially just after the etch is terminated. 3he second group consists of surface materials that have an electronegativity value higher than that of Sn ( χ Sn -χ M < 0).In this group, incomplete Sn etching is observed.Moreover, the thickness of the remaining Sn increases for surfaces with higher electronegativity values, indicating that Sn is more strongly bound to these surfaces.It is remarkable that we find a simple linear relationship between the Sn thickness and electronegativity for χ Sn -χ M < 0. A possible explanation for this result (see below) is that the surface material modifies the electronegativity of the first few layers of the Sn.For this to occur, the two layers should not be significantly intermixed.
To determine if the Sn intermixes with the surface layer, angle resolved XPS analysis was performed on the etched Ru sample.This showed minimal intermixing between Ru and Sn.For Au and Ag, angle resolved XPS was inconclusive.
The XPS spectra, however, show that, for Ag, the Sn 3d 5/2 spectral line shifts from 484.78 eV (Sn as deposited) to 484.91 eV (after etching with H • ), while the expected chemical shift (CS) for Sn 3d 5/2 spectral line upon Ag-Sn alloy formation is 1 eV. 14,15Likewise, the Ag 3d 5/2 spectral line shifts from 368.22 eV (Ag as deposited) to 368.32 eV (after etching with H • ), which is also less than the expected 1 eV shift.These measurements confirm that Ag and Sn do not significantly intermix.
In contrast, the XPS spectra for Sn and Au are suggestive of intermixing.The expected CS for the Au 4f 7/2 spectral line upon Au-Sn alloy formation is 0.45 -1.1 eV, 16 while for the Sn 3d 5/2 spectral line the expected CS is 0.1 -0.4 eV .The Sn 3d 5/2 spectral line after etching with H • was measured to be 485.29 eV, a CS of 0.34 eV, while the Au 4f 7/2 spectral line shifts from 84.06 eV to 84.6 eV (0.4 eV).These results are consistent with intermixing.
Two notable exceptions from the second group are carbon and boron.It is known from the literature that carbon can be successfully etched from silicon using hydrogen reactive species. 17,18Thus, because the Sn layer grows in islands, 3 hydrogen radicals also have access to the underlying carbon surface before the Sn is completely removed.An XPS analysis of a carbon sample before and after Sn deposition and etching confirmed that the carbon layer thickness was reduced by approximately 25% during Sn etching.As a result, the etching process may remove carbon from around the Sn, allowing the Sn to also be etched.It is also possible that hydrogen radicals may react to create more complex organo-metallic volatiles, which effectively allows Sn to be etched together with carbon.
Boron can be etched by hydrogen radicals, but only at elevated temperatures. 19XPS analysis confirmed that boron was not etched during hydrogen radical exposure, eliminating both the possibility of boron hydride formation and of volatile metallo-complexes.This illustrates that, although electronegativity is generally predictive (especially for the first group), detailed electronic structure also plays a role.
Our results show that the etch rate, near the interface between two materials, can be suppressed or enhanced, depending on the electronegativity difference.The suppression is sometimes sufficient to prevent the last few nanometers of material from being etched, as seen in Fig. 4. It should be noted that, for Ag and Ru this certainly includes Sn that is not in direct contact with the surface material.

IV. DISCUSSION
During etching, an electron must be donated for hydrogen to form a stable bond.However, for Sn, bound to a highly electronegative surface, the valence electrons have probably been donated to surrounding atoms, making it impossible for Sn hydride to form.
Electronegativity also explains why the remaining thickness of Sn increases as the electronegativity difference becomes more negative.Analogous to the case for oxides 20 the effective electronegativity of Sn at the interface is increased when it binds to a strongly electronegative surface (and, likewise, decreased when it binds to a weakly electronegative surface).Thus, the first monolayer of Sn donates its electrons, which are delocalized and are more likely to be found in the surface material.In doing so, the monolayer of Sn becomes more electronegative because it is now electron poor.However, it is not as electronegative as the original surface.Similarly, the second monolayer of Sn donates electrons to the first Sn layer, becoming slightly more electronegative.This process continues, until, at some layer thickness, the electronegativity of the uppermost Sn returns to its natural value.
For Ru and Ag, where XPS confirms that there is minimal intermixing, this screening length is indicated by the thickness of the remaining Sn (4.1 and 2.3 nm), while for Au, it is likely that a similar screening length is present, however, intermixing precludes quantification.

V. CONCLUSIONS
The remaining amount of Sn after etching with hydrogen radicals depends on the substrate onto which Sn was deposited.Transition metals with electronegativity values that are lower than or equal to Sn ( χ Sn -χ M ≥ 0) can be fully etched from Sn.The remaining Sn thickness is found to be less than 0.08 nm on Sc, Ta, Ti, Nb, V, Cr, Al, and Si surfaces.
Incomplete Sn etching was observed when Sn was initially deposited onto materials with an electronegativity value higher than that of Sn ( χ Sn -χ M < 0).Moreover, the thickness of the remaining Sn increases with more negative electronegativity differences.This indicates that, not only is Sn bound more strongly to the underlying surface material, but, near the interface, the Sn -Sn bonds are stronger than those found in bulk Sn.
It is known that, even though electrons are delocalized in a metal bond, the continuity of the Fermi level requires that the energy levels and occupancy of the electronic states changes within a few nanometers of the interface.These changes make the Sn surface (when it is within a few nanometers of the surface material) less reactive and more difficult to etch, and can even prevent etching all together.This raises an interesting new approach to surface passivation that warrants further investigation.

FIG. 2 .
FIG. 2. Delta as a function of wavelength during Sn deposition (a) and hydrogen radical etching (b) on a silicon wafer.Delta values for a clean silicon wafer are presented as a dashed line in figure 2(b).

FIG. 3 .
FIG. 3. Delta value for 530 nm during Sn deposition and hydrogen radical etching for a) silicon wafer and b) Ru layer.

FIG. 4 .
FIG.4.Remaining Sn after etching with hydrogen radicals as a function of difference in electronegativity between Sn and the surface material.

TABLE I .
Electronegativity values deposition method, thickness of surface material and native oxide thickness on each sample for test materials used in this work.

TABLE II .
Deposited Sn thickness (XRF) for all surface materials.