Open Access
Add to BibSonomy
Abstract
This research compares the use of NF3 and SF6 process gases for the removal of the native oxide from Al mirrors and their subsequent fluorination using low-temperature electron-beam generated plasmas. This single step process produces a stoichiometric AlF3 layer of controllable thickness which provides an excellent passivation layer for Al mirrors for far-ultraviolet optics applications. We show that NF3 produces more atomic F content within the plasma than SF6 for similar process conditions, allowing faster oxide removal and fluoride film growth. Al mirrors passivated in either SF6 or NF3 were subsequently studied as a function of gas flow concentration. We found that above a threshold value of atomic F content in the plasma (≈2 × 1018 m−3) samples processed with SF6 or NF3 exhibit similar optical and structural properties regardless of process parameters.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
A novel passivation procedure utilizing plasma-based fluorination of bare aluminum (Al) has recently been developed [1–4]. The oxide removal and simultaneous fluoride film growth is accomplished using an electron-beam generated plasma produced in a fluorine-containing working gas. In this environment, F-based ions and radicals simultaneously etch the native oxide layer while promoting the formation of an AlF3 passivation layer with tunable thickness. Importantly, the treatment does not require high substrate temperatures or ultra-high vacuum environments. The passivation process is carried out in the Large Area Plasma Processing System (LAPPS) at the U.S. Naval Research Laboratory (NRL) [5,6] using bare, aluminum coated glass samples prepared at Goddard Space Flight Center (GSFC) coating facilities. This new fluoride passivation process has demonstrated the following characteristics:
- • FUV reflectivity: Al mirrors with enhanced reflectivity have been validated at different key far-ultraviolet (FUV) wavelengths: 48% at 102.6 nm (H Lyman β), 53% at 103.5 nm (center of O VI doublet) [3], and >80% at wavelengths >110 nm, including 91% reflectivity at 121.6 nm (H Lyman α) [2]. To date, this is one of the few passivation techniques for Al, along with the works reported in Refs. [7,8], which consistently delivers reflectance above 90% at the reference FUV wavelength of Hydrogen Lyman α (121.6 nm) without using elevated substrate temperatures.
- • Uniformity: The plasma uniformity demonstrated by the LAPPS system [9] enables mirror reflectance non-uniformity of <3% for wavelengths between 110 nm and 200 nm and <1% at wavelengths above 200 nm on sample coupons tested on the LAPPS processing stage distributed to simulate a larger 150 mm diameter flat mirror [10,11]. This reflectance uniformity could likely be improved upon further with the inclusion of standard plasma processing procedures such as stage rotation, and chamber design to optimize gas flow over the substrate.
- • Polarization studies made on passivated mirrors distributed on a 150 mm diameter area showed an average phase retardance (Δs-Δp) of 0.82° and normalized diattenuation (Rs-Rp)/(Rs + Rp) of 0.11% at an angle of incidence of 12° in the 200-3000 nm spectral range [10,11]. We demonstrated that these polarization aberrations are largely driven by the Al layer, and the AlF3 protective layer does not contribute significantly to the polarization budget.
- • Environmental stability: Samples stored for 6-10 months in environments with relative humidity of ≈30% showed an average reflectance degradation at their peak FUV wavelengths of ≈ 4.2% [2]. In addition, accelerated environmental tests demonstrated that samples could survive one week at a constant temperature (25°C) and humidity (60%), with an average FUV reflectance degradation of <1% [10,11].
These recent studies on the passivation of thin Al films with electron-beam generated plasmas have demonstrated the feasibility of a room-temperature process for effectively fluorinating the surface of Al mirror samples with good uniformity of the fluoride layer, and high reflectivity across the entire ultraviolet, optical, infrared (UVOIR) bandpass. The films produced with this method have also shown lower induced polarization phase and amplitude differences between orthogonal polarizations, and good durability and environmental stability. All these milestones were achieved using SF6 plasmas.
This manuscript reports on the use of NF3 as an alternative to SF6 as the process gas. The main motivation for use of NF3 is that it has the potential to generate a higher of atomic fluorine density due to its lower average binding energy (66.4 kcal/mol) when compared to SF6 (78.3 kcal/mol) [13]. A higher density would lead to an increase in the flux of reactive fluorine to the surface, which could minimize the influence of background contaminates (e.g. O, C, etc.) and increase the fluoride growth rate. Both would be advantageous for development of the passivation process for large-area mirrors. Consequently, this research aims to present a comparison between Al mirrors passivated with plasmas produced in mixtures of argon and SF6 or argon and NF3 to assess whether NF3 can offer similar results in terms of optical quality and nanostructure of the passivation layer, but with a significantly faster passivation process. To this end we make use of in situ ellipsometry capability fitted on the LAPPS reactor, which allows for real-time monitoring of film thickness and optical properties during fluoride growth. This capability provides nanometer scale thickness control of the AlF3 protective layer with plasma exposure time and enables a high level of reproducibility. Importantly, this limits the risk associated with thickness control of the AlF3 layer which ultimately determines the wavelength coverage of the mirror (i.e., lower end cut-off wavelength). As such the uncertainties related to the AlF3 layer thickness (such as interferential effects) are mitigated, allowing for a more straightforward comparison between samples fabricated with different process parameters.
This manuscript is organized into two main sections. The first section describes the process characterization of electron beam-generated plasmas produced in varying concentrations of either NF3 or SF6 using similar process parameters. In the second section, we examine the similarities and differences between NF3 and SF6 produced films in terms of growth rate, structure, chemistry, and optical performance.
2. Experimental equipment
2.1 Fabrication of bare Al mirrors
The fabrication of bare aluminum samples was carried out in a 1-meter diameter coating chamber at GSFC described in Ref. [14]. All bare Al mirrors were deposited on soda-lime glass slides of 50 mm x 50 mm area with the conditions stated in Ref. [14] for Al, but for the present investigation the chamber baked out prior to the aluminum coating deposition was omitted. Most of the bare Al coatings shown in this research were fabricated in the same coating run with samples distributed over an area of 200 mm x 160 mm. Based upon prior characterization of the chamber, the Al composition and film thickness are anticipated to be consistent between samples within a deposition run.
2.2 LAPPS reactor: passivation of Al samples
LAPPS makes use of a linear hollow-cathode electron source to generate sheet-like electron beams with typical current densities of 1-5 mA/cm2, and beam energies between 1-5 keV [5,6]. The system consists of the linear hollow cathode (length =16.2 cm), an entrance aperture through which the electron beam is injected, a termination anode, and magnetic field coils. Materials are introduced on a processing stage that is oriented parallel to the direction of beam propagation and below the beam volume. In this work the standoff distance between the beam axis and substrate is 2.5 cm. A co-axial magnetic field of 150 Gauss is used to collimate the e-beam and thus ensure uniformity along its length [15]. The chosen magnetic field strength leaves the plasma ions un-magnetized while aiding in electron confinement. These parameters are sufficient to produce uniform plasma sheets compatible with a processing stage that accepts 150 mm diameter substrates. The processing stage can either be left at ground (unbiased) resulting in substrate bombardment by very-low-energy ions (< 5 eV) or biased using either DC or RF voltage to raise the ion energy. Figure 1 illustrates the basic design of an e-beam based plasma processing system together with the orientation of the in situ ellipsometry components.
Fig. 1. (a) Side on view of the LAPPS reactor setup showing the orientation of the substrate and stage in relation to the magnetized electron beam. (b) Rotated side on view of the LAPPS reactor showing the orientation of the ellipsometer light source and detector relative to the sample. Optical emission spectroscopy access is shown for monitoring of F atom density.
Importantly, in highly electronegative gases such as SF6 and NF3, LAPPS enables the production of plasmas dominated by negative ions rather than electrons. These, so-called ion-ion plasmas enable the extraction of either positive or negative ion species from the system. Early work in Ar/SF6 plasmas show that for 50:50 mixtures (by pressure), both positive and negative ions can be delivered to the surface in equal amounts and with energies that are commensurate with the extraction bias [16–18]. For the present work a DC bias of +20 V was used to preferentially extract negative species from the plasma. This was found to be critical to the effective growth of thick, stoichiometric AlF3 layers on Al films.
The Ar/SF6 and Ar/NF3 gas mixtures were introduced into the reactor using MKS mass flows controllers. The reactor base pressure was nominally ≈5 × 10−7 torr. For this work, the flow fraction of the fluorine containing gas was varied between 2% and 50% with the remainder being Ar. The hollow cathode current was held constant at 20 mA, which is achieved by allowing the hollow cathode voltage to vary, thus producing a beam of constant current but varying energy. Monitoring of the atomic F density as a function of process conditions was performed using optical emission spectroscopy; spectral measurements in the Visible-NIR (630–830 nm) were performed using an Ocean Insight HR2000 + spectrometer with a 1200 g/mm grating providing 0.1 nm spectral resolution.
2.3 Safety and environmental information
NF3 is more hazardous than SF6 but significantly less so than HF; The US Occupational Safety and Health Administration (OSHA) states that the permissible airborne exposure limits in any 8-hour work shift of a 40-hour work are 6000 mg/m3 (SF6), 29 mg/m3 (NF3), and 2 mg/m3 (HF) [19]. NF3 is reactive and SF6 is an inert gas. Both SF6 and NF3 are denser than air, so there is a risk of asphyxiation hazard in enclosed spaces. In terms of relative global warming potential (GWP), SF6 has a GWP 23,500 times greater than CO2, and NF3 has a GWP 17,200 times greater than CO2.
2.4 Optical and structural characterization
The LAPPS processing reactor incorporates a J.A. Woollam M-2000 in situ spectroscopic ellipsometer for real-time monitoring of film thickness. The ellipsometer is set at a fixed 60° angle of incidence, and the operational spectral range is 193 nm – 1000 nm. The characterization of the passivation layer with in situ ellipsometry was done by modeling the optical properties (n,k) of the passivation layer with a Bruggeman effective medium approximation (EMA), which self-consistently fits the measured Ψ and Δ parameters based on a mixing of the optical constants of two pre-selected materials. In this case, the optical constants used were for amorphous Al2O3 and previously reported optical constants for AlF3 grown via exposure to the LAPPS plasma [3]. The optical constants for the Al native oxide layer were modeled using a Cauchy approximation, n(λ) = 1.751 + (6.32 × 10−3)/λ2-(1.0152 × 10−4)/λ4, assuming negligible absorption over the wavelength range measured.
Far ultraviolet reflectance was performed with the same instrument described in Ref. [2], using similar measurement configuration. X-ray photoelectron spectroscopy (XPS) was performed in a Thermo Fischer Nexus system using a monochromatic Al Kα source, flood gun for charge control, and a 400 µm spot size. Depth profiling was done with a MAGCIS dual-mode Ar ion source set in monoatomic mode at 3 kV and low current. X-ray reflectivity (XRR) measurements were performed on a Rigaku 3 kW monochromated Smartlab diffractometer (Cu Kα wavelength = 1.54059 Å). Atomic Force Microscopy (AFM) measurements were conducted using a Park Systems NX-20 AFM in non-contact mode utilizing a PPP-NCHR cantilever with a tip of <7 nm radius. Each sample was measured with a resolution of 512 pixels for a 1 µm x 1 µm size region scanned at 0.2 Hz and a 10 µm x10 µm size region scanned at 0.4 Hz.
3. Results and discussion
3.1 Plasma properties and passivation rate with SF6 and NF3 chemistries
The line ratio technique was used to measure ratio of excited atomic F, undergoing the $2{s^2}2{p^4}({{}_{}^3P} )3p \to 2{s^2}2{p^4}({{}_{}^3P} )3s$ (703.7 nm) transition, to excited Ar undergoing the $3{s^2}3{p^5}({}_{}^2P_{1/2}^{^\circ})$ $4p \to 3{s^2}3{p^5}({{}_{}^2P_{1/2}^{^\circ} } )4s$ (750.4 nm) transition, and then calculate the density of ground state F in the system based on the known density of the ground state Ar population. Further details on this method and its application in electron beam generated plasmas can be found in previous work by Boris et al. [20].
In Fig. 2, the atomic F density within the reactor for varying NF3 and SF6 flow fractions are shown. The results indicate the atomic fluorine density in NF3 containing plasmas is at least ≈ 4x greater than that for SF6 containing plasmas, except at the lowest flow fractions of F containing gas. For these measurements the chamber pressure varied between 25 - 35 mTorr depending on the flow fractions of Ar and either SF6 or NF3. This pressure variance is likely due to the difference in pumping speed between the two gases.
As expected, the higher atomic F content generally present in the Ar/NF3 lead to faster growth of AlF3 films compared with similar conditions in Ar/SF6 mixtures. An example of this behavior is shown in Fig. 3 where the temporal growth profiles of AlF3 films grown in 50% SF6 and 50% NF3 mixtures are compared, along with growth rates at the varying relative flow conditions. Figure 3(a) shows the approximately linear increase in thickness, followed by a plateau resulting from turning off the plasma once the desired AlF3 film thickness was reached. It is notable that film growth in the NF3 case is nearly 3x faster than in the SF6 case. The more rapid film growth in NF3 mixtures was observed under all conditions with the exception of the lowest concentrations of NF3 and SF6 (2% by flow) where the growth rate of the film was comparable for the two gases. This was not the only notable observation when flow fractions were low. Indeed, films grown with either gas showed the highest growth rates, but also the largest oxygen content and the lowest reflectivity.
Fig. 3. (a) The in-situ ellipsometer measurements for a sample processed with 50% SF6 and 50% NF3 showing film thickness as a function of process time. Note that the final thickness of both samples is ≈22 nm, with the 50% NF3 case requiring substantially less growth time. (b) Growth rates of the AlF3 films as a function of relative SF6 or NF3 flow.
Importantly, under these 2% flow conditions, the measured current to the stage was larger than the cases with higher flow fractions of NF3 and SF6. This is characteristic of a plasma that is not operating in an ion-ion plasma mode [18], where the concentration of negative ion forming gases are too low [20,21] and electrons are readily extracted from the plasmas. In conjunction with the growth rates and materials properties, it is evident that the process results are a function of both the flux of reactive fluorine and charged particles. In particular, the materials properties are improved when ion-ion plasmas are used.
At higher SF6 flow fractions of 10%, 30%, and 50% the growth time for a 21 nm passivation layer was 17.4 minutes, 22.8 minutes, and 28.1 minutes respectively. In contrast for the higher NF3 flow fractions the growth time for a 21 nm passivation layer was much less at 11.1 min, 11.0 minutes, and 11.2 minutes for the 10%, 30%, and 50% NF3 cases, respectively. The growth rates for these cases are shown in Fig. 3. The process times and thicknesses of all samples determined with the in-situ ellipsometer are included in Tables 1 and 2, where all relevant sample properties are summarized.
Table 1. Summary of the principal morphology values obtained in this section (SF6 samples)
Table 2. Summary of the principal morphology values obtained in this section (NF3 samples)
3.2 Passivation using SF6 or NF3: optical performance and microstructure comparison
This subsection compares the optical and structural properties of the samples processed with SF6/Ar (or NF3/Ar) gas flow fractions of 2%, 10%, 30%, and 50%. As was mentioned in section 2.2 a positive stage bias of +20 V was employed for all samples aimed at providing a constant negative ion energy (≈20 eV) to all the samples processed. The results are presented in Fig. 4(a) (for SF6) and 4b (for NF3). The substrate temperature was generally constant at 23-28 °C (depending on ambient conditions), monitored with an in-situ thermocouple. Interestingly, and in contrast to samples presented in Refs. [2–4], no inductively coupled plasma (ICP) source was used to augment atomic fluorine production. This is an important development, as the ability to grow AlF3 films without the use of the ICP source simplifies the process.
Fig. 4. FUV reflectance of samples processed with (a) constant ion energy, but different SF6/Ar flow ratios, (b) constant ion energy, but different NF3/Ar flow ratios. (c) UV/Optical/IR reflectance of all samples (including non-passivated bare Al).
In terms of gas (SF6 or NF3) flow fraction optimization, the best sample using 121.6 nm wavelength as a metric is A2 (SF6−50%) with R≈91%. Slightly lower Lyman-α reflectivity was found for samples Al10 (NF3−10%), Al11 (NF3−30%), and A1 (NF3−50%), all of them with a R≈85%, and then, samples Al12 (SF6−30%) and B2 (SF6−10%), with R < 85% and R ≤ 80% at 121.6 nm, respectively. The lowest Lyman-α reflectance was found for samples B1 (SF6−2%) and Al9 (NF3−2%), which were fabricated with the lowest fluorine containing gas flow ratio (2%). In both of these samples R < 65% at 121.6 nm. In contrast, the samples with higher reflectance (> 85%) are those fabricated with SF6 and NF3 flow ratios of 10% or more. The results further confirm the observation that film properties are better when both atomic F and negative ions are delivered to the surface, which is likely when large concentrations of either SF6 or NF3 are employed.
The samples reported in Ref. [2] presented a broadband reflectivity dip at ≈150-155 nm, attributed to the excitation of surface plasmons at the Al/AlF3 interface [22]. This excitation is driven by the effect of spatial structures in roughness smaller than the incident wavelength on the Al surface and leads to an additional absorption rather than a scattering of the light [23]. In this work, the mirrors presented in Fig. 4 have a barely noticeable Al plasmon effect. This fact hints to a significant improvement in terms of mirror roughness, which will be discussed below.
The reflectance of all samples in the UV/Optical/IR range shown in Fig. 4(c) does not show significant differences between samples processed with SF6 and NF3. Interestingly, the reflectance of bare Al (oxidized) is <1% greater than AlF3-passivated Al above 450 nm. This relative difference increases in the 450-200 nm range, before becoming increasingly negative (i.e., passivated Al reflects more than bare Al) with decreasing wavelength below ≈200 nm, where the thin native oxide layer on the bare Al starts absorbing radiation. This data is not shown in neither Fig. 4(a) nor Fig. 4(b).
XPS with depth profiling measurements (Fig. 5) were performed to probe the elemental composition and bonding states of the outermost 30-35 nm of the samples fabricated with different gas concentrations shown in Fig. 4. The aluminum, fluorine, and oxygen contents, bonds, and stoichiometry were derived from the measurements by analyzing the Al2p, F1s, and O1s peak position variation.
Fig. 5. XPS with depth profiling showing the F, Al, C, and O concentration throughout the fluoride layer of all samples passivated with similar ion energy (20 eV) but increasing gas flow ratios: 2%, 10%, 20%, and 50%.
We found that oxygen content seems to be correlated with a threshold amount of F. Samples processed with high F content (50% SF6 concentration or 10% or higher NF3 concentration) show very little oxygen in AlFx film (< 2 at%), and the films are extremely uniform in composition and bonding. Interestingly, the 50% SF6 case has comparable F content to the 10% NF3 case. Again, this agrees well with the atomic F densities presented in Fig. 1. The F/Al ratio of the samples with almost no O in the AlF3 layer is very consistent and close to 3, so these AlF3 layers are stoichiometric. At conditions with lower F content in either SF6 or NF3, there appears to be incomplete O removal from the film. In these, the O content is quite high and not consistent throughout the AlFx film. Regarding contaminants, no N or C was detected beyond the surface in any sample, but we still see small amounts S and Sn at the AlF3/Al interface, even for the samples processed with NF3. This result is quite interesting as it suggests that the reactor (in particular, the cathode) may have S- and Sn-containing atoms or molecules adsorbed/absorbed on/in the walls, and these molecules are desorbed whenever plasma is generated, no matter which process gas is used. We also measure a significant amount of O (≈ 10 at%) in the as-grown Al layer. The most probable reason is that this O was incorporated during the Al deposition at GSFC. As mentioned above, the vacuum chamber where Al samples were deposited was not baked out to evacuate water vapor prior to the Al evaporation. Stoichiometry and O content within the AlF3 layers are summarized in Tables 1 and 2.
The main conclusion from the XPS analysis is that there are no significant variations in terms of oxygen content, stoichiometry, and contaminants in the AlF3 layers among the samples processed using SF6 or NF3 with high F content. However, it is important to note that the residual concentration of O within the AlF3 layer decreases as the concentration of SF6 or NF3 increases. Among the best samples, A2 exhibits the lowest O content (1 at%), followed by A1 and Al11 (1.3 at%), and then Al10 with 1.9 at%. The impact of the O content within the AlF3 layer on the FUV reflectivity will be further investigated below.
XRR was performed to compare densities and verify the thickness of the AlF3 layers of samples processed with either SF6 or NF3. Figure 6 displays the XRR of samples shown in Fig. 4. For the sake of clarity, reflectance curves and models are multiplied for a factor of 10 with respect to the previous sample, starting from the curves (experimental and model) Al12 which are multiplied by 10°, Al11 curves are multiplied by 101, Al10 are multiplied by 102 and so forth, until A1 curves which are multiplied by a factor of 107.
Fig. 6. (a) XRR of samples processed with SF6 and NF3 (constant ion energy (20 eV), increasing gas flow ratio) as a function of the incident angle. Note that XRR curves are multiplied by a factor of 10n for clarity. (b) Density values of the AlF3 layer, compared with tabulated data.
The XRR models were built and fitted using a genetic algorithm with IMD software [24]. The fitting range was from 0.2 to 1.3 degrees; measurements below 0.2 degrees are expected to differ from real data because the beam section is half-cut by the sample at zero degrees incidence. Sample A12 was modelled starting from 0.25 degrees. A possible interface between substrate-Al was not considered in the model and neither was density nor stoichiometry gradients within films. However, the models accounted for the small density of particles found on the surface of samples during the AFM analysis by using a thin artificial layer on top of the AlF3, since XRR is sensitive to the lateral averaged density profile along the vertical direction. In this artificial layer, the optical constants of AlF3 were mixed with those of air, using the particle area ratio found in the AFM analysis as the volume ratio for the mixing of the optical constants.
The parameters obtained from XRR models are film density, thickness, and surface/ interface roughness. The results are summarized in Tables 1 and 2. The average thickness of the AlF3 layer is 22.2 nm with a standard deviation of 0.6 nm. This value is close to the designed thickness (21 nm) and confirms the efficacy of the in-situ ellipsometer for the thickness control of the protective layer. From the shortwave cut-off wavelengths shown in Figs. 4(a) and 4(b), we observe that the AlF3 layer is slightly thicker in A2 (50% SF6) than in A1 (50% NF3), and XRR models confirm that sample A2 has a slightly thicker AlF3 (23.4 nm) layer than A1, Al11 (30% NF3, 20 eV), and Al10 (10%NF3, 20 eV), which are 22.7 nm, 21.8 nm, and 22.4 nm, respectively. The impact of these small differences on the FUV reflectivity of mirrors will be discussed below. The average thickness of the Al layer is 76.6 nm with a standard deviation of 3.9 nm. Figure 6(b) presents the density of the AlF3 layer derived from models. The average density was found to be ≈3.1 g/cm3, which is at the higher end of the tabulated AlF3 bulk density range (2.88-3.10 g/cm3). This high density is expected as this room-temperature passivation process has demonstrated FUV reflectance and stability only compatible with very dense AlF3 films [2]. The two samples with the highest amount of O in the AlF3 layer, Al9, and B1, present a noticeably higher density of the passivation layer than other samples, which have a relatively constant density. The roughness values obtained from the XRR analysis will be discussed together with the data obtained from the AFM measurements.
The main conclusion from the XRR characterization is that there are no noticeable differences in terms of density, roughness, and layer thickness among the samples processed with either SF6 or NF3 with high F content.
The roughness for all of the samples was characterized by measuring the root mean square (RMS) of AFM scans over areas of 10 µm x 10 µm and 1 µm x 1 µm. The values are presented in Fig. 7. We also measured a bare Al sample (untreated) for comparison purposes, and the roughness is displayed in Fig. 7(a) with a blue-dashed line. As mentioned above, all samples presented a small fraction of their surface (generally < 1%) covered with particles, but sample A12 had an anomalous number of particles (8.6%), therefore for this sample the particles were artificially removed from measurements during the roughness calculation for a fair comparison with the other samples.
Fig. 7. (a) Roughness of the Al mirrors passivated with SF6 or NF3 using a constant energy of 20 eV. (b) The power spectral densities (PSD) as a function of spatial frequencies of all samples plotted along with a global PSD model. The PSD of a bare Al sample and of a bare glass substrate are also included. (c) AFM topographs (10 µm x 10 µm size) of the bare glass substrate, of aluminum, and of samples A1 (50% NF3) and A2 (50% SF6).
The average RMS roughness for all samples from the 10 µm x 10 µm scans is around 0.81 nm, and around 0.77 nm from the 1 µm x 1 µm scans. The roughness of the bare Al (oxidized) is 0.78 nm for the 10 µm x 10 µm and 0.75 nm for the 1 µm x 1 µm scans. The roughness of the bare glass substrate is 0.17 nm for the 10 µm x 10 µm and 0.13 nm for the 1 µm x 1 µm scans. We note that the roughness of bare Al is comparable with those of the passivated samples. We measured the roughness uniformity of sample A2 at five different locations, and the dispersion was only 0.03 nm so the small differences in roughness among samples shown in Figs. 7(a) and 7(b) are probably intrinsic to the different process parameters or variations in substrate roughness. Nevertheless, these differences are so small that we assume all samples have a similar roughness, so we can draw the important conclusion that the surface roughness of the AlF3 layer seems to be constant for all samples and independent of process gas (SF6 or NF3) and gas flow concentration (2% to 50%). The spatial frequency distribution of the roughness seems to be independent of the gas used (NF3 or SF6) and process parameters, as can be seen in Fig. 7(b) where all power spectral densities (PSDs) are plotted together, showing a good overlap. The “ABC” PSD model with a fractal term [25] was fit to all experimental PSDs to calculate a global value for the correlation length (τ), which can be crudely related to an averaged crystallite size (diameter) of all samples, and we obtained τ ≈ 55 nm.
These results are quite remarkable, as they show that the AlF3 passivation layer is conformal to the underlying Al layer and does not add roughness to the coating. This is a significant improvement when compared to other passivation techniques with AlF3, where surface roughness is reported to be > 1 nm [26–29]. The roughness values determined by AFM and reported in Ref. [2] were about 1.6 nm and independent of process parameters, except for higher ion energies (>200 eV). Under the present process conditions the roughness values of most of the samples in this work are almost half those values. Since we concluded before that the AlF3 passivation layer does not add roughness to the system, we suspect that these lower roughness values are because the Al used in this research may be smoother than the Al used in Ref. [2]. The Al deposition process has undergone a number of variations with respect to the previous work that may impact the Al surface roughness: new and smoother glass substrates from a different vendor (see Fig. 7(c)), higher Al deposition rate (≈17-18 nm/s), and a thinner Al (76 nm average) instead of the 87 nm-thick Al films used in Ref. [2], determined with XRR (not shown in this work). In addition, a new plasma reactor with lower base pressure was used to process these samples. Although we have not yet isolated which one of the variation(s) in the aforementioned processes caused the significant reduction in the Al roughness, the most plausible contributions are the substrates and the thickness of the Al, since there is a well-established correlation between increased surface roughness and Al layer thickness during deposition (see Fig. 2.80 in Ref. [30]).
We also determined the AlF3 surface and Al/AlF3 interface roughness from XRR measurements. We found that the AlF3 surface roughness of all samples is 0.91 ± 0.07 nm (higher than the average surface roughness determined with AFM), and the Al/AlF3 interface has an average roughness of 0.48 ± 0.018 nm (slightly lower than the roughness of the bare Al by AFM). The sample Al12 was excluded from these averages. We find that the agreement between the roughness values determined by AFM and XRR is acceptable, and the differences have two possible explanations. The first is that the roughness from XRR measurements is obtained from models, where the independent parameters are cross correlated to a certain extent. On the other hand, AFM is a direct measurement of the surface topography, so we give more weight to the values obtained through this technique. The second explanation is the different bandwidth of the power spectral density covered by these two methods. As explained in Ref. [31], the high frequency information in AFM scans is limited by the tip radius and the lowest spatial frequency on the surface is limited by the maximum scan range of AFM (10 µm x 10 µm). In contrast, the high spatial frequency in XRR is the inverse of about λ/2 (0.077 nm) and the lowest spatial frequency is limited by the coherence length (λ2/2Δλ) of the X-ray projected on the surface, which in turn is limited by the beam divergence of the X-ray source. Therefore, the characterization tool covering a larger bandwidth should result in a higher effective RMS roughness. Since the low frequency part is dominant in the power spectral density (see Fig. 7(b)), the technique which extends to the lower frequencies usually gives rise to a larger effective RMS roughness. The angular divergence, Δθ, of the XRR tool is 0.003°, and using Δλ/λ= Δθ cotθ, we calculate Δλ to be ≈1.4·10−7 nm. Hence, the coherence length is ≈84 µm, which means that XRR extends to lower spatial frequencies than AFM.
This detailed knowledge of the layer thicknesses, chemical composition, roughness and grain size, and existing optical properties of the Al [32] and AlF3 [3] layers allowed us to explain the reflectivity dispersion shown in Fig. 4 among samples processed with high F concentration and also calculate the reflectance loss budget using the formalism described in Ref. [33].
- • One of the main conclusions drawn from this research is that samples processed with higher atomic F content in plasma (A2, A1, A11, A10) do not show significant differences in terms of nanostructure and optical quality. However, there is a slight reflectivity dispersion among these samples at the reference wavelength of 121.6 nm. Based on XRR, XPS, and AFM measurements, we note that the stoichiometry, AlF3 density, surface roughness, and grain size are qualitatively similar for these aforementioned samples, hence these characteristics cannot explain the dispersion in R. Therefore, the reflectivity dispersion must be attributed to variations in the thickness and/or contamination of the AlF3 layer. Using calculations based on the Fresnel formalism, we find that the reflectivity dispersion at 121.6 nm among all samples shown in Fig. 4 is less than 0.4% due to small differences in the thickness of the AlF3 layer, so this hypothesis is also discarded. Hence, the reflectivity dispersion is likely caused by variations in the O content within the AlF3 layer. To validate this hypothesis, the reflectivity of all samples is plotted as a function of the O at% content within the AlF3 layer at different wavelengths (Fig. 8(a), filled circles) and compared to models (Fig. 8(a), dashed lines) in which the optical constants of AlF3 are combined with those of Al2O3 at different % to create a rough optical model of the AlFxOy layer. Note that the calculations at 140 nm and 160 nm wavelengths are plasmon-corrected, where the effect of the Al surface plasmon calculated in Fig. 8(b) has been considered. Interestingly, this simplified model reproduces the experimental data fairly well, and hence demonstrates the significant impact of small quantities of O in the AlF3 layer in the shorter wavelengths of the FUV range due to the strong absorption of Al2O3. For instance, the difference in O content between 0% and 2% in the AlF3 layer at 122 nm wavelength results in a decrease in reflectivity from ≈89% to ≈85%. This negative effect is mitigated at wavelengths above 160 nm, where Al2O3 becomes relatively transparent. Consequently, strategies to mitigate O in the AlF3 layer should be considered among the most effective ways to enhance the optical quality of the mirrors presented in this study.
- • Reflectance loss budget: Using one of the best samples A2 (50% SF6) with the lowest content of oxygen in the AlF3 layer, we computed the individual absorption of the Al and AlF3 layers, the transmission losses, roughness scattering losses (using the averaged PSD model shown in Fig. 7(b), and the Al surface plasmon absorption losses were calculated as unity minus all other losses added together. The results are displayed in Fig. 8(b). We observe that in the short FUV wavelengths (< 115 nm), the losses are dominated by the AlF3 absorption, but this absorption decreases with wavelength and becomes only 2.4% at 122 nm. Above 114 nm wavelength, the Al losses become dominant, with a relatively constant value of 7.1-7.4% throughout the FUV range. This absorption is intrinsic to the Al layer, and there is little we can do to significantly reduce it other than reducing the amount of oxygen within the layer by achieving a lower water partial pressure in the chamber prior to the Al deposition. Interestingly, the scattering losses (specular and non-specular) are very small, only 0.86% at 122 nm wavelength. The losses due to the excitation of the Al surface plasmon are negligible below 140 nm and noticeable only in the 140-180 nm range and are peaked at about 153 nm with 7.9%. The transmission losses are negligible throughout the FUV range (<0.04%) due to the thickness of the Al layer. Overall, the performance of the mirrors presented in Fig. 4 is very satisfactory, but there is still room for improvement via oxygen mitigation and roughness, as shown in Figs. 8(a) and 8(b).
Fig. 8. (a) Variation of reflectivity of the mirrors show in Fig. 4 as a function of oxygen concentration within the AlF3 layer determined in the XPS analysis, for a few selected wavelengths. The dots are experimental data, and the dashed lines are models in which the optical constants of AlF3 are mixed with Al2O3 with different % content. (b) The exemplary sample A2, processed with 20 eV and 50% SF6, along with the calculated reflectance losses budget.
4. Conclusions
In previous works, we demonstrated the feasibility of a room-temperature plasma process using electron beam-generated plasmas produced in Ar/SF6 backgrounds for effectively fluorinating the surface of Al mirror samples, resulting in native oxide removal and the formation of an AlF3 passivation layer with tunable thickness. It does not require elevated substrate temperatures or ultra-high vacuum conditions. This plasma treatment has been used to demonstrate high FUV reflectivity (R > 90% at λ=121.6 nm), good coating uniformity of the fluoride layer, low coating-induced polarization aberration, and improved durability.
In this work, we studied NF3 as a possible alternative to SF6. First, we showed that the atomic F density is generally larger in plasma produced in Ar/NF3 backgrounds than those produced in Ar/SF6 backgrounds, leading to a significantly faster passivation for Al in cases where flow fractions of fluorine containing gas are >10%. Then, we examined Al mirrors passivated in SF6 or NF3 containing backgrounds and compared the optical and structural properties of films grown in varying concentrations of these two gases. We found no significant differences in terms of FUV reflectance, stoichiometry, oxygen content within the AlF3 layer, density of the AlF3 layer, and surface roughness among samples fabricated with high concentrations (50% by flow) either process gas. However, at lower concentrations of SF6 (30%, 10%, and 2% by flow) increased oxygen content in the film was observed using XPS depth profiling. At lower concentrations of NF3 only the 2% case exhibited elevated oxygen content in the AlF3 layer. The broader concentration range over which almost complete oxygen removal was observed in the NF3 case is likely due to the higher atomic F density generated by NF3 containing plasmas. We conclude that Al mirrors can be passivated faster with NF3 than with SF6 without loss of optical performance or nanostructure degradation.
Funding
National Aeronautics and Space Administration (20-APRA20-0093 / N0017322GTC0044); U.S. Naval Research Laboratory (NRL Base Program).
Acknowledgements
This work was partially supported by the Naval Research Laboratory Base Program through the Office of Naval Research. LRM acknowledges the support of CRESST II cooperative agreement, which is supported by NASA under award number 80GSFC21M0002.
Disclosures
The authors declare no conflict of interest.
Data availability
Data underlying the results presented in this paper may be obtained from the authors upon reasonable request.
References
1. M. A. Quijada, D. R. Boris, J. del Hoyo, E. J. Wollack, A. C. Kozen, S. Walton, and V. Dwivedi, “E-beam generated plasma etching for developing high-reflectance mirrors for far-ultraviolet astronomical instrument applications,” Proc. SPIE 10699, 103 (2018). [CrossRef]
2. L. V. Rodriguez de Marcos, D. R. Boris, E. Gray, J. G. del Hoyo, A. C. Kozen, J. G. Richardson, S. G. Rosenberg, S. G. Walton, V. Wheeler, E. J. Wollack, J. M. Woodward, and M. A. Quijada, “Room temperature plasma-etching and surface passivation of far-ultraviolet Al mirrors using electron beam generated plasmas,” Opt. Mater. Express 11, 740–756 (2021). [CrossRef]
3. L. V. Rodríguez-de Marcos, D. R. Boris, J. Del Hoyo, Virginia Wheeler, Jeffrey M. Woodward, J. G. Richardson, S. G. Walton, E. J. Wollack, and M. A. Quijada, “Advanced AlF3-passivated Aluminum mirrors for UV astronomy,” Proc. SPIE 11820, 1 (2021). [CrossRef]
4. D. R. Boris, L. V. Rodriguez de Marcos, J. Del Hoyo, V. D. Wheeler, S. G. Walton, E. J. Wollack, and M. A. Quijada, “Plasma based production of AlF3-passivated aluminum mirrors for UVOIR astronomy,” Proc. SPIE 12181, 121813C (2022). [CrossRef]
5. W. M. Manheimer, R. F. Fernsler, M. Lampe, and R. A. Meger, “Theoretical overview of the large-area plasma processing system (LAPPS),” Plasma Sources Sci. Technol. 9(3), 370–386 (2000). [CrossRef]
6. R. A. Meger, R. F. Fernsler, M. Lampe, and W. Manheimer, “Large area plasma processing system,” U.S. Patent US5874807A (1999).
7. M. A. Quijada, L. V. Rodriguez de Marcos, J. G. Del Hoyo, E. Gray, E. J. Wollack, and A. Brown, “Advanced Al mirrors protected with LiF overcoat to realize stable mirror coatings for astronomical telescopes,” Proc. SPIE 12188, 66 (2022). [CrossRef]
8. M. A. Quijada, J. G. Del Hoyo, E. Gray, J. G. Richardson, A. Howe, L. Rodriguez de Marcos, and D. A. Sheikh, “Influence of evaporation rate and chamber pressure on the FUV reflectance and physical characteristics of aluminum films,” Proc. SPIE 11819, 18 (2021). [CrossRef]
9. S. Rauf, A. Balakrishna, A. Agarwal, L. Dorf, K. Collins, D. R Boris, and S. G Walton, “Three-dimensional model of electron beam generated plasma,” Plasma Sources Sci. Technol. 26(6), 065006 (2017). [CrossRef]
10. M. A. Quijada, D. R. Boris, L. V. Rodriguez de Marcos, V. D. Wheeler, J. G. del Hoyo, M. Batkis, C. Bos, S. G. Walton, and E. J. Wollack, “Environmental and polarization characterization of E-beam plasma-based AlF3-passivated Aluminum mirrors for astronomical telescopes,” Submitted to Proc. of SPIE12676 (2023).
11. See figures 10, 11 (uniformity), and 13 (polarization) in http://www.astrostrategictech.us/pdf/projectfiles/Reports/17-SAT17-0017_FR_Quijada_Manuel_2022-09.pdf
12. . Unpublished data.
13. D. C. Hays, K. B. Jung, Y. B. Hahn, E. S. Lambers, S. J. Pearton, J. Donahue, D. Johnson, and R. J. Shul, “Comparison of F2-Based Gases for High-Rate Dry Etching of Si,” J. Electrochem. Soc. 146(10), 3812 (1999). [CrossRef]
14. L. Rodríguez-de Marcos, B. Fleming, J. Hennessy, D. Chafetz, J. Del Hoyo, M. Quijada, M. Bowen, D. Vorobiev, and B. Indahl, “Advanced Al/eLiF mirrors for the SPRITE CubeSat,” Proc. SPIE 12188, 70 (2022). [CrossRef]
15. G.M. Petrov, D.R. Boris, E.H. Lock, T.B. Petrova, R.F. Fernsler, and S.G. Walton, “The influence of magnetic field on electron beam generated plasmas,” J. Phys. D: Appl. Phys. 48(27), 275202 (2015). [CrossRef]
16. S.G. Walton, D. Leonhardt, R.F. Fernsler, and R.A. Meger, “The extraction of positive and negative ions from electron beam-generated plasmas,” Appl. Phys. Lett. 81(6), 987–989 (2002). [CrossRef]
17. S.G. Walton, D. Leonhardt, R.F. Fernsler, and R.A. Meger, “On the extraction of positive and negative ions from electron beam-generated plasmas,” Appl. Phys. Lett. 83(4), 626–628 (2003). [CrossRef]
18. S.G. Walton and R.F. Fernsler, “Symmetric extraction of positive and negative ions from electron beam-generated ion–ion plasmas,” Plasma Source Sci. Technol. 18(2), 022001 (2009). [CrossRef] \
19. . https://www.osha.gov/chemicaldata
20. D. R. Boris, T. B. Petrova, G. M. Petrov, and S. G. Walton, “Atomic fluorine densities in electron beam generated plasmas: A high ion to radical ratio source for etching with atomic level precision,” J. Vac. Sci. Technol., A 35(1), 01A104 (2017). [CrossRef] \
21. G. M. Petrov, D. R. Boris, T. B. Petrova, and S. G. Walton, “One-dimensional Ar-SF6 hydromodel at low-pressure in e-beam generated plasmas,” J. Vac. Sci. Technol. A 34(2), 1 (2016). [CrossRef]
22. J. G. Endriz and W. E. Spicer, “Study of Aluminum Films II. Photoemission studies of Surface-Plasmon Oscillations on Controlled-Roughness Films,” Phys. Rev. B 4(12), 4159–4184 (1971). [CrossRef]
23. R. P. Feuerbacher and W. Steinmann, “Reflectance of evaporated aluminium films in the 1050-1600 Å region, and the influence of the surface plasmon,” Opt. Commun. 1(2), 81–85 (1969). [CrossRef]
24. D. L. Windt, “IMD - Software for modeling the optical properties of multilayer films,” Comput. Phys. 12, 360–370 (1998). [CrossRef]
25. G. Palasantzas, “Roughness spectrum and surface width of self-affine fractal surfaces via the K-correlation model,” Phys. Rev. B 48(19), 14472–14478 (1993). [CrossRef]
26. S. Wilbrandt, O. Stenzel, H. Nakamura, D. Wulff-Molder, A. Duparré, and N. Kaiser, “Protected and enhanced aluminum mirrors for the VUV,” Appl. Opt. 53(4), A125–A130 (2014). [CrossRef]
27. N. Gutiérrez-Luna, B. Perea-Abarca, L. Espinosa-Yáñez, C. Honrado-Benítez, T. de Lis, L. Rodríguez-de Marcos, J. A. Aznárez, and J. I. Larruquert, “Temperature Dependence of AlF3 Protection on Far-UV Al Mirrors,” Coatings 9(7), 428 (2019). [CrossRef]
28. J. Hennessy, A. D. Jewell, K. Balasubramanian, and S. Nikzad, “Ultraviolet optical properties of aluminum fluoride thin films deposited by atomic layer deposition,” J. Vac. Sci. Technol., A 34(1), 01A120 (2016). [CrossRef]
29. J. J. Hennessy, K. Balasubramanian, C. S. Moore, A. D. Jewell, S. Nikzad, K. C. France, and M. A. Quijada, “Performance and prospects of far ultraviolet aluminum mirrors protected by atomic layer deposition,” J. Astron. Telesc. Instrum. Syst. 2(4), 041206 (2016). [CrossRef]
30. R. R. Willey, “Practical production of optical thin films,” Ed. Willey Optical, 5th edition (2019).
31. H.-C. Su, C.-H. Lee, M.-Z. Lin, and T.-W. Huang, “A comparison between X-ray reflectivity and atomic force microscopy on the characterization of a surface roughness,” Int. J. Nanosci. 02(04n05), 343–348 (2003). [CrossRef]
32. E. Shiles, T. Sasaki, M. Inokuti, and D. Y. Smith, “Self-consistency and sum-rule tests in the Kramers–Kronig analysis of optical data: application to aluminum,” Phys. Rev. B 22(4), 1612–1628 (1980). [CrossRef]
33. E. Kretschmann and E. Kröger, “Reflection and transmission of light by a rough surface, including results for surface-plasmon effects,” J. Opt. Soc. Am. 65(2), 150–154 (1975). [CrossRef]
