Finishing of metal optics by ion beam technologies

Abstract. Ultraprecise mirror devices show considerable potential with view to applications in the visible and the ultraviolet spectral ranges. Aluminum alloys gather good mechanical and excellent optical properties and thus they emerge as important mirror construction materials. However, ultraprecision machining and polishing of optical aluminum surfaces are challenging, which originates from the high chemical reactivity and the heterogeneous matrix structure. Recently, several ion beam-based techniques have been developed to qualify aluminum mirrors for short-wavelength applications. We give an overview of the state-of-the-art ion beam-processing techniques for figure error correction and planarization, either by direct aluminum machining or with the aid of polymer or inorganic, amorphous surface films.


Introduction
Metal optics has attained increasing interest for satisfying the growing needs of high-performance components for short-wavelength optical systems. Aluminum is a widely used mirror construction material because of its low price, good machinability, and a detailed metallurgical knowledge. Technical aluminum alloy materials, such as AL6061, show an excellent shape stability, which is necessary for robust and durable optical devices. 1,2 Optical aluminum surfaces for use in the infrared spectral range are usually fabricated by single-point diamond turning (SPDT). [3][4][5] However, for short-wavelength applications in the visible (VIS) or even the ultraviolet (UV) range, the requirements on the optical surface quality increase. Apart from a high figure accuracy to allow for best imaging properties, very low surface roughness and waviness values need to be reached. Otherwise, scattering and diffraction losses decrease the optical device efficiency. For VIS and UV spectral range applications, root mean square (rms) roughness values of 1 nm and <0.3 nm need to be targeted, respectively. 6 However, by SPDT, a typical roughness of 2-to 3-nm rms is achievable. 1,7,8 An additional issue is the periodic turning marks with a pitch of 1 to 10 μm distance and up to a few ten nanometers in height. As a result, undesired diffraction effects appear in UV/VIS range applications.
Topography errors on optical surfaces are usually categorized by the spatial frequency of their occurrence (Fig. 1). 9,10 Low-spatial frequency (LSF) errors are found for surface errors below the cutoff frequency f f separating the figure and the waviness. On the contrary, high-spatial frequency (HSF) error comprises surface errors larger than the cutoff frequency f c of waviness and roughness. In particular, microroughness is also included in the HSF region, which is separated from roughness by the cutoff frequency f s . In between the LSF and the HSF, the mid-spatial frequency (MSF) errors are situated. Standard cutoff frequency values are given in Deutsche Industrie-Norm ISO 10110-8:2012-02 as f f ¼ 400 m −1 , f c ¼ 12.5 mm −1 , and f s ¼ 400 mm −1 . However, those values strongly depend on the specific application and, in particular, the applied wavelength of light. The ISO reference wavelength is 546.07 nm. The shorter the wavelength in application, the higher are the cutoff frequencies. Hence, the cutoff frequencies are illustrated as ranges in Fig. 1.
In optical device finishing, the following surface errors are targeted: (1) LSF errors are surface figure errors, i.e., a reduced perfection in the surface shape. Figure errors cause imaging errors as distortions and can be corrected by ion beam figuring (IBF). (2) Regular or irregular error features in the MSF-to-HSF range are waviness and roughness. Prominent feature types are the turning marks from SPDT, which result from the feed of the diamond tool. Furthermore, chatter marks resulting from mechanical vibrations of the turning machine are apparent. Unfortunately, latter are situated in the MSF error range and cannot be corrected so far. Ion beam planarization (IBP) is a common technique to improve the waviness and roughness. (3) Surface errors, which are smaller than the wavelength of light, are microroughness features. As a result of microroughness, the optical surface exhibits a diffusive shine, the so-called halo. 11 Microroughness can also be faced by IBP. For any surface-finishing technique, it is desired to not only improve the surface in the specific spatial frequency range but also take care that the surface is not diminished in another spatial frequency range.
Ion beam technologies offer particular solutions for figure error correction as well as for surface smoothing to qualify optical surfaces (Fig. 2). However, machining of optical aluminum surface is challenging, and not all of the techniques illustrated in Fig. 2 are currently technically mature enough to meet the requirements for UV/VIS range applications. This paper is intended to give an overview of the state-ofthe-art ion beam machining techniques available for aluminum mirror finishing, in particular, IBF [ Fig. 2 Fig. 2(c)] are presently under development but need to be improved to be applicable to aluminum optics. In Sec. 2, the recent progress in IBF by reactive process control is focused. In Sec. 3.1 polymer deposition techniques as standard spin-coating and innovative plasma-jet deposition strategies are compared. To achieve ultrasmooth optical surfaces, the application of an amorphous nickel-phosphorus (NiP) thin film is an established technique. Recent developments are treated in Sec. 3.2.

Figure Error Correction by Deterministic Ion
Beam Technology IBF is a well-established technique for direct figure finishing of optical surfaces. 6,12,13 As a device-finishing technique, IBF is a slow but most accurate and controllable process. Hence, figure error correction is almost aimed at the nanometer scale as the ultimate surface figuring step in the optical device fabrication chain.
The ion beam width is small compared to the optical surface to be machined. Usually low-energy argon or xenon ions (E < 1.5 keV) are applied in a pure physical sputter erosion process. [14][15][16][17][18][19] The reactive ion beam etching (RIBE) technology allows further degrees of freedom since chemical interactions between the ion beam and the substrate material are applied to increase the process' efficiency, e.g., by an increasing etching rate due to the formation of volatile chemical species 20,21 or to improve the surface quality by a chemical modification of the substrate surface during ion irradiation. 22,23 In principle, the process sequence for figure error correction is as follows: first, a dwell-time algorithm is applied to transfer the figure error profile into a motion map. Afterward, the ion beam is moved deterministically along the device surface following the simulated motion profile. This approach allows the figure error correction of huge and diversely shaped mirror devices. The pathway is s-like in standard operation, but this not a necessity. The pathway can be adapted to the geometry of the special figure-removal profile.  For example, for the figure error correction of a radial symmetric surface error, it can be more advantageous to use sample rotation to reduce the dwell-time problem to a linear motion in radial direction. 24 The smaller the tool size, the higher is the flexibility in the device shape and the higher are the spatial frequency features to be reached during correction. Thus, the deterministic technology approach allows the machining of strongly curved or even free-form optics.
With a view to short-wavelength imaging systems, mirror devices often exhibit a complex figured surface with a spherical, aspherical, or free-form shape. 25 For ion beam generation, a 13.56-MHz transformercoupled plasma (TCP)-type ion beam source is used. 26 Inert (Ar) or reactive (O 2 , N 2 ) process gas is applied at a beam voltage of ≤1.5 kV (Fig. 3). The ion extraction and beam formation is performed by a triple grid system. The grids are spherically concave-shaped. As a result, the ion beam exhibits a constricted region. The working distance is chosen as the point of maximum constriction, i.e., at the smallest beam width. A higher working distance allows better flexibility in sample positioning in front of the ion source during ion-beam processing. The beam constriction and the working distance can be tuned by adjusting the grid curvature. 27 With a curvature radius of 150 mm and a screen grid opening of 10 mm, a narrow ion beam with a full width at half maximum (FWHM) of ∼5 mm and a strong ion current density of 8.1 mA∕cm 2 are obtained at the working distance of 32 mm [ Fig. 3 The aluminum surface composition is characterized by time-of-flight secondary ion mass spectrometry (TOF-SIMS). For topography analysis, aspherical stitching interferometry (ASI, QED technologies), white-light interferometry (WLI), and atomic force microscopy (AFM) are used. To allow a quantitative topography evaluation, the power spectral density (PSD) function is calculated from the WLI and AFM measurements and rms roughness values are determined within the reliable spatial frequency range, which depends on the particular WLI and AFM measurement conditions (see Ref. 23, for the details on the measurement equipment, PSD evaluation, and roughness calculation).
Usually the figure error correction of optical devices is performed by IBF with Ar gas. However, if this process is applied to aluminum surfaces, a strong degradation is obtained. 28 Defects become enlarged and a huge surface roughening develops. As reasons, a high Al surface mobility and preferential sputtering are addressed. 23 Recently, it has been shown that the application of reactively driven, low-energy ion beam tools allows the direct figure error correction of mirror optics made from aluminum technical alloy materials AL6061 and AL905. 23 Nontoxic gases, such as oxygen and nitrogen, are used for ion-beam processing. As a result of the low-energy reactive ions impacting the aluminum, the surface is chemically modified [ Fig. 4(a)]. By the interplay between near-surface ion implantation and sputter erosion, quasistationary surface conditions develop. Hence, a characteristic surface layer is formed with 12-to 15-nm thickness [ Fig. 4(b)]. The erosion process rests on the pure physical sputtering, enabling a high degree of process control. Furthermore, the sputtering process occurs within this surface layer exclusively. It is noteworthy that no volatile species is formed during RIBE processing of aluminum surfaces.
The surface layer is very advantageous for the ion beam machining process, because it inhibits the Al-surface mobility efficiently. Furthermore, structural and compositional inhomogeneities of the aluminum are passivated. As a result, the surface roughness is almost preserved during RIBE figuring [ Fig. 5(a)]. This conception holds up to 400 nm in depth for the oxygen process 23 and up to 1 μm for the nitrogen process [ Fig. 5(b)]. With increasing depths beyond these limits, etch pits emerge, which are related to the heterogeneous Al alloy matrix. In particular, in Al6061, the preferential sputtering of precipitates of Mg 2 Si and the    Fig. 4 (a) Sketch: during RIBE treatment a quasistationary nearsurface layer is formed as a result of the interplay between ion implantation and sputtering. This surface layer passivates structural and compositional inhomogeneities of the Al matrix. (b) TOF-SIMS depth profiles revealing the formation of the near-surface oxide or nitride layer after RIBE treatment with 1.5-keV oxygen or nitrogen ions, respectively. Note, after ion processing the samples are exposed to the lab environment. Thus, the Ar ion-treated surface exhibits a native oxide layer.
quaternary phase Al 5 Cu 2 Mg 8 Si 6 have been identified to cause etch pits. 23 The width of the etch pits ranges within a few micrometers for Al6061, while for Al905 almost submicrometer sized etch pits are apparent. Further investigations are necessary to clarify the etch pit formation in Al905 surfaces. However, for a detailed analysis, it is necessary to know the main alloy constituents. Rapidly solidified aluminum (RSA) from RSP technology is applied. Diverse composition sets are found in literature for the RSA alloy materials. The following composition sets are sorted by descending atomic fraction: Gubbels et al. 29 report on (1) Al6061 containing Mg, Si, and Cu, and (2)  With the RIBE figuring technology, it is now possible to correct the figure error of aluminum mirror surfaces without degradation. Usually surface figuring is an iterative technique, i.e., the figure error is reduced in a stepwise procedure starting from the correction of coarse surface errors with lower spatial frequency. For that reason, aperture-less ion beam tools with a beam width in the range of a few millimeters (as seen in Fig. 3) are applied. The subsequent refinement of the surface figure is achieved by lowering the ion beam width in the next iteration steps. Depending on the figure requirements defined by the application, typically 1 to 4 iteration steps need to be performed. For the present, an ultimate beam width of FWHM ∼ 0.6 mm is achievable by applying a contraction aperture, 27 defining the upper spatial frequency limit for figure error correction at ∼1.6 mm −1 (Fig. 1). Typical removal rates for the nitrogen ion beam process at 1.5 keVon aluminum are 0.80 mm 3 ∕h for an ion beam tool with FWHM ∼ 5 mm and 0.003 mm 3 ∕h for FWHM ∼ 0.6 mm. The lower the beam width, the lower is the removal rate and thus the higher are the process accuracy and controllability. Figures 6 and 7 show an example on the coarse figure error correction of a 50-mm diameter planar Al905 disc optics as a first step in the iterative process sequence in RIBE finishing. The figure error after SPDT is radially symmetric, exhibiting a hill of about 220 nm in the center of the sample [ Fig. 7(a)]. The task was the reduction of this figure error by 100 nm due to the removal of a best-fit paraboloidalshaped profile. For this reason, the aperture-less ion beam tool, which is depicted in Fig. 3, is used in deterministic operation via the dwell-time approach. Figure 6(a) shows a footprint etching in Al905. The footprint profile reproduces the Gaussian tool function excellently in shape and size. Hence, the reactive ion beam exhibits a proportional material removal, which is an essential requirement for a deterministic machining scheme. A crossed machining procedure with two perpendicular runs has been chosen [ Fig. 6(b)]. First, a parabolic correction with a 50-nm center removal is performed. Then a second correction is done by an additional perpendicular run. The deterministic machining is performed by s-like scan paths with an equal line pitch of 0.5 mm to allow a sufficient scan line overlapping. On each scan line the drive axis velocity is held constant, but the velocity is varied from line to line to merge the individual line removal characteristics into a parabolic removal profile. A dwell-time simulation is calculated to determine the drive axis velocity profile resulting in the best-fit parabolic removal [ Fig. 6(c)].    The deposition of polymer film masks for use in the mask transfer technique has been investigated by atmospheric plasma-jet deposition with methane as a film-forming agent. 31 The plasma-jet system employed for thin film deposition has been originally developed for atmospheric pressure plasma etching, where very high spatiotemporal stability of the plasma discharge is required. [32][33][34] High frequency power is used to ignite and sustain the plasma. The plasma-jet source consists of a coaxial conductor system [ Fig. 8(a)]. The inner conducting electrode forms at the same time as the feeding gas tube. The body of the plasma source acts as outer conductor of the coaxial system. The system is tuned in a manner that maximum electric field strength is ensured at the outlet of the inner tube electrode. There, the jet-like plasma of ∼10-mm length is discharged. The source is mounted on a three-axis computerized numerical control stage to enable a relative motion over the substrate surface. The plasma jet is fed by 900 sccm helium as carrier gas. The layer-forming precursor is methane with a flow rate of 5 sccm. A peripheral nitrogen-shielding gas flow of 350 sccm prevents the entrainment of surrounding air into the plasma and ensures spatial stability of the jet. Plasma excitation is performed by microwave at 2.45 GHz, applying short pulses of 5 μs at a repetition rate of 6 kHz and peak power of 200 W. On an average, a power of 6 W is dissipated, which maintains a surface temperature slightly above the room temperature. Under static conditions, where the plasma source is not moved, thin film deposition occurs on substrates in the form of a Gaussian footprint exhibiting a FWHM of ∼1 mm and a maximum deposition rate of 160 nm∕s. Homogeneous film deposition is performed by raster path motion on a 20 × 20 mm 2 area on silicon substrate. Total thickness of the films is determined by reflectometry to be ∼190 nm. The film composition strongly depends on the process gas composition and the plasmajet parameters. [35][36][37] The x-ray photoelectron spectroscopy (XPS) analysis at the deposited polymer films has revealed a composition relation of C∶N∶O ≈ 4∶2∶1 from C 1s peak asymmetry evaluation [ Fig. 8(b)]. Furthermore, strong hydrogen group bands as NH and CH x are observed in the Fourier-transform infrared spectroscopy (FTIR) fingerprint [reference plot in Fig. 9(a)].
The polymer films are machined by RIBE applying the nitrogen process. The aperture-less configuration (Fig. 3) is used with a beam width of FWHM ∼10 mm at 600 eV and FWHM ∼17 mm at 1200 eV ion energy. The polymer film samples are covered by a graphite hard mask with a circular 13 mm opening to obtain a distinct machining field edge. A homogeneous etching is performed by s-like scanning of a 21-mm-sized scan field over the mask opening with 3 mm∕s scan velocity and 3 mm scan line pitch. The machining duration is 65 s. The effect of RIBE on the polymer structure is analyzed by FTIR. In particular, the choice of the ion energy considerably influences the polymer matrix [ Fig. 9(a)]. After RIBE at 600 eV, the general FTIR footprint is maintained corresponding to the initial polymer film before RIBE (reference plot). In contrast, after RIBE at 1200 eV, the dominating FTIR peaks in the range of 1700 to 2700 cm −1 are strongly diminished. The corresponding vibrational bands refer to multiple bound fractions within the polymer matrix. This result indicates either a higher degree of crosslinking or matrix fragmentation by release of lessbound groups or chain cracking and thus polymer degradation caused by the ion irradiation. In Fig. 9(b), the etching rate during RIBE machining is focused. For the usual case of sputter erosion as a pure physical process, it is expected that the higher the ion energy, the higher the etching rate would be. This situation holds for photoresist and aluminum. However, the polymer films exhibit an opposite behavior. In accordance to the FTIR results, it is suggested that the crosslinking of the polymer, which is intensified by a higher ion energy, results in a higher resistivity in RIBE processing.
To achieve a higher RIBE resistivity, diverse postdeposition treatment procedures have been tested: (1) exposure to 365 nm UV light for 45 min (UVA), (2) exposure to 172 nm UV light for 5 min (UVC), and (3)  (b) (a) Fig. 8 (a) Microwave-driven plasma-jet tool used for local polymer deposition. Helium is supplied through the inner capillary,while a cold plasma is generated by 2.45-GHz microwave heating. Methane acting as the polymer-forming agent is admixed to the nitrogen shield gas, which homogeneously surrounds the helium plasma torch. (b) XPS analysis of the polymer film revealing a composition relation of C∶N∶O ≈ 4∶2∶1.
10 min at 100°C and 10 min at 200°C. For all treatment procedures, an increased RIBE resistivity, i.e., a reduction of the etching rate, is obtained [ Fig. 9(b)]. However, the treatments cause a shrinkage of the deposited film thickness (Table 1). An ideal postdeposition treatment would result in marginal polymer shrinkage, but strong decrease of the polymer etch rate in the RIBE process. As seen in Table 1, shrinkage and rate decrease develop with similar tendencies for the UVC treated and the annealed case. Hence, both effects level out each other to a considerable degree. The overall etching time necessary to remove the complete polymer film on the machined area A can be calculated from the deposited film thickness d 0 , shrinkage s, and the etching rate R after postdeposition treatment: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 1 ; 6 3 ; 3 8 The relative etching rate decrease r refers to the etching rate R ref of the untreated polymer film. As a measure, an effective process efficiency yield f can be defined as E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 2 ; 6 3 ; 3 0 9 f ¼ The f-values are summarized in Table 1 for the different treatment procedures. As a result, there is a ∼20% effective increase in process efficiency for UVC irradiation and annealing, compared to the untreated polymer case, meaning that the overall etching time is prolonged for both kinds of treatment to a very similar extent. Hence, the resistance of the polymer film is effectively increased, thereby lowering the etch selectivity. In contrast, for the UVA treatment, a negative efficiency yield is found. Thus, the UVA treatment is disadvantageous for process improvement. The effective selectivity values in Table 1 include the effect of shrinkage due to post-treatment: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 3 ; 3 2 6 ; 4 5 2 The selectivities reported are far beyond unity. Thus, the present plasma-jet-deposited layers are not appropriate for IBP processing. However, these films are interesting to use in the figure transfer technique. The figure error topography has to be converted into a modulated polymer thickness map profile. This map profile is given by the figure error profile, which has to be multiplied by the reciprocal of the effective selectivity value, which is given as the stretching factor m in Table 1. Hence, the thickness profile has to be stretched by a factor of 5 to 7, compared to the actual figure error profile. In principle, the height accuracy should be improved due to the stretching. However, one issue in film topography deposition is the realization sufficient steep slopes to correctly reproduce the error topography. But a stretching of the profile involves a stretching of the slopes particularly. The steeper the slopes, the smaller the   Fig. 9 (a) FTIR measurement of untreated polymer films before and after RIBE machining. Multiple bonds within the polymer matrix are destroyed by ion beam treatment at 1200 eV. (b) Etching rate during RIBE machining with respect to the ion energy. The etching rate shows a general trend for the investigated materials: polymer films > photoresist > aluminum. Table 1 Evaluation of an effective process efficiency yield and an effective process selectivity value with respect to the applied postdeposition treatment procedure. The deposited film thickness prior to the treatments is ∼190 nm. The effective process efficiency yield is reduced by the thickness shrinkage during the postdeposition treatment and is increased by the etch rate decrease in RIBE machining. UVC and annealing are roughly equally advantageous.
Polymer treatment Thickness (shrinkage s) RIBE depth (rate decrease r ) Effective process efficiency yield f Eff. selectivity (stretching m) plasma-jet tool width to be chosen should be. The dependence of the tool width on the deposition rate follows a squarish relation. As a result, the profile stretching prolongates the deposition time by two factors: (1) corresponding to the stretching, a higher film thickness is needed to completely capture the error heights by the RIBE process and (2) the plasma-jet tool size has to be reduced to meet the requirements on the increased slope values, which are necessary to achieve an accurate reproduction of the stretched figure error profile. An important property of either resist-related ion beam technique is the surface roughness. In Table 2, the surface roughness is summarized for the different postdeposition procedures. In general, the polymer films are very smooth with R q ∼ 0.5 nm after deposition. The UVA treatment even slightly improves the roughness. After UVC irradiation and annealing, the initial roughness is preserved. During RIBE machining, the untreated polymer film increases in roughness to minor extent. All postdeposition-treated films exhibit the initial roughness of ∼0.5 nm also after RIBE instead. Hence, postdeposition treatment is advantageous to maintain the surface roughness during RIBE machining. In general, from the geometrical point of view, a higher profile stretching should result in a reduction of the surface roughness. However, comparing the stretching factors in Table 1 and the roughness values after RIBE in  Table 2, such a trend cannot be observed in the present study. In contrast, the roughness values for all post-treated procedures are almost equal. Note, in addition to the geometrical influence resulting from the thickness stretching, structural effects within the polymer matrix can also contribute to the ultimate roughness after RIBE. Further experiments are necessary to account for structural changes in relation to diverse post-treatment processing procedures and the subsequent behavior under ion irradiation. Those should include: (a) the state of cross linking, (b) polymer matrix degradation effects, (c) the release of less-bound groups by energy impact, (d) reactive ion implantation and incorporation into the matrix, and (e) stoichiometry changes as a result of the component-wise different sputtering yields.
As mentioned earlier, so far the plasma-jet-deposited layers are no option for use in direct IBP. Spin-coated photoresist layers [see Fig. 9(b)] exhibit a higher selectivity of ∼0.6 related to aluminum in the ion energy range of 0.9 to 1.5 keV. However, these selectivity values are also even far from unity. Improved planarization layer materials need to be found to establish an efficient, direct IBP approach.

Ultrasmooth Optical Surfaces with Aid of an Isotropic Surface Layer
At present, the direct smoothing of optical aluminum surfaces by IBP, magnetorheological finishing (MRF), or chemical mechanical polishing (CMP) is not sufficient to reach a surface roughness of ≤1.0 nm rms, [38][39][40] which is a necessity for its application in the UV/VIS spectral range. For this reason, an alternative technological solution has been established. The aluminum device is coated with an amorphous film to allow an improved surface smoothing. Rather than the heterogeneous matrix of an aluminum alloy, a structurally and chemically isotropic material allows a much better machinability in SPDT as well as in ion beam machining. Amorphous materials, such as a-Si or NiP, are common materials in optics fabrication. [41][42][43][44] The a-Si can be produced by magnetron sputtering or plasma-enhanced chemical vapor deposition. 45 The NiP is deposited in an electroless approach from an electrolytic bath. 46 In contrast to galvanic deposition, no electric current is applied and the nickel is provided from the bath or by a salt. To increase the reflective surface properties of both amorphous materials, usually a further metallization layer (e.g., Au, Ag) is needed. A major drawback of NiP-coated optics is the mismatch in the coefficients of thermal expansion (CTE) between coating and aluminum. Especially under thermal load conditions, severe issues can arise, e.g., bending due to the bimetallic effect and NiP delamination. 47 For the same reason, lightweight device architectures with low material thickness are strongly limited, if NiP coatings are necessary. 38 Alternative aluminum alloy compositions as AlSi40 with improved CTE matching are under consideration. 48 However, for the present, NiP is the material of choice for fabricating UV/VIS mirror optics, because it is well-shapeable and reveals ultrasmooth surfaces after IBP. 49,50 Main goals for IBP are the reduction of the regularly aligned and distributed turning marks resulting from SPDT as well as the broadband smoothing of the optical surface in the roughness and the microroughness spatial frequency range (see Fig. 1). For the study, typical NiP samples after SPDT with turning marks with spacings of 1.5, 3.5, 6, and 25 μm and heights of 10, 20, and 60 nm are applied. A spray-coating resist (SX AR-PC 5000/22.4, Allresist GmbH, Germany) with thickness ranging between 170 and 300 nm is used as planarization layer. The higher the resist layer thickness, the better is the smoothing behavior. 50 Ar ion processing is performed by a Kaufman-type broadbeam ion source with 700-V beam voltage. A two-grid extraction system with 180-mm grid opening is applied. The corresponding Ar ion beam has a beam width of 120 to 150 mm (FWHM) and a beam current of 70 mA. Prior to IBP processing, the angle dependencies of the etching rates of photoresist and NiP are analyzed [ Fig. 10(a)]. The working point for IBP is defined by the intersection point of both curves to achieve a selectivity near unity. Hence, following a planarization angle of 35 deg with an etching rate of 10 to 12 nm∕ min is applied for ion-beam processing.
The topography is tracked by AFM analysis during all stages of IBP processing [ Fig. 10(b)]. An NiP surface with turning marks having 1.5 μm spacing and 20-nm height shows a rms roughness of ∼6.4 nm after SPDT. A considerable surface smoothing toward ∼2.0 nm rms is achieved by deposition of the photoresist planarization layer. By Table 2 Effect of the postdeposition procedures on the rms roughness (R q ) of the polymer film before and after RIBE machining. The roughness values are determined by WLI in the spatial frequency range of ð3.6 − 60Þ · 10 −3 μm −1 . means of argon ion beam treatment, this smooth topography is transferred into the NiP surface. As a result, the smooth topography of the planarization layer is fully maintained in the NiP after IBP. Hence, the IBP process exhibits a roughness decrease of about 60% to 70% during a single-process cycle.
In this study, the resist is completely removed after 30min ion-beam processing. To attain more detailed information about process stability and controllability, the temporal evolution during Ar ion treatment is examined at NiP surfaces with turning marks having 6-μm spacing and 60-nm height [ Fig. 11(a)]. Up to 60 min of processing time after complete resist removal, the surface roughness is kept constant at the value of ∼6.4 nm rms. A more detailed view on the surface topography reveals no apparent qualitative degradation [ Fig. 11(b)].
A much improved surface smoothing can be obtained by iterative IBP processing (Fig. 12). In addition to the significant reduction of the turning mark features, the broadband roughness is considerably decreased after each IBP run cycle [ Fig. 12(a)]. As a result, a decrease of ∼88% in surface roughness is obtained after four to five IBP run cycles [ Fig. 12(b)]. Note, the strongest roughness reduction with ∼70% shows already after the first IBP run cycle. During second to forth IBP run cycle, the improvement is already at 20% to 25% per cycle. Beyond four IBP runs, the surface topography reaches an ultimate smoothness. Further improvement by IBP is not obtained. With a residual roughness of 0.7 to 0.8 nm rms, IBP reveals similar surface qualities as a recently reported NiP smoothing approach, which is based on a combination of MRF and postpolishing. 51 In addition, a very low surface roughness of 0.3 nm rms for NiP has been reported by a combination of fluid jet polishing and bonnet polishing. 52 However, the results are based on WLI measurements, which are limited in the range of microroughness, and thus are not fully comparable to this study. An alternative approach to NiP could be a coating of pure aluminum on the optical surface, which can be smoothed more easily <1.0 nm rather than technical aluminum alloy material by conventional techniques as CMP. 53 Future work has to be done to test this conception also with IBP. However, for complex figured mirror surfaces, a homogeneous aluminum coating could become a challenging issue. In summary, ion beam technologies exhibit the following diverse capabilities in finishing of aluminum-based optics.
1. Direct figure error correction of optical aluminum surfaces: IBF with a reactive process control is a most promising technique for a direct figure correction approach on optical aluminum surfaces. RIBE allows an improved process control for aluminum surface machining. The use of oxygen or nitrogen process gas permits etching depth up to 400 nm and 1 μm, respectively, while the surface topography is almost preserved in its initial state. Beyond these depth limits, surface degradation by etch pit formation is observed. These etch pits are related to the heterogeneous aluminum alloy matrix. In Al905, smaller etch pits are obtained than in Al6061 material. As a result of the interplay between implantation of the low-energy ions and sputter erosion, the etching process is based on the formation of a quasistationary passivation layer in the 12-to 15-nm near-surface region. Narrow ion beam tools with 0.6 mm up to a few millimeters in width are available to successfully perform figure error correction without surface roughening. Corresponding to the etching selectivity of <1, the figure mask profile has to be stretched compared to the figure error profile. Since the error profile is converted into a local thickness modulation within the figure mask, the transfer can then be done by use of a broad beam ion source. Further studies are necessary to evaluate this approach in optical device fabrication. 3. IBP of NiP-based optical surfaces: Surface smoothing by IBP is reached by the topography leveling properties of the resist planarization layer. A proper estimation of the geometric conditions during ion-beam processing is necessary to allow a loss-less transfer of the smooth planarization layer topography into the NiP surface. The smoothing effect is improved by usage of increased resist layer thickness and iterative IBP processing. Consequently, the surface roughness can be reduced by up to ∼88% after four to five IBP run cycles. The residual surface roughness of 0.7-to 0.8-nm rms after IBP permits high-performance optical applications in the visible spectral range.  Diamond turned 1st run 2nd run 3rd run Fig. 12 The topography is improved by iterative IBP processing. (a) Evolution of the PSD function before and after the iterative steps (turning marks: 6 μm spacing/20 nm height). A broadband decrease of the PSD functions is observed with increasing iteration number. Furthermore, the peak features representing the turning marks diminish progressively. (b) The rms roughness decreases exponentially with increasing iteration number (turning marks: 1.5 μm spacing/20 nm height). After a fivefold IBP iteration sequence, the surface roughness decreases from initially ∼6.5 to ∼0.8 nm after IBP. 50 Jens Bauer received his diploma degree in physics and finished his PhD at the Institute of Inorganic Chemistry at the University of Leipzig in 2009. Afterward, he worked in institutional research as well as in several companies with focus on nanomaterial and thin film technology. Since 2014, he has been investigating optical device finishing by ion beam technologies at the Leibniz Institute of Surface Engineering (IOM) in Leipzig. As an author, he has contributed to 27 peer-reviewed papers.
Frank Frost received his doctorate from the University of Leipzig in 1998. His research focuses on fundamental investigations of the manufacturing of nano-and microstructures using (reactive) ion beam etching (RIBE), pattern formation due to ion beam driven self-organization, the use of ion beam processes for smoothing and planarization down to sub 0.1 nm surface roughness level, and the implementation of these technologies in production practice. He has authored/coauthored more than 130 papers. Thomas Arnold has been active in the area of ultraprecision surface machining for more than 20 years. His current research interests include technology development for optical freeform surface fabrication by ion beams as well as chemically reactive plasma jets targeting form accuracy in the nanometer range. He is the author of 30 peerreviewed papers. Since 2014 he has been endowed professor for ultraprecision surface machining using ions and plasmas at Technische Universität Dresden, Germany.