1. Introduction

The needs of space-based remote sensing sensitivity and resolution drive telescope apertures to be ever larger. For instance, to achieve a 1-m ground resolution from a geosynchronous orbit of the earth, the clear aperture of the telescope should be as large as 20 m [1], which is hardly achieved in the currently used conventional reflective optical imaging systems. Principally, weight is a crucial factor for any object sent into space because the mass of space-based optics increases non-linearly with diameter, with the realistic mass growth going at least as the square of the diameter when the basic optic and all necessary support and control hardware are included. The requirement of excessively rapid mass growth of the optics make the systems become increasingly difficult to fabricate and increasingly costly to develop, launch, and deploy [2]. Limited by the ability of current rocket transportation, the relative weight of large optics over a given area must be reduced. An alternative approach in which the lightweight optical membrane is used as a diffractive transmission element has been previously proposed, offering a significant relaxation in the control requirements on the membrane surface figure [3]. Compared to traditional large aperture reflective elements, the mass of the active region becomes negligible and the overall system mass growth is slower than the square of the diameter due to the lightweight membrane materials.

Although diffractive membrane optics for large aperture telescopes has exhibited a promising perspective, fabrication of a single precision diffractive membrane optic, especially a large aperture element, still is a challenging task. First of all, the novel membrane materials which only have a tiny distortion relative to the desired geometrical form, typically no more than 0.04λ (rms, the root mean square deviation of the collected wave-front) [4], were hardly obtained directly from the manufacturers. Furthermore, to meet the optical performance objectives, the membrane needs to be held dimensionally stable like a rigid substrate and remains lightweight, so that a designed frame, which holds the membrane around its entire boundary, has to be used to minimize the distortion of the etched pattern [5]. In this membrane to frame mounting process, the residual tension would be introduced into the final product, and the manufacturing loads would lead to a changed wave-front error, or even unacceptable optical performance degradation [5]. Therefore, how to obtain an optical membrane with the desired form accuracy became a chief problem to be solved.

Analogous to the traditional rigid mirror fabrication, the original substrate needs to be figure-corrected and finished via a series of ultra-precision surface manufacturing technologies. Generally, the ultra-precision optical fabrication chain consist of a first rough cut by rigid-tool grinding, followed by a re-roughing procedure to reduce machining marks (lapping) and final advanced polishing techniques, such as CNC polishing, Magneto-Rheological Finishing (MRF) and Ion Beam Figuring (IBF), which can achieve figuring and smoothing of the surface down to nanometer level. Unfortunately, even the state-of-the-art processes are hardly applied to the figure correction and finishing of the novel lightweight, flexible membrane material, due to the conflict between their removal mechanism and material properties. First, contact processes, including grinding, polishing and MRF, which mainly depend on the interaction between stiff abrasives or fluid and rigid substrates to keep a certain extent of determinism in removal processes [6], show unpredictable behaviors of the tool and consequent slower convergence, i.e. more iterations and larger removal depth, while figuring and finishing the flexible membrane material via these contact processes would lead to a lower level deterministic control and then an unacceptable, excessive removal depth. An excess of removal depth, which are caused by the increased number of required figuring iterations, conflicts with the ultrathin characteristic of the lightweight material, which means that a convergent figure correction is hardly achieved in the removal depth of several tens of micrometers, i.e. the thickness of the optical membrane, and optical segmented membrane mirrors, which are necessary for ultra-large aperture membrane telescope, for instance, a 20 m aperture system, become impossible in the fine co-phasing stage due to the requirement of uniform thickness for each segmented mirror. Second, IBF is a very slow figuring method due to its poor material removal rates and point-to-point mode [7]. In the time-consuming process, the effort of increasing the material removal rate using higher energy sources would induce more heat onto the substrate. Furthermore, the polymeric membrane has much higher sensitivity of the optical and mechanical performance to heat than traditional inorganic materials due to the difference between their composition, structures and properties. In our early experiments, the optical performance of the membrane, for instance, transparency, suffered the unacceptable degradation, even as low energies as possible in an acceptable figuring time for large aperture membrane optical elements. Therefore, new figuring technologies need to be investigated and developed for the figure correction and finishing of the novel flexible membrane optical elements.

2. Process flow

Here, we proposed a fully deterministic, rapid figuring procedure with high position accuracy, high material removal rates and nanometer level repeatability for large aperture membrane optical elements, which based on an iterative procedure by means of Reactive Ion Etching (RIE) where different distributions of masking layers allow the membrane material with a controlled depth in various selected areas to be removed. RIE, as a widely used pattern transfer process in semiconductor industries, has exhibited many advantages such as high selectivity, nanometer level etching accuracy and high anisotropy in material removal [8]. Accurate, stable etching rates, i.e. removal rates, can be achieved by optimized reactor configurations and recipes in RIE. The accurate positioning of removal area depends on the alignment of positioning marks in a similar way to photolithography, which can achieve micrometer level alignment accuracy and reduce the iterations remarkably.

The schematic of the RIE figuring procedure was shown in Fig. 1. A membrane was mounted onto a frame, as shown in Fig. 1(a). The transmitted wave-front of the lightweight membrane optical substrate was measured by an interferometer, after a series of customized position marks was added onto the backside of the membrane, as shown in Fig. 1(b). The position, shape and size of the position marks were included in the obtained transmitted wave-front map, which could be integrated into the design of the photomask and finally was patterned onto a chrome binary mask shown in Fig. 1(d). Generally, the transmitted wave-front error can reveal the variation of the membrane thickness, supposing that the values of the membrane refractive index over the full aperture have a negligible fluctuation. Therefore, the map of the full aperture could be divided into two kinds of regions: thinner regions and thicker regions, i.e., protected regions and removed regions, if a boundary corresponding to a chosen thickness was defined, and then the distribution of protected regions and removed regions was patterned onto the chrome binary mask together with the alignment marks. Subsequently, a layer of photoresist was uniformly coated onto the surface of the membrane substrate, as shown in Fig. 1(c), and an alignment procedure and subsequent UV exposure similarly to conventional photolithography were achieved. After developing, the distribution of protected regions and removed regions was patterned into the resist layer, as shown in Fig. 1(e). The resist in removed regions was removed and the substrate in removed regions became uncovered, while the resist in protected regions was remained as a masking layer. In the next step, as shown in Fig. 1(f), the membrane was introduced into a RIE reactor and a certain amount of membrane material in removed regions was removed in several tens of seconds. Similarly to a conventional RIE process, the sufficient thickness of the masking layer kept the material in protected regions out of reaction. Finally, the residual resist was stripped and the transmitted wave-front over the full aperture was measured again, shown in Fig. 1(g). The measured result determined whether the figure correction procedure needed to be finished or continue with a new loop from Fig. 1(a) until the process convergence was achieved.

 

Fig. 1. Schematic of one loop in RIE figuring. (a) A membrane was mounted onto a frame; (b) The transmitted wave-front of the membrane substrate was measured; (c) A layer of photoresist was uniformly coated onto the surface of the membrane substrate; (d) the distribution of protected regions and removed regions was patterned onto a chrome mask; (e) UV exposure and developing were achieved; (f) The substrate was figure-corrected in a RIE reactor; (g) The transmitted wave-front of the membrane substrate was measured again.

Download Full Size | PDF

3. Results and discussion

Figure 2 shows the transmitted wave-front of a 400 mm aperture membrane substrate before and after RIE figuring. Starting from an initial figure error of 105 nm rms, shown in Fig. 2(a), convergence is attained after 15 iterations. Final deviation from the desired form is ∼17 nm rms, as shown in Fig. 2(b). The effective figuring time per iteration was 30 seconds and total effective figuring time was less than 7.5 minutes. More details are reported in Table 1.

 

Fig. 2. The transmitted wave-front of a 400 mm aperture membrane substrate (a) before and (b) after 15 RIE figuring iterations.

Download Full Size | PDF

Table 1. Summary table of RIE figuring results

View Table

Reduced surface roughness or low roughness deterioration is one of important characteristics for an ideal figure correction of optical surfaces. Investigations of roughness for the 400 mm aperture figure-corrected membrane substrate were carried out and no significant degradation was evidenced. The surface roughness assessments were carried out by means of a 3D optical surface profiler (NewView 7300 by Zygo) before and after RIE figuring. The measurement field of view was 0.28 mm×0.28 mm with a 50X magnification lens. Figure 3(a) and Fig. 3(b) show illustrative snapshots of the figure-corrected membrane substrate before and after RIE figuring. A schematic of roughness investigation pattern was shown in Fig. 3(c). This comprised 37 snapshots spaced 50 mm from one another and arranged to cover the figuring area. Characteristic pre- and post-figuring roughness for the 400 mm aperture membrane ranged from ∼0.8 to 1.7 nm RMS, as Fig. 3(d) shown. Although only negligible fluctuation on the value of roughness was ever detected and no actual correlation of roughness to the depth of removed material appears evident from experimental results, novel smoothing or planarization techniques for flexible membrane were expected to be developed further for ultra-smooth surfaces.

 

Fig. 3. Results of the roughness investigation for the 400 mm aperture membrane substrate before and after RIE figuring. Typical surface topographies of the 400 mm aperture membrane in a 0.28 mm×0.28 mm area (a) before and (b) after RIE figuring; (c) Schematic of roughness investigation pattern; (d) Scatter diagram of roughness values for the 400 mm aperture membrane before and after RIE figuring.

Download Full Size | PDF

For current figuring technologies, the accuracy of the information about the position of the workpiece with respect to the tool, but also to the interferometer field of view can significantly influence the result of surface machining processes, where some coordinate measuring machine arms were provided to help registering the workpiece. However, RIE figuring is an exception to this principle, because the masking layer covered the surface of the workpiece tightly with overmatched alignment accuracy and thus no other alignment steps and no extra mechanical devices were needed. Generally, the alignment accuracy of photolithography is ∼1 µm while the measurement resolution is usually about several hundreds of micrometers. The transformation from the transmitted wave-front image to the mask design depended completely on computer image processing and the photomask fabricated by laser writing had usually a position accuracy of several hundreds of nanometers. There was no need of alignment between the workpiece and the reactor because of the parallel removal manner in RIE figuring. In a word, in RIE figuring, with respect to the figuring tool (the RIE reactor), the workpiece had only a requirement with position accuracy of several tens of millimeters (manual alignment); with respect to the interferometer field of view, the alignment accuracy between the workpiece and the field of view was ∼1 µm, which was much less than the measurement resolution, and the measurement became the bottleneck stage preventing a more rapid production of high form accuracy optics.

The influence of heat effect to position accuracy, especially for a large component, has been paid particular attention in current figuring technologies, when processing durations in the order of several hours are involved. In these precision machining systems, overheating occurs due to frictions and energy dissipation from engines. Additional misalignments could be introduced which progressively caused by thermal fluctuations of the inner chamber environment and then expansion of the supporting structure. If the misalignement resulted into lateral shifts of the component, additional residual errors would arise, thus slowing down the rate of convergence. However, RIE figuring has little heat effect in single iteration due to its short effecting etching time, mild chemical reactive mechanism and accessory water cooling system. Therefore, no thermally compensated device or machine coordinate system was required in RIE figuring.

Besides photolithography, RIE figuring is compatible with various positioning methods, masking-layer-forming processes and masking layer materials, including free-standing attached membrane, mechanical auxiliary positioning and even manual tailoring, which would lead various evaluation of costs, efficiency and position accuracy for RIE figuring. In our early experiments, there was no significant difference among them in terms of RIE figuring results, except for the possible increased iterations. Therefore, RIE figuring can meet various potential requirements with different cost and efficiency goals by choosing appropriate positioning method, masking-layer-forming process and masking layer material. Notably, photolithography provides the highest position accuracy among all candidates but dissatisfactory cost control and process simplicity for real practice, although the superiority of parallel removal mechanism on efficiency relative to the conventional point-to-point mode will become increasingly apparent with the increase of mirror aperture up to meter class. Several attempts to lower the cost and improve the efficiency have been proposed and investigated, including cheaper photomask materials and fabrication, mask fabrication process independent of the entire figure correction and maskless lithography as the masking-layer-forming technique. Film photomask, a kind of flexible film for screen process printing or Printed circuit board (PCB) fabrication, also can be used as a masking-layer-forming method in RIE figuring, which would lower the cost of photomask fabrication down to less than two hundredths of that of quartz photomask fabrication with a slight deterioration of resolution and position accuracy, due to its flexibility and manufacturing process. Generally, the film photomask has a resolution of about 5∼30 µm, which is still far less than the measurement resolution. Although the flexible photomask would easily cause the loss of position accuracy and need a designed auxiliary setup to improve alignment accuracy, there was no remarkable influence to be found in our relevant experiments. Furthermore, all 15 distributions of protected regions and removed regions, i.e., the patterns to be fabricated onto photomasks, could be obtained from the first transmitted wave-front just at the start of the entire RIE figuring process, due to the nanometer level removal accuracy and sufficient alignment accuracy, if the process parameters are stable. Obviously, the initial transmitted wave-front could determine the value of removal depth on different areas and then each distribution of masking layers in loops, if the slight figuring error can be ignored, which lead the photomask fabrication process become independent of the entire RIE figuring process. That’s one of the reasons that the photomask fabricating time had not been listed in Table. 1. If this situation is considered, “iterations” maybe become inaccurate and “loops” or “stages” would be more appropriate, although 1-2 extra iterations would be possibly required. When the measurement time becomes negligible due to the dramatically decreased iterations, the total process time would be reduced down to ∼12 hours level according to the data in Table. 1, which can almost meet the 10 hours target in meter-class optical fabrication. Due to the similar figuring time for mirrors with any size in RIE figuring, which arises from the parallel removal mechanism, a less than 10 hours duration for meter-class optical fabrication could be expected after a further optimization of process parameters and facilities. Additionally, maskless lithography has been considered as the masking-layer-forming technique, which can directly generate the pattern onto the photoresist layer without any photomask, provide a resolution about 10∼50 µm and potential higher alignment accuracy than conventional photolithography. Photomask fabrication would become unnecessary when RIE figuring combines with maskless lithography. That’s another reason that the photomask fabricating time had not been listed in Table. 1.

As an optical membrane substrate for large aperture diffractive membrane optics, a geometrical form accuracy of no more than 0.03λ rms, similar thicknesses of around 25 µm among segmented mirrors and sufficient transmittance at operating wavelengths are known requirements. After RIE figuring, the form accuracy of the membrane has met the requirement (Fig. 2). The largest removal depth in the whole sequence was ∼450 nm, which was ∼2% membrane thickness. The thickness of the membrane with negligible change means that the mechanical performance of the membrane has no significant change after the RIE figure correction and the thicknesses between different segmented mirrors can easily be kept similar (∼100 nm deviation) combining with slight thinning. The transmittance of the membrane has no remarkable deterioration and still exceeds 80% after the whole sequence, not as apparently as in earlier IBF figuring where the deterioration of the transmittance could be observed by eyes. The influence of larger removal depth in RIE figuring to the optical membrane need be investigated further and new requirements for large aperture diffractive membrane optics would be possibly proposed with the deepening of the project process.

4. Experimental details

As Fig. 1 shown, a homemade Polyimide membrane, which has a thickness of about 25 µm, Young’s modulus of above 7 GPa and a transmittance of more than 80% at 632.8 nm, was firstly fixed onto a 400-mm aperture metal frame by glues and mechanical clamp devices. The transmitted wave-front of the lightweight membrane was measured by an interferometer (Zygo GPI600). Subsequently, a layer of AZ3100 photoresist was spun-coated onto the membrane fixed on the frame in a customized, 550-mm aperture spin-coater (Lebo Science). The spin speed was 1000 rpm. Exposure performed on a customized mask aligner with a 600 mm × 600 mm aperture (Institute of Optics and Electronics, CAS) and a designed auxiliary setup was used to improve exposure quality by keeping the out-plane deformation of the membrane in an acceptable range and provided extra stability, which made the membrane on the frame performed like a rigid substrate in exposure. The thickness of the photoresist layer was about 1 µm, exposure dose was 4.5 mJ/cm⁻³ and exposure time was around 30 s. After about 40 s development, the membrane was dried by nitrogen. RIE figuring performed in a customized, 650-mm aperture capacitively coupled plasma (CCP) reactor (Beijing jinshengweina Technology Co., Ltd). The etching gases were oxygen and trifluoromethane in a ratio of 10:1. The flowing rate of the etching gases was set to 190 sccm, the RF power was 800 W and the chamber pressure was 1.0 Pa. Additionally, a designed auxiliary setup was used to improve the etching uniformity.

5. Conclusions

In conclusion, RIE figuring is a means of parallel removal different from the conventional point-to-point mode, which leads to less figuring time and huge potential in the rapid figure correction for meter-class optics. Based on the mature processes including RIE and photolithography, RIE figuring constitutes an ideal figuring alternative, combining the advantage of a non-contact tool with very high material removal rates, nanometer level repeatability, mild removal mechanism and extra high position accuracy. A 400 mm aperture membrane substrate has been figure-corrected from the initial transmitted wavefront error of 105 nm rms to the final figure error of ∼17 nm rms in total effective figuring time of less than 7.5 minutes. Theoretically, RIE figuring can be applied to the figure correction of flat surfaces and curved surfaces with a small rise of arch, and most of optical material can be removed easily in RIE plasma, including quartz and SiC. The RIE figuring of optical elements with various materials, even metal, will be investigated in future.