Abstract
The plasma polishing process is one of the non-conventional techniques used to remove material at the atomic level from the substrate. During the polishing of the fused silica substrate, the process parameters, namely radio-frequency (RF) power, pressure ratio (SF6/O2), and total pressure of the plasma chamber, are investigated and optimised for material removal rate (MRR) and % change in surface roughness (% ΔRa) using response surface methodology. The optimum values obtained for MRR and % ΔRa are 0.012 mm3/min and 3.59, at RF power of 60 W, pressure ratio of 3, and total pressure of 14.3 mbar. The experimental results reveal that surface roughness slightly increases from 0.344 to 0.356 μm after plasma processing at optimised process conditions. Moreover, the plasma-processed fused silica substrate is characterised using field emission scanning electron microscopy and energy dispersive X-ray spectroscopy, which depict the presence of silicon, oxygen, and fluorine on the processed substrate.
Keywords
Introduction
The demand for finishing complex and freeform optical products is motivated to create innovative manufacturing techniques. Industries constantly focus on developing time-saving and cost-effective processes, particularly for finishing optical products [1]. Nowadays, many modern optics industries require high surface quality and minimum surface defects but face challenges in producing them. Xie et al. [2] reported that optical substrates with exceptional physical properties are widely used in aerospace, aviation, electronics, and other fields. There are always some defects in the final surface in conventional processes, such as lattice disturbance, surface defects, dimensional inaccuracy, micro-cracks, etc. Zhang et al. [3] brought down the surface roughness (Ra) by more than 1 nm after plasma polishing and finally reached 0.63 nm, which proves the capability of atmospheric pressure plasma processing (APPP) to achieve high-quality ultra-smooth surfaces. The modern optics manufacturing industry demands glossy surfaces, which take the surface roughness to the nanometre level, and nominal surface or subsurface defects become the key features [4]. Subsurface defects are severe for brittle and hard materials, i.e. fused silica, zerodur, ceramics, and glass. Hsu et al. [5] reported that plasma polishing can enhance polymer surface properties such as permeability, wettability, conductivity, and adhesion. Shi et al. [6] reported improved process efficiency of silicon carbide (SiC) using ion-enhanced atmospheric pressure plasma machining (IAPPM). Sulfur hexafluoride (SF6) is used as a reactive gas, and the achieved material removal rate (MRR) is 3 μm/min. Li et al. [7] used the atmospheric pressure plasma processing (APPP) process. They reported that the difference in topography between the opaque and transparent areas is not affected by chemical composition. Furthermore, the root-mean-square surface roughness (Rq) of the processed surface was reduced to 2.2 nm from its initial value of 38 nm. Also, the comprehensive investigation of surface topography, analysis of opacification phenomenon, chemical composition, and quantitative roughness has been characterised at different areas on the plasma-processed surface. Su et al. [8] reported that etched pits became more significant and increased in quantity. Further removal causes the etched pits to coalesce, resulting in irregular concave–convex structures. Zhuang et al. [9] studied the polishing of planar optical components to avoid uneven MRR while enhancing the finishing efficiency. Zhang et al. [10] achieved 0.631 nm surface roughness (Ra) and 32 mm3/min removal rate on silicon wafers. They utilised a low-temperature plasma chemical process to remove material at the atomic scale while avoiding surface/subsurface defects. Dev et al. [11] revealed that after plasma processing, fused silica's power spectrum density characteristics show an 80% reduction in higher spatial wavelength, indicating an improvement in the surface figure. The low-pressure (10−2–0.5 mbar) plasma method can reduce the subsurface damage on freeform and complex surfaces. The different working gases are successfully used in atmospheric pressure plasma to reduce the subsurface damages while achieving an ultralow surface finish [12–14]. APPP processes are limited to small apertures and do not apply to freeform surfaces. Yao et al. [15] proposed that the material removal mechanism from zerodur is based on mechanical action and chemical reaction using APPP. The breadth of the material removal function profile is mainly affected by chemical machining. The chemical property of the plasma jet was examined by atomic emission spectroscopy. They also analysed the surface roughness variation of the plasma-processed surface with different process parameter conditions [15]. Yadav et al. [16] proposed the mechanism of plasma polishing for several kinds of sophisticated freeform and aspheric optical materials, which are mostly employed by the optical industries. Zhang et al. [17] studied the effect of process parameters (i.e. radio-frequency (RF) power and gas composition ratio) on surface roughness and removal rate using optical emission spectroscopy (OES). Gerhard et al. [13] observed that waviness and surface roughness is significantly reduced after plasma treatment. Su et al. [18] examined the surface quality of APPP and bonnet polishing (BP) processes. The etched pits and concave–convex structure were generated during APPP, and BP was used to remove the peaks and etched pits, improving smoothing efficiency. Zhang et al. [19] reported that atmospheric pressure plasma polishing is efficient for producing damage-free ultra-smooth surfaces because of its chemical nature. The APPP technique improves mechanical qualities, particularly residual stress, which, around 4.2 GPa, decreases after 60 s of machining. The topography is also more regular, reaching surface roughness (Ra) below 0.5 nm [19]. A regression model has been developed utilising the central composite rotatable design of response surface method. The purpose of the regression model is to establish a correlation between the input and output parameters [20]. Li et al. [21] reported the effect of reactive radicals and substrate temperature on surface finish using plasma-induced atom migration manufacturing (PAMM). It can generate a smoother surface on silicon dioxide (SiO2) substrate, decreasing surface roughness (Ra) from 86.57 to 0.15 nm. PAMM achieves a damage-free surface with high efficiency and low cost without removing the material from the substrate surface during surface smoothing. This technique can fabricate large optical components having excellent optical performance and long operational products [21]. Yadav et al. [22] revealed the distribution of electron density, temperature, and potential in the plasma chamber for two separate substrates of fused silica using multiphysics simulation.
A new non-contact plasma-assisted atomistic machining process is conceived to finish the complex surface of fused silica. This process improves surface finish while removing material atom by atom from damaged surface/subsurface layers without changing the surface topography. Dev et al. [11] reported a qualitative process for analysing micro-cracks on the surface of an optical substrate by laser illumination. They concluded that the surface finish and surface integrity could improve using a medium-pressure plasma process and achieved material removal rate up to 0.008 mm3/min [11]. A medium-pressure plasma process has been developed to overcome the earlier problems, which can remove the surface and subsurface damages. However, the removal rate of fused silica substrates and surface characteristics depend on the process parameters and their levels, which demand parametric investigation and finding optimum conditions.
The current research article aims to establish a non-contact finishing process to improve the surface quality and remove the strained layer of fused silica while achieving the required surface finish. It removes the material from the surface and subsurface layer without any redeposition of contamination during plasma processing. The novelty of the present method combines the advantage of low-pressure plasma etching by ions having isotropic material removal on the substrate surfaces and the benefit of an APPP, for instance, chemical etching rather than physical bombardment by species. This method attains a defect-free, very fine-polished surface on an optical substrate. The present article focuses on the effect of input parameters, for instance, RF power, pressure ratio (SF6/O2), and total pressure of the process chamber on output responses, i.e. MRR and % change in surface roughness (% ΔRa). The process parameters (processing time and gas composition (He:(SF6+ O2))) have been kept constant during all experiments. An experimental analysis is performed to determine significant parameters and their influence on MRR and % ΔRa using the response surface method (RSM). The RSM model is experimentally validated at the optimum parameter condition for % ΔRa and MRR. Furthermore, the microstructure and chemical compositions of the fused silica are revealed by FESEM and EDX analysis, respectively, before and after processing.
Materials and methods
Figure 1 presents the schematic representation of the developed plasma setup where RF power, vacuum pump, process chamber, and reactive and process gases are used to remove the surface defects and enhance the surface quality on the optical substrate. The process chamber during processing is presented in Figure 2. RF power is operated at 40.68 MHz with a 200 watts power amplifier having a matching network. He and O2 are used as process gases, and sulfur hexafluoride (SF6) is used as reactive gas. The different gases are admitted into the process chamber through a controlled mass flow metre which independently controls the flow of gases (standard cubic centimetres per minute, sccm) from different gas cylinders. A process chamber of zerodur material has been used for plasma generation. The chamber is sealed with a cap having optical transparency in the 300–1200 nm wavelength range. The outer part of the process chamber acts as a dielectric barrier for the electrodes.
Schematic representation of developed setup. Photographic view of the plasma chamber.
The principle of this process is controlled by the chemical interaction between the radicals, ions, species, and substrate surfaces while performing atomic-level material removal. The samples are prepared using grinding followed by the abrasive polishing method with Al2O3 as abrasive particles. The chemically driven plasma polishing generally improves the subsurface defects. During finishing, the process and reactive gases are admitted into the process chamber. The RF power is used to ionise the gases in the process chamber. The gases reach the excited state to produce high energy and high-density reactive fluorine radicals (F*), electrons, and ions, which finally generate ions, species, and reactive radicals, as shown in Figure 3(a,b). Then the produced reactive radicals interact with the substrate surface (i.e. Si atom) and generate silicon tetrafluoride (SiF4). The produced SiF4 is volatile and comes out of the plasma chamber without contaminating the substrate surface. Here MRR is obtained using a chemical reaction according to Eqs. (1) and (2) [23].
(a) plasma species inside process chamber and (b) principle of generation of reactive species during plasma polishing.
Figure 4 presents the flow chart of the plasma process. Initially, the gases come out from the gas cylinders through pipes. Then mass flow metres are used to control the flow of each gas separately. Further, the gases are mixed as per optimised parameters and allowed to enter the process chamber. RF power is used to ionise the gases, and the generated species react with the substrate's surface atoms and produce the volatile product.
Plasma process flow chart.
Process parameters with range and fixed parameters.
, 
= observed data points, 
= mean, 
= number of observations) analysis is performed, as presented in Table 2.
Experimental values of responses with experimental runs as per DOE.
Results and discussion
Model fit summary for MRR.
*Sq – Sequential, *Adj – Adjusted, *Pred – Predicted, *LoF – Lack of fit.
Model fit summary for % ΔRa.
*Sq – Sequential, *Adj – Adjusted, *Pred – Predicted, *LoF – Lack of fit.
ANOVA for MRR.
*SS – Sum of squares, *DF – Degree of freedom, R2 – Coefficient of determination = 0.9812, *MS – Mean square, *Significant (p-value < 0.05), ** Not significant (p-value > 0.05).
ANOVA for % ΔRa.
*SS – Sum of squares, *DF – Degree of freedom, *MS – Mean square, R2 – Coefficient of determination = 0.9791, *Significant (p-value < 0.05), ** Not significant (p-value > 0.05).
Tables 5 and 6 provide the developed statistical model, including the significant parameters and model adequacy. ANOVA measures the % contribution of every independent variable. It represents the relative effect that a process parameter has on the response. The highest contribution towards the MRR (Table 5) can be seen as RF power (A), followed by pressure ratio (SF6/O2). The contributions of these two parameters are 39.8% and 30.1%, respectively. Similarly, the highest contribution towards the % ΔRa (Table 6) can be seen as the interaction of RF power and pressure ratio (AB), followed by total pressure (C). The contributions of these two parameters are 48.5% and 17.9%, respectively.
Furthermore, the percentage of significance for individual parameters and their combined effects are calculated from the F-test values for both responses, illustrated in Figure 5(a,b), respectively. The percentage contribution of parameters is evaluated using the F/ΣF values for each factor [25]. The parameters such as RF power (A) and combined effect of RF power and pressure ratio (AB) have the highest contribution to MRR and % ΔRa, respectively. Regression analysis is carried out to establish a relationship between process parameters and the responses. ANOVA results from Tables 5 and 6 revealed that the empirical models are significant. The regression equations for MRR and % ΔRa have been derived using the quadratic approximation as per the model adequacy test and are presented in Eqs. (5) and (6), respectively.
Percentage contribution of process parameters on (a) MRR and (b) % ΔRa. Comparison between experimental and predicted values of (a) MRR and (b) % ΔRa.
It is required to analyse the effect of each parameter simultaneously on outcomes and obtain the better and most possible solution to achieve the desired output response. The plasma intensity increases with RF power, which can be thought of as an increase in the electron density of the discharge. Changes in plasma discharge mode affect plasma temperature, resulting in visible changes in plasma dynamics [26]. Yao et al. [15] also reported that the depth and width are increased using an atmospheric pressure plasma jet (APPJ) machining with the increased RF power. When the power exceeds 360 W, abrupt changes in depth are obtained, caused by increased plasma temperature. Thus, RF power has a significant impact on the removal function.
Initially, the samples are prepared using grinding followed by the abrasive polishing method with Al2O3 as abrasive particles. In conventional polishing, material removal is realised through mechanical and chemical effects. The cracks are closed after the conventional mechanical machining. The cracks get filled up / closed due to the plowing action by abrasives during conventional polishing, and the surface becomes smooth. This pre-polished surface is further plasma processed to achieve a highly polished surface uniform over the entire workpiece surface.
During plasma processing, the reactive fluorine radicals react with the substrate surface. The fluorine radicals react with the sidewalls of the micro-cracks. Thus, the micro-cracks start opening up, forming etched pits. After etching away a thin layer by the plasma, cracks get opened up, and the isolated etched pits start creating. Therefore, the surface finish is degraded as these etched pits appear over the entire surface of the workpiece. These pits make the surface dramatically rougher.
The maximum depth of these cracks on the substrate surface is around 30 μm. The current plasma process can remove the material maximum of 5–10 μm depth from the surface; hence, it can't remove all the surface cracks on the substrate. Xin et al. [27] also observed that the roughness evolves with the etching depth, and the roughness evolution is a single-peaked curve. This curve results from opening and coalescing surface cracks and fractures on a polished fused silica surface. With further material removal from the workpiece surface, the surface cracks got polished, and the surface became smoother with reduced surface roughness. In atmospheric pressure plasma, the polishing is conducted on a single spot, unlike the uniform polishing requirement over the entire surface in the medium-pressure plasma process (i.e. in the present study). Hence, a material removal depth of 300 µm by atmospheric pressure plasma is easily achieved, which is very difficult in medium-pressure plasma. Therefore, the material removal rate has been considered the main objective of the current plasma process.
Figure 7(a,d) show the variation of MRR and % ΔRa with the increased power. The % ΔRa increases with RF power, which aligns with prior research [17] as discussed above. The increasing trend of MRR has been found with the variation of pressure ratio between 2 and 3, as shown in Figure 7(c). This could be because more atoms are produced with the increase in SF6 gas, which leads to more materials being removed from the surface. The occurrence of oxygen will endorse the separation of SF6 into SF5 and F*. Moreover, adding the process oxygen with SF6 gas can help increment the reactive radicals’ intensity and further higher MRR. Figure 7(f) shows the variation of % ΔRa where its value decreases with increased SF6/O2 pressure ratio [28]. It occurs due to chemical machining, which increases the speed of chemical reactions at a higher pressure ratio. As the pressure ratio increases, the presence of O will promote the dissociation of SF6 into SF5 and F*, and O will associate with SF5 and prevent the recombination of F* and SF5. So it can be seen that with the pressure ratio increase, the roughness variation decreases. Because of the increase of the SF6 and O2 flow rate, the generated O* react with the fused silica surface to form SiO = SiO aggregates, which will reduce the etching rate of the silica. So roughness variation decreases with the increase of the pressure ratio. Figure 7(b,e) show that MRR is slightly raised and % ΔRa is decreased with the total pressure. The plasma intensity increases in bulk can be expected with increasing the total pressure, at constant power. It may be due to the electron's increased mean free path as there is more chance to gain enough energy between collisions to perform ionisation and excitation.
MRR and % ΔRa variation with power (a, d), total pressure of the chamber (b, e), and pressure ratio of SF6/O2 (c, f), respectively.
Statistical analysis (shown in Tables 5 and 6) reveals that the input variables have a cumulative effect on the output responses. 3D response surface plots are required to understand combined parametric effects on output responses. The interaction of the pressure ratio and power, the total pressure of chamber and power, total pressure and pressure ratio, and their consequences over the MRR and % ΔRa are illustrated in Figure 8(a–c) and (d–f), respectively. From Figure 8(a,b), it is realised that with increased pressure ratio and power, MRR increases; MRR also increases with decreased total pressure and increased power. MRR increases with a lower value of total pressure and a higher value of pressure ratio, as illustrated in Figure 8(c). The ions/radicals density increases with the increased power, which leads to higher MRR. As stated earlier, a higher concentration of ions/radicals causes more etching. Whenever the power increases, the particle collisions are strong enough to etch a greater amount of material. Even at reduced power, effective collisions become lower due to lesser particle energy leading to smaller MRR. The variation of % ΔRa with power, pressure ratio, and the total pressure is emphasised using 3D plots, as shown in Figure 8(d–f). At higher power and lower pressure ratio, % ΔRa becomes maximum (Figure 8(d)). The combined power and total pressure effect on % ΔRa are insignificant, as shown in Figure 8(e). At lower total pressure and pressure ratio, the % ΔRa is maximum (Figure 8(f)) due to the higher energy of radicals and ions generated inside the plasma chamber.
Combined effect of power and pressure ratio (a, d), power and total pressure (b, e), and pressure ratio and total pressure (c, f) on MRR and % ΔRa, respectively.
A scale-free value termed desirability is created from an estimated response using the desirability function technique [29,30]. The goals of optimisation can be utilised to either maximise, minimise, or obtain the desired response value [31]. Different desirability functions might be used depending on the purpose. Derringer and Suich [29] proposed a particular function for the transformation of the responses (Xi) to the desirability di(Xi). As a result, two transformations are suggested in Eq. (7).
The one-sided transformation is utilised to minimise or maximise Xi; the two-sided transformation (i.e. Eq. (7)) is used to acquire the objective value ti for Xi, where ui and li are the upper and lower limits of the responses. The superscripts s and t in Eq. (7) correspond to the weighted factors. s and t are the parameters that govern the shape of di(Xi). For s= t= 1, the desirability function increases linearly to ti. The overall desirability function Y in Eq. (8) is defined as the geometric mean of the specific desirability functions of individual responses, i.e. di(Xi), where n is the number of responses. The optimal solutions are obtained by maximising Y.
Multi-response optimisation using the RSM technique is used in the current study to find the optimal values of the process parameters. Here, desirability is used as a measure to optimise the responses (MRR and % ΔRa) in the medium-pressure plasma polishing process. During optimisation, the main objective was to maximise MRR while minimising % ΔRa. The input variables are kept ‘in range’ during optimisation, as presented in Table 7. Results and plots obtained using the optimisation study are displayed in Figure 9. According to Figure 9, it can be observed that at 60 W power, pressure ratio of 3, and total pressure of 14.3 mbar, the optimised MRR and % ΔRa values are 0.012 mm3/min and 3.59, respectively. The maximum desirability value is 0.7490 at optimal parametric conditions.
Optimisation plots for MRR and % ΔRa. Range of process parameters for MRR and % ΔRa.
Experimental validation of RSM model at optimised parameter conditions.
The surface roughness profiles of the substrate were measured using a 3D optical profiler before and after plasma processing at a constant processing time of 50 min. Figure 10(a–d) represents the surface roughness before and after post-processing using medium-pressure plasma polishing under optimum condition, i.e. power of 60 W, pressure ratio (SF6/O2) of 3, and total pressure of 14.3 mbar. The processed substrate is slightly degraded from its initial surface roughness (Ra) value of 0.344 µm to the final Ra value of 0.356 µm without any etched pits. During plasma polishing, although surface roughness increases, subsurface defects diminish. The rate of surface roughness increases with increased MRR. Hence, the plasma polishing process aims to achieve the optimum material removal rate.
1D surface roughness profiles (a) before and (c) after plasma processing; 2D surface roughness profiles (b) before and (d) after plasma processing.
FESEM plays an essential role in visualising the shape of the microstructure on the fused silica. The investigation has been carried out on the optimum process parameters, i.e. power of 60 W, pressure ratio of (SF6/O2), and total pressure of 14.3 mbar for the microstructure analysis. The microstructure of fused silica is coarse before processing. It has been changed to finer after plasma processing at optimum conditions, as illustrated in Figure 11(b). EDXS results show that silicon (Si) and oxygen (O) elements appear on the surface before processing, as presented in Figure 12(a). The element fluorine (F) appears on the processed surface, as seen in Figure 12(b). The presence of O2 gas may cause the breaking of SF6 into SF5 and F radicals, showing evidence of chemical reactions occurring during plasma processing.
FESEM images of fused silica (a) before and (b) after plasma processing. EDX analysis of fused silica (a) before and (b) after plasma processing.
Conclusions
The present study proposes a medium-pressure plasma processing as a non-contact type plasma-assisted atom-by-atom material removal technique for an optical material, i.e. fused silica. The initial studies are focused on the variation of MRR and % ΔRa with process parameters. Regression and statistical analysis for MRR and % ΔRa with different power, pressure ratio, and total pressure are performed to find a suitable relationship between input and output parameters. The following are the critical findings of the present research study.
The plasma process can be used to polish the optical components. It is an effective process to remove the material from the substrate surface. MRR and % ΔRa are significantly affected by power, pressure ratio, and total pressure of the process chamber. A low-pressure ratio and low power with constant total pressure during the reaction resulted in lower MRR and % ΔRa. The maximum value of MRR achieved throughout all experiments is 0.013 mm3/min (at power = 50 W, pressure ratio = 4, and total pressure = 12.5 mbar). The maximum achieved value of % ΔRa is 21.36 (at power = 60 W, pressure ratio = 2, and total pressure = 10 mbar). The RF power has the highest contribution (39.7%), followed by the pressure ratio (30.18%) for MRR. In the case of % ΔRa, the combined effect of power and total pressure of the plasma chamber has the highest contribution (47.5%). The desirability approach has been employed to optimise the parametric conditions. The values achieved after the optimisation study of MRR and % ΔRa are 0.012 mm3/min and 3.59 for power, pressure ratio, and total pressure of 60W, 3, and 14.3 mbar, respectively. The obtained desirability is 74.90%, which is acceptable. After plasma processing at the optimum parameter, the measured surface roughness value is 0.356 µm, slightly higher than its initial Ra value of 0.344 µm. Surface contamination and etched pits have not been observed after the plasma processing. The elements silicon (Si), oxygen (O), and fluorine (F) appear on the plasma-processed surface, showing the reaction occurring during plasma processing.
Footnotes
Disclosure statement
No potential conflict of interest was reported by the author(s).