Improvement of Laser Damage Resistance of Fused Silica Using Oxygen-Aided Reactive Ion Etching

Abstract

Reactive ion etching (RIE) with fluorocarbon plasma is a facile method to tracelessly remove the subsurface damage layer of fused silica but has the drawback of unsatisfactory improvement in laser damage resistance due to the induction of secondary defects. This work proposes to incorporate O₂ into the CHF₃/Ar feedstock of RIE to suppress the formation of secondary defects during the etching process. Experimental results confirm that both the chemical structural defects, such as oxygen-deficient center (ODC) and non-bridging oxygen hole center (NBOHC) defects, and the impurity element defects, such as fluorine, are significantly reduced with this method. Laser-induced damage resistance is consequently greatly improved, with the 0% probability damage threshold increasing by 121% compared to the originally polished sample and by 41% compared to the sample treated with conventional RIE.

1. Introduction

Fused silica is a frequently used optical material in high-power laser systems due to its excellent performance in transmitting light and resisting laser-induced damage. The damage threshold of fused silica to an UV ns-scale laser reaches an impressive magnitude of 100 J/cm², theoretically. However, damage always happens at the rear surface at a fluence of lower than 20 J/cm² in practice, which seriously limits the power advancement of the laser system [1,2]. This has been proved to be caused by the surface and subsurface defects (SSDs) induced by mechanical grinding and polishing. As such, getting rid of these damage precursors through post-treatments is of great importance to improve the laser-induced damage resistance of the fused silica optics [3,4].

In the past decades, several techniques, including HF-based wet chemical etching (e.g., advanced mitigation process, AMP, or dynamic chemical etching, DCE, with the aid of ultrasonic oscillation), plasma-based dry etching (e.g., reactive ion etching, RIE, based on fluorine-containing plasma), and their combinations have been developed to remove SSDs of fused silica optics [5,6,7]. Wet chemical etching can effectively eliminate contaminations, impurity elements, and electronic structure defects presented in the SSD layer of fused silica optics and greatly improve their laser-induced damage threshold (LIDT). However, this is accomplished with the sacrifice of a deterioration in their surface morphologies due to the exposed and enlarged scratches brought about by the isotropic etching. Further, the usage of highly corrosive and toxic hydrofluoric acid means the implementation of this technique has high risks in safety.

As an anisotropic etching technique, plasma-based dry etching, e.g., RIE, is a potentially more advantageous approach, with the benefits of simplicity and safety in its implementation. It can tracelessly remove the SSD layer, including scratches buried in it, producing quite a smooth surface without deterioration in surface quality [6,8]. The drawback limiting its application as an independent technique is the unsatisfactory improving effect on damage-resistant performance, which is caused by the secondary defects induced in the etching process. Our group has proposed to conduct a shallow DCE as a following step to address this issue [7]. In this combined technique, DCE is proved to be quite effective in eliminating the RIE-induced secondary defects, with the damage-resistant performance being greatly improved. However, the issue of high risks in safety brought about by DCE is introduced again. Furthermore, the combination of dry etching and wet chemical etching makes the implementation quite long and complicated. As such, developing a more concise and efficient method that reduces the secondary defects of RIE to ensure a more significant improvement in damage-resistant performance is still necessary.

From previous research, the secondary defects induced by RIE treatments primarily encompass chemical structural defects, such as oxygen-deficient centers (ODCs) and non-bridging oxygen hole centers (NBOHCs), and trace impurity elements, like fluorine [9]. Extensive mechanistic studies on the etching of SiO₂ with fluorocarbon plasmas also indicate that the etching process is accompanied by the disruption of Si-O bonds and the formation of fluorocarbon polymers [10,11,12]. From the studies concerning point defects in fused silica, it is apparent that oxygen plays a crucial role in the formation and transformation of chemical structural defects. Almost all types of chemical structural defects are associated with the loss and liberation of oxygen atoms [13,14,15,16,17]. With this consideration, this study proposes to incorporate a certain amount of O₂ in the feedstock of RIE to generate a plasma rich in reactive oxygen species. This may aid in suppressing the generation of secondary defects, thereby enhancing the improvement of the damage-resistant performance of fused silica.

2. Methods and Experiments

The procedures and the underlying mechanisms of the two RIE processes are schematically illustrated in Figure 1. The experiment employed three fused silica samples (Corning 7980, NY, USA) with dimensions of 50 mm × 50 mm × 5 mm which were mechanically grinded and polished by the same vendor. They were all ultrasonically cleaned as received with an aqueous solution of Micro90 to remove potential surface contaminants like dust, debris, and oily residues. After air drying, one sample (Sample #0) was left untreated to serve as an original control; one (Sample #1) was subjected to conventional RIE of 1 μm with a mixture of CHF₃ and Ar as feedstock; and one (Sample #2) was similarly etched but instead used a modified RIE technique with a mixture of CHF₃, Ar, and O₂ as feedstock. The two RIE processes were conducted within the same parallel-plate plasma discharge apparatus. Except for the feedstock (containing O₂ or not), all other plasma and etching parameters of them were exactly the same, where the input power was 200 W, the bias voltage was 625 V, the working pressure was 20 mTorr, the base pressure was 1 × 10⁻⁵ mTorr, the gas flow rates of CHF₃ and Ar were 72 SCCM and 5 SCCM, respectively, and the etching depth was 1 mm. Additional O₂ with a gas flow rate of 5 SCCM was incorporated in the feedstock of the modified RIE. Prior to the conduction of the etching, the etching rates were calibrated to determine the etching durations. For a 1 μm etching depth, the durations for Samples #1 and #2 were 28 min and 30 min, respectively. After etching, the samples were ultrasonically cleaned and air dried once again, and then subjected to damage performance testing and other characterizations. All cleaning, etching, and drying processes were conducted in Class 100 clean environment.

For the assessment of the damage-resistant performance of the samples, 1-on-1 laser-induced damage tests [18,19] were conducted using an Nd:YAG tripling pulse laser with a wavelength of 355 nm, pulse width of 5 ns, and repetition rate of 1 Hz. In the tests, the laser beam waist was positioned on the rear surface (i.e., the treated surface) to achieve a high laser fluence. The energy and spatial profile of the beam were monitored with an energy meter and a beam profile, respectively, by picking off a fraction of the beam. The beam exhibited a near-Gaussian distribution with an efficient diameter of approximately 1.2 mm and a quality factor (ratio of the peak energy to the average energy of the beam spot) of 2.94. A long-focal-distance microscopic CCD camara was used to observe the damage performance of the samples in situ. For the 1-on-1 laser-induced damage test, 20 testing sites with the distance between two adjacent sites of 3 mm on the sample were irradiated one by one by the laser shoot with a certain fluence. Damage initiation probability at this irradiation fluence was obtained by calculating the number of damaged sites.

Cathodoluminescence (CL) was employed to characterize the distributions of chemical structural defects in the near-surface regions of these samples. Upon electron beam irradiation, emissions of luminescence from defect structures in the samples will be induced. The spectral peak positions of the luminescence reveal the nature of the defects, while the intensity reflects the defect concentration. By modulating the energy of the electron beam, the depth distribution of the defects can be obtained. More detailed information about CL can be referred to in reference [20]. In the present research, the detecting depth was varied by adjusting the electron beam voltage. For different samples, the measurement parameters were kept consistent to ensure comparability of the results.

To analyze the near-surface distributions of Si, O, and potentially introduced F in the samples, Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) was employed. The primary ion beam was Bi¹⁺ with energy of 30 keV and incidence angle of 45°, covering an area of 200 µm × 200 µm. The sputtering ion beam was Cs⁺ with energy of 2 keV, yielding a sputtering rate of approximately 0.204 nm/s.

3. Results and Discussion

Figure 2 illustrates the damage probability versus laser irradiation fluence for each sample. It is evident from this figure that samples treated with either of the two RIE processes exhibit superior resistance to laser damage than the virgin one. Compared to the latter, the 0% probability damage threshold for that treated with conventional RIE increases by about 57% (from 8.9 to 14.0 J/cm²), while the 100% probability damage threshold increases by about 113% (from 11.5 to 24.5 J/cm²). Notably, in comparison, the modified RIE with oxygen added in the source gas results in a more pronounced enhancement in laser damage resistance. The 0% probability damage threshold rises to approximately 20 J/cm² (from 8.9 to 19.7 J/cm²), denoting a remarkable 121% increment relative to the virgin sample and a 41% increment (from 14.0 to 19.7 J/cm²) compared to the conventional RIE-treated one.

Figure 3 shows the CL spectra of the samples under irradiation of electron beam with varying detecting depth. The depth value was estimated from the beam voltage with the database provided by the CL instrument manufacturer. Evidently, all three samples show fluorescence peaks at 2.7 eV and 1.9 eV, characteristic peaks of fused silica, corresponding to ODC and NBOHC defects, respectively [20]. The fluorescence intensity of the virgin sample remains relatively invariant with the detecting depth increasing from 100 nm to 2 μm, which implies that the ODC and NBOHC defects in the originally polished sample locate up to 100 nm beneath the surface, predominantly within the polishing redeposition (Beilby) layer. Contrarily, samples treated with either RIE methodology exhibit an increased fluorescence intensity with increasing detecting depth. The increment corresponding to the depth from 1 µm to 2 µm is less pronounced than that corresponding to the depth from 500 nm to 1 µm, suggesting that the ODC and NBOHC defects in these samples primarily locate within 2 µm beneath the surface. These results verify that RIE induces chemical structural defects, including ODCs and NBOHCs, over a more extended depth, while it removes the SSDs of the originally polished sample.

With an identical detecting depth (1 µm), the CL spectra of the three samples are compared, as shown in Figure 4. It is obvious that the concentrations of ODCs and NBOHCs in the conventional RIE-treated sample increase dramatically compared to that in the virgin sample, further demonstrating the substantial induction of chemical structural defects near the surface in conventional RIE. In comparison, the concentrations of defects induced by the modified RIE process remarkably decrease, suggesting that the incorporation of oxygen in the fluorocarbon plasma feedstock efficiently suppresses the generation of secondary chemical structural defects. This might be a contributing factor to the more pronounced enhancement in the laser damage resistance of the sample treated with modified RIE. It should be noted, however, that the concentration of ODCs and NBOHCs in the sample treated with modified RIE is still higher than that of the virgin sample. It is possible to further attenuate the concentration of secondary chemical structural defects by modulating the proportion of oxygen added to the feedstock.

Figure 5 illustrates the depth-dependent changes in the signal intensity ratio of O to Si for each sample. Due to the saturation of the ¹⁶O signal, its isotope ¹⁸O was analyzed. In ToF-SIMS, the signal intensity of a certain element is proportional to its concentration, making the O-to-Si intensity ratio representative of their stoichiometry. It is evident from Figure 5 that the curves of the three samples almost coincide. Herein, the comparison between the conventional RIE-treated sample and the virgin one reveals no significant alteration in the O/Si stoichiometry in the sample during the etching process with CHF₃/Ar plasma, which agrees well with the results in our previous research [9]. The comparison of samples treated with modified RIE and conventional RIE indicates that the incorporation of oxygen in the feedstock of plasma introduces no additional O into the substrate of the fused silica.

It is widely accepted that the etching of SiO₂ using fluorocarbon plasma operates through an ion-enhanced reaction, since the inherent reactivity between SiO₂ and F atoms is minimal [21,22]. In the “damage” mechanism model, ion bombardment modifies the SiO₂ network, either transforming it into a Si-rich layer or disrupting the Si-O bonds, both exponentially enhancing the reactivity of the altered material layer with F atoms [21]. The results in Figure 5 suggest that the RIE process does not markedly alter the O/Si stoichiometry, and thus are in support with the latter aspect of the damage mechanism. This also agrees with the CL results shown in Figure 4, where ion bombardment during the conventional RIE process damages the Si-O bonds, resulting in a substantial increase in ODC and NBOHC concentrations. It is worth mentioning that although the introduction of oxygen in the feedstock of fluorocarbon plasma effectively diminishes the damage to Si-O bonds, it does not markedly influence the apparent etching rate, as shown in the first paragraph of Section 2.

Figure 6 illustrates the distribution of F versus depth near the surface region for the three samples. The intensity of the F signal was normalized with the silicon particle number at the corresponding depth as a standard. Considering the relationship of the three curves, we can divide them into three intervals. In the 0 to ~10 nm depth range, named the “environment influencing zone” (denoted in gray), all three samples exhibit a pronounced decrease in F concentration as the depth increases. Given that the originally polished sample has no procedure in the processing that might introduce F, we deduce that the detected F primarily originates from the environment and might not significantly affect the substrates of the samples. The ~10 nm to ~25 nm depth interval is the “transition zone” (denoted in blue). Here, the black curve representing the virgin sample converges to a negligible value and remains stable with increased depth, suggesting a near-zero concentration of F. Within this zone, both the conventional RIE-treated Sample #1 and the modified RIE-treated Sample #2 exhibit higher F concentrations than the virgin sample. Comparatively, Sample #2 has a significantly lower F concentration than Sample #1, which rapidly descends to a value similar to the virgin sample by the end of this zone (~25 nm in depth). Beneath the depth of ~25 nm, in the “core zone”, the blue curve representing Sample #2 perfectly aligns with the black curve of the virgin sample, maintaining a near-zero F concentration. In contrast, Sample #1 still contains a certain amount of F, which gradually decreases with increasing depth.

It is evident from these results that the RIE process introduces F into the near-surface region of fused silica to a specific depth. However, compared to the conventional RIE, the modified RIE with oxygen incorporated in the feedstock considerably mitigates both the concentration and penetration depth of the F introduced into the sample. The penetration depth reduces from over 70 nm in the former to approximately 15 nm in the latter. Interpretations of this phenomenon can be obtained from previous research. Cardinaud et al. [10] suggested the formation of a SiO_xF_y layer containing fluorocarbon on the substrate surface during the etching of SiO₂ using CHF₃ plasma. Takahashi et al. [23] pointed out that the addition of oxygen in the fluorocarbon plasma suppressed the formation of fluorocarbon. Efremov et al. [24] noted that in a CHF₃/Ar/O₂ plasma, an increase in O₂ proportion inhibited the generation of fluorocarbon polymers, reducing the polymer film thickness. The decrease in the penetration depth and concentration of F might be another contributing factor to the superior damage-resistant performance of the sample treated with modified RIE compared to the sample treated with conventional RIE.

From all the experimental results shown above, we can obtain an overview of the oxygen-aided RIE proposed here with consideration of the effect on improving the laser-induced damage resistant performance of fused silica optics. It is evident that the oxygen-aided RIE has a pronounced effect on improving the LIDT of the sample to a much larger extent compared to the conventional RIE, as the damage test result in Figure 2 shows. From the CL and ToF-SIMS results, we can obtain an outline of the possible underlying mechanisms, as follows: although the conventional RIE would not alter the O/Si stoichiometry of the material (Figure 5), ion bombardment during this process would damage the Si-O bonds and induce chemical structural defects including ODCs and NBOHCs as a result while it removes the SSD layer of the originally polished sample (Figure 3 and Figure 4). Besides, fluorine impurity element defects would be introduced with a penetration depth of more than 70 nm (Figure 6). These factors lead to a limited improving effect of conventional RIE on the damage-resistant performance. The incorporation of O₂ in the feedstock of the plasma in the proposed oxygen-aided RIE can help to suppress the generation of ODCs and NBOHCs (Figure 4). At the same time, it can reduce both the concentration and penetration depth of the introduced impurities of element F (Figure 6). Both of these two aspects may contribute to the superior damage-resistant performance of the sample treated with this method.

4. Conclusions

A modified RIE technique with O₂ incorporated into the feedstock of the plasma has been developed to improve the laser-induced damage resistance of fused silica optics. Compared to the conventional RIE, employing CHF₃/Ar plasma, this method suppresses the generation of secondary chemical structural defects as well as the introduction of impurities of element F near the surface of the sample, and thereby significantly improves the damage-resistant performance. The 0% probability damage threshold of the sample treated by this method increases by 121% compared to that of the virgin sample and by 41% compared to that of the sample treated by the conventional RIE. By optimizing the proportion of O₂ added in the feedstock and other plasma discharge parameters, it is anticipated that a more significant improvement in damage resistance can be achieved. This method retains all the technical advantages of the existing dry etching technique and admirably addresses the issue of insufficient improvement in damage resistance without increasing the complexity of the procedures, presenting a highly promising, traceless, environmentally friendly, and efficient approach to improving the laser damage resistance of fused silica.