Silicon Oxide Etching Process of NF₃ and F₃NO Plasmas with a Residual Gas Analyzer

Abstract

The use of NF₃ is significantly increasing every year. However, NF₃ is a greenhouse gas with a very high global warming potential. Therefore, the development of a material to replace NF₃ is required. F₃NO is considered a potential replacement to NF₃. In this study, the characteristics and cleaning performance of the F₃NO plasma to replace the greenhouse gas NF₃ were examined. Etching of SiO₂ thin films was performed, the DC offset of the plasma of both gases (i.e., NF₃ and F₃NO) was analyzed, and a residual gas analysis was performed. Based on the analysis results, the characteristics of the F₃NO plasma were studied, and the SiO₂ etch rates of the NF₃ and F₃NO plasmas were compared. The results show that the etch rates of the two gases have a difference of 95% on average, and therefore, the cleaning performance of the F₃NO plasma was demonstrated, and the potential benefit of replacing NF₃ with F₃NO was confirmed.

1. Introduction

Since the 1980s, the use of NF₃ plasmas in semiconductor, display, and solar cell processing applications has been investigated [1]. NF₃ plasma is used to etch various thin films under reactive ion etching (RIE) conditions [2,3] or to clean a plasma-enhanced chemical vapor deposition (PECVD) chamber [4,5]. Cleaning the PECVD chamber is performed by supplying ions and radicals for cleaning through a remote plasma source (RPS) [6] or by directly supplying ions and radicals through an in situ plasma discharge [7]. In addition, NF₃ is attracting attention as a new etching technology, such as cryogenic electron beam induced etching (EBIE) [8,9] and highly selective etching [10]. NF₃ has a high etch rate, etching efficiency, and a relatively high chemical stability [2,11]. Accordingly, the use of NF₃ is significantly increasing every year. However, NF₃ is a greenhouse gas with a very high global warming potential of 16,100 and a lifespan of 500 years [12]. The contribution of NF₃ to the radiative forcing in the Earth’s atmosphere is very small, ~0.01%, in 2011, but the use of NF₃ is increasing every day; therefore, this number continues to increase [13]. The share of NF₃ in fluorinated gases increased from 13–28% in 2005 to 17–36% in 2010, and NF₃ is currently the most widely used and released fluorinated gas [1,14]. Thus, it was included in the second commitment period of the Kyoto Protocol as the seventh greenhouse gas whose emissions are to be regulated [15].

Studies have been conducted to replace NF₃ [16,17]. Among them, F₃NO was considered a candidate gas to replace NF₃ [18]. Similar to NF₃, F₃NO does not contain perfluorocarbons. In addition, because its molecule has an N=O bond, its atmospheric life is relatively short, and thus, its contribution to global warming is expected to be less than that of NF₃. The etch rate of F₃NO is almost the same as that of NF₃. However, further studies on F₃NO have not been conducted, and information on the mechanism occurring in the F₃NO plasma during the cleaning process is insufficient. Therefore, research on the F₃NO plasma to replace NF₃ is urgently needed.

The present study focused on evaluating and analyzing the properties of F₃NO plasma. Accordingly, the cleaning abilities of NF₃ and F₃NO were compared, and the characteristics of the F₃NO plasma were analyzed. The cleaning ability of the F₃NO plasma was evaluated by etching a SiO₂ film previously deposited on a Si wafer sample. Although the etching time of the small sample and the cleaning time of the PECVD chamber may not coincide, it was considered suitable for the evaluation of the cleaning ability of the gas. To analyze the etch mechanism, the etching plasma was diagnosed using a residual gas analyzer (RGA) and a high-voltage probe. The reactions in the F₃NO plasma were predicted by comparing the types and intensity of ions generated in the NF₃ and F₃NO plasmas. In addition, the etch rates of the SiO₂ thin film using the NF₃ and F₃NO plasmas were compared to confirm whether the F₃NO plasma can replace NF₃.

2. Materials and Methods

The plasma etching equipment used in this study is shown in Figure 1. We manufactured the RIE equipment with a direct capacitively coupled plasma chamber to measure the plasma etching characteristics. The detailed geometry of the device has been shown in a previous study [19]. A coolant path was formed in the electrode, which was designed to maintain a constant temperature (15 °C) during the process.

The sample used for etching was a SiO₂ thin film deposited on a Si wafer. The size of the sample was 30 mm × 30 mm, and the thickness of the thin film was 2 μm. NF₃ and F₃NO were used as process gases for etching. The F₃NO used in the experiments was manufactured and supplied by SK Materials (Sejong, Korea) and the National NanoFab Center (Daejeon, Korea). The purity of the provided F₃NO gas was 99.995%. The injected process gas was controlled using mass flow controllers. The process flow rate was fixed at 120 sccm. After the injection of the process gas, the process pressure was adjusted using a butterfly valve. The working pressure range was 130–270 mTorr. This working pressure range was set to confirm the possibility of cleaning in situ with a small amount of gas in the chamber while reducing the amount of cleaning gas without any additional high flow MFC configuration. After reaching the process pressure, the discharge power was applied through the RF power generator. The RF input power range was 240–400 W, and the process time was fixed at 3 min.

Mass spectrometry measurements were performed using an RGA (RGA 300, SRS, Sunnyvale, CA, USA). As shown in Figure 1, the RGA is equipped with a differential pump system. The pressure of the differential pump system was fixed at 0.8 mTorr regardless of the process pressure. At this time, the ionization energy of the RGA was fixed at 70 eV. The DC self-bias (DC offset) was measured using a high-voltage probe to determine the characteristics of the NF₃ and F₃NO plasmas. The etching rate of the thin film was evaluated after etching by measuring its thickness using the spectroscopic reflectometry method via the S-TRC series (Wonwoo Systems Co., Ltd., Seoul, Korea) [20].

3. Results and Discussion

To compare the characteristics of the NF₃ and F₃NO plasmas and to understand the characteristics of the latter, a DC offset measurement during the RIE plasma discharge was performed (Figure 2). The F₃NO plasma showed a similar DC offset value very close to that of the NF₃ plasma.

The NF₃ plasma has a very high electronegativity [21,22]. Therefore, when compared with fluorocarbon plasmas, the NF₃ plasma shows a relatively low DC offset value. Plasma with a high electronegativity tends to be unstable and easily collapses. In this case, the NF₃ plasma is discharged only in a part of the chamber, and the plasma collapses in other parts; therefore, these parts may not be exposed to the plasma. Alternatively, a constant discharge may not occur but may flicker and cause discharge.

The comparison results of the DC offset of the NF₃ and F₃NO plasmas show that the electronegativity of the F₃NO plasma can also be quite considerable. In the case of the NF₃ plasma, an inert gas (e.g., He, Ar) was diluted and discharged to increase the stability and uniformity of the plasma and enhance etch rates [23,24,25,26]. Inert gases rarely participate in chemical reactions in the plasma but can have a great influence on the discharge process. The DC offset result indicates that the characteristics of the F₃NO plasma can also be comparable to those of the NF₃ plasma. Thus, in future studies, dilution with an inert gas, such as Ar, is recommended.

Figure 3 shows the current of F+ ions and etch products (SiF, SiF₂, and SiF₃) generated from the NF₃ and F₃NO plasmas when etching SiO₂ thin films. As the pressure and power increased, the intensity of the peak of F ions and the etch product also increased. At this time, when the pressure or discharge power was low, more F ions were generated in the F₃NO plasma than in the NF₃ plasma. In addition, the lower the discharge power, the higher the generation of etch products in F₃NO than in NF₃. These findings confirmed that the etching of F₃NO occurs more actively at low power and pressure. In addition, as the power increases, the intensity of F ions in NF₃ rapidly increases, whereas in the F₃NO plasma, the F ions gradually increase even if the discharge power of F increases. As will be shown further, this trend is similar for other ions as well. Consequently, the plasma density of F₃NO reacts more slowly to the change of the discharge power compared to NF₃.

Among the ions generated in the NF₃ and F₃NO plasmas during the silicon oxide thin film etching, O and O₂ were more common in the F₃NO plasma under all conditions (Figure 4). Compared with the F₃NO plasma, O and O₂ ions hardly occurred in the NF₃ plasma. In the case of the F₃NO plasma, O and O₂ ions were simultaneously generated by O contained in F₃NO and O₂ gas already present in the chamber, whereas the NF₃ plasma generated O and O₂ ions only from O₂ gas existing in the base. Therefore, when the pressure increased, the intensity of O and O₂ ions generated in the NF₃ plasma was almost unchanged, whereas in the F₃NO plasma, when the pressure increased, O and O₂ ions appeared to rapidly increase compared to the NF₃ plasma.

Figure 5 shows the peak intensities of the major ions generated in the NF₃ and F₃NO plasmas during the etching of a silicon oxide thin film. N, N₂, and F₂ ions occurred more in F₃NO at low discharging powers and more in NF₃ at higher discharging powers. Conversely, NF₂ ions occurred more in the NF₃ plasma under all conditions.

Figure 6 shows the intensity of NO and NO₂ ions generated in the NF₃ and F₃NO plasmas during the etching of silicon oxide thin films. When the discharge power was low, the NO ions generated in the F₃NO plasma were more than those of the NF₃ plasma (numerical value). However, as the discharge power increased, the number of NO ions in the F₃NO plasma gradually decreased, whereas those in the NF₃ plasma rapidly increased and became almost the same as the peak intensity of NO ions generated in the F₃NO plasma. Conversely, many more NO₂ ions occurred in the F₃NO plasma under all conditions.

$$\mathrm{In}\mathrm{the}{\mathrm{NF}}_3\mathrm{and}{\mathrm{F}}_3\mathrm{NO}\mathrm{plasmas},\mathrm{NO}\mathrm{was}\mathrm{produced}\mathrm{through}\mathrm{the}\mathrm{following}\mathrm{reaction}:{\mathrm{N}}_2\left({\mathrm{A}}^3{\mathsf{\Sigma }}_u^{+}\right)\mathrm{O}\to \mathrm{NO}+\mathrm{N}.$$

(1)

Metastable N₂ is produced mostly by collisions with a high-energy electron, making the mechanism more significant in an electronegative gas discharge, such as NF₃ and F₃NO. At a higher discharge power, higher energy electrons are supplied, thus increasing the density of NO ions [27]:

N + O₂ → NO + O.

(2)

A reaction involving atomic nitrogen that resulted in a different density of NO ions in the F₃NO plasma and the NF₃ plasma was more important in an oxygen-rich plasma [28,29]. As shown in Figure 4a, because F₃NO is an oxygen-rich plasma, reaction (2) became significant, and a large amount of NO ions were produced.

In the F₃NO plasma, NF is decomposed through the following reaction.

NF₂ + O → F + FNO.

(3)

The species produced in the primary reactions led to secondary reactions which formed NO₂ ions through the following exothermic reactions:

FNO + O → F + NO₂,

(4)

FO + NO → F + NO₂,

(5)

N₂O + NO → N2 + NO₂,

(6)

NO + O₃ → O₂ + NO₂.

(7)

In the case of the NF₃ plasma, the above reaction is not significant because the number of O ions is remarkably small, but F₃NO causes a more significant reaction. Therefore, the number of NF₂ and NF ions are fewer in the F₃NO plasma than in the NF₃ plasma, whereas ions such as N, N₂, and F are present in similar amounts in the F₃NO plasma and the NF₃ plasma.

In the NF₃ plasma, NO ions may be generated through the bonding of O and N ions, which are etching by-products, or through a process in which N ions form Si–O–N bonding on the surface of the silicon oxide thin film [30]. In the F₃NO plasma, NO is formed by the N and O ions contained in F₃NO. In the case of F₃NO, this condition becomes the main mechanism. NO formed in this way forms NO₂ through the following reaction:

NO + O + M → NO₂ + M.

(8)

In the NF₃ plasma, because the number of O ions was remarkably small, this oxidation reaction occurred only to a small extent. Conversely, in the F₃NO plasma, as the number of O ions was much larger, the extent of the oxidation reaction was more significant. When the discharge power was increased, the number of NO formed was almost unchanged, but the peak of NO ions decreased due to the considerable oxidation of NO.

Figure 7 shows the SiO₂ etch rate according to the discharge power and process pressure. The results imply that the lower the pressure and discharge power, the higher the etch rate of F₃NO. This is because when the pressure and discharge power is low, chemical etching occurs more easily because the number of F ions generated is greater in F₃NO. Conversely, the DC offset at a low power has a small absolute value for the NF₃ and F₃NO plasmas; therefore, the ion bombardment energy does not have a significant effect on etching. When the discharge power was increased, the DC offset value of the NF₃ plasma became larger than that of the F₃NO plasma, so the etch rate of NF₃ also became higher. When the discharge pressure and discharge power increased, the intensity of F ions also increased as NF₃ increased, so the etch rate decreased in F₃NO. Furthermore, the etch rate of silicon oxide during F₃NO plasma etching was approximately 95.0% of the rate during NF₃ plasma etching.

We performed SEM measurements to determine whether O or N ions present in the plasma had a negative effect on the etching quality. Figure 8 shows a SiO₂ surface of SEM images after the etching process at 400 W discharge power and 270 mTorr pressure. For accurate SEM measurements, a platinum coating was applied on the surface by sputtering. The round shape particles in the figures are platinum nanoparticles from the platinum coating. The size of these nanoparticles is in order of several nanometers. Besides platinum nanoparticles, no other structures such as cracks or holes were found on the surface. No significant difference was observed between the SEM images of the unprocessed and processed surface of SiO₂. Therefore, it was confirmed that O or N ions in F₃NO did not have a negative effect on the etching quality.

Figure 9 shows EDS spectra of the SiO₂ surface without process and after etching at a 400 W discharge power and 270 mTorr pressure with the NF₃ and F₃NO plasmas. The C peak in the EDS spectra was caused either by carbon contamination or by the window in the detector. Except for carbon and platinum (from the platinum coating), no peaks other than O and Si were found in the EDS spectra. This indicates that nitrogen, oxygen, or NO did not chemically contaminate the SiO₂ surface during the etching process.

The mass ratio of silicon and oxygen is noticeable in the EDS spectra. The mass ratio of silicon and oxygen on the SiO₂ surface is almost the same when the process is not performed and when the NF₃ etching process is performed. However, the mass ratio of O appears less on the SiO₂ surface after F₃NO etching. This may be caused by the following reaction on the SiO₂ surface during F₃NO etching.

O(s) + NO(g) → NO₂(g).

(9)

As above, NO ions absorb O in the SiO₂ surface to decrease surface oxidation [31]. Therefore, the F₃NO etched SiO₂ surface has a smaller oxygen mass ratio than the NF₃ etched SiO₂ surface. Moreover, this de-oxidation process increases the Si etching rate, especially during etching with F₃NO at low pressure and low power with a significant quantity of NO ions.

4. Conclusions

The DC offset was measured during NF₃ and F₃NO plasma discharges. Compared with the DC offset of the NF₃ plasma, the F₃NO plasma showed an almost similar DC offset value. This finding confirms that the F₃NO plasma, similar to the NF₃ plasma, can have a very high electronegativity. Moreover, the NF₃ plasma, similar to the F₃NO plasma, may also exhibit unstable or non-uniform characteristics. Therefore, the diluent of an inert gas into the F₃NO plasma can be effective.

The ions generated in the NF₃ plasma and the F₃NO plasma during the etching of the SiO₂ thin film were measured through the RGA. In the case of F ions, when the discharge power and discharge pressure were low, more F₃NO plasmas were generated than NF₃ plasmas. The result was the same for the etching by-products (SiF, SiF₂, and SiF₃). In addition, as the power increased, the intensity of F ions in NF₃ rapidly increased, whereas in the F₃NO plasma, the ions of F gradually increased even if the discharge power of F increased. This result showed a similar trend for other ions afterward, implying that the plasma density of F₃NO reacts more slowly to the change of the discharge power compared to NF₃. Ions O and O₂ generated during the plasma discharge were much more significant in the F₃NO plasma than in the NF₃ plasma, which is attributed to the O ions contained in F₃NO. Furthermore, the intensity of O ions can affect the etching mechanism of the F₃NO plasma. F₂, N, N₂ ions occur more in F₃NO at a low discharge power and occur more in NF₃ at a high discharge power. In contrast, NF₂ ions are much higher in NF₃ ions under all conditions. O ions in F₃NO cause a reaction to decompose NF₂. When the discharge power is low, the NO ions of F₃NO are generated in higher amounts compared to the NF₃ plasma. However, as the discharge power increases, the number of NO ions in F₃NO gradually decreases, whereas the NO ions in the NF₃ plasma rapidly increase, and the intensity of the peak of NO generated in F₃NO becomes almost similar. Conversely, NO₂ ions occur more in the F₃NO plasma under all conditions. In the F₃NO plasma, many NO ions are generated in the process of decomposing NF₂. However, when the pressure increases, NO ions are oxidized by O ions to form NO₂ ions, and thus the number of NO ions decreases. Through this oxidation reaction, many more NO₂ ions are generated in the F₃NO plasma than in the NF₃ plasma.

The SiO₂ etch rates of the NF₃ and F₃NO plasmas were compared. The results show that the lower the pressure and discharge power, the higher the etch rate of F₃NO. This is because the intensity of F ions is higher in the F₃NO plasma at low pressure. As the discharge power increases, the intensity of F ions in the NF₃ plasma increases. Therefore, the etch rate of the NF₃ plasma increases. The etch rate of silicon oxide during F₃NO plasma etching was approximately 95.0% of the NF₃ plasma etching rate.

To compare the etch qualities, SEM measurements were performed. There was no difference between the unetched and etched SiO₂ surface with the NF₃ and F₃NO plasmas. Therefore, we found that N or O ions in F₃NO did not negatively affect the etch quality. The results of this study confirm the cleaning properties of F₃NO. Nonetheless, the limitation of this study is that only NF₃ and F₃NO plasmas were compared. In addition, EDS measurements were performed in parallel to assess the possibility of chemical contamination of the surface by ions in F3NO and phenomena occurring on the SiO₂ surface during etching. As a result of the measurement, no chemical contamination was observed during etching with NF₃ plasma or F₃NO. Unlike NF₃ plasma etching, it was observed that the mass ratio of oxygen of the SiO₂ surface decreased during F₃NO plasma etching. This may be attributed to the de-oxidation process of the SiO₂ surface by NO ions.

The characteristics of the F₃NO plasma were identified through these results, and the potential for replacing F₃NO with NF₃ was confirmed. Further studies will be needed when inert gases, such as Ar or He, are used as diluent. In addition, higher pressures need to be evaluated for the cleaning ability.