Characterization of SiO₂ Plasma Etching with Perfluorocarbon (C₄F₈ and C₆F₆) and Hydrofluorocarbon (CHF₃ and C₄H₂F₆) Precursors for the Greenhouse Gas Emissions Reduction

Abstract

This paper proposes the use of environmentally friendly alternatives, C₆F₆ and C₄H₂F₆, as perfluorocarbon (PFC) and hydrofluorocarbon (HFC) precursors, respectively, for SiO₂ plasma etching, instead of conventional precursors C₄F₈ and CHF₃. The study employs scanning electron microscopy for etch profile analysis and quadrupole mass spectrometry for plasma diagnosis. Ion bombardment energy at the etching conditions is determined through self-bias voltage measurements, while densities of radical species are obtained using quadrupole mass spectroscopy. The obtained results compare the etch performance, including etch rate and selectivity, between C₄F₈ and C₆F₆, as well as between CHF₃ and C₄H₂F₆. Furthermore, greenhouse gas (GHG) emissions are evaluated using a million metric ton of carbon dioxide equivalent, indicating significantly lower emissions when replacing conventional precursors with the proposed alternatives. The results suggest that a significant GHG emissions reduction can be achieved from the investigated alternatives without a deterioration in SiO₂ etching characteristics. This research contributes to the development of alternative precursors for reducing global warming impacts.

1. Introduction

The semiconductor industry has extensively embraced plasma technology for semiconductor device manufacturing [1,2]. In the etching process, plasma has become crucial for nanoscale material processing, enabling precise and deep etch profiles to meet the requirements of scaling down of transistor feature sizes [3]. As a results, plasma processing has become indispensable for the production of next-generation semiconductor devices, including high-aspect-ratio contact (HARC) hole.

Plasma etching utilizes various gases or gas mixture tailored to the target materials, such as semiconductors [4,5], metals [6,7,8], and dielectrics [9,10,11,12]. For dielectrics, such as SiO₂ and Si₃N₄, fluorocarbon-based molecules, such as C₄F₈ and CHF_3, are widely adopted due to their polymerizing properties, which are particularly advantageous for achieving anisotropic etch profiles. During the etching process, fluorocarbon films passivate the sidewalls of the dielectric trench, preventing them from reacting with reactive etchants. However, the energetic plasma ion bombardment renders the fluorocarbon film unable to protect the trench bottom, resulting in an anisotropic etch profile [13,14,15]. Since anisotropy is a crucial characteristic in semiconductor manufacturing, large quantities of perfluorocarbon (PFC) and hydrofluorocarbon (HFC) precursors, such as C₄F₈ and CHF₃, are extensively utilized in the semiconductor industry.

However, the utilization of the PFC and HFC precursors is subject to strict regulations due to their significant greenhouse effect [16,17]. Global warming potential (GWP) is a metric used to quantify the relative warming impact of greenhouse gases (GHGs) on the Earth’s climate system over a specified time horizon, typically, 20, 100, or 500 years, relatively to that of CO₂ [18]. The GWP₁₀₀ (GWP over a 100-year time horizon) of C₄F₈ and CHF₃ is reported as 9540 and 12,400, respectively, indicating that these precursors have a greenhouse effect over 10,000 times stronger than CO₂ [19]. Given the substantial global efforts to combat global warming in recent decades [20,21], the emission of these high-GWP gases is facing increasingly stringent restrictions in the industry [22].

In this context, researchers worldwide have been actively engaged in developing alternative plasma etching precursors that process low GWPs while demonstrating comparable etch performance to conventional ones. Sung et al. investigated the etching characteristics of SiO₂ using the C₆F₆/Ar/O₂ gas mixture in both capacitively coupled and inductively coupled plasma etching systems, comparing the etch results obtained from these different plasma systems [23]. Kim et al. studied SiO₂ etching using C₄H₃F₇O isomers and compared them with C₄F₈ through X-ray photoelectron spectroscopy, quadrupole mass spectroscopy, and scanning electron microscopy [24]. Furthermore, Lim et al. compared the etch results with a low-GWP precursor, C₆F₁₂O, to those with the conventional precursor, CF₄ [25]. Although many papers, including the aforementioned ones, have documented the development of alternative low-GWP precursors in the field of semiconductor manufacturing [26,27,28,29,30], there are still numerous potential precursors that need to be investigated regarding their qualification as promising alternatives to conventional high-GPW precursors.

In this work, we propose the use of environmentally friendly alternatives, C₆F₆ and C₄H₂F₆, as PFC and HFC precursors, respectively, for SiO₂ plasma etching, in lieu of the conventional precursors C₄F₈ and CHF₃, which are the most representative etching gases for dielectric etching [3]. C₄F₈ and CHF₃ are selected as the reference precursors for PFC and HFC precursors, respectively, instead of other alternative low-GWP precursors, such as C₄F₆ (GWP < 1), to excavate additional alternatives that possibly show different etching performance from C₄F₆, which would provide a wide range of options to, for example, process engineers. Among various candidates as an alternative low-GWP precursor, the investigation into C₆F₆ and C₄H₂F₆ is reported here. It should also be noted that CHF₃, an H-containing fluorocarbon precursor, is normally used for nitride etching. To the best of our knowledge, more than one material are etched simultaneously in an etching process and specifically in dielectric etching, etching selectivity of nitride-to-oxide or vice versa is essential. We thus consider that evaluating the etching performance of HFC precursors, such as CHF_3, on SiO₂ is quite necessary. Additionally, we also consider that an HFC precursor should be compared with another HFC precursor, not a perfluorocarbon precursor like C₄F₈, to reduce any complexity that might be induced by H atoms included in HFC precursors but not in PFC ones. A low-GWP candidate, C₄H₂F₆, was therefore compared with CHF₃ to determine whether C₄H₂F₆ was appropriate as an alternative to CHF₃ or not.

We employ scanning electron microscopy and plasma diagnostics to analyze the etching results. Ion bombardment energy is determined by measuring self-bias voltages, while densities of various radical species are obtained through quadrupole mass spectroscopy. The obtained results include a comprehensive comparison of the etch performance, encompassing etch rate and selectivity, between C₄F₈ and C₆F₆, as well as between CHF₃ and C₄H₂F₆. Furthermore, we evaluate the greenhouse gas emissions using a million metric tons of carbon dioxide equivalent, which demonstrates significantly lower emissions when substituting conventional precursors with the proposed alternatives. The experimental setup and procedure details are provided in the subsequent sections.

2. Experiment

2.1. Processing Chamber Setup

The schematic diagram of the plasma chamber used for SiO₂ etching is presented in Figure 1. Radiofrequency (RF) power of 13.56 MHz is applied to the electrode, which has a diameter of 208 mm covered with a 2 mm-thick ceramic plate, through an impedance matching box. While being injected into the chamber via a 1/4” SUS tube, the flow rate of Ar is controlled by a mass flow rate controller (MFC) (1179A611522K, MKS Instruments Inc., Andover, MA, USA) and that of PFC and HFC precursors is controlled by the different model of an MFC (M3030VA, Line-Tech, Daejeon, Republic of Korea). The investigated PFC precursors are octafluorocyclobutane (C₄F₈) and hexafluorobenzene (C₆F₆), while the investigated HFC precursors are trifluoromethane (CHF₃) and hexafluoro-isobutylene (C₄H₂F₆). The specific material information on the investigated precursors is summarized in

Table 1. The atmospheric lifetime and GWP100 of the investigated PFC and HFC precursors.

Name	Molecular Formula	Atmospheric Lifetime (Years)	GWP100	Reference
Octafluorocyclobutane	C4F8	3200	9540	[19]
Hexafluorobenzene	C6F6	0.23	7	[23]
Trifluoromethane	CHF3	222	12,400	[19]
Hexafluoro-isobutylene	C4H2F6	0.03	2.8	[31]

.

The flow rate of Ar and a fluorocarbon precursor are fixed at 20 sccm and 10 sccm, respectively. It should be noted that the canisters containing C₆F₆ and C₄H₂F₆ are heated to $130 °\mathrm{C}$ in order to vaporize them, and are in the liquid phase at room temperature. The RF power applied ranges from 300 W to 500 W and the pressure varies from 10 mTorr to 30 mTorr (

Table 2. The experimental conditions for the examined etching processes.

Power (W)	Pressure (mTorr)	Duration (min)	Gas Flow Rate (sccm)	Electrode Temperature (°C)
300–500	10–30	10	20 for Ar 10 for FC	10

). When the RF power varies, the pressure is regulated by a throttle valve; when the pressure changes, the RF power is fixed at 400 W. The chamber is evacuated by a turbomolecular pump, with an oil rotary pump serving as a backing pump, to achieve a base pressure below 10⁻⁴ Torr.

2.2. Plasma Diagnosis

We utilized a high-voltage (HV) probe (P5100, Tektronix, Inc., Beaverton, OR, USA) and a residual gas analyzer (RGA) (SRS-200, Stanford Research Systems, Sunnyvale, CA, USA) to analyze the SiO₂ etching results obtained with different precursors. The HV probe is connected to the transmission line responsible for RF input power positioned between the electrode and the matching box. This measures the self-bias voltage of the electrode, providing an estimation of the ion bombarding energy during plasma processing [32]. RGAs are widely employed plasma diagnostics tools that offer information about gas-phase species in plasmas [33,34,35,36]. With an RGA, gas-phase atoms or molecules ionized by the collision with 70 eV energetic electrons and then filtered based on their mass as they pass through a quadrupole mass filter operating in a cross-linked manner [37]. This filtering process enables the RGA to generate a signal intensity profile as a function of atomic mass [38]. The RGA, located on a chamber port, has a separate vacuum space from the main chamber via an orifice with the diameter of approximately 100 $\mathsf{\mu }\mathrm{m}$, which helps in maintaining the RGA pressure at approximately two orders of magnitude lower than the main chamber pressure [39]. The orifice is considered to directly contact to the plasma diagnosed as the chamber wall. This dedicated vacuum system allows gas-phase atoms and molecules to traverse the RGA without experiencing collisions with background gas, thereby increasing data reliability.

It should be noted that the impact of electrons of 70 eV would result in a measurement error from the dissociative ionization reaction of a parent molecule [40]; specifically, if C₄F₈ collides with 70 eV electrons, CF₂⁺ ions can be generated and they are almost indistinguishable from the CF₂⁺ ions generated in plasma in QMS analysis. Reducing the noise from dissociative ionization will be addressed in future work.

2.3. Sample Preparation

We prepared SiO₂ coupon wafers with a masking layer of amorphous carbon (ACL) that has bar-type deep-trench patterns. Figure 2 presents the cross-sectional view of the mask pattern before etching with a width of 170 nm, a depth of 1500 nm, and a pitch of 530 nm. This cross-sectional image of the not-etched SiO₂ sample serves as a reference for comparing the images of the samples after etching. All the images presented in this paper were acquired using a scanning electron microscope (SEM) (SU7000, Hitachi, Ltd., Tokyo, Japan).

3. Results and Discussion

3.1. SiO₂ Etching with PFC Precursors (C₄F₈ and C₆F₆)

Figure 3 illustrates the etch results obtained with different PFC precursors (C₄F₈ and C₆F₆) at varying RF power levels ranging from 300 W to 500 W, while maintaining a fixed pressure of 20 mTorr. The resulting etch rates and selectivity for the C₄F₈ and C₆F₆ cases are summarized in the right-most column of Figure 3. The etch rate of ACL is calculated by measuring the difference between the lengths from the ACL-SiO₂ border line to the top surface of ACL before and after etching, divided by the etching duration of 10 min. Similarly, the etch rate of SiO₂ is determined by dividing the etch depth, defined as the length between the border line and the minimum position of the etched SiO₂ trenches, by the etching duration of 10 min. The etch selectivity is then obtained by dividing the SiO₂ etch rate by the ACL etch rate.

Regarding the SiO₂ etch rate, there is no significant difference observed between C₄F₈ and C₆F₆, although the difference becomes more pronounced with increasing RF power. It is noteworthy that, despite the significantly higher GWP₁₀₀ of C₄F₈ (9540) compared with C₆F₆ (7) by over 1000 times [23], the similar etch rate implies that adopting the low-GWP alternative precursor can result in a substantial reduction in the greenhouse effect without compromising etch throughput.

As depicted in Figure 3, it can be observed that the SiO₂ and ACL etch rates with C₄F₈ increase almost linearly with increasing RF power, leading to a nearly constant selectivity. On the other hand, with C₆F₆, it is notable that the ACL etch rate abruptly increases with increasing RF power while the SiO₂ etch rate shows a slight linear increase. Consequently, this rapid increase in ACL etch rate causes the selectivity to drop below 1. Such low selectivity with C₆F₆ can be attributed to the mask faceting, which worsens with higher RF power. Comparing the etch profiles obtained with C₄F₈ and C₆F₆, it is evident that mask faceting is more pronounced with C₆F₆ than with C₄F₈. For both C₄F₈ and C₆F₆ cases, it is commonly observed that, as RF power increases, mask faceting tends to deteriorate. This deterioration results in bowing of the mask, referring to a concave etch profile caused by the lateral, as well as vertical etching of the trench sidewall due to ion bombardment on the facet [3]. According to material sputtering theory, the sputtering yield is highest at incident angles ranging from 40° to 60°, rather than 90° [41]. This behavior is clearly reflected in the etch results shown in Figure 3. Comparing the transition of mask height and faceting, it is observed that, while the mask height barely changes, the mask faceting evolves rapidly with increasing RF power. Mask faceting induces mask bowing, which eventually affects the SiO₂ etch profile as the concave mask profile expands across the mask-SiO₂ border. Such mask faceting occurs more intensely with C₆F₆ than with C₄F₈, as presented in Figure 3. Notably, an increase in RF power from 400 W to 500 W resulted in a significant rise in the mask etch rate. This could be attributed to the left and right mask facets meeting during etching, forming a spear-like mask head. As discussed earlier, ion bombardment is more effective on a tilted surface compared with a flat surface, and thus the spear-like mask head created by the joint of the left and right facets could lead to the rapid mask etching.

To analyze the differences in etching behavior between C₄F₈ and C₆F₆ precursors, we investigated changes in ion bombardment energy and radical density. In a capacitively coupled plasma etching system, such as the one used in this work, where the samples are loaded on the bottom electrode, also as in this work, the powered electrode (bottom electrode) voltage primarily consists of a negative DC component that accelerates the bombarding ions, allowing them to reach even the bottom of deep trenches [41]. Therefore, it is crucial to measure the self-bias voltage of the powered electrode to determine the ion bombardment energy under specific plasma conditions. We employed a high-voltage probe to measure the self-bias voltage on the powered electrode.

Figure 4a shows the measured self-bias voltage (left y-axis) and the CFx/F (x = 1, 2, 3) ratio (right y-axis) as a function of applied RF power using C₄F₈ and C₆F₆. The CFx/F ratios are calculated by using the FC radical signals represented in Figure 4b where CFx refers to the sum of the CF, CF₂, and CF₃ signal and is divided by the F signal at different RF power conditions. Note that the parameter, a CFx/F ratio, is adopted to interpret the chaotic contribution of numerous radical species to etch results in a simple manner. More detailed analysis of the measured radical densities will be followed separately. It can be observed that, as the RF power increases, the absolute value of the self-bias voltage increases, indicating an increase in ion bombardment energy. Notably, there is little difference between the self-bias voltages with C₄F₈ and C₆F₆, suggesting that the ion bombardment energy is very similar for both precursors. Furthermore, the absolute value of a self-bias voltage is known to be proportional to the plasma electron density. In CCP at a constant RF power, the dominant current path in the bulk plasma is the conduction current by plasma electrons so that, if the electron density decreases, the voltage between plasma and the electrode increases to meet the constant input power, a product of current and voltage [41]. In this context, the almost identical self-bias voltages at different RF powers may indicate that the ion density arriving at the etched surface is also very similar. In this case, the etch results are primarily influenced by the density of radical species [42]. As shown in Figure 4a, the CFx/F ratios at different RF powers are higher with C₄F₈ than C₆F₆, which indicates that polymerization occurs more with C₄F₈ than C₆F₆. This might lead to the better passivation of trench sidewalls during etching with C₄F₈, as shown in Figure 3.

In plasma etching, the synergetic effect between ions and radicals plays a significant role [43]. Therefore, measuring the radical composition is just as essential as investigating the ion bombardment energy. Radicals are generated through the dissociation of parent molecules by electron impact in plasma etching. Complex molecules like C₄F₈ undergo dissociation into various species with a wide range of molecular masses, including light molecules, such as CF and CF₂, as well as heavy molecules, such as C₂F₃ and C₃F₅ [44]. This wide range of species has led to the extensive use of mass spectroscopy for investigating FC plasma chemistry [38].

As depicted in Figure 4b, while low-mass radicals, such as F, CF, and CF₂, show no significant dependence on the power variation, a notable difference in high-mass radical signal between C₄F₈ and C₆F₆ is observed. Several high-mass species, such as C₂F₄, C₂F₅, and C₃F₅, exhibit higher densities with C₄F₈ compared with C₆F₆, primarily due to the dominant dissociation pathways of C₄F₈ in plasma [44]. Apart from these radicals, high-mass species are observed to be predominantly generated through the dissociation of C₆F₆. This may contribute to intense faceting of the ACL mask during etching with C₆F₆ by inducing the creation of high-mass ions that could transfer larger momentum to the ACL mask compared with low-mass species ions. The identification of the existence of such high-mass ions, however, is not conducted in this work due to the unverified ion measurement ability of the QMS equipment used. The modification of the equipment for reliability ion density measurement will be addressed in future work.

Figure 5 illustrates the etch results obtained with different PFC precursors (C₄F₈ and C₆F₆) at varying pressure ranging from 10 mTorr to 30 mTorr, while maintaining a fixed RF power of 400 W. The resulting etch rate and selectivity for the C₄F₈ and C₆F₆ cases are summarized in the right-most column of Figure 5. When etching with C₄F₈, a notable trend is observed where the SiO₂ etch rate increases with increasing pressure, while the ACL etch rate decreases. As a result, there is a significant increase in selectivity. Conversely, when using C₆F₆, the SiO₂ etch rate initially rises with increasing pressure but eventually starts to decrease at higher pressure, while the ACL etch rate shows a linear increase. Consequently, this leads to a decrease in selectivity. These findings suggest that the optimal SiO₂ etching window for C₆F₆ exists at lower pressures than that of C₄F₈, as the etch rate and selectivity with C₆F₆ can be comparable to those achieved with C₄F₈.

Regarding mask faceting, the etch profiles of the mask with both C₄F₈ and C₆F₆ show a relatively constant extent of faceting as the pressure increases. This is in contrast to the changes observed with variations in RF power, as depicted in Figure 3. This characteristic of mask faceting is advantageous for SiO₂ etching, as it induces rapid etching of the mask and leads to bowing in the etch profile of both the mask and the substrate.

Figure 6a illustrates the relationship between the measured self-bias voltage (left y-axis) and the CFx/F (x = 1, 2, 3) ratio (right y-axis) and pressure with C₄F₈ and C₆F₆. As the pressure increases, the self-bias voltage also increases, indicating a decrease in ion bombardment energy. With the change in pressure from 10 mTorr to 20 mTorr, the self-bias voltages show minimal difference between C₄F₈ and C₆F₆. However, as the pressure continues to rise, the disparity in self-bias voltage between C₄F₈ than C₆F₆ becomes more pronounced, with C₄F₈ exhibiting a larger increase. Since the absolute value of the self-bias voltage with C₄F₈ at 30 mTorr is lower than that with C₆F₆, the ion density and energy reaching the substrate are considered to be less, compared with the C₆F₆ case. This finding will aid in the interpretation of the etching discrepancy observed with C₄F₈ and C₆F_6, along with measurements of radical species.

Figure 6b demonstrates the changes in radical species signal as pressure increases for C₄F₈ and C₆F₆. As the pressure increases, the F signal is found to increase for both cases. However, the SiO₂ etch rate with C₆F₆ does not follow the F signal trend regarding the pressure change, which implies that the contribution of radical species to etching should be addressed through comprehensive consideration of various radical species. There is a significant disparity in high-mass radical signal between C₄F₈ and C₆F₆, as observed in Figure 4b. Additionally, the dependence of radical species signal on pressure is much substantial compared with its dependence on RF power.

Considering the slight decrease in self-bias voltage and the simultaneous rapid increase in radical species signal with varying pressure, it is expected that the SiO₂ etch rate would decrease as the deposition rate increases more than the etch rate [42]. However, contrary to this expectation, Figure 5 reveals that the SiO₂ etch rate with C₄F₈ actually increases as the pressure rises. This anomalous behavior can be attributed to the increasing difficulty of radical transport toward the trench bottom as the trench’s aspect ratio rises. Since a SiO₂ etch rate increases when the depositing FC film is very thin and eventually decreases and saturates when the FC film becomes significantly thick [45], the monotonic increase in the SiO₂ etch rate with C₄F₈ depicted in Figure 5 appears to be caused by the insufficient radical supply [46].

Overall, a comparison of ACL-masked SiO₂ etch results with C₄F₈ and C₆F₆ under variations in RF power and pressure reveals that C₆F₆ exhibits superior etch rate and selectivity. This finding is even more remarkable considering that C₆F₆ has a GWP₁₀₀ over 1000 times lower than that of C₄F₈. While C₆F₆ demonstrates a nearly identical self-bias voltage to that of C₄F₈, indicating similar ion bombardment energy, the plasma radical chemistry significantly differs. However, ACL etching with C₆F₆ experiences mask faceting, leading to rapid mask etching and decreased selectivity. Thus, it is necessary to identify a more optimal etching window for C₆F₆ to enhance performance for both ACL and SiO₂ etching processes.

3.2. SiO₂ Etching with HFC Precursors (CHF₃ and C₄H₂F₆)

Figure 7 presents the etch results obtained with different HFC precursors (CHF₃ and C₄H₂F₆) at varying RF power levels from 300 W to 500 W, while maintaining a fixed pressure of 20 mTorr. The resulting etch rates and selectivity values for C₄F₈ and C₆F₆ are summarized in the right-most column of Figure 7. It is commonly observed that as RF power increases, mask faceting tends to deteriorate, as evidenced in the etch results with C₄F₈ and C₆F₆. While the use of CHF₃ produces nearly vertical ACL etch profiles, the utilization of C₄H₂F₆ at an RF power of 500 W results in a distorted ACL etch profile. This distortion appears to be induced by the rough mask surface, which causes random ion scattering. However, when the RF power is set below 500 W, sharp and vertical trench etching can be achieved without any defects, such as bowing.

Regarding the etch rate and selectivity, CHF₃ exhibits a SiO₂ etch rate lower than that of ACL. As RF power increases, this trend is reversed, leading to a selectivity slightly above unity. On the other hand, with C₄H₂F₆, both the SiO₂ and ACL etch rates gradually increase as RF power increases. The etch rate of SiO₂ remains slightly higher than that of ACL, while the selectivity remains above unity, even at an RF power of 500 W when the mask sidewall is distorted, possibly due to random ion scattering. Although the SiO₂ etch rate with C₄H₂F₆ is slightly lower than that with CHF₃, the use of C₄H₂F₆ yields a comparable etch selectivity with that of CHF₃. Additionally, with minor optimization of the processing pressure, such as decreasing the pressure as demonstrated in Figure 7b, a fine etch profile could be achieved.

Figure 8a depicts the measured self-bias voltage (left y-axis) and the CFx/F (x = 1, 2, 3) ratio (right y-axis) as a function of applied RF power using CHF₃ and C₄H₂F₆. As the RF power increases, both CHF₃ and C₄H₂F₆ show an increase in the absolute value of the self-bias voltages, indicating an increase in the ion bombardment energy. Notably, the absolute value of the self-bias voltages with CHF₃ at different RF powers is approximately twice that with C₄H₂F₆, indicating more intense ion bombardment during etching with CHF₃ compared with C₄H₂F₆. This higher ion energy can be advantageous for HARC etching, where sufficient ion energy is required to reach the trench bottom. Etching with C₄H₂F₆ would therefore require additional biasing to enhance the ion bombardment in order to match that achieved with CHF₃.

Figure 8b presents the signal of various radical species resulting from the dissociation of CHF₃ and C₄H₂F₆, which varies with increasing RF power. As the RF power increases, CHF₃ and C₄H₂F₆ show different trends of radical signal variation species by species. Specifically, F from CHF₃ decreases with increasing RF power, while that from C₄H₂F₆ increases slightly. Those trends, however, do not follow the etch rate trend that changes with an increase in RF power, which might be attributed to a complex contribution of other FC radicals to the etch rate. When comparing CHF₃ and C₄H₂F₆, it is noteworthy that C₄H₂F₆ yields carbon-rich radical species that contain multiple carbon atoms in their chemical structure, while CHF₃ generates them to a lesser extent. This difference is expected since the generation of such carbon-rich radicals from CHF₃ dissociation requires additional recombination processes between carbon-containing species such as CF₂ and CHF₂, which have relatively low reaction rates. However, this particular behavior does not appear to have a significant impact on ACL-masked SiO₂ etching. Despite the prevalence of carbon-rich radicals in the process, the use of C₄H₂F₆, as shown in Figure 7, does not demonstrate substantial advantages in terms of ACL mask protection or SiO₂ trench profiles.

Nevertheless, it may be meaningful to focus on the difference in radical compositions between CHF₃ and C₄H₂F₆, particularly in relation to the major dissociation fractions of CHF₃, rather than solely considering the evident difference in plasma chemistry between the two gases. This is because the major species resulting from CHF₃ dissociation could play a critical role in determining the etch results. In Figure 8b, it can be observed that C₄H₂F₆ dissociation yields comparable densities for most major species compared with CHF₃ fractions. However, the densities of C₂H₂ and CHF₂ are significantly lower with C₄H₂F₆ than with CHF₃. Furthermore, it is found that the densities of major species are barely affected by RF power variation, suggesting the need for another control parameter to manipulate plasma conditions.

Figure 9 illustrates the etch results obtained with different HFC precursors (CHF₃ and C₄H₂F₆) at varying pressures from 10 mTorr to 30 mTorr, with a fixed RF power of 400 W. The resulting etch rate and selectivity for the CHF₃ and C₄H₂F₆ cases are summarized in the right-most column of Figure 9.

When etching with CHF₃, an interesting trend is observed. The ACL etch rate initially decreases and then rebounds as the mask faceting reduces, while the SiO₂ etch rate monotonically increases. This behavior leads to a minimum selectivity at the intermediate pressure of 20 mTorr. In contrast, when using C₄H₂F₆, ACL etching continuously increases with enhanced mask faceting as the pressure rises, resulting in significant bowing at the mask sidewall. At the same time, SiO₂ etch rate shows minimal change with increasing pressure, leading to a substantial decrease in etch selectivity.

These findings highlight the importance of pressure control in achieving desirable etch results. With CHF₃, an optimal pressure range can be identified to balance ACL etch rate and SiO₂ etch rate, thereby maximizing selectivity. On the other hand, C₄H₂F₆ exhibits challenges in maintaining selectivity as pressure increases, leading to distorted mask sidewalls and reduced etch performance.

Overall, understanding the pressure-dependent behavior of different HFC precursors is crucial for optimizing the etching process and achieving the desired etch profiles and selectivity.

Figure 10a illustrates the measured self-bias voltage (left y-axis) and the CFx/F (x = 1, 2, 3) ratio (right y-axis) as a function of pressure for CHF₃ and C₄H₂F₆. It is observed that with increasing pressure, both CHF₃ and C₄H₂F₆ show no significant change in the self-bias voltages. However, it is important to note that CHF₃ consistently provides higher absolute self-bias voltages compared with C₄H₂F₆ across a wide range of pressures. This indicates the ion bombardment during etching is stronger with CHF₃ than with C₄H₂F₆ owing to higher ion density and energy. This difference in ion bombardment strength may be linked to the higher SiO₂ etch rate observed with CHF₃ compared with C₄H₂F₆.

In Figure 10b, the signal of various radical species resulting from the dissociation of CHF₃ and C₄H₂F₆ is presented as a function of increasing pressure. As previously observed in Figure 8b, carbon-rich radical species containing multiple carbon atoms in their chemical structure are significantly observed with C₄H₂F₆, while they are barely present with CHF₃. Such carbon-rich species are believed to react with ACL, enhancing its etching rather than passivating the ACL surface. Additionally, the key etchants of FC plasmas, CF₂ and CHF₂, are observed to exhibit reversed relative densities depending on the precursor used. In the case of CHF_3, the CF₂ signal is lower than that of CHF₂, whereas CF₂ exceeds CHF₂ with C₄H₂F₆. This reversal in CF₂ and CHF₂ densities between the two precursors would be useful in selective etching of SiO₂ over Si₃N₄ [47].

3.3. Evaluation of the Greenhouse Effect

Overall, our results demonstrate that the conventional FC gases of C₄F₈ and CHF₃ exhibit a more vertical etch profile and a higher SiO₂ etch rate across different power and pressure conditions. On the other hand, the alternative gases of C₆F₆ and C₄H₂F₆ exhibit improved etch selectivity. However, when evaluating the etch performance of low GWP gases, it is crucial to assess the reduction in GHG emissions achieved by their implementation.

To quantify the environmental impact, we employ the million metric tons of carbon dioxide equivalents (MMTCDE) metric, which allows for a comparison of GHG emissions based on their GWP. This metric is particularly useful when a GHG is emitted in significantly smaller quantities compared with the reference amount of 1 ton. The carbon dioxide equivalent for a gas is derived as below.

$$MMTCDE=\frac{Q\times GW{P}_{100}}{{10}^9}$$

(1)

In Equation (1), Q refers to the mass of a GHG emitted, measured in kg. To calculate Q for each PFC or HFC gas, we multiply the flow rate of the FC gases used for etching (10 sccm) by the processing time (10 min) to obtain the volumetric amount of the gas used. This value is then converted to mass in kg, taking into account the atomic mass unit of the gas. It should be noted that, in calculating the MMTCDE, we employ the mass of mother gas species, such as C₄F₈ and CHF₃, rather than their fractions, including CF and CF₂, that are observed by the RGA. This is based on a presumption that such reactive species would have relatively short lifetimes and thus have less affect global warming at the atmosphere compared with the chemically stable non-reactive molecules.

Figure 11 illustrates a comparison of the estimated MMTCDEs for the conventional gases (C₄F₈ and CHF₃) and the alternative gases (C₆F₆ and C₄H₂F₆). The MMTCDE comparisons under the other experimental conditions show a nearly identical result so that we introduce only the result under etching conditions of 20 mTorr pressure and 400 W RF power for succinctness. Notably, the MMTCDEs associated with the alternative gases are approximately three orders of magnitude lower than those with the conventional gases, despite the etch results showing little difference. This observation suggests that the investigated alternative gases, C₆F₆ and C₄H₂F₆, hold significant promise as candidates for reducing GHG emissions in the semiconductor industry.

It should be noted, however, that the above MMTCDE evaluation may not be sufficient since it is conducted only in the processing chamber, not at the exhaust line of the chamber. Even though low-GWP precursors such as C₆F₆ and C₄H₂F₆, which are investigated in this work, are utilized in plasma processes, undesired high-GWP molecules such as C₂F₆ with 11,000 GWP can be created and exhausted into the atmosphere, which may reduce the benefits from the low-GWP precursors employment. Additionally, the MMTCDE evaluation with QMS could also include a certain level of inaccuracy due to the dissociative ionization by high-energy electron impacts. Thus, evaluating MMTCDE at the exhaust line of the vacuum chamber with non-invasive diagnostic tools would lead to more rigorous analysis on the global warming effects.

4. Concluding Remarks

In this study, we have explored the etch performance of C₆F₆ and C₄H₂F₆ as alternatives to the conventional etch gases, C₄F₈ and CHF₃, respectively, with the goal of reducing GHG emissions during plasma etching. Given the significantly lower GWP of the alternative gases, it is encouraging to investigate whether they can deliver comparable etch performance to the conventional gases.

To evaluate their performance, we conducted etching experiments using gas mixtures of different FC gases under various conditions of RF power and pressure. The obtained results were carefully analyzed using SEM images. Specifically, we compared the etch results of C₄F₈ with those of C₆F₆, and the etch results of CHF₃ with those of C₄H₂F₆.

Our findings indicate that the alternative gases demonstrated comparable etch performance to the conventional gases across various conditions. This suggests that the alternative gases hold promise for utilization in the next-generation etch processes with reduced GHG emissions. Importantly, our MMTCDE estimation revealed that replacing the conventional gases with the alternative gases could result in GHG emissions reduction by several orders of magnitude.

The outcomes of this study are expected to contribute to the wider adoption of low-GWP gases in semiconductor manufacturing. By incorporating these alternative gases into the etch processes, the industry can effectively reduce its environmental impact while maintaining the desired etch performance. Further research and development efforts should be directed towards optimizing the use of low-GWP gases and promoting their integration into commercial semiconductor fabrication processes.