Plasma Etching of SiO₂ Contact Holes Using Hexafluoroisopropanol and C₄F₈

Abstract

This study presents the feasibility of the use of hexafluoroisopropanol (HFIP) as a substitute to perfluorocarbon (PFC) for the plasma etching of SiO₂ to confront the continuous increase in demand for PFC emission reduction. SiO₂ etching is conducted in HFIP/Ar and C₄F₈/Ar plasmas, respectively, and its characteristics are compared. The SiO₂ etch rates in the HFIP/Ar plasma are higher compared with those in the C₄F₈/Ar plasma. The thickness of the steady-state fluorocarbon films formed on the surface of SiO₂ are lower in the HFIP/Ar plasma compared with in the C₄F₈/Ar plasma. Higher SiO₂ etch rates and thinner fluorocarbon films in the HFIP/Ar plasma are attributed to the oxygen atoms in HFIP, which generate oxygen radicals that react with the fluorocarbon films to turn into volatile products. Due to the higher dissociation of C-F bonds in CF₄ compared with in HFIP, the etch rates of SiO₂ in the C₄F₈/Ar plasma increase more rapidly with the magnitude of the bias voltage compared with those in the HFIP/Ar plasma. The etch profiles of the 200 nm diameter SiO₂ contact holes with an aspect ratio of 12 show that fairly anisotropic SiO₂ contact hole etching was achieved successfully using the HFIP/Ar plasma.

1. Introduction

Selective etching of the desired area with high aspect ratios is vital for developing next generation ultra-large-scale integrated devices. In particular, high-aspect-ratio SiO₂ contact hole etching is one of the critical processes for the fabrication of multilevel interconnects [1]. Fluorocarbon-containing plasmas are widely used for the etching of SiO₂ because fluorocarbon polymer films are piled up on the surface of the substrate in the fluorocarbon-containing plasmas, which protects the sidewalls of the contact holes from etching, thereby, resulting in anisotropic etch features.

The fluorocarbon species primarily used for the SiO₂ etching are perfluorocarbons (PFCs), such as CF₄ and C₄F₈. [2,3,4]. PFCs are known to be environmentally unfavorable because they have high global warming potential (GWP) as well as a long atmospheric lifetime [5,6]. According to the sixth assessment report of the Intergovernmental Panel on Climate Change, GWP₁₀₀ values (defined as a relative GWP value for the period of 100 years with respect to CO₂) of CF₄ and C₄F₈ are 7380 and 10,200, respectively.

The United Nations Framework Convention on Climate Change adopted the Kyoto Protocol in 1997 and designated CO₂, CH₄, NO₂, hydrofluorocarbons (HFCs) and SF₆ as greenhouse gases as well as PFCs. With continuing efforts to confront climate change issues, the UNFCCC drew up the Paris agreement in 2015 stating that each country should set its own goals for reducing greenhouse gases and implement them, thus, resulting in a continuous increase in the demand for PFC emission reductions.

To reduce PFC emissions, several strategies, such as alternatives, process optimization, recycling/recovery and abatement, have been attempted. The use of alternative materials to PFCs is thought to reduce PFC emissions by employing environmentally benign chemistries with low GWPs. Process optimization through increasing the efficiency of PFC-consuming processes can reduce PFC usage and, correspondingly, PFC emissions. Recycling/recovery and abatement of PFCs from exhaust streams can also control PFC emissions. Among these, the use of lower-GWP alternatives to PFCs would be promising for reducing the emissions of high-GWP etchants.

Earlier studies on the use of PFC alternatives focused on unsaturated fluorocarbons [7] and hydrofluorocarbons [8] because the unsaturated sites and the hydrogen atoms lead to an increase in the reactivity of the molecule to hydroxyl group. Recently, other studies in chemistries, such as hydrofluoroethers and perfluoroalkyl vinyl ethers were investigated as alternatives to PFCs, and their etch characteristics were reported [9,10,11,12].

It was suggested that the presence of oxygen atoms in either hydrofluoroethers or perfluoroalkyl vinyl ethers would make oxygen species (oxygen radicals and ions) available in the plasma, which serve the substrate (e.g., SiO₂) etching [13]. Considering the role of oxygen atoms, fluorinated alcohols (perfluorinated or partially fluorinated alcohols) are also worth studying as an alternative to PFCs. However, limited reports on the use of fluorinated alcohols for plasma etching were found.

In this work, hexafluoroisopropanol (HFIP) was examined for the plasma etching of SiO₂, and its etch features were compared with those in C₄F₈. HFIP is partially fluorinated alcohol, whose GWP is ~190, which is notably lower compared with those of PFCs. The etching features of SiO₂ were explored by varying the bias voltage. The thickness of the steady-state fluorocarbon film piled on the surface of the substrate and the optical emission intensity of fluorine radicals were investigated to explain the etch characteristics. The etching of a SiO₂ contact hole with a diameter of 200 nm and an aspect ratio of 12 was also conducted in each plasma.

2. Materials and Methods

2.1. Etching Method

Figure 1 shows an inductively coupled plasma (ICP) apparatus for the etching of SiO₂. The planar induction coil was positioned on top of a dielectric window (quartz). A plasma was ignited in a reaction chamber through applying a 13.56-MHz radio-frequency (RF) power (source power) to the coil. Another separate 13.56-MHz RF power (bias power) was applied to the electrode to independently control the bias voltage of a sample on the electrode from the source power. The plasma chamber was made of stainless steel. The electrode temperature was controlled using a chiller. A bias voltage applied to the electrode was monitored on the readout unit of the bias power generator.

Either HFIP or C₄F₈ was used as the discharge medium.

Table 1. The properties of HFIP and C₄F₈.

Name	Hexafluoroisopropanol (HFIP)	Octafluorocyclobutane
Molecular formula	C3H2F6O	C4F8
Structural formula
Boiling point (°C)	59	−5.8
GWP	190	9540

shows the properties of HFIP and C₄F₈. HFIP had a boiling temperature of 59 °C. Thus, it was contained in a canister because it was a liquid at a room temperature. HFIP was vaporized by heating the canister at 75 °C to introduce it to the plasma chamber in the gaseous phase. The piping between the canister and the plasma chamber was also heated by a heating jacket to minimize the residual liquid droplets. HFIP and C₄F₈ were introduced to the reaction chamber with Ar, respectively.

SiO₂ etching was performed in HFIP/Ar and C₄F₈/Ar plasmas, respectively. The pressure in the plasma chamber was 10 mTorr, and the source power was set to 250 W. The bias voltage ranged from −400 to −1200 V. The flow rates of HFIP and C₄F₈ were 10 sccm. The Ar flow rate was 20 sccm in both HFIP/Ar and C₄F₈/Ar plasmas, resulting in total flow rates of 30 sccm in both plasmas.

Etching was conducted using either a blanket or a patterned sample. The blanket sample was a p-type Si substrate covered with a 500 nm thick SiO₂ film, and it was used to obtain the etch rates of SiO₂. The sample was in a rectangular shape (10 mm long and 5 mm wide), and it was placed on the electrode. A Si substrate consisting of a SiO₂ film with a 200 nm diameter hole pattern was also used as a pattern sample for SiO₂ contact hole etching.

The top and cross-sectional SEM images of the patterned sample are shown in Figure 2. A 2400 nm thick SiO₂ layer was thermally grown on a Si substrate. Then, a 1350 nm thick amorphous carbon layer (ACL) was deposited on the SiO₂ layer by chemical vapor deposition, and a hole pattern with a 200 nm diameter was delineated. The patterned samples were provided by the Korea Semiconductor Industry Association through the pattern-wafer support program.

2.2. Materials Characterization

The etch rates of the blanket samples were acquired using a thickness meter (model SpectraThick 2000-Deluxe)(K-MAC, Daejeon, Korea), which measures the thickness of the SiO₂ film on the Si substrate. The thickness of the steady-state fluorocarbon film piled on the SiO₂ surface was analyzed through X-ray photoelectron spectroscopy (XPS, ThermoFisher, K-Alpha, Waltham, MA, USA.), which had an X-ray source of 1486.6 eV and was generated from a moveable Al node at 15 kV. The relative amount of radicals in the plasma was obtained utilizing optical emission spectroscopy (OES, GetSpec, 2048TEC, Dresden, Germany). The etch features of the patterned samples were obtained using field-emission scanning electron microscopy (FE-SEM, Hitachi, S-4800, Krefeld, Germany).

3. Results and Discussion

Figure 3 shows the SiO₂ etch rates in HFIP/Ar and C₄F₈/Ar plasmas as a function of the bias voltage. The etch rates in both HFIP/Ar and C₄F₈/Ar plasmas monotonically increased with the magnitude of the bias voltage. This behavior is a typical characteristic of plasma etching because the energy of the incident ions bombarding the substrate increases with the magnitude of the bias voltage. The SiO₂ etch rates in the HFIP/Ar plasma were faster compared with those in the C₄F₈/Ar plasma under all etching conditions.

Higher etch rates in HFIP/Ar plasmas compared with in C₄F₈/Ar plasmas can be attributed to the different chemistry of the discharge materials. HFIP contains oxygen atoms, while C₄F₈ does not contain them, meaning that oxygen radicals are only formed in the HFIP/Ar plasma. A thin fluorocarbon film (a so-called steady-state fluorocarbon film) is subject to being piled on the surface of the substrate during etching in fluorocarbon-containing plasmas, such as HFIP and C₄F₈.

Oxygen radicals react with the fluorocarbon films into volatile reaction products, such as CO, CO₂ and COF₂. Therefore, the steady-state fluorocarbon films on the surface of SiO₂ would be thinner in the HFIP/Ar plasma compared with in the C₄F₈/plasma. Since fluorocarbon films act as etch resistance against radicals and ions during etching, SiO₂ is more vulnerable to the etching reaction in the HFIP/Ar plasma compared with in the C₄F₈/plasma. Therefore, the SiO₂ etch rates in HFIP/Ar plasmas were faster compared with those in C₄F₈/Ar plasmas.

The characteristics of the steady-state fluorocarbon films in each plasma were analyzed to support this explanation. Figure 4 shows the thickness of the steady-state films piled on the SiO₂ surfaces at bias voltages of −400, −800 and −1200 V in HFIP/Ar and C₄F₈/Ar plasmas, respectively. The thickness of the steady-state fluorocarbon film piled on the SiO₂ surface was obtained using the silicon 2p X-ray photoelectron spectra before and after etching [14]. The thickness of the fluorocarbon film (d) is expressed as

$$\mathrm{d}=\mathsf{\lambda } \cos \mathsf{\varphi } \ln \frac{I_0}{I}$$

(1)

where I and I₀ are the intensities of the Si 2p peak in the presence and absence of the fluorocarbon film, respectively. λ is the inelastic mean free path of Si 2p electrons in the fluorocarbon film and assumed to be 2.5 nm [14]. ϕ is the angle of electron emission with respect to the surface normal, and it was 45° in this study.

The film thickness decreased with increasing the magnitude of the bias voltage in both HFIP/Ar and C₄F₈/Ar plasmas. The thickness of the steady-state fluorocarbon film in the HFIP/Ar plasma was lower than that in the C₄F₈/Ar plasma at all bias voltages. As mentioned above, the fluorocarbon film piled on the surface of the substrate acts as an etch resistance during etching in fluorocarbon-containing plasmas, meaning that as the thickness of the steady-state film on the surface of the substrate decreases, the substrate is more vulnerable to the etching reaction. Thus, the SiO₂ etch rates were faster in the HFIP/Ar plasma.

Figure 4 also shows that the difference in the thicknesses of the steady-state fluorocarbon films between HFIP/Ar and C₄F₈/Ar plasmas decreased with the increase in the magnitude of the bias voltage. This behavior is similar to the difference in the SiO₂ etch rates between the two plasmas (see Figure 3). It was 2580 Å/min at a bias voltage of −400 V, monotonically decreased with the increasing magnitude of the bias voltage and was 1300 Å/min at a bias voltage of −1200 V. Therefore, we concluded that the thickness of the steady-state fluorocarbon film piled on the surface of SiO₂ directly affects the etch rate of SiO₂ in HFIP/Ar and C₄F₈/Ar plasmas.

One can argue why the SiO₂ etch rates in the C₄F₈/Ar plasma increased more rapidly with the magnitude of the bias voltage compared with those in the HFIP/Ar plasma. This can be attributed to the difference in the C-F bond energy between HFIP and C₄F₈. C₄F₈ comprises carbons forming a cyclic ring, and the C-F bond energy in C₄F₈ is 4.4 eV [15]. Comparatively, carbons are linearly connected in HFIP, and the C-F bond in the linear carbons is usually stronger than that in the cyclic carbons.

For example, the bond energies of C-F in CF₄, C₂F₆ and C₃F₈ are 5.6, 5.4 and 5.3 eV, respectively [16]. The C-F bond energy in HFIP is expected to be similar to that in C₃F₈ since HFIP consists of three linearly-located carbons although they are not exactly the same [17]. As the bias voltage increases, the dissociation of C-F bonds is subject to increase in both HFIP and C₄F₈.

Since the C-F bond energy in C₄F₈ is lower than that in HFIP, the dissociation of C-F bonds occurs more vigorously in C₄F₈ than in HFIP, thereby, resulting in a higher increase in the formation of fluorine radicals (which are responsible for SiO₂ etching by producing volatile reaction products, such as SiF_x and COF₂) in C₄F₈/Ar plasmas compared with in HFIP/Ar plasmas. Thus, the SiO₂ etch rates increased more rapidly with the magnitude of the bias voltage in the C₄F₈/Ar plasma compared with those in the HFIP/Ar plasma.

A higher increase in the production of fluorine radicals when increasing the magnitude of the bias voltage in C₄F₈/Ar plasmas compared with in HFIP/Ar plasmas was investigated using OES. Figure 5 shows the optical emission intensities of the fluorine radicals generated in HFIP/Ar and C₄F₈/Ar plasmas, respectively, as a function of the bias voltage. Fluorine radicals were identified as lines at 685.6 and 703.7 nm [18].

The fluorine-radical intensities in the HFIP/Ar plasma were greater compared with those in the C₄F₈/Ar plasma. As the magnitude of the bias voltage increased, the intensity of fluorine radicals increased slightly in the HFIP/Ar plasma and increased rapidly in the C₄F₈/Ar plasma. Therefore, the OES measurements support the higher increase in the SiO₂ etch rates with the magnitude of the bias voltage in the C₄F₈/Ar plasma compared with those in the HFIP/Ar plasma.

Figure 6 shows the SEM images of the 200 nm diameter SiO₂ contact holes etched in HFIP/Ar and C₄F₈/Ar plasmas at various bias voltages. The SEM images were taken after the fluorocarbon films were removed by ashing. For SEM measurements, the samples were coated with Pt for 20 s. In all cases, etching was done for 12 min. As expected, the etch depth of the SiO₂ contact hole increased with the magnitude of the bias voltage in both HFIP/Ar and C₄F₈/Ar plasmas.

Furthermore, the etch depths of the contact hole in the HFIP/Ar plasma were greater compared with those in the C₄F₈/Ar plasma at all bias voltages. This agrees with the results from higher SiO₂ etch rates in HFIP/Ar plasmas compared with those in C₄F₈/Ar plasmas (see Figure 3). At bias voltages higher than −800 V, SiO₂ contact holes with an aspect ratio of 12 were completely etched in the HFIP/Ar plasma.

Comparatively, even at the highest magnitude of the bias voltage (−1200 V) used in this study, SiO₂ contact holes were not etched through in the C₄F₈/Ar plasma. In the HFIP/Ar plasma, sidewall necking below the top of the hole was improved by increasing the magnitude of the bias voltage, and fairly anisotropic SiO₂ contact hole etching was successfully achieved at a bias voltage of −1200 V. Finally, the use of HFIP (whose GWP is significantly lower than that of PFC) for the plasma etching of SiO₂ contact holes was realized.

4. Conclusions

HFIP was examined for the plasma etching of SiO₂, and its etch characteristics were compared with those in C₄F₈. The SiO₂ etch rates in the HFIP/Ar plasma were faster compared with those in the C₄F₈/Ar plasma under all etching conditions used in this study. This was attributed to oxygen atoms in HFIP, which produced oxygen radicals. Since oxygen radicals react with the fluorocarbon film piled up on the surface of the substrate into volatile products, the steady-state fluorocarbon films on the surface of SiO₂ were thinner and more vulnerable to the etching reaction (a correspondingly higher SiO₂ etch rate) in the HFIP/Ar plasma compared with in the C₄F₈/Ar plasma.

XPS analysis confirmed that the thicknesses of the steady-state fluorocarbon films piled on the surface of SiO₂ in the HFIP/Ar plasma were lower compared with those in C₄F₈/Ar plasma. The etch rates in both HFIP/Ar and C₄F₈/Ar plasmas monotonically increased with the magnitude of the bias voltage. However, due to the higher dissociation of the C-F bonds in CF₄ compared with in HFIP, the SiO₂ etch rates in the C₄F₈/Ar plasma increased more rapidly with the magnitude of the bias voltage than did those in the HFIP/Ar plasma.

The etch profiles of the 200 nm diameter SiO₂ contact holes showed that the etch depths of the hole in the HFIP/Ar plasma were deeper compared with those in the C₄F₈/Ar plasma at all bias voltages, thus, agreeing with the results from the higher SiO₂ etch rates in HFIP/Ar plasmas compared with those in C₄F₈/Ar plasmas. In the HFIP/Ar plasma, the 200 nm diameter SiO₂ contact holes with an aspect ratio of 12 were completely etched at bias voltages greater than −800 V, and fairly anisotropic SiO₂ contact hole etching was achieved successfully at a bias voltage of −1200 V. Finally, this study realized the use of HFIP (whose GWP is significantly lower than that of PFC) for the plasma etching of SiO₂ contact holes.