Etching of Si₃N₄ induced by electron beam plasma from hollow cathode plasma in a downstream reactive environment

Angstrom level precision etching processes are essential for the fabrication of semiconductor devices with nanometer critical dimensions.^1–4 Traditional plasma-based dry etching techniques utilize ion bombardment to achieve material removal either physically⁵ or by synergistic interactions with added reactive chemical species.^6–15 We have developed a new electron-based etching technique utilizing a hollow cathode (HC) plasma discharge for beam electron extraction with low and adjustable electron energy for surface etching. Briefly, as discussed in our previous work,¹⁶ the work of other electron beam plasma etching research groups studying electron beam-induced plasma and plasma etching differs from the present work in several respects. In most of the studies, the electron beam is parallel to the wafer surface, and the carriers of energy/momentum that is transferred to the material surface are still ions based on RF biasing.^17,18 The present work differs in that it uses (1) beam electrons traveling directly toward the surface to activate surface reactions and (2) a chemically reactive environment generated by an additional remote plasma source that allows great flexibility in terms of nature of the chemical radical fluxes to the surface. In contrast, other work that uses direct exposure of a surface to energetic electrons to initiate etching reactions is mostly based on the use of readily available precursors. An example is XeF₂ used in combination with an electron beam.¹³ The current approach is based on an argon HC discharge where the acceleration voltage is employed to accelerate beam electrons toward a material surface in a reactive remote plasma environment that supplies reactive radicals to the surface. This hollow cathode electron beam (HCEB)/remote plasma etching system enables operation in a relatively high-pressure (approximately mTorr) reactive gas environment and permits electron enhanced chemical etching. The HCEB source is fundamentally different from traditional field emission electron beam (EB) sources operating in a high vacuum nonreactive environment. For the HCEB source, the interaction of electrons with the downstream gaseous environment changes their energies. The present article provides results of a systematic parametric study and improves the understanding of this system.

This article is organized as follows: in the first part, the authors discuss the precursor composition dependence of Si₃N₄ etching in a CF₄/O₂ remote plasma etching environment and identify process conditions for which the Si₃N₄ etching rate (ER) can be minimized. Additionally, the electron enhanced etching of Si₃N₄ with varying acceleration voltages is studied. In the second part, the authors study electron enhanced chemical surface etching of Si₃N₄ for the conditions where the HCEB source is combined with the effluent of the CF₄/O₂ remote plasma source. The impacts of varying acceleration voltages, discharge current, main chamber pressure, and operation distance are examined using optical emission for discharge characterization and ellipsometry for surface etching. In the last part, the authors report optimized process conditions and demonstrate electron controlled Si₃N₄ etching.

The experiments were performed in an electron etching system, consisting of a hollow cathode (HC) EB source on top of a reaction chamber connected to an electron cyclotron wave resonance (ECWR) remote plasma source that supplies reactive species on the side [Fig. 1(a)]. This system has been described in a previous publication.¹⁹ The base pressure achieved before processing was in the 1 × 10⁻⁶Torr range. The pressure in the HC source (approximately Torr level) is significantly higher than the chamber pressure below because of a choked flow connection. The pressure in the reaction chamber during processing was kept at 2–100 mTorr using a butterfly valve in the pumping line. The ECWR remote plasma source consists of a COPRA DN160 ECWR plasma beam source running at 13.56 MHz radio frequency with a modified neutralization plate.²⁰ A 400 W source power level and 1.8 mT static magnetic fields were applied in the ECWR work for all experiments requiring the injection of reactive radicals.²⁰ The CF₄/O₂ remote plasma is a well-developed method to generate reactive neutral radicals (F radical) for Si₃N₄ etching.²¹ 

Argon gas was admitted to the HC discharge using a mass flow controller. The HCEB source to substrate distance can be controlled with an adjustable bellows tube over a range of 70 mm. A direct current (DC) power supply providing constant discharge current (I_D) was used to generate the main HC plasma discharge between the cathode and the source anode. The diameter of the anode hole is 1.57 mm. The detailed beam current characterization is covered in Sec. III A of Ref. 16. A second DC power supply provided an acceleration voltage (V_E) to the anode plate to accelerate a beam of electrons through the center hole (10 mm diameter) of the grounded acceleration plate toward the grounded substrate. The acceleration plate to HC anode plate distance is 10 mm.

The equivalent circuit diagram is shown in Fig. 1(b). The arrows depict the direction in which electrons move and is opposite to the conventional current direction. The HC discharge provides an electron current I_D. Typical EB source operating conditions are 10 SCCM Ar carrier gas flow for a constant 1.5 A discharge current (I_D). The net DC current measured at the grounded substrate is denoted with I_B.

The beam electron and the secondary plasma generated have been characterized using a Langmuir probe (LP) provided by plasma sensors. The LP used in this study is designed to collect highly accurate measurements of the electron energy probability function (EEPF), electron density (n_e), and electron temperature (T_e) by fast sweep speeds, ion bombardment cleaning, and electron heating.^19,22,23

The Ar emission at wavelengths of 750.39 nm was measured by optical emission spectrometer (OES) as the energy of the extracted BE was varied. F and O emissions were also monitored at wavelengths of 703.75 nm and 777.19 nm, respectively.

Sample surface etching and modification were monitored in real time with an in situ ellipsometer setup that used an HeNe laser (λ = 632.8 nm). The etch rate for each condition is measured for at least 30 s.

After processing, selected samples were transferred under vacuum to a Vacuum Generators ESCALAB MK II surface analysis system for x-ray photoelectron spectroscopy (XPS) measurements. High-resolution scans of the Si2p, C1s, O1s, N1s, and F1s binding energy regions were obtained at 20 eV pass energy at an electron take-off angle of 90° and 20° with respect to the sample surface. The spectra were fitted using least-squares fitting routines with 70/30% Gaussian/Lorentzian line shapes at fixed full width at half maximum for each spectrum after Shirley background subtraction.^9,12,14

As a basis of studying the electron enhanced chemical etching, Si₃N₄ surface etching in the remote plasma environment without injection of beam electrons was studied. Remote plasma etching results for Si₃N₄ as a function of varying CF₄/O₂ gas flow ratio are shown in Fig. 2. Operating conditions were variable CF₄/O₂ gas mixtures with a total gas flow rate of 50 SCCM into the ECWR chamber, a source power of 400 W, and 1.8 mT DC magnetic fields. Simultaneously, 60 SCCM Ar gas flowed through the HCEB system, which protected the HC system from the reactive radicals generated by the remote plasma and provided consistent conditions in the main chamber relative to the HCEB operation. The main chamber pressure is held at 20 mTorr, which is higher than the pressure used for HCEB etching (3.5 mTorr) and enhances the etching rates due to the remote plasma only. Figure 2 shows that the Si₃N₄ etching rate (ER) initially increases with precursor O₂ concentration added to CF₄/O₂ and peaks at 20% O₂. This classic effect is due to oxidation of the CF₄ gas, which provides more F atoms for material removal. With more O₂ in the precursor gas mixture, the F will be diluted and the surface reacted F will drop.²⁴ Gas phase F and O densities were evaluated using optical emission spectroscopy. For this, F and O emission intensities were normalized relative to Ar emission under the same conditions. The O emission increases with O₂ flow, whereas the F emission shows a peak at 20% O₂ concentration, which is consistent with the surface reacted F trend. However, at 40% O₂ concentration the surface reacted F rate drops faster than gas phase F, indicating that a passivation or oxidation layer on the surface suppresses the etching of Si₃N₄. At 80% O₂ concentration, the Si₃N₄ etching rate has dropped to the baseline level even though there is still available gas phase F indicated by the OES measurements. The lack of Si₃N₄ etching can be explained by oxygen-related surface passivation. This is the optimal condition to study for electron enhanced etching, since for this situation spontaneous etching due to reactive radicals produced by the remote plasma has become negligible, whereas the electrons can be used to activate surface etching with residual F atoms still available.

To further confirm the surface passivation, Si₃N₄ samples after remote plasma etching using different O₂/O₂ + CF₄ processing gas ratios were transferred under vacuum into a surface analysis system and analyzed by XPS. Figure 3 shows several important surface chemistry data obtained by XPS analysis of Si₃N₄ processed using different O₂ concentrations in the remote CF₄/O₂ plasma. A comparison with data obtained with a pristine Si₃N₄ sample is also shown. The electron emission angle is 20° relative to the sample surface, which enables an analysis of the top 2–3 nm from the surface. With the addition of O₂ in plasma, there is no surface carbon contamination shown in C1s spectra. An SiOF layer was observed in Si2p spectra, which is the oxidation passivation layer as we discussed above.

By applying an attenuation model to SiN intensities in Si2p spectra with two different probe angles, the thickness of the surface SiONF layer can be obtained.²⁵ The Si₃N₄ ERs are plotted versus angle-resolved XPS (AR-XPS) passivation layer thicknesses (Fig. 4). The data show that as the passivation layer thickness increased, the Si₃N₄ ER is reduced. Again, the 80% O₂ concentration condition for CF₄/O₂ shows baseline level etching with sufficient SiONF passivation layer thickness. This fairly thick passivation layer which still contains F atoms is a promising basis of applying the HCEB to induce a surface reaction. The F atoms in this layer are expected to react and produce surface etching as energy is delivered to this surface upon beam electron injection.

Based on the background remote plasma results, two representative CF₄/O₂ gas flow ratios were studied in HCEB enhanced etching conditions. The electron etching condition is the standard condition for most cases reported in our previous publication.¹⁶ To reduce the interaction of the beam electrons with the gas atoms and molecules, for this work the pressure in the main chamber was reduced to 3.5 mTorr, which required to reduce both the Ar and CF₄/O₂ flows accordingly. The Ar flow rate in the HC is 10 SCCM with a discharge current of 1.5 A. The source to sample distance is 5 cm. The CF₄/O₂ gas mixture with a total gas flow rate of 10 SCCM is injected into ECWR chamber with a source power of 400 W and a DC magnetic field of 1.8 mT. The main chamber pressure was held at 3.5 mTorr for all experiments.

Figure 5(a) shows the ER results by varying the acceleration voltage for two CF₄/O₂ gas mixtures using either 20% or 80% O₂. As described in previous publication,¹⁶ when V_E is greater than the gas ionization threshold, plasma-related optical emission can be observed along with an enhancement of the substrate etching rate. For the 20% O₂ concentration case, the Si₃N₄ ER is over 5 times greater than for 80% O₂ with some spontaneous etching without V_E. This enhanced ER indicates that for the 80% O₂ condition, the F etchant density is the limiting factor for surface etching. By increasing the F radical concentration, we could dramatically increase the Si₃N₄ ER. The F and O emission intensities normalized by the Ar emission in the HCEB secondary plasma conditions are shown in Fig. 5(b). For the same gas composition, the F and O emissions are unchanged with V_E. Only the gas composition can change the radical density in the chamber environment.

As discussed in our previous publication,¹⁶ while the exact surface interaction mechanism by which the injected electrons initiate substrate etching is unclear at this time, e.g., the role of beam electrons versus energetic ion bombardment from the secondary plasma, it is clear that the root cause of the observed etching is the injection and acceleration of the beam electron.

In the following studies, we only focus on the F deficient condition (80% O₂ concentration) for which spontaneous etching becomes negligible. We use Ar emission to characterize the plasma density.

The corresponding EEPF in 80% O₂ concentration condition with varying V_E measured by the Langmuir probe (LP) is shown in Fig. 6. As discussed in a previous publication,¹⁶ the LP EEPF result for HCEB condition consists of two major peaks: one peak located at lower energy represents electrons due to the beam electron-induced plasma with fairly low average electron temperature; the peak at a high electron energy that shifts strongly with V_E and corresponds to the anisotropic primary beam electron with low to no in-elastic collisions. The beam electron peak energy is proportional to the acceleration voltage V_E. In Fig. 6, the corresponding beam electron energy increases with V_E from 40 to 80 V with a similar peak area. This finding agrees with the results reported for the Ar HCEB in our previous publication.¹⁶ 

By comparing the LP data with the ER data shown in Fig. 5(a) at V_E in 60–80 V range, we note that the ER for the CF₄with 80% O₂ case has reached a saturation value and does not increase further even though the beam electron energy increases. This confirms the conclusion above, that the etching rate in that acceleration voltage range (60–80 V) is limited by the F etchant. Conversely, for the CF₄ with 20% O₂ case, which is characterized by a high gas phase F density, the Si₃N₄ ER continues to increase in the same acceleration voltage range. In Sec. III C, the other HCEB parameters will be studied in the same fashion with varying acceleration voltages.

To further characterize the behavior of HCEB enhanced etching, the Ar emission intensity and Si₃N₄ ER at various acceleration voltages were evaluated for the other HCEB source parameters, i.e., discharge current (I_D), pressure in the main chamber, and acceleration plate to substrate distance. The Ar emission intensity characterizes excited Ar atoms and our related publication¹⁶ shows that this behaves very similar to HCEB induced plasma density. The variation of Ar emission intensity and Si₃N₄ ER as a function of acceleration voltage (V_E) with three different discharge currents (I_D) are shown in Figs. 7(a) and 7(b), respectively. With increased discharge current I_D [Fig. 7(a)], the Ar emission intensity increases for acceleration voltages greater than 50 V. This indicates that the HCEB induced secondary plasma density increases with the total electron current controlled by I_D, consistent with the expectations. In contrast to the Ar emission intensity, the Si₃N₄ ER increases as I_D was changed from 1 to 1.5 A at the same V_E, but then stays the same when I_D is increased further to 2 A. On the other hand, the ER for the same I_D shows a saturation at high acceleration voltage regions, suggesting some other parameters limiting the ER other than V_E or beam electron energy.

To further investigate this phenomenon, Si₃N₄ ER is plotted versus Ar emission intensity at three different V_E (Fig. 8). It clearly shows that the Si₃N₄ ER at same V_E increases initially with Ar emission (or plasma density) and saturates at 2 A discharge current. This plot suggests that the ER depends on both V_E and plasma density, but that ultimately both plasma density and V_E are no longer limiting the etching rate for these conditions. This high V_E and plasma density regime likely corresponds to a limitation by the arrival of F radicals at the surface and corresponds to neutral limited etching as discussed in Fig. 5. Figure 8 strongly suggests that the delivery of both beam electron energy, flux and reactive F neutrals to the surface are required to enable the electron-neutral synergy etching.

The variation of Ar emission intensity and Si₃N₄ ER as a function of acceleration voltage (V_E) with three different main chamber pressures are shown in Figs. 9(a) and 9(b), respectively. With increased chamber pressure, the ionization in the main chamber increases and leads to an increase of Ar emission. The EEPF data obtained by LP measurements at high main chamber pressure conditions show an increase of the low-energy HCEB plasma electrons along with a drop of the high energy beam electrons. Since the low-energy HCEB plasma electron density and Ar emission intensity are both representing the plasma density, the increase of the low-energy electron distribution is consistent with the increase of the Ar emission intensity at high main chamber pressure condition. For the same V_E, the Si₃N₄ ER increases with pressure up to 10 mTorr and may be explained by the increase in plasma density, which causes more energy delivery to the Si₃N₄ surface, e.g., in the form of more low ion bombardment to the negatively charged dielectric substrate surface. At 20 mTorr, an increase of the Si₃N₄ thickness is seen, which could be related to redeposition. When the main chamber pressure increases to 20 mTorr, the high density of HCEB plasma will sputter the HC source metal and this causes metal deposition on the surface. Both of these high-pressure effects are unwanted for HCEB enhanced etching; thus, a low main chamber pressure is preferred.

As the distance between the HCEB source acceleration plate and the substrate electrode is varied (the acceleration plate to HC source anode distance is fixed), the relative importance of electron interactions with gas phase species or surface species changes. However, with the CF₄/O₂ processing gas, this change is relatively small, which may be due in part to electron attachment to the electronegative gas species from the CF₄/O₂ mixture. The attachment of low-energy electrons for process conditions that include CF₄/O₂ makes it more difficult to detect discharge differences for these conditions over the 5 cm range. This may explain why in this case a change of the acceleration plate - substrate electrode distance from 5 cm to 1.5 cm [Fig. 10(a)] leaves the Ar emission the same. However, the Si₃N₄ ER increases dramatically when the distance is as short as 1.5 cm [Fig. 10(b)]. In that case, the sample surface will collect more electrons and show higher ER results.

After investigating the impact of the precursor chemistry composition, acceleration voltage, discharge current, main chamber pressure and the operation distance, suitable conditions for HCEB enhanced etching were identified. To avoid spontaneous etching from background remote plasma, a high oxygen concentration in the CF₄/O₂ mixture is required to form an SiONF passivation layer on the Si₃N₄ surface. As the key parameter of ER control, the acceleration voltage needs to form an energetic beam electron with energy delivery to the reacted surface that is high enough to lead to observable surface etching. The discharge current is desired in the saturation region and enables sufficient energy delivery to the surface so that material etching is limited by other factors. In order to minimize secondary plasma production and ion bombardment impact from plasma, the main chamber pressure is preferred to be low but high enough so that the reactive radical delivery is sufficient. A small source to sample distance can minimize the electron loss to surrounding walls and enhance the beam electron etching efficiency. With the above considerations, suitable Si₃N₄ HCEB etching conditions can be realized using 80% O₂ in CF₄/O₂ mixture, 50 V acceleration voltage, 1.5 A discharge current, 3.5 mTorr main chamber pressure, and 2.5 cm source to sample distance. An exemplary Si₃N₄ thickness time evolution at this condition is shown in Fig. 11(a) before and after beam electron injection. No spontaneous etching is observed without energizing the electrons from the HC electron source. After applying an acceleration voltage V_E = 50 V, Si₃N₄ is etched at a steady level by electron-beam-reactive radial synergy effect with an etch rate of 6.18 nm/min. The normalized etched thickness distribution measured by ex situ ellipsometry and the electron saturation current measured for the same lateral position obtained using the Langmuir probe are shown in Fig. 11(b). Optical photographs of etched sample surfaces showing etched thickness distribution by optical interference are shown on the top part of Fig. 11(b). Both of the results suggest a strong correlation of the material etching with the beam electrons irradiation.

The surface chemistry of Si₃N₄ samples for the optimized etching conditions with and without HCEB activation has been investigated and selected results are shown in Fig. 12. Both Si₃N₄ samples were processed in the same low chamber pressure at 3.5 mTorr. The remote plasma condition was running at a higher HCEB source to sample distance at 5 cm in order to get higher clearance for sample vacuum transfer. Since the HCEB is not on in this case, this distant difference has no/little impact to the surface chemistry. Since the etched surface is not a uniform distribution, we only used the 90° photoelectron probe angle to minimize the area being analyzed. For the Si₃N₄ etching conditions used, the remote plasma chemistry effect has been reduced due to lower chamber pressure. The low HCEB source to sample distance generates locally a higher Ar partial pressure and dilutes the radical density near the sample surface. As a result, a traceable amount of carbon is shown in the C1s spectra relative to remote plasma only. The HCEB etched Si₃N₄ sample surface also shows a decreased XPS intensity of Si2p, N1s, and O1s and a stronger intensity in F1s relative to the remote plasma etched Si₃N₄. This surface chemistry change may indicate an electron-stimulated removal of the nitrogen and oxygen from the SiONF passivation layer with precursor attachment. It is consistent with the two-step etching mechanism reported in focused electron beam etching of SiO₂ that involves electron-stimulated desorption of oxygen from the SiO₂ matrix followed by the precursor-mediated removal of elemental silicon.²⁶ 

In this work, a systematic investigation of the parameter space available for an etching approach based on beam electron controlled Si₃N₄ etching has been performed by characterizing the combination of an HCEB source for energetic beam electron generation with a remote plasma source for reactive radical supply. The O₂ concentration of the CF₄/O₂ processing gas mixture strongly impacts the remote plasma etching rate of Si₃N₄. A SiONF passivation layer is formed on the surface for O₂-rich conditions, which reduces the Si₃N₄ ER in a simple remote plasma environment. Combining the HCEB source with the remote plasma, a strong precursor dependence has been observed in beam electron controlled etching. In this case, the acceleration voltage controls the energy of beam electrons and provides an effective method to control surface etching. At this time, the detailed sequence of events operative in the etching mechanism is unclear. While the increase of the electron energy is ultimately responsible for initiating surface etching, at present we cannot rule out a role of ions from the simultaneously produced secondary plasma in plasma-surface interaction mechanisms. The impact of other parameters, including discharge current, process chamber pressure, and HCEB source to sample distance, were studied by evaluating Ar optical emission and ER for various acceleration voltages. After combining a suitable set of process parameters, a localized beam electron controlled Si₃N₄ etching without spontaneous etching has been demonstrated with electron-neutral synergy effect with plasma-surface interaction.

The authors gratefully acknowledge financial support of this work by the Carl Zeiss SMS, and additional support by the U.S. Department of Energy (DOE) (No. DE-SC0001939). The authors thank Plasma Sensors for Langmuir Probe support. The authors thank A. Knoll, P. Luan, A. Pranda, K. Y. Lin, and A. Perruccio for helpful discussions and collaboration.