A material etching system was developed by combining beam electron injection from a direct current hollow cathode (HC) electron source with the downstream reactive environment of a remote CF4/O2 low temperature plasma. The energy of the injected beam electrons is controlled using an acceleration electrode biased positively relative to the HC argon discharge. For an acceleration voltage greater than the ionization potential of Ar, the extracted primary electrons can produce a secondary plasma in the process chamber. The authors characterized the properties of the secondary plasma by performing Langmuir probe measurements of the electron energy probability function (EEPF) 2.5 cm below the extraction ring. The data indicate the existence of two major groups of electrons, including electrons with a primary beam electron energy that varies as the acceleration voltage is varied along with low energy electrons produced by ionization of the Ar gas atoms in the process chamber by the injected beam electrons. When combining the HC Ar beam electron with a remote CF4/O2 electron cyclotron wave resonance plasma, the EEPF of both the low energy plasma electron and beam electron components decreases. Additionally, the authors studied surface etching of Si3N4 and polycrystalline Si (poly-Si) thin films as a function of process parameters, including the acceleration voltage (0–70 V), discharge current of the HC discharge (1–2 A), pressure (2–100 mTorr), source to substrate distance (2.5–5 cm), and feed gas composition (with or without CF4/O2). The direction of the incident beam electrons was perpendicular to the surface. Si3N4 and polycrystalline silicon etching are seen and indicate an electron-neutral synergy effect. Little to no remote plasma spontaneous etching was observed for the conditions used in this study, and the etching is confined to the substrate area irradiated by the injected beam electrons. The electron etched Si3N4 surface etching rate profile distribution is confined within a ∼30 mm diameter circle, which is slightly broader than the area for which poly-Si etching is seen, and coincides closely with the spatial profile of beam electrons as determined by the Langmuir probe measurements. The magnitude of the poly-Si etching rate is by a factor of two times smaller than the Si3N4 etching rate. The authors discuss possible explanations of the data and the role of surface charging.
Angstrom level precision etching processes are essential to scale down semiconductor devices to several 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.6,7 Recently, a new ion-based etching technique called atomic layer etching8,9 has been developed for use in semiconductor fabrication,10 which is based on temporal separation of chemical reactant supply and ion bombardment induced etching into two sequential steps. The ion bombardment energy is set to a level between the chemical enhanced etching energy threshold and physical sputtering energy threshold to ensure self-limited etching. These ion energies are much lower than used in traditional continuous wave plasma etching, thus reducing substrate damage and increasing the materials’ etching selectivity.11–14 However, when etching materials with low physical sputtering threshold energies, e.g., Si or SiGe, difficulties are encountered since the lowest possible ion energies15 are determined by both the plasma potential16 and sheath potential and typically above 15–20 eV for inductively coupled plasma.9 This causes difficulties with respect to achieving high etching selectivity for materials with low etching energy threshold, e.g., Si or SiGe,15 and the momentum transfer of ions causes damage and surface roughness of the etched material surface.17,18 An etching method with low energy and momentum transfer to a surface is attractive.
Electron induced etching is a potential approach for low damage selective etching applications.7,19–21 Compared to ion mass, the electron mass is negligible, which dramatically reduces the momentum transfer to the surface, leading to no surface displacement damage. Similar to ion assisted etching, electron assisted etching in a reactive gaseous environment was discussed by Coburn and Winters in their 1979 article where they used XeF2 to etch Si3N4, SiO2, and SiC.7 Focused electron beam (FEB) assisted etching was reported later using XeF2 precursor to etch Si, SiO2, Si3N4, TaN, and AlGaAs.21–23 However, for FEB research, electron energies are typically above 1 keV. At these energies, the interaction of primary electrons with surface atoms is weak. Most surface reactions are induced by secondary electrons near the surface in the energy range below 100 eV. Furthermore, FEB electron sources are normally based on thermionic electron emission and require very low background pressure for stable operation. Higher pressure operation is preferred to ensure better throughput for chemical and electron beam (EB) exposure. For these reasons, it is attractive to examine alternative approaches using an electron beam source that supplies electrons at low energy to the surface and is simultaneously exposed to reactive species at fairly high gas pressure.
Hollow cathode (HC) EB sources24–32 allow us to inject electron beams into a fairly high-pressure gas (up to 100 s of mTorr) and cover an area that can be fairly large (∼cm2). Gershman and Raitses33 performed studies of hollow cathode and anode EB plasma sources for which they demonstrated the presence of ∼15 eV beam electrons by Langmuir probe (LP) and optical emission spectroscopy measurements. They stated, “These electrons become available for surface modification and radical production outside of the source” which “opens exciting opportunities for future exploration.”
Material etching using an HC EB plasma has been studied by Walton and Boris et al.20,34–37 and Rauf and Dorf et al.19,38–40 In their work, the EB is parallel to the substrate surface and a beam dump is used for the electrons after passage through the plasma volume. Surface processing is controlled via radio frequency biasing of the substrate by adjusting the energy of the low energy ions produced by the EB plasma. The potential of utilizing plasma-produced EBs injected into process chambers for direct control of surface processes has hardly been studied. This contrasts with work in the focused electron beam community for which surface reactions are stimulated by energetic electrons in nonplasma environments.
In the present work, the authors have used an HC EB source with low and adjustable electron energy that allows electron beam injection into a relatively high-pressure (∼mTorr) environment, thus allowing combination with a remote plasma source for reactive radical generation. The combination of these sources was evaluated for the purpose of achieving electron enhanced chemical etching.
This article is organized as follows: in the first part, the authors discuss the development and characterization of the Ar hollow cathode electron beam source by measuring electrical current at the sample position and optical emission from the secondary plasma region. In the second part, the beam electron and the secondary plasma generated were characterized by Langmuir probe measurements for both pure Ar and when combining with the effluent from a CF4/O2 remote plasma. In the final part, the authors study electron enhanced chemical surface etching of Si3N4 and polycrystalline Si (poly-Si) for the conditions where the HC electron beam source is combined with the effluent of the CF4/O2 remote plasma source.
The experiments were performed in an electron etching system, consisting of a HC EB source on top of a reaction chamber and an electron cyclotron wave resonance (ECWR) remote plasma source on the side [Fig. 1(a)]. The base pressure achieved before processing was in the 1 × 10−6 Torr range. The pressure in the HC source (Torr level) is significantly higher than the chamber pressure below because of a choked flow connection. The reaction chamber pressure was kept at 2–100 mTorr during processing 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.41 This neutralization plate consists of an electrically grounded aluminum plate covered by Kapton tape and a quartz plate to protect the plate from plasma erosion. Direct exposure of the substrate and reaction chamber to the ECWR plasma is prevented by this plate, and only neutral chemical radicals can diffuse from the ECWR chamber to the processing chamber, creating a remote plasma process condition.42–44 The ECWR effect requires an additional static magnetic field to utilize the interaction of an electromagnetic wave with a plasma.41 A 400 W source power level and 1.8 mT static magnetic fields were applied to the ECWR system for all experiments requiring chemical radicals.41 As shown in Fig. 1(a), the EB source consists of a 1.57 mm diameter direct current (DC) HC tube, a 34.35 mm diameter discharge tube, an anode plate with a hole in it, a separation mesh, and an electron acceleration plate. The latter is located below the HC source inside the vacuum chamber above a water cooled (10 °C) and electrically grounded stainless steel (SS) substrate. The electrons are extracted from the HC plasma through a 1.57 mm diameter hole in the anode. An SS mesh with high transmission is attached to the anode plate to suppress extraction electric field penetration into the HC plasma region. This type of HC electron beam setup is well developed and characterized, and we refer the reader to descriptions and applications where similar setups have been employed.24–32
Argon gas was admitted to the HC discharge using a mass flow controller. The HC EB source to substrate distance can be varied using an adjustable bellows tube over a range of 70 mm. A DC power supply that provides a constant discharge current (ID) was used to generate the main HC discharge between the cathode and the source anode. A second DC power supply provided the acceleration voltage (VE) to the anode plate for acceleration of beam of electrons through the center hole of the grounded acceleration plate toward the grounded substrate. The distance of acceleration plate to HC anode plate is 10 mm.
The equivalent circuit diagram is shown in Fig. 1(b). The arrows depict the direction in which electrons move and are opposite to the conventional current direction. The HC discharge provides an electron current ID. Typical HC EB source operating conditions are 10 SCCM Ar carrier gas flow for a constant 1.5 A ID. Part of ID flows through the HC anode back to the discharge power supply (Ia). The extraction current IE penetrates through the central hole of the HC anode and moves toward the acceleration plate. With a positive bias voltage on the acceleration plate, electrons are accelerated, and then move toward the substrate after having gained additional energy. A portion of the extraction current flows back to the power supply through the acceleration plate (IA), while the rest of the electrons are transmitted through the aperture. These primary beam electrons move toward the substrate and, if sufficiently energetic, cause additional ionization as they move through the Ar gas in the main chamber. The net DC current measured at the grounded substrate is denoted with IB.
A photograph of the EB plasma is shown in Fig. 1(c). The top bright region is the Ar discharge located between the hollow cathode anode and the acceleration plate. An energetic electron beam is extracted through the center aperture of the acceleration plate, reaching the grounded SS substrate. The cone shaped beam shown underneath the acceleration plate is due to a lens focusing effect of the acceleration aperture. By adjusting the acceleration plate to EB source distance and the aperture opening size, the focus point can be optimized.
The electron beam and the secondary plasma generated have been characterized using a 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 (ne), and electron temperature (Te) by fast sweep speed, ion bombardment cleaning, and electron heating.45–47 The probe tip and reference probe were located in the center below the acceleration plate and above the substrate electrode. The cylindrical tungsten probe tip was 4.8 mm long with a 0.025 mm radius and for these measurements was located 25 mm beneath the acceleration plate. We should mention that the interpretation of the LP data for this situation is complex owing to the fact that typical assumptions used for interpretation of LP data, e.g., isotropic distribution of electron energies, are not justified for the present situation at lowest gas pressure when the electron free path λe(ɛ) is larger than the distance between acceleration electrode and probe d = 25 mm. At higher gas pressure when λe(ɛ) < d, the accelerated electrons form an electron swarm rather than an electron beam, and LP procedure is quite applicable. Nevertheless, at low gas pressure, these measurements provide a qualitative overview of the behavior of hot electrons and their influence on secondary plasma production and interactions.
In addition, an optical emission spectrometer (OES) setup is attached to the chamber and allows us to measure plasma optical emission intensity and gas phase chemical composition in real time. The Ar I and Ar II emissions at wavelengths of 750.39 and 472.69 nm, respectively, were measured as the energy of the extracted HC EB was varied. For conditions when CF4 and O2 reactive gases were used, F and O emissions were monitored at wavelengths of 685.60 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 ellipsometer is an automated rotating compensator ellipsometer working in the polarizer-compensator-sample-analyzer configuration at an incident angle of ≈73.5°.
III. RESULTS AND DISCUSSION
A. Parametric studies of the hollow cathode EB source
In order to utilize the HC EB source for etching studies, characterization of the electron source behavior, such as electron-gas interaction, EB induced secondary plasma production, electron energy control, and other aspects, is essential. For these HC EB source characterization experiments, only the EB HC source and the Ar gas flow into the main chamber were used. No Si3N4 or poly-Si samples were present on the SS substrate electrode, and no additional CF4/O2 gas flow was used. Electrons passing through the hole in the anode were accelerated by the acceleration voltage.27 It is well known that high-energy electrons can be scattered by gas molecules through elastic and inelastic collisions. When the energy of an electron is higher than the gas excitation/ionization threshold, inelastic collision will take place with a certain cross section and this additional ionization can produce a plasma in the main chamber for the present situation. A direct way to characterize the HC EB and the secondary plasma produced, the electrical current at the substrate electrode, and optical emission characterization was measured for varying acceleration voltages, VE. This is shown in Fig. 2 for operating conditions of 10 SCCM Ar flow rate, 2 mTorr main chamber pressure, 1.5 A discharge current ID in the HC source, and 5 cm acceleration plate to substrate distance.
When VE is greater than 20 V, the electron energy is greater than the Ar ionization energy threshold (15.76 eV). Consistent with this, a strong increase of all measured currents is shown in Fig. 2, along with a significant rise of both Ar I (neutral Ar) and Ar II emission (from excited Ar+ ions). This current/emission increase saturates at VE of about 50 V and is limited by the total discharge current ID in the HC setup and the setup geometry. As shown in Fig. 2(a), measurement of the emission current IE, the acceleration plate current IA, and the net measured substrate current IB increases with acceleration voltage VE. The sum of IA + IB is shown in Fig. 2(a) and is equal to the measured extraction current IE. This matched current indicates that electrons leaving the HC or produced by additional ionization in the main chamber are mostly collected by the anode, acceleration plate, and the substrate. Electron loss to the chamber walls or other grounded surfaces is limited, consistent with a strong degree of directionality toward the substrate electrode. The measured currents provide an estimate of the net electron current, and the influence of the ion current to these surfaces, e.g., the substrate electrode surface, is not known. A contribution of the latter appears unavoidable since these surfaces are in contact with the discharge region.
The behavior of the Ar I and Ar II optical emission intensities reflects the atomic Ar and Ar+ ions, respectively.30 The production rate of excited atomic and ionized Ar+ is proportional to the excitation and ionization cross sections, respectively. Comparing the emission intensities of Ar I (750.39 nm) and Ar II (472.69 nm) in Fig 2(b), the Ar II emission (ionized Ar) peaks at a higher energy than Ar I (atomic Ar), which agrees with the electron impact ionization and excitation cross section trends.28
To further characterize the behavior of HC EB, the electrical current response at 50 V acceleration voltage was evaluated for the other HC EB source parameters, i.e., ID, pressure in the process chamber, and acceleration plate to substrate distance. The variation of emission current IE, acceleration plate current IA, and net measured substrate current IB as a function of these parameters is shown in Figs. 3(a)–3(c), respectively. With increased discharge current ID [Fig. 3(a)], all other currents increase linearly. This indicates that ID controls the total HC EB current, consistent with the expectations.
With increased chamber pressure, ionization in the main chamber increases and both emission current and acceleration plate current increase. This indicates that a portion of the low energy electrons created in the proximity of these electrodes is collected. On the other hand, the net measured substrate current IB decreases with pressure. This indicates that more electron-neutral collisions reduce the number of accelerated electrons from reaching the substrate electrode, leading to a reduced IB.
As the distance between the extraction electrode and substrate electrode is varied, the relative importance of electron interactions with gas phase species or surface species changes. By changing this distance [Fig. 3(c)], IA stays the same due to an unchanged discharge condition between the HC EB source anode and the acceleration plate. However, the net measured substrate current IB and emission current IE drop slightly when increasing the distance. This reduction of current is caused by more electron-neutral collisions taking place along the path toward the collecting substrate electrode, thus preventing those electrons from reaching the substrate electrode. This current reduction is more pronounced at higher pressure (10 mTorr)
B. Langmuir probe measurements of hollow cathode EB
The beam electrons and the secondary plasma generated were further characterized using an LP provided by Plasma Sensors to determine EEPF, ne, and Te.45–47 As mentioned before, the interpretation of the LP data for the current situation ideally should take into account that typical assumptions, e.g., isotropic distribution of electron energies, used for interpretation of LP data are not justified. Here, we restrict ourselves to a qualitative overview of the behavior of the beam electrons and their influence on secondary plasma production and interactions with the gas and sample surface. The LP measurements of HC electrons in many references focus on the low electron energy portion of the EEPF. Here, this corresponds to the distribution without additional electron heating, and the electron energy distribution of the fast beam electrons tends toward this with an increase of collisions.25,35,48 The anisotropic beam electron EEPF has recently been measured and discussed by Gershman, Kraus, and Raitses for their HC plasma source with a beam electron energy of about 15 eV.33,49
In this section, EEPF measurements performed with pure Ar HC EB are described first. The change of HC EB EEPF when CF4/O2 is added is subsequently discussed.
The results of the Langmuir probe measurements demonstrate the impact of the beam electrons injected from the HC source into the process chamber and are shown in Fig. 4. The standard HC operating conditions were fixed at 10 SCCM Ar, 5 cm operation distance, 3.5 mTorr main chamber pressure, 1.5 A discharge current, and 50 V acceleration voltage. The following experiments were studied by varying one of the above parameters. The measured EEPF versus acceleration voltage VE is shown in Fig. 4(a), discharge current ID in Fig. 4(b), and pressure in the process chamber in Fig. 4(c), respectively.
In Fig. 4(a), the pressure in the process chamber was held constant at 3.5 mTorr. For an acceleration voltage of 0 V, the measured EEPF indicates that electrons passed from the hole in the anode and the center hole of the acceleration plate to the position of the LP. They are characterized by a very low electron energy and density Ne. A strong increase in electron density is seen as the acceleration voltage is increased, which can be explained by beam electrons in the reaction chamber and ionization of the Ar gas by beam electrons. For example, for an acceleration voltage of 30 V, the data show that the Ar HC EB EEPF contains two major peaks for a pressure of 3.5 mTorr. A possible explanation is that the peak at ∼10 eV is due to electrons that have lost energy by inelastic scattering with gas molecules. This low electron energy peak may be the isotropic HC EB induced plasma peak, which was reported widely in other publications.25,35,48 A second peak at ∼30 eV and higher—with the position depending on acceleration voltage—likely represents the primary beam electrons that have lost very little or no energy in inelastic collisions. Figure 4(b) shows the EEPF with increasing ID. Higher ID will increase electron densities for all energies with unchanged primary beam electron peak energy. At high ID conditions, an additional electron distribution is shown in between the two major peaks (in about 30–60 eV). This mid-energy range electron distribution relates to either insufficient acceleration from the extraction plate or the final state of beam electrons after several collisions. The impact of pressure in the main chamber on EEPF has also been evaluated and is shown in Fig. 4(c). As the pressure in the main chamber is varied, strong changes in EEPF have been observed. These changes are consistent with the qualitatively expected behavior considering the electron inelastic mean free path and energy dependence of the Ar ionization cross section. Since the beam electron component disappeared with increased pressure, we only focused on the low energy plasma peak in the EEPF. Due to a greater number of inelastic collisions between gas and electrons for high-pressure operating conditions, high-energy electrons will lose their energies by exciting/ionizing Ar atoms, leaving a sharper EEPF peak in the lower energy region and no more primary beam electrons.
If for simplicity we apply a regular Maxwellian distribution model for an isotropic plasma to the low energy portion of the EEPF, which as mentioned may not be the case, the corresponding electron temperature and density can be calculated and is shown in Figs. 5(a)–5(c) for varying VE, ID, and pressure, respectively. The electron temperature is unchanged at 4–5 eV for various acceleration voltages. The electron density increases initially and saturates at high VE. The discharge current ID controls the total available electrons for extraction and acceleration. For the same acceleration voltage, ne increases with ID. Due to more collisions with more available electrons, Te drops with ID. As discussed above, the higher collision rate with increased pressure will reduce the electron temperature Te and result in a higher electron density ne. We operated the standard process at a relatively lower pressure (3.5 mTorr) than the pressure used often for electron beam generated plasma studies reported by other groups.48,50 This may explain why the measured Te is higher than the typical value of ∼1 eV.
The beam electron peak energy is plotted versus acceleration voltage in Fig. 6(a) with two reference energy lines (VE and VE + VD). The beam electron peak energy is higher than the acceleration voltage and indicates that the electrons may gain some additional energy from the top discharge (voltage between HC and anode, VD). This may be due to less effective decoupling of the extraction electrode from the HC source as the acceleration voltage is increased to very high values. With an increased acceleration voltage, the lower energy peak is almost unchanged, but the higher energy peak moves linearly with acceleration voltage. The peak area of EB plasma and the primary beam peaks are shown in Fig. 6(b) with varying VE. Both peak areas show saturation when VE is greater than 30 V, indicating that after a certain acceleration voltage, electron density and electron energy can be controlled separately.
The EEPF areas for three different energy ranges are calculated and shown in Fig. 7(a) with ID. The electrons in all three energy ranges increase linearly with ID. With linear fittings, the slopes are calculated and shown in Fig. 7(b). This slope indicates the electron production efficiency in the specific energy range. Like the EEPF area, the slope in EB plasma region (0–30 eV) shows the highest value. This indicates that the majority of the additional electrons from ID lose energy as a result of multiple inelastic collisions (the energy loss for an electron colliding with Ar atoms is characteristic and give rise to a specific energy loss depending on the Ar energy states corresponding to the excitation/ionization). The electrons in the middle energy range (30–60 eV) represent electron experiencing insufficient collisions and only losing some energy. The increase of EEPF area in this middle energy range provides some indirect evidence for the assumption that the primary beam electron is anisotropic if it experiences no inelastic collisions with the Ar gas in the main chamber.
The combination of an HC EB source with a remote plasma radical source was used to investigate electron related surface processes. Before describing these results, we briefly review the Langmuir probe measurements performed on these discharges. In Fig. 8, the results of Langmuir probe measurements are shown for this combination, demonstrating the impact of beam electrons injected from the HC source into the process chamber. The HC operating conditions were fixed at 10 SCCM Ar, 3.5 mTorr with 1.5 A discharge current, and 50 V acceleration voltage. A processing gas mixture of CF4/O2 (1:4) was injected into the ECWR remote plasma source using different gas flow rates relative to the Ar flow from the HC EB source. The CF4/O2 ratio was chosen so that spontaneous etching of Si3N4 or poly-Si under remote plasma operating conditions was negligible.51,52
Figure 8 shows the measured electron energy probability function versus processing gas flow rate for this situation. The data are characterized by a higher noise level than the pure Ar data (no ECWR source), which may be due to electrons from the ECWR leaking into the main chamber and interacting with the HC EB. Only qualitative analysis of these data will be discussed here. As seen for pure Ar operating conditions, we observe also here two major peaks in the EEPF representing plasma electrons and beam electrons. With a small admixture of the CF4/O2 processing gases, both peaks drop dramatically due to the reduction in the overall plasma density related to the injection of the electronegative processing gases. A possible explanation is the attachment of low energy electrons to electronegative gas species. When the processing gas flow rate is at 2 SCCM and greater, the EEPF distributions remain fairly similar. An increase in the higher energy peak was observed, and the low energy peak remains at the same height. This may relate to more processing gas interacting with the HC source and changing the plasma. Since the ionization cross sections for the CF4/O2 gas mixture are different from that of Ar, the high-energy beam electron peak position varies. For this reason, the processing gas flow rate was fixed at 10 SCCM for further tests.
C. Characterization of the combination of remote plasma radical source with a hollow cathode electron beam source for material etching
To explore the electron-radical synergy effect using this HC EB source at surfaces, Si3N4 and poly-Si samples were exposed to the HC EB and fluxes of F and O radicals generated by the ECWR CF4/O2 remote plasma. As described above, a neutralization plate made of an electrically grounded aluminum plate covered with Kapton tape and a quartz plate to protect the plate from erosion allows only neutral radicals to diffuse from the ECWR plasma into the main chamber.42–44 A 400 W source power level and 1.8 mT static magnetic field were applied for all ECWR experiments. In previous work, remote plasma etching or chemical dry etching of Si3N4 and poly-Si etching using CF4/O2 chemistries have been studied.42–44 Atomic oxygen oxidizes the CF4 precursor to provide F radicals as an etchant in the downstream chamber. On the other hand, the oxygen will also oxidize the Si3N4 and poly-Si surfaces and slow down the etching of these materials, giving smaller etching rates (ERs). The amount of oxygen relative to CF4 is key for controlling the substrate ER for the remote plasma operating conditions. When about 20% O2 is used, the F radical generation rate is maximized and enables the highest ER of these materials. On the other hand, for O2-rich operating conditions, e.g., 80%O2/20%CF4, both Si3N4 and poly-Si etching can be suppressed by surface oxidation. This regime of suppressed etching is preferred for the present electron enhanced etching study since it is desirable to reduce spontaneous Si3N4 and poly-Si etching without electron irradiation. In the following tests, 2/8 SCCM CF4/O2 gas flows were used to generate remote radicals in the ECWR remote plasma source.
When the HC EB source acceleration voltage, VE, is increased, discharge-related optical emission along with material etching is observed. The Si3N4 and poly-Si ER results are shown in Fig. 9(a). Figure 9(a) shows that Si3N4 and poly-Si can be etched by the HC EB in a remote CF4/O2 plasma environment. The F (685.60 nm), Ar (750.39 nm), and O (777.19 nm) emission intensities measured by OES are shown in Fig. 9(b). When the extraction energy is high enough to ionize the Ar/O2/CF4 gas mixture (about VE = 20 V for Ar), secondary plasma emission can be observed with a relatively stable F and O concentration. However, the increase in ER with VE in the 40–70 V range indicates that the etching behavior is not controlled by the radical density. At these conditions, arrival of beam electrons at the substrate surface appears to be the limiting factor for observing substrate etching due to an electron-neutral synergy effect.
A higher ER is seen for Si3N4 than for poly-Si. Differences in surface oxidation relative to volatile product formation may be responsible for this difference. Even without electron bombardment, the poly-Si etching rate is strongly suppressed because of surface oxidation (the process conditions involve CF4/80%O2 to prevent spontaneous etching). Surface oxidation and reduction of etching are much stronger for poly-Si than for Si3N4, and it is possible that for poly-Si surface oxidation may be assisted by electrons.
Another difference is that the Si3N4 surface is expected to be at floating potential, and for the same conditions, the poly-Si surface was grounded. It is possible that the increased negative surface charge for the Si3N4 may enhance positive Ar+ ion bombardment from the HC EB plasma region and induce more significant etching of the Si3N4. Some ion bombardment is unavoidable even at grounded conducting surfaces for this setup, since Ar ions of the electron beam generated plasma are accelerated toward the sample surface by the sheath potential. The sheath potential should be controlled by the fairly high electron temperature Te of the secondary plasma (∼5 eV), beam electrons, and surface electrical properties.
In order to connect the observed etching reaction with HC EB induced surface chemistry, we scanned the etched thickness profiles as a function of position for the Si3N4 and poly-Si samples after etching using ellipsometry [Fig. 10(a)]. The experimental conditions used were the same as used for the LP measurements with a 2.5 cm source to sample distance, 3.5 mTorr main chamber pressure using CF4/O2 at 2/8 SCCM flow rates, 1.5 A discharge current, and 50 V acceleration voltage. The normalized ERs and the electron saturation current measured for the same lateral position obtained using the Langmuir probe are shown in Fig. 10(b). A circular etched shape is seen for both Si3N4 and poly-Si surfaces. The center of the spot shows the highest electron saturation current intensity and corresponds to the highest surface ER. The ER located away from the HC EB (e.g., at 30 mm) is negligible. This suggests that the observed etching behavior is an HC EB controlled localized phenomenon, and no spontaneous etching is seen for the oxygen-rich radical flux for the remote plasma operating conditions used here. The normalized etch depth of Si3N4 in Fig. 10(b) shows a broader lateral profile than obtained with poly-Si. It is possible that this is related to surface charging differences, with increased surface charge accumulation for Si3N4 repelling electrons from the center and producing positive ion bombardment to maintain the steady-state surface charge. The thickness etched in the case of poly-Si is only 10 nm, and it is possible that this little material removal limits the sensitivity of the ER measurement in this case as measurements are performed away from the center. Figure 10(b) shows that most of the profile normalized EB current and etch depth are within a 30 mm diameter range and agree very well with each other.
IV. SUMMARY AND CONCLUSIONS
In this work, an HC EB etching system has been developed that is based on an Ar-based DC HC electron beam source and electron beam injection into the downstream reactive environment of a remote CF4/O2 plasma. The acceleration voltage provides electron acceleration toward the substrate and can be used to independently control the primary electron beam energy. An HC EB induced secondary plasma can be produced when the primary electron beam energy is greater than the Ar ionization energy threshold. Characterization results of the properties of the secondary plasma and surface etching of Si3N4 and poly-Si as a function of typical process parameters, including acceleration voltage (0–70 V), discharge current of the HC discharge (1–2 A), pressure (2–100 mTorr), source to substrate distance (2.5–5 cm), and feed gas composition (with or without additional CF4/O2 remote plasma) have been presented. The EEPF measured by Langmuir probe suggests two major components for the current setup: primary beam electrons and low energy electrons produced by ionization of the Ar in the chamber. After combining the HC Ar EB with a CF4/O2 ECWR remote plasma, both the low energy plasma electron and beam electron components decreased in the EEPF. An electron-neutral synergy etching effect has been observed for Si3N4 and polycrystalline silicon. Because of the finite extent of the area exposed to beam electrons, the etching is confined to an area directly below the injected electron beam within an ∼30 mm diameter circle. This etched circular area is slightly broader for Si3N4 than that for poly-Si but agrees well with the LP data as a function of position. Negligible remote plasma spontaneous etching was observed outside of the circular area coinciding with the location of beam electrons. The etching rate for poly-Si is smaller than that for Si3N4. A possible explanation may be additional ion bombardment caused by the accumulation of a negative surface charge for the Si3N4 surface, which maintains a steady-state surface charging level, and this being reduced for poly-Si.
The authors gratefully acknowledge the financial support of this work by Carl Zeiss SMS, and additional support by U.S. Department of Energy (No. DE-SC0001939). The authors thank A. Knoll, P. Luan, Pranda, K. Y. Lin, and A. Perruccio for helpful discussions and collaboration. They also thank Y. Raitses from Princeton Plasma Physics Laboratory for helpful discussions.