- Email: [email protected]
Improvement of SiO2 pattern profiles etched in CF4 and SF6 plasmas by using a Faraday cage and neutral beams
Surface & Coatings Technology 193 (2005) 75 – 80 www.elsevier.com/locate/surfcoat
Improvement of SiO2 pattern profiles etched in CF4 and SF6 plasmas by using a Faraday cage and neutral beams Jae-Ho Mina, Gyeo-Re Leea, Jin-kwan Leea, Chang-Koo Kimb, Sang Heup Moona,* a
School of Chemical Engineering and Institute of Chemical Processes, Seoul National University, San 56-1, Shillim-dong, Kwanak-ku, Seoul 151-744, Republic of Korea b Department of Chemical Engineering, Ajou University, Suwon 443-749, Republic of Korea Available online 18 October 2004
Abstract This study reports on a new plasma etching technique, using a Faraday cage and neutral beams, that improves the etch profile of an SiO2 trench in CF4 and SF6 plasmas. A Faraday cage, with a top plane made of conductive grids, was placed on the cathode of a transformercoupled plasma etcher such that the initial direction of ions passing through the grids traveling inside the cage is maintained. Neutral beams were generated by reflecting the ions on the surfaces of 58-slanted conductive reflectors located under the top grid of the Faraday cage. Three sets of etch experiments were conducted to observe the individual effects of the Faraday cage and neutral beams on the etch profiles: conventional plasma etching (case I), etching using a Faraday cage (case II), and etching using a Faraday cage and neutral beams (case III). For case II, faceting of the mask and substrate and narrowing of the line spacing were drastically suppressed. This is because the electric potential in the Faraday cage is uniform, and, therefore, the distortion of the electric field near the convex corner of a microfeature is eliminated. For case III, faceting and narrowing were further reduced because the direction of neutral beams is unaffected by negative charges accumulated on the sidewall during this process. The increase in the neutral to ion flux ratio via the use of a Faraday cage contributes to the evolution of different etching profiles depending on CF4 and SF6 plasmas. This can be attributed to differences in the angular dependence of the etch yield for the two plasmas. D 2004 Elsevier B.V. All rights reserved. Keywords: Plasma etching; Faraday cage; Neutral beam; Faceting
1. Introduction Plasma etching is widely used for the patterning of thin solid films due to its anisotropic etching capability. However, there are two factors in plasma etching which contribute to the evolution of a nonvertical etch profile. One is the local distortion of electric fields at the convex corners of the microfeatures [1,2]. This field distortion enhances the ion flux at the corner to produce faceting of the mask and substrate, and narrowing of the line spacing [3,4]. The other is the accumulation of negative charges on the surface of insulating sidewalls due to the combination of lessanisotropic electron flux and highly anisotropic positive
* Corresponding author. Tel.: +82 2 880 7409; fax: +82 2 875 6697. E-mail address: [email protected] (S. Heup Moon). 0257-8972/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2004.08.153
ion flux [4,5]. This localized charging of the sidewall surface additionally bends the trajectory of the ions, leading to bowing, microtrenching, an RIE lag, and to electrical damage. The above two factors cannot be overcome by simply adjusting process variables including the source power, bias voltage, pressure, substrate temperature, and others. In this study, a new etching technique using a Faraday cage and neutral beams was used to minimize or avoid faceting of the mask and substrate and narrowing of the trench line spacing. For conventional neutral beam etching (NBE) [6], a neutral beam is generated from an ion beam by charge exchange reactions in the gas phase and residual ions are eliminated by a retarding grid. This process has the disadvantage of low etch rates and is affected by gas chemistry to a significantly different extent from that of conventional plasma etching. In comparison with the conventional NBE, the plasma etching
76
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
technique employed here, using a Faraday cage and neutral beams, has the advantage of showing relatively high etch rates because radicals in the plasma also contribute to the etching. Another advantage is that the energy of neutral beams and the density of reactive radicals can be independently controlled by adjusting the bias voltage and the source power in a transformer-coupled plasma (TCP) etcher.
2. Experiment Etching experiments were carried out using a TCP etcher, as shown in Fig. 1(a). The etching system has been described in detail in our previous paper [7]. Therefore, only the most important dimensions will be presented in this paper: (i) the inner diameter of the reaction chamber was 30 cm, (ii) the cathode diameter was 12.5 cm, and (iii) the spacing between the 0.8-cm-thick dielectric window under the coil and the cathode was 3.5 cm. The substrate for the etching was a 2-Am-thick SiO2 film thermally grown on a P-type Si wafer. The film was masked with a 1500-2-thick Al line. The line width and line space of the Al mask were 0.6 m and 1 Am, respectively. Three samples of the substrate, cut into a 55 mm2 square, were placed on the cathode and were etched under three different conditions, i.e., cases I, II, and III, as shown in Fig. 1(b). In case I, the substrate is exposed directly to the plasma. This corresponds to the conventional plasma etching process, which is subject to the electric field distortion at the convex corners of microfeatures and charging of the sidewalls. In case II, the substrate is placed in a Faraday cage. A Faraday cage is simply a closed box consisting of a conductor. The cage with copper sidewalls and a roof made of a brass wire grid was bolted to the stainless-steel cathode to make an enclosed Faraday cage. When the substrate is located in the cage, the electric field near the convex corners
of the microfeatures is not distorted by the external field because the electric potential in the cage is unaffected by outside voltages and is uniform in the cage [8–14]. Therefore, ions enter perpendicular to the sheath formed along the top grid plane of the cage and reach the substrate with their initial direction maintained. However, even when the Faraday cage is used, the problem of local electric fields caused by negative charges accumulating on the surface of insulating sidewalls continues to exist. In case III, the substrate is placed in a Faraday cage with specially-designed neutralizing reflectors. The grid plane of the Faraday cage is slanted at a 108 angle with respect to the horizontal cathode, and the neutralizing reflectors (arranged in parallel to one another) are slanted at an angle of 58 with respect to the grid plane. Ions passing through the grid of the Faraday cage are neutralized by the grazing angle collision of ions with the reflector surface. As a result, neutralized ions or neutral beams, which impinge perpendicular to the substrate surface, are obtained. CF4 and SF6 were separately used as discharge gases in experiments concerning the above three cases. The process conditions were as follows: the pressure was 5 mTorr, the gas flow rate 4 sccm, the source power 300 W, the bias voltage 400 V, and the substrate temperature was 15 8C. The high bias voltage (400 V) was selected to effectively observe faceting. Because the etch rates were different for the above three cases, the etch time was selected such that an etch depth of 1 Am was obtained for all cases. In another set of experiments using a CF4 plasma, an extended etch time, corresponding to 150% of the above cases, was used to observe changes in the etch profile with time. The etch rates of blank SiO2 substrates were determined by optically measuring changes in the thickness after etching (SpectraThick 2000-Deluxe). The etch profiles of substrates patterned with Al lines were obtained by scanning electron microscopy (SEM).
Fig. 1. Schematic diagrams of the system used in this study. (a) A transformer-coupled plasma (TCP) etcher. (b) Three cases of etch experiments: conventional plasma etching (case I), etching using a Faraday cage (case II), and etching using a Faraday cage and neutral beams (case III).
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
3. Results and discussion 3.1. Etch rates for the cases I, II, and III Fig. 2(a) shows the etch rates of blank SiO2 substrates in CF4 and SF6 plasmas for cases I, II, and III. The etch rates obtained in the Faraday cage (cases II and III) are much smaller than those obtained in the case where no Faraday cage is employed (case I). The difference in the etch rates is attributed to the following factors. First, the plasma density is lower for cases II and III than for the case I because the presence of the Faraday cage, which functions as a cathode of the plasma etcher, reduces the distance between the cathode and the dielectric window under the RF coil, and, as a result, the plasma density is decreased. Second, the ion flux on the substrate for cases II and III is decreased by the blockage of grid wires of the Faraday cage. The fraction of open grid area, i.e., the grid transparency, used in this work was about 50% with a grid pitch of 0.67 mm and a grid diameter of 0.2 mm. Third, the etch rates for case III are the lowest among the three cases because the reflectors placed under the grid further impede the flux of radicals and ions on the substrates. The reflectors also reduce the ion energy, and the loss of ion energy by the collision of ions with reflectors is demonstrated in Fig. 2(b). Fig. 2(b) shows the etch rates of blank SiO2 substrates in a CF4 plasma for the three cases as a function of bias voltage. For a better comparison, the etch rate was normalized by the rate at 600 V. In the case of chemical sputtering, etching is dependent on the energy transferred from ions to the surface such that the etch yield is proportional to the square root of the ion energy [11]. Assuming that the plasma potential is about 30 V, the relation between the etch rate and the bias voltage is given by the dotted line in Fig. 2(b). For both cases I and II, the normalized etch rates (NER) agree fairly well with the dotted line, indicating that the use of a Faraday cage does
77
not diminish the ion energy. On the other hand, for case III, the normalized etch rates are smaller than those for cases I and II, suggesting that the ion energy is decreased by the collision of ions with reflectors. 3.2. Etch profile improvement by use of a Faraday cage and neutral beams Fig. 3 shows the cross-sectional SEM images of patterns obtained in CF4 and SF6 plasmas for the three cases. In the figure, the term dnormal etchingT indicates the case where the etch depth of an SiO2 film is about 1 Am, while the term doveretchingT indicates the case where the etch time is 1.5 times longer than that used in normal etching. The SEM images confirm that the use of a Faraday cage and neutral beams effectively reduces the faceting of the mask and substrate, as well as narrowing the line spacing, such that the etch profile is significantly improved. For example, for overetching in a CF4 plasma, the line spacing near the bottom for case I is ca. 0.5 Am, which is much smaller than the original Al line spacing before etching (1 Am), while those for cases II and III are ca. 0.90 and 0.95 Am, respectively. To further investigate the effect of a Faraday cage and neutral beams, the degree of faceting ( f ) and the degree of narrowing (n) were defined as f ð% Þ ¼
W0 W x0 x 100 and nð%Þ ¼ 100; 2H x0
where W 0 is the width of the original mask before etching (0.6 Am), W the width of top surface of the line after etching, H the etched depth, x o the width of the original spacing between mask lines (1 Am), and x is the width of the trench line spacing near the bottom after etching. The values for f and n obtained in CF4 and SF6 plasmas for the three cases are shown in Fig. 4. In a conventional plasma etching, a microfeature is directly exposed to the plasma and develops a concentrated
Fig. 2. (a) Etch rates of blank SiO2 substrates in CF4 and SF6 plasmas for cases I, II, and III. (b) Changes in the normalized etch rate of an SiO2 substrate with bias voltage in a CF4 plasma.
78
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
Fig. 3. Cross-sectional SEM images of patterns etched in CF4 and SF6 plasmas for cases I, II, and III.
electric field at its convex corner. This is because the equipotential contours in the sheath are curved following the feature. The field distortion enhances the ion flux around the corner, thus accelerating faceting. This corresponds to case I (without the Faraday cage). For a CF4 plasma, as the etch time is increased, i.e., the etch depth is increased, f increases from 12.5% to 18% (case I in Fig. 4(a)). This is because, as the faceting proceeds, the electric field distortion near the convex corner becomes severe, and, as a result, the faceting is self-accelerating. The acceleration in field distortion near the convex corner makes the ion angular distribution inside the trench pattern broader, which decreases the ion flux on the bottom of the trench. Thus, the line spacing near the bottom is narrowed (case I in Fig. 4(b)). The use of a Faraday cage prevents the concentration of an ion flux at the convex corner of a microfeature, which leads to the drastic reduction in faceting (from 12.5% to 6% for normal etching, and from 18% to 9% for overetching) and narrowing (from
28% to 20% for normal etching, and from 4 8% to 20% for overetching). For case III (neutral beams with a Faraday cage), the faceting and narrowing are further reduced compared to case II (with Faraday cage only). The accumulation of negative charges at the surface of insulating sidewalls by collisions of less-anisotropic electrons should bend the ion trajectories, thus additionally contributing to faceting and narrowing. For case III, the charging effect is eliminated due to the neutralization of ions, and, as a result, the etch profiles with the lowest values of f and n are obtained. It is noteworthy that the f and n values for etch profiles obtained using a Faraday cage and neutralized ions are not increased after overetching. Similar to the case of a CF4 plasma, the etch profile for an SF6 plasma was also improved when a Faraday cage and neutral beams were used. However, the shapes of the patterns etched in an SF6 plasma were significantly different
Fig. 4. (a) The degree of faceting and (b) the degree of narrowing in CF4 and SF6 plasmas for cases I, II, and III. Insets in panels (a) and (b) define the degree of faceting and the degree of narrowing.
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
from those in a CF4 plasma (see Fig. 3). This can be attributed to the different etch characteristics between SF6 and CF4 plasmas as described below. 3.3. Angular dependence of the etch yield The increase in the f values after overetching in a CF4 plasma for case II (from 6% to 9%) cannot be explained by the self-acceleration of faceting originating from an increase in electric field distortion because the electric field is uniform inside the cage. This suggests that another factor also contributes to faceting. Because the substrate corner is rounded (even the edges of original mask), ions are incident on the rounded surface at different angles. The etch rates at positions with different ion incident angles are determined by the angular dependence of the etch rate or etch yield [15]. Fig. 5 shows the normalized etch rate (NER) and normalized etch yield (NEY) for SiO2 etching in CF4 and SF6 plasmas as a function of the ion incident angle. Here, the angular dependencies of the etch rates were obtained using a blank SiO2 substrate etched in a Faraday cage [11–13]. The angle of ions incident on the substrate could be accurately controlled by adjusting the angle of a sample holder placed inside the cage. The NERs (h) are defined as the etch rates normalized by the rate at 08, and the NEYs (h) were obtained by dividing the NERs (h) by cosh. If ion-enhanced chemical etching is a major contributor to etching, the etch yield would decrease monotonically with the ion incident angle [15,16]. On the other hand, if physical sputtering is a major contributor to etching, the etch yield (or even the etch rate) would reach a maximum at angles in the range between 408 and 608 [17,18]. Fig. 5 suggests that, under the process conditions employed in this study, physical sputtering is a major contributor to the SiO2 etching in a CF4 plasma. Once faceting starts to occur in a CF4 plasma, an increase in the
79
angle between the incoming ions and the substrate surface leads to increased faceting. Therefore, the f values increase with etch time for case II (see Fig. 4(a)). In contrast to the case of a CF4 plasma, the NEYs for an SF6 plasma remain close to 1 and are independent of the ion incident angle. This suggests that faceting occurs to a smaller extent for an SF6 plasma than for a CF4 plasma, as shown in Fig. 4(a). 3.4. Change in the neutral to ion flux ratio inside the Faraday cage The Faraday cage affects the neutral to ion flux ratio as well as the electric field near the substrate surface. The ion flux on the substrate is blocked by the grid of a Faraday cage due to the directionality of ions. On the other hand, the decrease in the neutral flux by the cage is much smaller than that in the ion flux because the neutrals move isotropically and circumvent the blockage by grid wires. As a result, the neutral to ion flux ratio is increased inside the Faraday cage, which contributes to the evolution of different etch profiles depending on the CF4 and SF6 plasmas. For a CF4 plasma, the sidewall surface is passivated with a fluorocarbon polymer layer due to the deposition of neutral CFx species [19]. Therefore, an increase in the neutral to ion flux ratio due to the use of a Faraday cage additionally suppresses faceting. For an SF6 plasma, the suppression of faceting due to the deposition of neutrals does not occur because the plasma does not contain polymer precursors. On the other hand, the use of a Faraday cage induces an undercut for an SF6 plasma because an increase in the relative concentrations of F radicals contributes to spontaneous etching, which proceeds without the assistance of ion bombardment (cases II and III in Fig. 3). The spontaneous etching characteristics of an SF6 plasma are also observed in Fig. 5(a). The NER at 908 for an SF6 plasma is about 0.1, which indicates that the
Fig. 5. The angular dependencies of (a) the normalized etch rates and (b) normalized etch yields of blank SiO2 substrates in CF4 and SF6 plasmas. The dashed curve in panel (b) represents the cosine distribution.
80
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
lateral etch rate is higher by 10% than the vertical etch rate when the Faraday cage is used.
4. Conclusions A new etching technique, using a Faraday cage and neutral beams, improves the etch profile of an SiO2 pattern obtained in CF4 and SF6 plasmas. In both plasmas, the degree of faceting of the mask and substrate, as well as the degree of narrowing of the line spacing, was significantly reduced by the use of the Faraday cage. This is attributed to the suppression of ion-flux focusing on the convex corner of features. The use of neutral beams in conjunction with a Faraday cage further reduced the faceting and narrowing in both plasmas because the neutral beams were not affected by negative charges accumulating on the sidewalls.
Acknowledgements This work was supported by the Brain Korea 21 project, the Center for Ultramicrochemical Process Systems (CUPS) designated by KOSEF, and National Research Laboratory program of Ministry of Science and Technology, Korea. One of the authors (C.-K. Kim) appreciates the financial support from the Basic Research Program of the Korea Science and Engineering Foundation (grant No. R01-2003000-10103-0), and the Ajou University Research Facilities Support Program in the year 2003.
References [1] S.G. Ingram, J. Appl. Phys. 68 (1990) 500. [2] B. Chapman, Glow Discharge Processes, Wiley, New York, 1980. [3] D.M. Manos, D.L. Flamm, Plasma Etching, Academic Press, San Diego, CA, 1989. [4] J.C. Arnold, H.H. Sawin, J. Appl. Phys. 70 (1991) 5314. [5] M. Schaepkens, G.S. Oehrlein, Appl. Phys. Lett. 72 (1998) 1293. [6] T. Yunogami, K. Yokogawa, T. Mizutani, J. Vac. Sci. Technol., A 13 (1995) 952. [7] B.-O. Cho, S.-W. Hwang, I.-W. Kim, S.H. Moon, J. Electrochem. Soc. 146 (1999) 350. [8] G.D. Boyd, L.A. Coldren, F.G. Storz, Appl. Phys. Lett. 36 (1980) 583. [9] B.-O. Cho, S.-W. Hwang, J.-H. Ryu, S.H. Moon, Rev. Sci. Instrum. 70 (1999) 2458. [10] J.-H. Ryu, B.-O. Cho, S.-W. Hwang, S.H. Moon, C.-K. Kim, Korean J. Chem. Eng. 20 (2003) 407. [11] B.-O. Cho, S.-W. Hwang, G.-R. Lee, S.H. Moon, J. Vac. Sci. Technol., A 18 (2000) 2791. [12] G.-R. Lee, S.-W. Hwang, J.-H. Min, S.H. Moon, J. Vac. Sci. Technol., A 20 (2002) 1808. [13] S.-W. Hwang, G.-R. Lee, J.-H. Min, S.H. Moon, Electrochem. SolidState Lett. 6 (2003) G12. [14] J.-H. Min, S.-W. Hwang, G.-R. Lee, S.H. Moon, J. Vac. Sci. Technol., A 20 (2002) 1574. [15] H.F. Winters, J.W. Coburn, T.J. Chuang, J. Vac. Sci. Technol., B 1 (1983) 469. [16] S. Tachi, S. Okudaira, J. Vac. Sci. Technol., B 4 (1986) 459. [17] T.M. Mayer, R.A. Barker, L.J. Whitman, J. Vac. Sci. Technol. 18 (1981) 349. [18] R.E. Lee, J. Vac. Sci. Technol. 16 (1979) 164. [19] D. Zhang, M.J. Kushner, J. Vac. Sci. Technol., A 19 (2000) 524.
Improvement of SiO2 pattern profiles etched in CF4 and SF6 plasmas by using a Faraday cage and neutral beams Jae-Ho Mina, Gyeo-Re Leea, Jin-kwan Leea, Chang-Koo Kimb, Sang Heup Moona,* a
School of Chemical Engineering and Institute of Chemical Processes, Seoul National University, San 56-1, Shillim-dong, Kwanak-ku, Seoul 151-744, Republic of Korea b Department of Chemical Engineering, Ajou University, Suwon 443-749, Republic of Korea Available online 18 October 2004
Abstract This study reports on a new plasma etching technique, using a Faraday cage and neutral beams, that improves the etch profile of an SiO2 trench in CF4 and SF6 plasmas. A Faraday cage, with a top plane made of conductive grids, was placed on the cathode of a transformercoupled plasma etcher such that the initial direction of ions passing through the grids traveling inside the cage is maintained. Neutral beams were generated by reflecting the ions on the surfaces of 58-slanted conductive reflectors located under the top grid of the Faraday cage. Three sets of etch experiments were conducted to observe the individual effects of the Faraday cage and neutral beams on the etch profiles: conventional plasma etching (case I), etching using a Faraday cage (case II), and etching using a Faraday cage and neutral beams (case III). For case II, faceting of the mask and substrate and narrowing of the line spacing were drastically suppressed. This is because the electric potential in the Faraday cage is uniform, and, therefore, the distortion of the electric field near the convex corner of a microfeature is eliminated. For case III, faceting and narrowing were further reduced because the direction of neutral beams is unaffected by negative charges accumulated on the sidewall during this process. The increase in the neutral to ion flux ratio via the use of a Faraday cage contributes to the evolution of different etching profiles depending on CF4 and SF6 plasmas. This can be attributed to differences in the angular dependence of the etch yield for the two plasmas. D 2004 Elsevier B.V. All rights reserved. Keywords: Plasma etching; Faraday cage; Neutral beam; Faceting
1. Introduction Plasma etching is widely used for the patterning of thin solid films due to its anisotropic etching capability. However, there are two factors in plasma etching which contribute to the evolution of a nonvertical etch profile. One is the local distortion of electric fields at the convex corners of the microfeatures [1,2]. This field distortion enhances the ion flux at the corner to produce faceting of the mask and substrate, and narrowing of the line spacing [3,4]. The other is the accumulation of negative charges on the surface of insulating sidewalls due to the combination of lessanisotropic electron flux and highly anisotropic positive
* Corresponding author. Tel.: +82 2 880 7409; fax: +82 2 875 6697. E-mail address: [email protected] (S. Heup Moon). 0257-8972/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2004.08.153
ion flux [4,5]. This localized charging of the sidewall surface additionally bends the trajectory of the ions, leading to bowing, microtrenching, an RIE lag, and to electrical damage. The above two factors cannot be overcome by simply adjusting process variables including the source power, bias voltage, pressure, substrate temperature, and others. In this study, a new etching technique using a Faraday cage and neutral beams was used to minimize or avoid faceting of the mask and substrate and narrowing of the trench line spacing. For conventional neutral beam etching (NBE) [6], a neutral beam is generated from an ion beam by charge exchange reactions in the gas phase and residual ions are eliminated by a retarding grid. This process has the disadvantage of low etch rates and is affected by gas chemistry to a significantly different extent from that of conventional plasma etching. In comparison with the conventional NBE, the plasma etching
76
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
technique employed here, using a Faraday cage and neutral beams, has the advantage of showing relatively high etch rates because radicals in the plasma also contribute to the etching. Another advantage is that the energy of neutral beams and the density of reactive radicals can be independently controlled by adjusting the bias voltage and the source power in a transformer-coupled plasma (TCP) etcher.
2. Experiment Etching experiments were carried out using a TCP etcher, as shown in Fig. 1(a). The etching system has been described in detail in our previous paper [7]. Therefore, only the most important dimensions will be presented in this paper: (i) the inner diameter of the reaction chamber was 30 cm, (ii) the cathode diameter was 12.5 cm, and (iii) the spacing between the 0.8-cm-thick dielectric window under the coil and the cathode was 3.5 cm. The substrate for the etching was a 2-Am-thick SiO2 film thermally grown on a P-type Si wafer. The film was masked with a 1500-2-thick Al line. The line width and line space of the Al mask were 0.6 m and 1 Am, respectively. Three samples of the substrate, cut into a 55 mm2 square, were placed on the cathode and were etched under three different conditions, i.e., cases I, II, and III, as shown in Fig. 1(b). In case I, the substrate is exposed directly to the plasma. This corresponds to the conventional plasma etching process, which is subject to the electric field distortion at the convex corners of microfeatures and charging of the sidewalls. In case II, the substrate is placed in a Faraday cage. A Faraday cage is simply a closed box consisting of a conductor. The cage with copper sidewalls and a roof made of a brass wire grid was bolted to the stainless-steel cathode to make an enclosed Faraday cage. When the substrate is located in the cage, the electric field near the convex corners
of the microfeatures is not distorted by the external field because the electric potential in the cage is unaffected by outside voltages and is uniform in the cage [8–14]. Therefore, ions enter perpendicular to the sheath formed along the top grid plane of the cage and reach the substrate with their initial direction maintained. However, even when the Faraday cage is used, the problem of local electric fields caused by negative charges accumulating on the surface of insulating sidewalls continues to exist. In case III, the substrate is placed in a Faraday cage with specially-designed neutralizing reflectors. The grid plane of the Faraday cage is slanted at a 108 angle with respect to the horizontal cathode, and the neutralizing reflectors (arranged in parallel to one another) are slanted at an angle of 58 with respect to the grid plane. Ions passing through the grid of the Faraday cage are neutralized by the grazing angle collision of ions with the reflector surface. As a result, neutralized ions or neutral beams, which impinge perpendicular to the substrate surface, are obtained. CF4 and SF6 were separately used as discharge gases in experiments concerning the above three cases. The process conditions were as follows: the pressure was 5 mTorr, the gas flow rate 4 sccm, the source power 300 W, the bias voltage 400 V, and the substrate temperature was 15 8C. The high bias voltage (400 V) was selected to effectively observe faceting. Because the etch rates were different for the above three cases, the etch time was selected such that an etch depth of 1 Am was obtained for all cases. In another set of experiments using a CF4 plasma, an extended etch time, corresponding to 150% of the above cases, was used to observe changes in the etch profile with time. The etch rates of blank SiO2 substrates were determined by optically measuring changes in the thickness after etching (SpectraThick 2000-Deluxe). The etch profiles of substrates patterned with Al lines were obtained by scanning electron microscopy (SEM).
Fig. 1. Schematic diagrams of the system used in this study. (a) A transformer-coupled plasma (TCP) etcher. (b) Three cases of etch experiments: conventional plasma etching (case I), etching using a Faraday cage (case II), and etching using a Faraday cage and neutral beams (case III).
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
3. Results and discussion 3.1. Etch rates for the cases I, II, and III Fig. 2(a) shows the etch rates of blank SiO2 substrates in CF4 and SF6 plasmas for cases I, II, and III. The etch rates obtained in the Faraday cage (cases II and III) are much smaller than those obtained in the case where no Faraday cage is employed (case I). The difference in the etch rates is attributed to the following factors. First, the plasma density is lower for cases II and III than for the case I because the presence of the Faraday cage, which functions as a cathode of the plasma etcher, reduces the distance between the cathode and the dielectric window under the RF coil, and, as a result, the plasma density is decreased. Second, the ion flux on the substrate for cases II and III is decreased by the blockage of grid wires of the Faraday cage. The fraction of open grid area, i.e., the grid transparency, used in this work was about 50% with a grid pitch of 0.67 mm and a grid diameter of 0.2 mm. Third, the etch rates for case III are the lowest among the three cases because the reflectors placed under the grid further impede the flux of radicals and ions on the substrates. The reflectors also reduce the ion energy, and the loss of ion energy by the collision of ions with reflectors is demonstrated in Fig. 2(b). Fig. 2(b) shows the etch rates of blank SiO2 substrates in a CF4 plasma for the three cases as a function of bias voltage. For a better comparison, the etch rate was normalized by the rate at 600 V. In the case of chemical sputtering, etching is dependent on the energy transferred from ions to the surface such that the etch yield is proportional to the square root of the ion energy [11]. Assuming that the plasma potential is about 30 V, the relation between the etch rate and the bias voltage is given by the dotted line in Fig. 2(b). For both cases I and II, the normalized etch rates (NER) agree fairly well with the dotted line, indicating that the use of a Faraday cage does
77
not diminish the ion energy. On the other hand, for case III, the normalized etch rates are smaller than those for cases I and II, suggesting that the ion energy is decreased by the collision of ions with reflectors. 3.2. Etch profile improvement by use of a Faraday cage and neutral beams Fig. 3 shows the cross-sectional SEM images of patterns obtained in CF4 and SF6 plasmas for the three cases. In the figure, the term dnormal etchingT indicates the case where the etch depth of an SiO2 film is about 1 Am, while the term doveretchingT indicates the case where the etch time is 1.5 times longer than that used in normal etching. The SEM images confirm that the use of a Faraday cage and neutral beams effectively reduces the faceting of the mask and substrate, as well as narrowing the line spacing, such that the etch profile is significantly improved. For example, for overetching in a CF4 plasma, the line spacing near the bottom for case I is ca. 0.5 Am, which is much smaller than the original Al line spacing before etching (1 Am), while those for cases II and III are ca. 0.90 and 0.95 Am, respectively. To further investigate the effect of a Faraday cage and neutral beams, the degree of faceting ( f ) and the degree of narrowing (n) were defined as f ð% Þ ¼
W0 W x0 x 100 and nð%Þ ¼ 100; 2H x0
where W 0 is the width of the original mask before etching (0.6 Am), W the width of top surface of the line after etching, H the etched depth, x o the width of the original spacing between mask lines (1 Am), and x is the width of the trench line spacing near the bottom after etching. The values for f and n obtained in CF4 and SF6 plasmas for the three cases are shown in Fig. 4. In a conventional plasma etching, a microfeature is directly exposed to the plasma and develops a concentrated
Fig. 2. (a) Etch rates of blank SiO2 substrates in CF4 and SF6 plasmas for cases I, II, and III. (b) Changes in the normalized etch rate of an SiO2 substrate with bias voltage in a CF4 plasma.
78
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
Fig. 3. Cross-sectional SEM images of patterns etched in CF4 and SF6 plasmas for cases I, II, and III.
electric field at its convex corner. This is because the equipotential contours in the sheath are curved following the feature. The field distortion enhances the ion flux around the corner, thus accelerating faceting. This corresponds to case I (without the Faraday cage). For a CF4 plasma, as the etch time is increased, i.e., the etch depth is increased, f increases from 12.5% to 18% (case I in Fig. 4(a)). This is because, as the faceting proceeds, the electric field distortion near the convex corner becomes severe, and, as a result, the faceting is self-accelerating. The acceleration in field distortion near the convex corner makes the ion angular distribution inside the trench pattern broader, which decreases the ion flux on the bottom of the trench. Thus, the line spacing near the bottom is narrowed (case I in Fig. 4(b)). The use of a Faraday cage prevents the concentration of an ion flux at the convex corner of a microfeature, which leads to the drastic reduction in faceting (from 12.5% to 6% for normal etching, and from 18% to 9% for overetching) and narrowing (from
28% to 20% for normal etching, and from 4 8% to 20% for overetching). For case III (neutral beams with a Faraday cage), the faceting and narrowing are further reduced compared to case II (with Faraday cage only). The accumulation of negative charges at the surface of insulating sidewalls by collisions of less-anisotropic electrons should bend the ion trajectories, thus additionally contributing to faceting and narrowing. For case III, the charging effect is eliminated due to the neutralization of ions, and, as a result, the etch profiles with the lowest values of f and n are obtained. It is noteworthy that the f and n values for etch profiles obtained using a Faraday cage and neutralized ions are not increased after overetching. Similar to the case of a CF4 plasma, the etch profile for an SF6 plasma was also improved when a Faraday cage and neutral beams were used. However, the shapes of the patterns etched in an SF6 plasma were significantly different
Fig. 4. (a) The degree of faceting and (b) the degree of narrowing in CF4 and SF6 plasmas for cases I, II, and III. Insets in panels (a) and (b) define the degree of faceting and the degree of narrowing.
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
from those in a CF4 plasma (see Fig. 3). This can be attributed to the different etch characteristics between SF6 and CF4 plasmas as described below. 3.3. Angular dependence of the etch yield The increase in the f values after overetching in a CF4 plasma for case II (from 6% to 9%) cannot be explained by the self-acceleration of faceting originating from an increase in electric field distortion because the electric field is uniform inside the cage. This suggests that another factor also contributes to faceting. Because the substrate corner is rounded (even the edges of original mask), ions are incident on the rounded surface at different angles. The etch rates at positions with different ion incident angles are determined by the angular dependence of the etch rate or etch yield [15]. Fig. 5 shows the normalized etch rate (NER) and normalized etch yield (NEY) for SiO2 etching in CF4 and SF6 plasmas as a function of the ion incident angle. Here, the angular dependencies of the etch rates were obtained using a blank SiO2 substrate etched in a Faraday cage [11–13]. The angle of ions incident on the substrate could be accurately controlled by adjusting the angle of a sample holder placed inside the cage. The NERs (h) are defined as the etch rates normalized by the rate at 08, and the NEYs (h) were obtained by dividing the NERs (h) by cosh. If ion-enhanced chemical etching is a major contributor to etching, the etch yield would decrease monotonically with the ion incident angle [15,16]. On the other hand, if physical sputtering is a major contributor to etching, the etch yield (or even the etch rate) would reach a maximum at angles in the range between 408 and 608 [17,18]. Fig. 5 suggests that, under the process conditions employed in this study, physical sputtering is a major contributor to the SiO2 etching in a CF4 plasma. Once faceting starts to occur in a CF4 plasma, an increase in the
79
angle between the incoming ions and the substrate surface leads to increased faceting. Therefore, the f values increase with etch time for case II (see Fig. 4(a)). In contrast to the case of a CF4 plasma, the NEYs for an SF6 plasma remain close to 1 and are independent of the ion incident angle. This suggests that faceting occurs to a smaller extent for an SF6 plasma than for a CF4 plasma, as shown in Fig. 4(a). 3.4. Change in the neutral to ion flux ratio inside the Faraday cage The Faraday cage affects the neutral to ion flux ratio as well as the electric field near the substrate surface. The ion flux on the substrate is blocked by the grid of a Faraday cage due to the directionality of ions. On the other hand, the decrease in the neutral flux by the cage is much smaller than that in the ion flux because the neutrals move isotropically and circumvent the blockage by grid wires. As a result, the neutral to ion flux ratio is increased inside the Faraday cage, which contributes to the evolution of different etch profiles depending on the CF4 and SF6 plasmas. For a CF4 plasma, the sidewall surface is passivated with a fluorocarbon polymer layer due to the deposition of neutral CFx species [19]. Therefore, an increase in the neutral to ion flux ratio due to the use of a Faraday cage additionally suppresses faceting. For an SF6 plasma, the suppression of faceting due to the deposition of neutrals does not occur because the plasma does not contain polymer precursors. On the other hand, the use of a Faraday cage induces an undercut for an SF6 plasma because an increase in the relative concentrations of F radicals contributes to spontaneous etching, which proceeds without the assistance of ion bombardment (cases II and III in Fig. 3). The spontaneous etching characteristics of an SF6 plasma are also observed in Fig. 5(a). The NER at 908 for an SF6 plasma is about 0.1, which indicates that the
Fig. 5. The angular dependencies of (a) the normalized etch rates and (b) normalized etch yields of blank SiO2 substrates in CF4 and SF6 plasmas. The dashed curve in panel (b) represents the cosine distribution.
80
J.-H. Min et al. / Surface & Coatings Technology 193 (2005) 75–80
lateral etch rate is higher by 10% than the vertical etch rate when the Faraday cage is used.
4. Conclusions A new etching technique, using a Faraday cage and neutral beams, improves the etch profile of an SiO2 pattern obtained in CF4 and SF6 plasmas. In both plasmas, the degree of faceting of the mask and substrate, as well as the degree of narrowing of the line spacing, was significantly reduced by the use of the Faraday cage. This is attributed to the suppression of ion-flux focusing on the convex corner of features. The use of neutral beams in conjunction with a Faraday cage further reduced the faceting and narrowing in both plasmas because the neutral beams were not affected by negative charges accumulating on the sidewalls.
Acknowledgements This work was supported by the Brain Korea 21 project, the Center for Ultramicrochemical Process Systems (CUPS) designated by KOSEF, and National Research Laboratory program of Ministry of Science and Technology, Korea. One of the authors (C.-K. Kim) appreciates the financial support from the Basic Research Program of the Korea Science and Engineering Foundation (grant No. R01-2003000-10103-0), and the Ajou University Research Facilities Support Program in the year 2003.
References [1] S.G. Ingram, J. Appl. Phys. 68 (1990) 500. [2] B. Chapman, Glow Discharge Processes, Wiley, New York, 1980. [3] D.M. Manos, D.L. Flamm, Plasma Etching, Academic Press, San Diego, CA, 1989. [4] J.C. Arnold, H.H. Sawin, J. Appl. Phys. 70 (1991) 5314. [5] M. Schaepkens, G.S. Oehrlein, Appl. Phys. Lett. 72 (1998) 1293. [6] T. Yunogami, K. Yokogawa, T. Mizutani, J. Vac. Sci. Technol., A 13 (1995) 952. [7] B.-O. Cho, S.-W. Hwang, I.-W. Kim, S.H. Moon, J. Electrochem. Soc. 146 (1999) 350. [8] G.D. Boyd, L.A. Coldren, F.G. Storz, Appl. Phys. Lett. 36 (1980) 583. [9] B.-O. Cho, S.-W. Hwang, J.-H. Ryu, S.H. Moon, Rev. Sci. Instrum. 70 (1999) 2458. [10] J.-H. Ryu, B.-O. Cho, S.-W. Hwang, S.H. Moon, C.-K. Kim, Korean J. Chem. Eng. 20 (2003) 407. [11] B.-O. Cho, S.-W. Hwang, G.-R. Lee, S.H. Moon, J. Vac. Sci. Technol., A 18 (2000) 2791. [12] G.-R. Lee, S.-W. Hwang, J.-H. Min, S.H. Moon, J. Vac. Sci. Technol., A 20 (2002) 1808. [13] S.-W. Hwang, G.-R. Lee, J.-H. Min, S.H. Moon, Electrochem. SolidState Lett. 6 (2003) G12. [14] J.-H. Min, S.-W. Hwang, G.-R. Lee, S.H. Moon, J. Vac. Sci. Technol., A 20 (2002) 1574. [15] H.F. Winters, J.W. Coburn, T.J. Chuang, J. Vac. Sci. Technol., B 1 (1983) 469. [16] S. Tachi, S. Okudaira, J. Vac. Sci. Technol., B 4 (1986) 459. [17] T.M. Mayer, R.A. Barker, L.J. Whitman, J. Vac. Sci. Technol. 18 (1981) 349. [18] R.E. Lee, J. Vac. Sci. Technol. 16 (1979) 164. [19] D. Zhang, M.J. Kushner, J. Vac. Sci. Technol., A 19 (2000) 524.