- Email: [email protected]
Control of the electron temperature by varying the resonance zone width in ECR plasma
Thin Solid Films 457 (2004) 59–63
Control of the electron temperature by varying the resonance zone width in ECR plasma N. Itagakia,*, H. Mutaa, N. Ishiib, Y. Kawaia a
Interdisciplinary Graduate School of Engineering Sciences, Kyushu University, Kasuga, Fukuoka 816-8580, Japan b Tokyo Electron Co. Ltd., Yodogawa, Osaka 532-0003, Japan
Abstract The electron temperature (Te) in an electron cyclotron resonance plasma is clarified to depend on the spatial profiles of the microwave-power absorption by both the electromagnetic-waves measurement and the simulation of microwave power absorption. It is found that Te is controlled by varying the magnetic field configuration andyor the microwave frequency since the power absorption profile is influenced by the effective resonance width. In fact, Te is observed to decrease with decreasing the magnetic field gradient at the resonance point for N2, Ar and O2yAr plasma. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Plasma processing and deposition; Electron cyclotron resonance plasma; Electron-temperature control
1. Introduction Electron cyclotron resonance (ECR) discharges have attracted considerable attention due to their advantages which allow their use in several industrial processes such as submicron etching, thin-film deposition, etc. w1,2x. A significant feature of ECR plasma sources is the high electron density (ne), which can be achieved at low gas pressures. Specifically, as the device and feature sizes decrease, an ECR plasma becomes indispensable for the etching process. Recently, great interest has been directed toward the electron temperature (Te) control in order to realize the less-damage processing, because high-Te plasma sources have been shown to cause serious problems such as substrate damage due to ion bombardment from plasma since the ion bombardment deteriorates the film quality in the case where the incident ion energy to the substrate, which usually depends on the electron temperature, is high. Moreover, in the case of plasma etching, charge build-up damage, low etching selectivity and local side etching are caused by the high degree of dissociation and charge accumulation on the substrate by high-energy electrons in the plasma w3–5x. Since the production of reactive species including plasma production or the decomposition of *Corresponding author. Tel.: q81-92-583-7651; fax: q81-92-5718894. E-mail address: [email protected] (N. Itagaki).
source gas molecules are triggered by the electron collisions, Te control is also essential to find out the best conditions necessary for qualified material processing. However, Te in a conventional ECR plasma produced by 2.45 GHz microwave is relatively high, and it is quite hard to control Te in a wide range. Up to now, several attempts for producing an ECR plasma with variable Te have been reported. In our previous experiment, the mean Te was decreased by pulse modulation of the incident microwave power, and high-quality amorphous Si thin films were obtained at room temperature using such control w6x. Fukumasa et al. observed that applying the magnetic filter, which is the localized transverse magnetic field of sheet type geometry and has an ability to reflect preferentially the high energy electrons and pass through low energy ones, leads to reduction of Te w7x. Kato et al. controlled Te by inserting a mesh grid into plasma w8x. However, the Te control with keeping the high electron density, which is earnestly required from industry, has not been achieved with those methods. Recently, the effects of the electromagnetic-wave frequency on the plasma parameters have been giving some attention and it has been reported that in the plasma produced by the microwaves of the frequency below 2.45 GHz, Te and the density of high energy electrons that contributes to ionization are low w9x. In our previous experiment, it was also observed that a
0040-6090/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2003.12.014
60
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
Fig. 2. The contour diagrams corresponding to the electric field strength made out of interferometer traces at the electron temperature of (a) 2 eV and (b) 7 eV. Here the electron density was approximately 2=1017 my3 for each case.
low-Te ECR plasma with high ne was produced by using 915 MHz microwaves and Te in a 915 MHz ECR plasma could be easily controlled by adjusting the external conditions such as incident microwave power, gas pressure and magnetic field configuration w10x. The numerical simulation of the microwave-power absorption suggested that Te in an ECR plasma changes with varying the spatial profile of the power absorption w11x; however, the mechanism of low-Te plasma production by using the microwaves of lower frequency has not still been clarified. In this study, we tried to experimentally clarify the relationship between Te and the powerabsorption profile by measuring the spatial profiles of the plasma parameters and wave patterns of the electromagnetic waves. The results are also compared with those of 2.45 GHz ECR plasma in order to investigate the effect of the microwave frequency on Te. In addition, based on the experimental and numerical results, a new simple method of Te control for an ECR plasma is proposed.
of stainless steel with an inner diameter of 290 mm and a length of 1200 mm. The chamber wall was grounded. The microwave was introduced through a quartz window and a substrate holder was placed approximately 550 mm from the window. The chamber was evacuated using a rotary pump and a 2000 lys turbomolecular pump to a base pressure of less than 2=10y6 Torr. N2, Ar or O2 gas was introduced into the chamber at a total flow rate of 50 sccm and the operating pressure was selected to be 5=10y3 Torr. Six magnetic coils with a thickness of 100 mm and an inner diameter of 320 mm were placed adjacent to the chamber to control the magnetic field distribution. The resonant magnetic field for a frequency of 915 MHz and 2.45 GHz were 0.0327 T and 0.0875 T, respectively, and the position was set at approximately 240 mm from the window, which is shown in Fig. 1b. Microwaves were converted from the rectangular TE10 mode to the circular TM01 or TE11 mode by a mode converter and were launched into the chamber through a waveguide uptaper w12x. The microwave power could be up to 5 kW. The plasma parameters were measured with a three-dimensional movable cylindrical Langmuir probe whose radius and length were 0.5 mm and 1 mm, respectively. The used loop antenna was made of a coaxial cable with heat-resistance (1000 8C) and the wave patterns were obtained by the interferometric method. A reference signal of electromagnetic waves from a directional coupler and signal of waves in plasma are introduced into a balanced mixer, whose output shows the phase difference of these waves depending on the positions of a loop antenna. The amplitudes of wave patterns are given with arbitrary units because the loop antenna was not calibrated.
2. Experiment
3. Results and discussion
A schematic diagram of the experimental apparatus is shown in Fig. 1a. The cylindrical chamber was made
In order to clarify the actual relationship between Te and the power-absorption profile in an ECR plasma, the
Fig. 1. Experimental apparatus: (a) the experimental setup, (b) the axial profiles of magnetic field at the center for 915 MHz ECR plasma.
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
Fig. 3. The spatial profiles of the amplitude of radial electric fields numerically obtained by assuming that the plasma profiles were similar to the experimental results at the electron temperature of (a) 2 eV and (b) 7 eV.
Fig. 4. The spatial distributions of the power absorption numerically obtained by assuming that the plasma profiles were similar to the experimental results at the electron temperature of (a) 2 eV and (b) 7 eV.
Fig. 5. Measured spatial profiles of the amplitude of the microwave electric field at the microwave frequency of (a) 915 MHz and (b) 2.45 GHz.
spatial wave patterns were firstly measured in a N2 plasma. Fig. 2 shows the spatial distributions of the amplitude of radial electric field for different Te, which was changed independently by adjusting the external conditions w11x. In this experiment, Te was confirmed
61
to be independent of both the longitudinal and radial positions for any external condition, which is considered to be due to the high thermal conductivity of electrons. The contour diagrams corresponding to the electric field strength were made using interferometer traces, and observed waves were confirmed to be the right-handed circularly polarized wave from the dispersion relations. As seen in Fig. 2a, the microwaves injected in TM01 mode, which has an electric field profile peaked near the waveguide wall, propagated not only in the vicinity of the chamber wall but also toward the center of the chamber, consequently, the microwave diverged along the z-axis at Tes2 eV. However, it was observed that the microwave gradually refracted toward the chamber wall along z-axis at Te of 7 eV, which is shown in Fig. 2b. These interferometer traces suggest that the spatial profile of the microwave power absorption at Tes2 eV was different from that at Tes7 eV, but do not show which way the power was really transported. Therefore, the spatial distribution of microwave electric fields and the power absorption corresponding to the interferometer traces were calculated to investigate the correlation between Te and the power-absorption profile, which was performed by using a simulation code ‘TASKyWF’ w13x. This is a code that uses the finite-element method to solve boundary value problems of the Maxwell equation for stationary oscillation of electromagnetic waves, and makes two-dimensional analysis of wave propagation in a partially filled plasma chamber possible. Fig. 3 and Fig. 4 show the spatial distributions of the amplitude of the radial electric field and the power absorption numerically obtained by assuming that the plasma profiles were similar to the experimental results. The contour diagrams correspond to the strength of the electric field and the power absorption. As seen in Fig. 3a and Fig. 4a, the microwave diverged along the z-axis at Tes2 eV, and the ones refracted toward the chamber wall at Tes7 eV, which was consistent with the experimental results. In the former case, the positions where the power absorption took place spread out in both of r and z directions, however, the power absorption profile peaked strongly near the window in the vicinity of the chamber wall in the latter case, which was seen in Fig. 3b and Fig. 4b. From these experimental and numerical results, it is indicated that Te in an ECR plasma depends on the power absorption profiles, that is, Te is low when the power absorption takes place in a wide area and Te increases when the power absorption is concentrated. As mentioned in Section 1, Te in a 2.45 GHz ECR plasma is higher than that in a 915 MHz ECR plasma, whose reason is considered as follows. Since the effective resonance zone width (Dzres) is in inverse proportion to yv, where v is the microwave angular frequency, Dzres becomes narrower as v becomes higher w14x. Therefore, the power absorption profiles at higher microwave frequency are expected to peak strongly and to be
62
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
Fig. 6. Measured spatial profiles of the amplitude of the microwave electric field for different magnetic field configurations. The gradient in the magnetic field strength at the resonance point was (a) 0.03 Tym and (b) 0.14 Tym.
Fig. 8. The dependences of (a) the electron temperature and (b) the positive ion density on the magnetic field gradient for different microwave powers in Ar plasma.
changed little with changing the external conditions. Consequently, Te becomes high and cannot be controlled in a wide range in a 2.45 GHz ECR plasma. Fig. 5 shows the measured spatial profiles of the electric-field amplitude at the microwave frequency of (a) 915 MHz and (b) 2.45 GHz, respectively. The gradient in the magnetic field strength at ECR point (±dBydz±res) was approximately 0.03 Tym for each cases, whose configuration is similar to the magnetic field configuration 1 in Fig. 1b. Here, the gas used was N2, and the incident power was 0.5 kW. As seen in Fig. 5, the microwaves gradually damped along the z-axis in the case of 915 MHz, while, the waves abruptly damped in the 2.45 GHz ECR plasma, which suggests that the area where the power absorption takes place in a 915 MHz ECR plasma is much larger than that in a 2.45 GHz ECR plasma. In our previous experiments, it was also observed that Te in an 915 MHz ECR plasma changed with changing the magnetic field configuration w10,11x, whose reason is considered in the same way. Dzres is not only proportional to vy1y2 but also)dBydz)resy1y2, so that Dzres
becomes narrower as ±dBydz±res becomes largerw14x. As a result, Te in the magnetic field configuration whose gradient is steep tends to be high. In fact, we observed that the microwaves gradually damped along the z-axis and propagated for several wavelengths up to the ECR point when ±dBydz±res was small, while the microwaves abruptly damped when ±dBydz±res was large, which is shown in Fig. 6. Since the relationship between Te and the power absorption profile was experimentally clarified in this way, we next tried to control Te by varying ±dBydz±res which determines Dzres as mentioned above. Here, the microwave frequency was 915 MHz and the measuring point was zs400 mm and rs0 mm. For different microwave powers, it was observed that Te increased while the positive ion density changed little with increasing ±dBydz±res from 0.03 to 0.14 Tym in N2, Ar and O2 y Ar plasmas, which are widely used in semiconductor processing, as shown in Figs. 7–9. In addition, it was confirmed that the positive ion density decreases linearly along the axis while Te was independent of the axial position except around the ECR point for any magnetic configuration type. This experimental result suggests that changing ±dBydz±res is a convenient method to
Fig. 7. The dependences of (a) the electron temperature and (b) the positive ion density on the magnetic field gradient for different microwave powers in N2 plasma.
Fig. 9. The dependences of the electron temperature and the positive ion density on the magnetic field gradient for different microwave powers in O2yAr plasma. Here O2 gas flow rate was 10%.
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
control Te because control of the magnetic coil current in order to change ±dBydz±res is easy. Thus, our experimental and numerical results suggest that Te in an ECR plasma can be controlled by changing the microwave frequency andyor ±dBydz±res because the power absorption profile is controlled by changing Dzres. Furthermore, N2, Ar and O2 gases were used to investigate Te control in our experiments, however, many kinds of reactive gases such as SiH4, CF4, C4F8, Cl2 and SF6 are used in the plasma processing and the plasma parameters depend on the gas species. It is necessary to clarify Te control in the reactive plasmas, which will be future work. 4. Conclusions In order to investigate the production mechanism of a low-electron-temperature ECR plasma, the relationship between the electron temperature and the power-absorption profile was studied. The spatial profiles of the electromagnetic wave patterns were measured for different electron temperatures, which suggested that the electron temperature changes with changing the spatial profiles of the power absorption. This dependence of the electron temperature on the power absorption profile was examined by numerical simulation, and it was confirmed that the spatial profile of the microwave power absorped by plasma has an effect on the electron temperature in an ECR plasma, that is, the electron temperature is low when the power absorption takes place in a wide area and the electron temperature increases when the power absorption is concentrated. Furthermore, our experimental and numerical results suggest that the electron temperature in an ECR plasma is controlled by changing the gradient in the magnetic
63
field strength near the resonant zone andyor the microwave frequency because the power absorption profile is changed with changing the effective resonance zone width. In fact, the electron temperature was observed to decrease with decreasing the magnetic field gradient at the resonance point for N2, Ar and O2 yAr plasma. Acknowledgments The authors would like to thank Prof. A. Fukuyama for the introduction of his calculation code. References w1x S. Matsuo, M. Kikuchi, Jpn. J. Appl. Phys. 22 (1983) L210. w2x K. Suzuki, S. Okudaira, N. Sakudo, I. Kanomata, Jpn. J. Appl. Phys. 16 (1977) 1979. w3x S. Samukawa, Y. Nakagawa, T. Tsukada, H. Ueyama, K. Shinohara, Jpn. J. Appl. Phys. 34 (1995) 6805. w4x S. Samukawa, Y. Nakagawa, T. Tsukada, H. Ueyama, K. Shinohara, Appl. Phys. Lett. 67 (1995) 1414. w5x S. Samukawa, Jpn. J. Appl. Phys. 33 (1994) 2133. w6x N. Itagaki, A. Fukuda, T. Yoshizawa, M. Shindo, Y. Ueda, Y. Kawai, Surf. Coat. Technol. 131 (2000) 54. w7x O. Fukumasa, H. Naitou, S. Sakiyama, J. Appl. Phys. 74 (1993) 848. w8x K. Kato, S. Iizuka, N. Sato, Appl. Phys. Lett. 65 (1994) 816. w9x K. Yokogawa, N. Itabashi, K. Suzuki, S. Tachi, Proceedings of the 43th National Symposium of American Vacuum Society, 1996, PS2-WeA5. w10x N. Itagaki, S. Kawakami, N. Ishii, Y. Kawai, Vacuum 66 (2002) 323. w11x N. Itagaki, S. Kawakami, N. Ishii, Y. Kawai, J. Vac. Sci. Technol. A20 (2002) 1967. w12x S. Samukawa, J. Vac. Sci. Technol. A11 (1993) 2572. w13x Y. Yasaka, A. Fukuyama, A. Hatta, R. Itatani, J. Appl. Phys. 72 (1992) 2652. w14x M.A. Lieberman, A.J. Lichtenberg, Principles of Plasma Discharges and Materials Processing, John Wiley and Sons, Inc, New York, 1994, Chapter 13.
Control of the electron temperature by varying the resonance zone width in ECR plasma N. Itagakia,*, H. Mutaa, N. Ishiib, Y. Kawaia a
Interdisciplinary Graduate School of Engineering Sciences, Kyushu University, Kasuga, Fukuoka 816-8580, Japan b Tokyo Electron Co. Ltd., Yodogawa, Osaka 532-0003, Japan
Abstract The electron temperature (Te) in an electron cyclotron resonance plasma is clarified to depend on the spatial profiles of the microwave-power absorption by both the electromagnetic-waves measurement and the simulation of microwave power absorption. It is found that Te is controlled by varying the magnetic field configuration andyor the microwave frequency since the power absorption profile is influenced by the effective resonance width. In fact, Te is observed to decrease with decreasing the magnetic field gradient at the resonance point for N2, Ar and O2yAr plasma. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Plasma processing and deposition; Electron cyclotron resonance plasma; Electron-temperature control
1. Introduction Electron cyclotron resonance (ECR) discharges have attracted considerable attention due to their advantages which allow their use in several industrial processes such as submicron etching, thin-film deposition, etc. w1,2x. A significant feature of ECR plasma sources is the high electron density (ne), which can be achieved at low gas pressures. Specifically, as the device and feature sizes decrease, an ECR plasma becomes indispensable for the etching process. Recently, great interest has been directed toward the electron temperature (Te) control in order to realize the less-damage processing, because high-Te plasma sources have been shown to cause serious problems such as substrate damage due to ion bombardment from plasma since the ion bombardment deteriorates the film quality in the case where the incident ion energy to the substrate, which usually depends on the electron temperature, is high. Moreover, in the case of plasma etching, charge build-up damage, low etching selectivity and local side etching are caused by the high degree of dissociation and charge accumulation on the substrate by high-energy electrons in the plasma w3–5x. Since the production of reactive species including plasma production or the decomposition of *Corresponding author. Tel.: q81-92-583-7651; fax: q81-92-5718894. E-mail address: [email protected] (N. Itagaki).
source gas molecules are triggered by the electron collisions, Te control is also essential to find out the best conditions necessary for qualified material processing. However, Te in a conventional ECR plasma produced by 2.45 GHz microwave is relatively high, and it is quite hard to control Te in a wide range. Up to now, several attempts for producing an ECR plasma with variable Te have been reported. In our previous experiment, the mean Te was decreased by pulse modulation of the incident microwave power, and high-quality amorphous Si thin films were obtained at room temperature using such control w6x. Fukumasa et al. observed that applying the magnetic filter, which is the localized transverse magnetic field of sheet type geometry and has an ability to reflect preferentially the high energy electrons and pass through low energy ones, leads to reduction of Te w7x. Kato et al. controlled Te by inserting a mesh grid into plasma w8x. However, the Te control with keeping the high electron density, which is earnestly required from industry, has not been achieved with those methods. Recently, the effects of the electromagnetic-wave frequency on the plasma parameters have been giving some attention and it has been reported that in the plasma produced by the microwaves of the frequency below 2.45 GHz, Te and the density of high energy electrons that contributes to ionization are low w9x. In our previous experiment, it was also observed that a
0040-6090/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2003.12.014
60
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
Fig. 2. The contour diagrams corresponding to the electric field strength made out of interferometer traces at the electron temperature of (a) 2 eV and (b) 7 eV. Here the electron density was approximately 2=1017 my3 for each case.
low-Te ECR plasma with high ne was produced by using 915 MHz microwaves and Te in a 915 MHz ECR plasma could be easily controlled by adjusting the external conditions such as incident microwave power, gas pressure and magnetic field configuration w10x. The numerical simulation of the microwave-power absorption suggested that Te in an ECR plasma changes with varying the spatial profile of the power absorption w11x; however, the mechanism of low-Te plasma production by using the microwaves of lower frequency has not still been clarified. In this study, we tried to experimentally clarify the relationship between Te and the powerabsorption profile by measuring the spatial profiles of the plasma parameters and wave patterns of the electromagnetic waves. The results are also compared with those of 2.45 GHz ECR plasma in order to investigate the effect of the microwave frequency on Te. In addition, based on the experimental and numerical results, a new simple method of Te control for an ECR plasma is proposed.
of stainless steel with an inner diameter of 290 mm and a length of 1200 mm. The chamber wall was grounded. The microwave was introduced through a quartz window and a substrate holder was placed approximately 550 mm from the window. The chamber was evacuated using a rotary pump and a 2000 lys turbomolecular pump to a base pressure of less than 2=10y6 Torr. N2, Ar or O2 gas was introduced into the chamber at a total flow rate of 50 sccm and the operating pressure was selected to be 5=10y3 Torr. Six magnetic coils with a thickness of 100 mm and an inner diameter of 320 mm were placed adjacent to the chamber to control the magnetic field distribution. The resonant magnetic field for a frequency of 915 MHz and 2.45 GHz were 0.0327 T and 0.0875 T, respectively, and the position was set at approximately 240 mm from the window, which is shown in Fig. 1b. Microwaves were converted from the rectangular TE10 mode to the circular TM01 or TE11 mode by a mode converter and were launched into the chamber through a waveguide uptaper w12x. The microwave power could be up to 5 kW. The plasma parameters were measured with a three-dimensional movable cylindrical Langmuir probe whose radius and length were 0.5 mm and 1 mm, respectively. The used loop antenna was made of a coaxial cable with heat-resistance (1000 8C) and the wave patterns were obtained by the interferometric method. A reference signal of electromagnetic waves from a directional coupler and signal of waves in plasma are introduced into a balanced mixer, whose output shows the phase difference of these waves depending on the positions of a loop antenna. The amplitudes of wave patterns are given with arbitrary units because the loop antenna was not calibrated.
2. Experiment
3. Results and discussion
A schematic diagram of the experimental apparatus is shown in Fig. 1a. The cylindrical chamber was made
In order to clarify the actual relationship between Te and the power-absorption profile in an ECR plasma, the
Fig. 1. Experimental apparatus: (a) the experimental setup, (b) the axial profiles of magnetic field at the center for 915 MHz ECR plasma.
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
Fig. 3. The spatial profiles of the amplitude of radial electric fields numerically obtained by assuming that the plasma profiles were similar to the experimental results at the electron temperature of (a) 2 eV and (b) 7 eV.
Fig. 4. The spatial distributions of the power absorption numerically obtained by assuming that the plasma profiles were similar to the experimental results at the electron temperature of (a) 2 eV and (b) 7 eV.
Fig. 5. Measured spatial profiles of the amplitude of the microwave electric field at the microwave frequency of (a) 915 MHz and (b) 2.45 GHz.
spatial wave patterns were firstly measured in a N2 plasma. Fig. 2 shows the spatial distributions of the amplitude of radial electric field for different Te, which was changed independently by adjusting the external conditions w11x. In this experiment, Te was confirmed
61
to be independent of both the longitudinal and radial positions for any external condition, which is considered to be due to the high thermal conductivity of electrons. The contour diagrams corresponding to the electric field strength were made using interferometer traces, and observed waves were confirmed to be the right-handed circularly polarized wave from the dispersion relations. As seen in Fig. 2a, the microwaves injected in TM01 mode, which has an electric field profile peaked near the waveguide wall, propagated not only in the vicinity of the chamber wall but also toward the center of the chamber, consequently, the microwave diverged along the z-axis at Tes2 eV. However, it was observed that the microwave gradually refracted toward the chamber wall along z-axis at Te of 7 eV, which is shown in Fig. 2b. These interferometer traces suggest that the spatial profile of the microwave power absorption at Tes2 eV was different from that at Tes7 eV, but do not show which way the power was really transported. Therefore, the spatial distribution of microwave electric fields and the power absorption corresponding to the interferometer traces were calculated to investigate the correlation between Te and the power-absorption profile, which was performed by using a simulation code ‘TASKyWF’ w13x. This is a code that uses the finite-element method to solve boundary value problems of the Maxwell equation for stationary oscillation of electromagnetic waves, and makes two-dimensional analysis of wave propagation in a partially filled plasma chamber possible. Fig. 3 and Fig. 4 show the spatial distributions of the amplitude of the radial electric field and the power absorption numerically obtained by assuming that the plasma profiles were similar to the experimental results. The contour diagrams correspond to the strength of the electric field and the power absorption. As seen in Fig. 3a and Fig. 4a, the microwave diverged along the z-axis at Tes2 eV, and the ones refracted toward the chamber wall at Tes7 eV, which was consistent with the experimental results. In the former case, the positions where the power absorption took place spread out in both of r and z directions, however, the power absorption profile peaked strongly near the window in the vicinity of the chamber wall in the latter case, which was seen in Fig. 3b and Fig. 4b. From these experimental and numerical results, it is indicated that Te in an ECR plasma depends on the power absorption profiles, that is, Te is low when the power absorption takes place in a wide area and Te increases when the power absorption is concentrated. As mentioned in Section 1, Te in a 2.45 GHz ECR plasma is higher than that in a 915 MHz ECR plasma, whose reason is considered as follows. Since the effective resonance zone width (Dzres) is in inverse proportion to yv, where v is the microwave angular frequency, Dzres becomes narrower as v becomes higher w14x. Therefore, the power absorption profiles at higher microwave frequency are expected to peak strongly and to be
62
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
Fig. 6. Measured spatial profiles of the amplitude of the microwave electric field for different magnetic field configurations. The gradient in the magnetic field strength at the resonance point was (a) 0.03 Tym and (b) 0.14 Tym.
Fig. 8. The dependences of (a) the electron temperature and (b) the positive ion density on the magnetic field gradient for different microwave powers in Ar plasma.
changed little with changing the external conditions. Consequently, Te becomes high and cannot be controlled in a wide range in a 2.45 GHz ECR plasma. Fig. 5 shows the measured spatial profiles of the electric-field amplitude at the microwave frequency of (a) 915 MHz and (b) 2.45 GHz, respectively. The gradient in the magnetic field strength at ECR point (±dBydz±res) was approximately 0.03 Tym for each cases, whose configuration is similar to the magnetic field configuration 1 in Fig. 1b. Here, the gas used was N2, and the incident power was 0.5 kW. As seen in Fig. 5, the microwaves gradually damped along the z-axis in the case of 915 MHz, while, the waves abruptly damped in the 2.45 GHz ECR plasma, which suggests that the area where the power absorption takes place in a 915 MHz ECR plasma is much larger than that in a 2.45 GHz ECR plasma. In our previous experiments, it was also observed that Te in an 915 MHz ECR plasma changed with changing the magnetic field configuration w10,11x, whose reason is considered in the same way. Dzres is not only proportional to vy1y2 but also)dBydz)resy1y2, so that Dzres
becomes narrower as ±dBydz±res becomes largerw14x. As a result, Te in the magnetic field configuration whose gradient is steep tends to be high. In fact, we observed that the microwaves gradually damped along the z-axis and propagated for several wavelengths up to the ECR point when ±dBydz±res was small, while the microwaves abruptly damped when ±dBydz±res was large, which is shown in Fig. 6. Since the relationship between Te and the power absorption profile was experimentally clarified in this way, we next tried to control Te by varying ±dBydz±res which determines Dzres as mentioned above. Here, the microwave frequency was 915 MHz and the measuring point was zs400 mm and rs0 mm. For different microwave powers, it was observed that Te increased while the positive ion density changed little with increasing ±dBydz±res from 0.03 to 0.14 Tym in N2, Ar and O2 y Ar plasmas, which are widely used in semiconductor processing, as shown in Figs. 7–9. In addition, it was confirmed that the positive ion density decreases linearly along the axis while Te was independent of the axial position except around the ECR point for any magnetic configuration type. This experimental result suggests that changing ±dBydz±res is a convenient method to
Fig. 7. The dependences of (a) the electron temperature and (b) the positive ion density on the magnetic field gradient for different microwave powers in N2 plasma.
Fig. 9. The dependences of the electron temperature and the positive ion density on the magnetic field gradient for different microwave powers in O2yAr plasma. Here O2 gas flow rate was 10%.
N. Itagaki et al. / Thin Solid Films 457 (2004) 59–63
control Te because control of the magnetic coil current in order to change ±dBydz±res is easy. Thus, our experimental and numerical results suggest that Te in an ECR plasma can be controlled by changing the microwave frequency andyor ±dBydz±res because the power absorption profile is controlled by changing Dzres. Furthermore, N2, Ar and O2 gases were used to investigate Te control in our experiments, however, many kinds of reactive gases such as SiH4, CF4, C4F8, Cl2 and SF6 are used in the plasma processing and the plasma parameters depend on the gas species. It is necessary to clarify Te control in the reactive plasmas, which will be future work. 4. Conclusions In order to investigate the production mechanism of a low-electron-temperature ECR plasma, the relationship between the electron temperature and the power-absorption profile was studied. The spatial profiles of the electromagnetic wave patterns were measured for different electron temperatures, which suggested that the electron temperature changes with changing the spatial profiles of the power absorption. This dependence of the electron temperature on the power absorption profile was examined by numerical simulation, and it was confirmed that the spatial profile of the microwave power absorped by plasma has an effect on the electron temperature in an ECR plasma, that is, the electron temperature is low when the power absorption takes place in a wide area and the electron temperature increases when the power absorption is concentrated. Furthermore, our experimental and numerical results suggest that the electron temperature in an ECR plasma is controlled by changing the gradient in the magnetic
63
field strength near the resonant zone andyor the microwave frequency because the power absorption profile is changed with changing the effective resonance zone width. In fact, the electron temperature was observed to decrease with decreasing the magnetic field gradient at the resonance point for N2, Ar and O2 yAr plasma. Acknowledgments The authors would like to thank Prof. A. Fukuyama for the introduction of his calculation code. References w1x S. Matsuo, M. Kikuchi, Jpn. J. Appl. Phys. 22 (1983) L210. w2x K. Suzuki, S. Okudaira, N. Sakudo, I. Kanomata, Jpn. J. Appl. Phys. 16 (1977) 1979. w3x S. Samukawa, Y. Nakagawa, T. Tsukada, H. Ueyama, K. Shinohara, Jpn. J. Appl. Phys. 34 (1995) 6805. w4x S. Samukawa, Y. Nakagawa, T. Tsukada, H. Ueyama, K. Shinohara, Appl. Phys. Lett. 67 (1995) 1414. w5x S. Samukawa, Jpn. J. Appl. Phys. 33 (1994) 2133. w6x N. Itagaki, A. Fukuda, T. Yoshizawa, M. Shindo, Y. Ueda, Y. Kawai, Surf. Coat. Technol. 131 (2000) 54. w7x O. Fukumasa, H. Naitou, S. Sakiyama, J. Appl. Phys. 74 (1993) 848. w8x K. Kato, S. Iizuka, N. Sato, Appl. Phys. Lett. 65 (1994) 816. w9x K. Yokogawa, N. Itabashi, K. Suzuki, S. Tachi, Proceedings of the 43th National Symposium of American Vacuum Society, 1996, PS2-WeA5. w10x N. Itagaki, S. Kawakami, N. Ishii, Y. Kawai, Vacuum 66 (2002) 323. w11x N. Itagaki, S. Kawakami, N. Ishii, Y. Kawai, J. Vac. Sci. Technol. A20 (2002) 1967. w12x S. Samukawa, J. Vac. Sci. Technol. A11 (1993) 2572. w13x Y. Yasaka, A. Fukuyama, A. Hatta, R. Itatani, J. Appl. Phys. 72 (1992) 2652. w14x M.A. Lieberman, A.J. Lichtenberg, Principles of Plasma Discharges and Materials Processing, John Wiley and Sons, Inc, New York, 1994, Chapter 13.