Langmuir Probes Lambda Resonatorl Plasma G Vinogradov

Probe diagnostics in a full wave resonator radio-frequency discharge G. K. Vinogradov,a) V. M. Menagarishvili, and S. Yoneyama MC Electronics Co., Ltd., 907-8, Shimoimasuwa, Shirane-cho, Nakakomagun, 400-02 Yamanashi, Japan

~Received 6 March 1997; accepted 23 March 1998! A full wave or lambda resonator ~l-R! is a capacitively balanced radio-frequency ~rf! inductive plasma source. It has three separate inductive excitation zones with opposite magnetic momenta strictly located at their axial positions. The 2 kW, 27 MHz, 1.4 Torr pressure discharge in oxygen was studied by means of movable single rf compensated fine cylindrical Langmuir probes. Flat wall probes were also used to reveal the distribution of positive ion flux and floating potentials on the chamber wall. The electron density in the l-R discharge varies from 107 – 108 cm23 at 5–10 mm distance from the wafer to 231011 cm23 in the central plasma toroid. One of the main problems of the probe operation is high gas temperature of ;1500 K in the plasma toroid and high-power dissipation on the probe surface. That is why not only fine cylindrical but also 100–130 mm diam spherical probes were used. The probe technique and preliminary results are presented. © 1998 American Vacuum Society. @S0734-2101~98!60103-2#

I. INTRODUCTION

II. EXPERIMENTAL DETAILS

There is a variety of inductive plasma sources utilized by industry in the last decade. Correspondingly, there are numerous publications about plasma parameters of the discharges. Mainly, the inductive discharges aimed to produce high-density plasmas at a very low pressure of about 1023 – 1024 Torr for etching and deposition microelectronics equipment. Plasma parameters in different plasma sources are essentially determined by the mechanisms of energy deposition. For instance, a conventional source with a cylindrical spiral and helicon sources with nonspiral applicators have different configurations of electromagnetic fields1,2 and, hence, different distributions of dissipated radio-frequency ~rf! power. Therefore, it is hardly possible to predict a priori all practical features of different plasma sources. A novel plasma source, a full wave helical resonator or lambda resonator ~l-R!, has been developed and manufactured for damage critical fast plasma processes. The first application is realized as a single 200 mm wafer quasidownstream asher l-Strip 3000. ~l-3000 is a product of MC Electronics Co., Ltd.! However, the explicit plasma physics of the discharge is unknown. The discharge has essentially a three-dimensional internal inductive/capacitive symmetric/antisymmetric orthogonal structure. To the best of our knowledge such a structure, having three independent plasma toroids with opposite magnetic momenta connected by capacitive currents,3 had not been reported previously. Here, we present preliminary results on probe diagnostics of the l-R discharge; mainly of its central high plasma density inductive zone. The measurements were carried out under conditions of relatively high gas pressure and high plasma density with very fine probes. The details of the probe measuring technique and experimental results are presented.

The l-R plasma source has been described previously.3,4 The basic configuration of the recent experimental setup is shown in Fig. 1. Both ends of the inductor coil are connected to the cylindrical grounded copper shield. The reactor dimensions and positioning of electrostatic probes is indicated. The downstream chamber is made of aluminum. The 0.005–2.25 kW rf power at 26–28 MHz frequency was supplied to the resonator by using a 50 V coaxial cable directly from the wideband tube amplifier ~IFI-410! with a sign wave signal generator HP-8648A. The rf power absorbed in the resonator was measured with a Bird reflectometer. The plasma source has energy efficiency up to ;99.9% and operates without any matching elements. It was directly measured neglecting heat losses in the coaxial; the resonator itself has a quality factor Q of 2360. So, high-energy efficiency is typical since the plasma is practically the only load absorbing rf power in the resonator. The device responsible for the main losses in any conventional inductively coupled plasma case is a matching box typically absorbing 20%–60% of the total rf power.5 A feedback loop control was implemented in order to keep the input rf voltage on the plasma source constant to suppress very low-frequency fluctuations of the discharge parameters brought about by heat-ionization instabilities. A wafer support platen is located at 120-mm-distance below the bottom ground end of the coils. Standard discharge conditions were: 1.4 Torr O2 pressure; 3 slm gas flow rate; and 2 kW discharge power at 27 MHz unless otherwise stated. Argon discharges were examined in several cases as well. Several kinds of electrostatic probes, as shown in Fig. 2, were used in this study. The cylindrical Langmuir probes were made of ~1–1.5!-mm-long 20-mm-diam Pt wire. The spherical Pt probes are of about 100-mm-diam and have a short 20-mm-diam leg. The Langmuir probes were inserted into the discharge tube from the downstream chamber. The probes can be moved along the tube radius by way of turning the probe input shaft having a sealing O ring or using a

a!

Author to whom all correspondence should be addressed; electronic mail: [email protected]

1444

J. Vac. Sci. Technol. A 16„3…, May/Jun 1998

0734-2101/98/16„3…/1444/5/$15.00

©1998 American Vacuum Society

1444

1445

Vinogradov, Menagarishvili, and Yoneyama: Probe diagnostics in a full wave resonator rf discharge

FIG. 1. Lambda resonator discharge apparatus, a schematic cross-sectional view. Langmuir probe position in the plasma toroid and flat wall probes on the bottom flange, chamber wall, and wafer platen are indicated.

horizontal step motor drive. A calibration of the probe position in the vicinity of the wall has been done to assure reasonable precision within the plasma-surface sheath area. The probes can touch the wall. The spherical Langmuir probes were used in the plasma toroid channel because it was difficult to maintain a high electron saturation current on the cylindrical probes at the plasma potential: they usually melted. The tiny sphere probes withstand a high-temperature environment much better.

FIG. 2. Fine platinum probes ~a! and ~b!: quartz capillaries are 70–90 mm diam; and flat wall probe ~c!, cross section. Probes are shown in different scales. JVST A - Vacuum, Surfaces, and Films

1445

The leg of the spherical probe was about 150 mm long in order to avoid any sphere shading by the quartz capillary. The leg area is about 10% of the sphere area and it increases the probe area. However, this short leg is essentially shaded by the capillary. So, we estimate the effective collecting ratio leg/sphere to be no more than 5% of the total. The experiments showed that the shorter legs noticeably decrease the probe current due to the shading effect. The spheres were almost perfect in shape, so we estimate an error on the area calculation to be less than 18%. The shading effect of the fine capillary on the shortest 1 mm cylindrical probes is less than 3% and can be neglected. About 60–80 mm diam quartz capillaries of the cylindrical probes were 15 mm long. This portion of the probe represents a coaxial capacitor with plasma as an external electrode. The estimated capacitance, neglecting the Debay sheath in the high plasma density toroid, is about 2–3 pF, which is high enough to additionally shunt the probe sheath at high rf and increase efficiency of the rf blocking filter. Small low-capacitance ~;0.02 pF! resistors of 10–20 kV value were used as the blocking filters in the probe line. Such filters were used by Godyak and Popov6 in a low plasma density rf discharge. The resistor filter has a negligible stray capacitance to ground, since it was inside the discharge volume, and a very small series capacitance of about 0.03 pF. The coaxial cable receives the probe signal from the filter and then goes outside to be terminated by several lowcapacitance LC filters in order to completely suppress the rf component but still keep the frequency response at a reasonable level while measuring a dc probe current of about 10 nA. Then, the signal comes to the data acquisition system having additional low-capacitance filters. The sensitivity of the probe measuring system is 5–7 nA at 30 s total sweep time in the ~2100 to 140! V range of probe voltages. Faster sweeps bring about a noticeable hysteresis to the I – V characteristics in the case of very low plasma density in the downstream chamber. The efficiency of the probe rf blocking filters is determined by the product of two independent factors: ~1! the ratio of the impedances of the filtering element in parallel with a stray capacitance to the probe capacitive sheath impedance; and ~2! the ratio of the electron temperature expressed in electronvolts to the amplitude of rf fluctuations of the plasma potential.6–8 We found that the blocking efficiency is high enough under our particular conditions. The capacitive sheath of the Langmuir probe at the floating potential was experimentally estimated using cylindrical probes of different length9 to be about 1 kV in a downstream afterglow plasma, while it decreases at least ten times in the high-density plasma toroid. The probes were operating under the favorable conditions of a very thin low impedance capacitive probe sheath with the rf voltage amplitude on the probe sheath of less than 0.5–1 V. For instance, the distance between the maximum and minimum of the second derivative of the probe I – V in a 1-kW, 2-Torr Ar discharge was 1.1 V, which is evidence of low rf distortion.10,11 The 5-mm-diam flat probe used in this study is similar to

1446

Vinogradov, Menagarishvili, and Yoneyama: Probe diagnostics in a full wave resonator rf discharge

1446

the one used for deposition and etching.9 Linear arrays of five and ten similar probes were used as well. The probes have 30 mm dielectric separator made of mica. Such a little gap prevents electric breakdowns through the gas by decreasing the P3d ~where P is the gas pressure and d is the gap! product in correspondence with Pashen’s law. The I – V characteristics were recorded using a data acquisition system based on a PC board DAP-1200e by Microstar Labs having an on-board programmable CPU with DASYLAB software by DASYTEC. In the case of differentiating, the I – V curves were smoothed using the Savitsky–Golay algorithm or adjacent averaging.

III. RESULTS AND DISCUSSION The l-R discharge plasma interacts with the current and voltage standing waves distributed along the inductor. Due to the antiphase voltage half waves, the capacitive current inductor ground at the discharge resonance frequency of 27 MHz is canceled in the central plane of the plasma source. This condition depends on the rf balancing the electrical length of the spiral transmission line to the one full wavelength. The remarkable plasma phenomena and the discharge structure including the appearance of ball plasmoids and plasma toroids were described in Refs. 3 and 4 for 10–50Torr-pressure argon discharges. The totally developed l-R argon discharge structure consist of three bright essentially contracted plasma toroids located at the center and end planes of the inductor and a diffuse capacitive plasma in the bulk. The capacitive plasma is generated by the currents flowing between the rf voltage standing-wave maxima. There is some interconnection between the inductive and capacitive currents, which can be even visualized under some conditions. The central toroid plays a role of a virtual ground electrode in the discharge. Thus, the l-R discharge plasma represents a large volume rigid three-dimensional structure. The central plasma toroid absorbs inductive power from a half-lambda central part of the coil. The two secondary inductive toroids can get only half the power in comparison with the central toroid since they are each inductively fed from the quarter-wave parts. The 27 MHz, 2 kW-rf-power, 1.4 Torr oxygen discharge produces a single inductive plasma toroid located at the center of the plasma source. Secondary side oxygen plasma toroids can be initiated at an essentially higher power of 4–4.5 kW in order to overcome very high resistive losses typical for oxygen plasmas. First probe measurements in the center of a 1 kW 1 Torr Ar discharge were carried out using the cylindrical 20 mm Pt probe. The electron density at the discharge axis was measured to be 4.531011 cm23, the electron temperature 1.7 eV, and the plasma potential was 11 V. However, it was difficult to measure the electron density in the plasma toroid channel because of the probe’s melting. Since the local electron density is about directly proportional to the specific discharge power, and the density in the toroid is about four times higher than on the axis, we assume the electron density in the J. Vac. Sci. Technol. A, Vol. 16, No. 3, May/Jun 1998

FIG. 3. Langmuir probe I – Vcharacteristic severely distorted above the plasma potential and having a ‘‘swan neck’’ shape; the first derivative is also shown.

plasma toroid of the 2 kW Ar discharge is to be of the order of 531012 cm23. Since we are using very fine probes typical for the probe diagnostics of 1–5 Torr pressure molecular glow discharges,12 it might not be a problem to use a reference electrode. There is '5000 cm2 area downstream metal chamber contacting to a large afterglow plasma. The typical ratio of the chamber area to that of the probe is on the order of 107 – 108 , which should be quite enough in comparison with the recommended ratio for electropositive gas discharges: 103 – 104 , 13 even taking into consideration a large difference of the local plasma density in the toroid and downstream chamber. The reference electrode for electronegative gas discharges may be smaller than that for electropositive plasmas since it must sink essentially a smaller ion current corresponding to the smaller electron saturation current. However, our chamber could not be used as the reference electrode in the case of measurements in the central highdensity plasma toroid: the probe I – V characteristics were severely distorted, as shown in Fig. 3. The I – V have noticeable hysteresis and a local maximum of the probe current above the plasma potential, which looks like a ‘‘swan neck.’’ The first derivative of the probe current on the probe voltage is essentially broadened and even looks rather like a second derivative crossing the voltage axis. We did not study the detailed mechanism of such strange behavior of the I – V characteristic. Certainly, the chamber ~reference! sheath itself is not responsible for the ‘‘swan neck.’’ If the conductivity of the chamber sheath was too low, the probe saturation current would be decreased due to an additional voltage drop on the reference sheath, but it never could be smaller at higher probe potentials than at lower. Another factor affecting the probe I – V is the finite plasma conductance between the probe and the reference.13 This effect, however, cannot create a maximum as well, because the plasma impedance between the probe and chamber is constant. Hence, there are some second-order factors involved. For example, the shape of the toroidal conductive

1447

Vinogradov, Menagarishvili, and Yoneyama: Probe diagnostics in a full wave resonator rf discharge

FIG. 4. Radial distributions of plasma space and floating potentials and efficient electron temperatures derived from the probe I – V ~see the explanation in the text!.

channel can be affected electrostatically so that it changes the interaction with the probe. The source plasma can be easily affected by a dc potential applied to the toroid as this plasma is almost dc insulated from ground. A 131 cm2 reference electrode made of Ni was inserted in the toroid and connected to ground through a series of rf chokes and a dc voltage source supplying 8–9 V dc potential in order to set the dc current biased by the reference to zero. The ‘‘swan neck’’ immediately disappeared from the I – V. It proves that the probe sheath itself is not responsible for the ‘‘swan neck.’’ All further measurements inside the plasma source were carried out with the same reference electrode immersed into the plasma toroid. Figure 4~a! shows radial distributions of V pl and V f potentials in the plasma toroid as measured with the spherical probe of about 130 mm diam. There is approximately a 5 V potential well inside the toroid. The floating potential V f has a similar radial dependence as the plasma space potential V pl . The potential well is generated by the requirement of the charge balance between the ions and mobile electrons diffusing from the toroidal channel. The plasma potential decreases close to the tube surface since the toroidal current tends to maximize the area of the absorbed rf magnetic flux, thus creating a very thin wall sheath. The floating potential V f on the wall contacting the toroid is about 14 V, hence, the electric sheath near the tube wall is about 16 V. The difference (V pl2V f ) does not noticeably depend on the radial position. Figure 4~b! shows the effective electron temperatures derived from the same probe I – V using different methods. JVST A - Vacuum, Surfaces, and Films

1447

FIG. 5. Second derivatives of the probe I – V characteristics measured at the top ~a! and bottom ~b! of the potential well at the plasma toroid plane of an oxygen discharge. The derivatives close to the discharge axis show a strong depletion of electrons having characteristic energies of about 4–6 and 8–10 eV, typical for oxygen discharges.

The results are not unusual for a non-Maxwellian electron energy distribution function ~EEDF!. The solid marks indicate the effective kT e calculated from the equation: (V pl2V f )5kT e 3ln(I e /I 1 ). 13 Even under a high negative ion density condition this formula seems to provide reasonable results. The lowest temperatures were determined from the slope of the logarithm of the electron current near the V f , and the highest temperatures were, typically at 3–5 V below V pl . The second derivatives ( @ m A/V2# ) of the probe current on the probe potential, which are directly proportional to the EEDF, clearly show that the EEDF is far from Maxwellian type. The EEDF is very broad and has more than one local maximum ~minimum!. Figure 5 shows examples of second derivatives obtained at the top edge and bottom of the potential well to show their evolution. The characteristic features of the second derivatives are well established with a large number of probe measurements at different locations over the plasma source. The derivatives in the well show electron depletion at about 14 and ~5–12! V probe potentials. The electrons diffuse from the toroid toward the discharge axis losing kinetic energy in numerous inelastic collisions,1 decreasing the low- and middle-energy part of the EEDF. The work on analyzing the EEDF is in progress. To estimate the electron density with the probes we used an effective electron temperature derived from the floating potential, which fit well between the limit values shown in

1448

Vinogradov, Menagarishvili, and Yoneyama: Probe diagnostics in a full wave resonator rf discharge

1448

power deposited here is very low so the plasma impedance is relatively high. IV. SUMMARY AND CONCLUSION

FIG. 6. Radial distribution of: ~a! electron and ion saturation probe currents; ~b! electron and ion densities and the ratio N 1 /N e in the central plasma toroid.

Fig. 4~b!. A systematic error of the electron density due to an error in kT e does not exceed 630%. The ion density exceeds the electron density 10–30 times in our oxygen discharge. Consequently, we used the ion temperature limit for the ion drift velocity to calculate the positive ion density from the ion saturation current.14,15 The ion temperature is assumed to be equal to ;0.1 eV. Figure 6~a! shows the probe saturation currents measured at V pl and V f . The shapes of both radial distributions are very similar. The electron and ion densities are shown in Fig. 6~b!. These values are typical for high-density plasmas. The axial electron density is decaying from 1011 cm23 in the plasma toroid down to 109 cm23 at the lower end of the inductor corresponding to a 121 mm distance from the wafer platen. The azimuthal distributions of the floating potential and ion saturation current density on the circular flange surface, which is 50 mm below the inductor, were measured by the flat probes. There is some systematic nonuniformity of both distributions. The surface floating potential varies from 21 to 14 V, while the ion current density from 10 to 2.5 uA/cm2. This area is very sensitive to any rf electric field originating from the plasma source, since the overall rf

J. Vac. Sci. Technol. A, Vol. 16, No. 3, May/Jun 1998

Single Langmuir probe measurements utilizing fine cylinder and sphere probes were carried out in a 235-mm-i.d. Lambda resonator rf inductive discharge having a push–pull antisymmetric capacitively balanced structure. The electron density in the central plasma toroid is about 231011 cm23, while the negative ion density exceeds 331012 cm23. The electrons generated in the high current density toroidal channel diffuse toward the discharge axis, thus generating a potential well. The EEDF is substantially non-Maxwellian with a high depletion of low- ~0–3 eV! and middle- ~;8–10 eV! energy electrons. There is a remarkable difference between the EEDF inside and outside the potential well. The oxygen plasma toroid is essentially dc insulated from the downstream ground chamber. It is very sensitive even to 100 mA current biased by a fine probe. The reference electrode must be inserted into the toroidal plasma channel in order to measure undisturbed Langmuir probe characteristics there. 1

M. A. Lieberman and A. J. Lichtenberg, Principles of Plasma Discharges and Materials Processing ~Wiley, New York, 1994!. 2 High Density Plasma Sources, edited by O. A. Popov ~Noyes, Park Ridge, 1995!. 3 G. K. Vinogradov and S. Yoneyama, Jpn. J. Appl. Phys., Part 2 35, L1130 ~1996!. 4 G. K. Vinogradov and S. Yoneyama, Proceedings of the 3rd International Conference on Reactive Plasma, Nara, Japan, January 21–24 1997, p. 221. 5 F. Bose, R. Patric, and H. P. Baltes, J. Vac. Sci. Technol. B 12, 2805 ~1994!. 6 V. A. Godyak and O. A. Popov, Sov. Phys. Tech. Phys. 22, 461 ~1977!. 7 A. Cantin and R. J. Gagne, Nuovo Cimento B 66, 193 ~1970!. 8 G. K. Vinogradov, Yu. A. Ivanov, and Yu. A. Lebedev, in Plasmachemical Reactions and Processes, edited by L. S. Polak ~Nauka, Moscow, 1977! @in Russian#. 9 G. K. Vinogradov, G. J. Imanbaev, and D. I. Slovetsky, High Energy Chem. ~USSR! 19, 370 ~1986!. 10 Yu. A. Ivanov, Yu. A. Lebedev, and L. S. Polak, J. Tech. Phys. 46 , 1459 ~1976!. 11 V. A. Godyak and S. N. Oks, Sov. Phys. Tech. Phys. 24, 784 ~1979!. 12 Yu. A. Ivanov, Yu. A. Lebedev, and L. S. Polak, Methods of Contact Diagnostics in Non-Equilibrium Plasma Chemistry ~Nauka, Moscow, 1981! @in Russian#. 13 L. Schott, in Plasma Diagnostics, edited by W. Lochte-Holtgreven ~AIP, New York, 1995!. 14 R. L. F. Boyd and J. B. Thompson, Proc. R. Soc. London, Ser. A 252, 102 ~1959!. 15 H. Amemiya, B. M. Annaratone, and J. E. Allen, Proceedings of the 3rd International Conference on Reactive Plasma, Nara, Japan, January 21–24 1997, p. 239.