Experimental investigation of the electrical characteristics and initiation dynamics of pulsed high-voltage glow discharge

Transcription

1 INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS J. Phys. D: Appl. Phys. 34 (2001) PII: S (01) Experimental investigation of the electrical characteristics and initiation dynamics of pulsed high-voltage glow discharge X B Tian and Paul K Chu 1 Department of Physics & Materials Science, City University of Hong Kong, 83, Tat Chee Avenue, Kowloon, Hong Kong paul.chu@cityu.edu.hk Received 1 November 2000 Abstract Pulsed high-voltage glow discharge, a process that falls between conventional plasma ion implantation and plasma nitriding, is a potentially effective surface treatment technique. The electrical characteristics and initiation dynamics of the process are experimentally investigated in this work. The discharge behaviour is found to depend on the applied voltage, gas pressure, pulse duration, and pulsing frequency. The voltage current characteristics basically obey the collisionless Child Langmuir relationship, and the discharge current varies substantially with the applied voltage and gas pressure in the high-voltage, high-pressure domain. The seed plasma before each glow discharge pulse affects the initiation mechanism regardless of whether it is sustained by an external plasma source or it is simply the residual plasma of the afterglow after the voltage pulse has been turned off. Our results also disclose that the presence of an initial plasma before the glow discharge is ignited has a large impact on the initiation time of the glow discharge but has only a slight effect on the steady-state current. 1. Introduction Plasma immersion ion implantation (PIII) in a pulsed highvoltage operation mode has emerged as an effective surface modification technique [1 5]. One of its advantages is the ability to treat irregularly-shaped components without the need for complex target manipulation as in conventional ionbeam ion implantation. In PIII, ions are implanted beneath the surface of the objects utilizing a high voltage applied to the objects at a low pressure as opposed to the low-voltage, high-pressure conditions employed in conventional pulsed discharges. The low pressure ensures that ions undergo no collisions and reach the sample surface without energy loss. However, the process must rely on one or more external plasma sources to provide the required plasma as it is very difficult to maintain a glow discharge at low pressure. Modulated power discharges, also called pulsed discharges, have advantages over continuous glow in many applications. For instance, it has been found that the discharge 1 Author to whom correspondence should be addressed. chemistry or properties of the deposited films can be altered, and the processing rate can be maintained despite the lower average power. Moreover, the process reduces dust particle formation, trench notching, as well as charging damage. Some of these advantages can be attributed to how the plasma is turned on and off [6]. The process is typically carried out under low-voltage, high-pressure conditions, and so it can be characterized as a hybrid process incorporating low-energy ion bombardment and ion sputtering. The surface adsorption and chemical reaction take place only on the surface or within a thickness of several atomic layers. Between the low-voltage ( several kv)/high-pressure (ten to thousands of Pa) conditions used in conventional plasma nitriding [7, 8] and the high-voltage (up to 100 kv)/lowpressure ( 10 2 Pa) conditions used in PIII [1 5, 9, 10], there exists a medium-voltage (tens of kv)/medium-pressure (0.1 to several Pa) process window to conduct pulsed high-voltage glow discharge treatment. In this process, the plasma is sustained by the glow discharge when a voltage is applied to the objects. In this work, we conduct experiments on a /01/ $ IOP Publishing Ltd Printed in the UK 354

2 Initiation dynamics of pulsed high-voltage glow discharge pulsed voltage glow discharge. One of the main advantages offered by this process is the ability to treat/implant a large sample or a large number of small samples in a fast and effective manner without external plasma sources. The process differs from conventional plasma nitriding or plasma immersion ion implantation (PIII) in the nitrogen transport mechanism and plasma generation method. This technique enables faster implantation of parts possessing irregular geometries on account of the smaller plasma sheath thereby resulting in more conformal implantation. However, although previously proposed to be a useful surface modification method [11 15], this technique has not been investigated in detail or systematically. In particular, there have been few studies on the mechanism such as the glow discharge dynamics that are critical in gaining an understanding of the process. In this paper, we report our systematic experimental investigation on the dynamic process of pulsed high-voltage glow discharge under the medium-voltage/medium-pressure conditions. 2. Experimental details The experiments were conducted in the PIII equipment in the Plasma Laboratory of the City University of Hong Kong [16], and the schematic of the instrument is shown in figure 1. The copper target platen is 140 mm in diameter and 25 mm thick, and supported by a 140 mm tall ceramic holder. The vacuum chamber is 1000 mm in diameter and 1200 mm in height. Four sets of filaments used to generate the seed plasmas are positioned symmetrically at the upper part of the vacuum chamber. An inductively-coupled radio-frequency plasma source is also present on the top of the vacuum chamber as an alternative seed plasma generation device. During the experiments, a hard-tube-based pulse modulator was used to produce a square voltage pulse with a rise-time of less than 2.5 µs. The chamber gas pressure was varied between 9 Pa and Pa, and hydrogen and nitrogen were chosen since they are frequently used in semiconductors and metals processing. The discharge voltage current transient characteristics were recorded using a capacitance divider (Pearson model 305A), Rogowski coil (Pearson model 110), and a digital oscilloscope (Tektronix TDS360). 3. Results During the pulsed high-voltage glow discharge process, parameters such as the applied voltage, gas pressure, pulsing frequency, and pulse duration have appreciable impacts on the behaviour. The evolution of the current versus time depicted in figure 2 shows the initial ignition stage of the glow discharge as voltage is applied to the sample and the ensuing steady-current stage. The high initial current, namely the displacement current, is not a conduction current and is induced by the capacitance effects of the PIII hardware. The magnitude is quite large, particularly under high-voltage and short rise-time conditions. At the initial time stage, there is no plasma in the chamber and no contribution to the initial current. In our experiments, the displacement current is generally 2.5 times the effective current [17]. After a time delay, the current gradually attains a constant value corresponding to the applied voltage. As shown in figure 2, the higher the applied voltage, Filament Matching box RF excited plasma Plasma diffusion Oil cooling Copper target Ceramics Coil Shielding sheet Vacuum chamber Cable PIII HV Figure 1. Schematic diagram of the vacuum chamber, sample holder, and plasma sources for our pulsed high-voltage glow discharge experiments. the shorter the delay time to reach the steady-state discharge current. With increasing voltage, the steady-state current goes up. As we will discuss in the subsequent sections, the timedependent characteristics of pulsed glow discharge also change with other experimental parameters Dependence on gas pressure As shown in figure 3, the gas pressure has a large influence on the glow discharge process. Once the glow discharge is initiated, the discharge current increases with the gas pressure. At the lower pressure limit of the discharge initiation, the steady-state discharge current shows a more abrupt change. As shown in figure 3(a), the discharge current abruptly jumps to 0.3 A for 20 kv hydrogen once the pressure rises beyond 0.35 Pa without undergoing a gradual increase. Based on our results, at a pressure of 0.4 Pa, an applied voltage of 10 kv with duration of 40 µs is sufficient to ignite the hydrogen plasma, but 15 kv and 60 µs is required for nitrogen. Thus, a hydrogen plasma is easier to ignite compared to a nitrogen plasma Dependence on discharge voltage As illustrated in figure 4, the discharge current increases with the applied voltage. The discharge behaviour is very sensitive to voltage variation especially at high voltage. The voltage current behaviour is also influenced by the pressure and gas species. For hydrogen, the rates that the pressure affects the current are calculated to be 1.1 A Pa 1 and 2.8 A Pa 1 for 15 kv and 25 kv applied voltage, respectively. In comparison, the rate for nitrogen is 0.8 A Pa 1 at 15 kv (figure 3(b)) Dependence on pulsing frequency The influence of the pulsing frequency on the discharge behaviour is not as significant. With increasing pulsing frequency, the steady-state current rises while the minimum 355

3 X B Tian and P K Chu Voltage ( kv ) (a) kv 2-15 kv 3-20 kv 4-25 kv 1-25 kv 2-20 kv 3-15 kv 4-10 kv kv-hydrogen 10 kv-hydrogen Col 11 vs Col 12 Col 14 vs Col 15 Gas pressure ( Pa ) 15 kv-nitrogen numerical fitting :I=k2*P 2 I=k1*P I=k1*P ( a ) ( b ) - Transition stage (b) Figure 2. Time-dependent applied voltage and hydrogen discharge current at a gas pressure of 7 Pa at different applied voltages under conditions of a pulse duration of 30 µs and a pulsing frequency of 150 Hz: (a) voltage; (b) discharge current Steady stage pulse duration (t i ) needed to ignite the discharge and the minimum pulse duration (t s ) to sustain the glow discharge once ignited become smaller (figure 5). It should be mentioned that t i >t s since the number of excited particles including ions or free electrons is small and the time to ignite the discharge is long (large t i ). In contrast, once the discharge has been ignited, a small pulse is sufficient to sustain it with the assistance of the existing excited particles. Therefore, a higher frequency may increase the ionization efficiency leading to a higher plasma density and consequently higher current [15]. Our observation is consistent with previous reports that the voltage to ignite the glow discharge decreases with increasing frequency in the pulsed low-voltage glow discharge mode such as that used in conventional plasma nitriding [18] Effects of the voltage and pressure in the transient region (ignition zone) The initiation dynamics vary significantly with the voltage, gas pressure, and pulse duration. For a small pulse duration, the ignition voltage (defined to be the voltage when the PIII current exceeds 50 ma) of the glow discharge is very high as shown in figure 6(a). When the pulse duration is increased, the Gas pressure ( Pa ) Figure 3. Steady-state discharge current as a function of gas pressure with a pulse duration of 30 µs: (a) hydrogen; (b) nitrogen. The best empirical fits are shown here with k1 and k2 being constants Pa 7 Pa 0.77 Pa I=k3*V 2.2 I=k4*V 2.2 I=k5*V Applied voltage ( kv ) Figure 4. Voltage-dependent discharge current at different hydrogen pressure. Also shown are the best empirical fits with k3, k4, and k5 being constants. voltage decreases precipitously, especially at low pressure. It thus appears that a low voltage is sufficient to ignite the glow discharge if the pulse duration is long. Figure 6(a) is indicative of the ignition limit of the glow discharge process, and glow discharge will commence for conditions above the curves. Figure 6(b) shows the discharge zone at a voltage of 15 kv. When the gas pressure is low, the pulse duration required to initiate the discharge increases. Our results indicate that the pulse duration required to initiate the glow discharge decreases with higher voltage or pressure. This behaviour/information is very useful in practical 356

4 Initiation dynamics of pulsed high-voltage glow discharge Frequency ( Hz ) Figure 5. Relationship between the minimum pulse duration needed to ignite the glow discharge or minimum pulse duration needed to sustain the discharge once ignited and the pulsing frequency for 7 Pa hydrogen gas. and represent the igniting pulse or sustaining pulse for a 10 kv discharge, respectively, whereas and represent the igniting pulse or sustaining pulse for a 20 kv discharge, respectively. Initiation voltage ( kv ) Gas pressure ( Pa ) Pulse duration ( µs ) No discharge zone Pulse duration ( µs ) Col 1 vs Col 2 Col 1 vs Col ( a 3 ) Col 10 vs Col 9 Col 14 vs Col 13 2 Pa 0.69 Pa Discharge zone Discharge zone ( b ) Figure 6. Discharge zone determined by pressure pulse duration for hydrogen: (a) voltage versus pulse duration; (b) pressure versus pulse duration. PIII processes. For instance, the glow discharge may take place during PIII (for example, at 95 µs pulse duration and 0.2 Pa pressure), and if this happens, the extra ions provided by glow discharge will be implanted. Consequently, the effective plasma density will probably increase contributing to a higher incident ion flux and rapid formation of the steady-state Child Langmuir sheath. It is also worth mentioning that, at low pressure, the glow discharge process seemingly cannot sustain itself and must be maintained by a primary plasma produced by an external plasma source. Thus, it will extinguish when the implantation voltage is turned off. Consequently, glow discharge may be difficult to observe in PIII experiments [19] even if it does exist Effects of seed plasmas Hot filament glow discharge and RF plasma sources were used to investigate the effects of the primary seed plasma on pulsed glow discharge. The experimental results shown in figure 7 are not as expected. The seed electrons do not influence the glow discharge behaviour significantly. As shown in figure 7(a), the time-dependent current is almost constant whether or not hot filaments are used to produce free electrons when the filament current is less than 20 A. Even when a dc voltage is applied between the filaments and vacuum chamber wall to generate a glow discharge, the current waveform hardly changes. In contrast, when the primary plasma is produced by an RF plasma source (50 W), there is no initial rise in the current waveform indicating early formation of the cathodic current (figure 7(b)). The steady-state current is similar to that without RF. The initial current can probably be attributed to the primary plasma and plasma-assisted ionization. Based on our findings, it is believed that there exists a density threshold beyond which electrons emitted by hot filaments or plasmas produced by RF source will accentuate the glow discharge. 4. Discussion Pulsed high-voltage glow discharge can be initiated under certain conditions, and the applied voltage, gas pressure, pulse duration, and pulsing frequency have a large effect on the voltage current behaviour. During dc glow discharge, the current density can be described by the space-charge-limited current law. Under the conditions of collisionless motion (unmagnetized ions and low gas pressure), the collisionless Child Langmuir law can be used to describe the flow of ions [20 22]: J = 4ε 0 9 ( ) 2q 1/2 Vc 3/2 (1) M L 2 where J is the cathodic current density, q is the ionic charge, M is the ionic mass, ε 0 is the permittivity of free space, V c is the cathode sheath voltage, and L is the cathodic fall length. In the case of mobility-limited motion, the current obeys the mobility-limited version of the Child Langmuir equation: J = 9ε 0 8 V 2 L 3. (2) Although collisions occur during the ion motion towards the cathode, the collisionless Child Langmuir law seems to be more suitable in describing our cases according to the experimental results and previous results [20, 21]. According to equation (1), the current increases potentially with the applied voltage, since J V 3/2. This is also indicated by the experiments of Matossian and Wei [15] in which they obtained a very large current of 150 A at a high pulsed voltage of 100 kv. However, the numerically fitted 357

5 X B Tian and P K Chu Without filament With filament With filament plus bias voltage With RF Without RF ( a ) ( b ) Figure 7. Effects of the preliminary plasma on the pulsed high-voltage discharge current for nitrogen: (a) electrons produced by hot filaments, pressure P = 0.45 Pa; filament current I = 15 A; bias voltage V = 80 V; (b) plasma produced by RF source, pressure P = 0 Pa, RF power = 50 W. curves shown in figure 4 indicate that J V n where n 2, and it seems to contradict equation (1). In fact, the cathodic fall length L depends very much on the discharge voltage, and L decreases with increasing voltage corresponding to L V ( ) [23]. Although the relationship in [23] was obtained under different conditions, for example at lower voltage, higher pressure, or for a different gas species, the trend could be similar, thereby supporting our observed relationships shown in figure 4. The cathodic fall length L is thus affected by not only the discharge voltage but also the gas pressure. As demonstrated previously [21], if the discharge voltage is held constant, the pressure dark space thickness product is nearly constant for a dc discharge. Therefore, when the pressure increases, the thickness of the dark space becomes thinner. Consequently, the collected current at the cathode is higher. For example, as shown in figure 3(b), when the pressure rises from 0.48 Pa to 1.8 Pa, the discharge current obeys the relationship J 1/L 2 P 2 for nitrogen. The initial dynamics of the glow discharge are related to the pulsing frequency, pulse voltage, and pressure. In fact, it very much depends on the density of the primary plasma just before each discharge pulse. As previously reported, [24, 25] the electron density is proportional to the discharge current, and the decay of the N + 2 density after the glow discharge can be described by [25] I t = I t=0 (1+αn e0 t) 2 (3) where t represents the time variable (t = 0 at the discharge cut-off), n e0 stands for the electron density at t = 0, and α is the recombination coefficient. The electron density thus decreases potentially with time. For a pressure condition of several Torr, the decay of the electrons requires a duration of several milliseconds [25]. Therefore, it is recommended that the post-discharge time during pulsed glow discharge must not exceed several milliseconds in order to limit the cooling of the neutral gas and to facilitate the discharge after the plasma has been turned off. A high voltage or high pressure may give rise to a high discharge current, and the electron density and residual electron concentration are high in the chamber. Thus, high frequency, high pressure, or high voltage naturally lead to a short delay time of glow discharge initiation. Our observation is in line with the results of Dhali and Low [26]. It may also be one of the reasons why a high frequency leads to a different discharge current reported in [15], in which it was thought to be related to the variation of the ionization mechanism of the secondary electrons at different pulsing frequencies. The external plasma sources seem to have a similar effect on the dynamics of the glow discharge during the afterglow. When there are seed plasmas, the discharge rapidly builds up as indicated in figure 7. These characteristics seem to be related to the plasma density and location although it is difficult to draw an unequivocal conclusion at this moment. Nevertheless, the plasmas near the anode, as in the case of hot filaments, have a smaller effect on the discharge than those near the cathode like the RF plasma sources. The free electrons emitted by the hot filaments have to stay near the chamber wall (anode) and far way from the target due to the dc bias. Consequently, the electrons have very little chance to collide with neutrals and contribute less to the gas ionization. However, it is not true in an RF plasma. As for the steady-state current, it is hardly influenced by the preliminary plasmas indicating that there is very little contribution to bulk ionization. 5. Conclusions The characteristics of pulsed voltage glow discharge are experimentally investigated. Our results show that the dynamics of this type of glow discharge depend considerably on the applied voltage, gas pressure, pulsing frequency, and pulse duration. The voltage current behaviour of the discharge obeys basically the collisionless Child Langmuir law. Consequently, the discharge current changes significantly with the applied voltage and pressure in the high-voltage, higher-pressure regime. The primary plasmas before each pulse determine the initiation behaviour. The plasmas produced by external sources change the initiation dynamics but have a small effect on the steady glow discharge. Even though pulsed high-voltage glow discharge has not been investigated extensively both theoretically and experimentally, its potential as a surface modification tool is expected to be quite far reaching since the process features a small plasma sheath and does not require external plasma sources, compared to conventional PIII nitriding. Acknowledgments This work was jointly supported by Hong Kong Research Grants Council CERG # or CityU 1003/99E 358

6 Initiation dynamics of pulsed high-voltage glow discharge and # or CityU 1032/00E. The authors also thank D T K Kwok, X C Zeng, and Y X Leng for valuable discussions. References [1] Conrad J R, Radtke J L, Dodd R A, Worzala FJandTranNC 1987 J. Appl. Phys [2] Qin S and Chan C 1992 IEEE Trans. Plasma Sci [3] Chu P K, Qin S, Chan C, Cheung N W and Ko P K 1998 IEEE Trans. Plasma Sci [4] Goeckner M J, Felch S B, Fang Z, Lenoble D, Galvier J, Grouillet A, Yeap GCF,Bang D and Lin M R 1999 J. Vac. Sci. Technol. B [5] Goeckner M J, Felch S B, Weeman J, Mehta S and Reedholm J S 1989 J. Vac. Sci. Technol. A [6] Overzet L J and Kleber J 1998 Time resolved measurements of pulsed discharge: the role of metastable atoms in the afterglow Electron Kinetics and Applications of Glow Discharges ed U Kortshagen and L D Tsendin (New York: Plenum) p 511 [7] Korhonen S, Sirvio E H and Sulonen M S 1983 Thin Solid Films [8] Verma R and Podob M 1985 Industrial Heating [9] Matossian J N and Geobel D M 1996 Surf. Coating Technol [10] Zeng Z M, Tang B Y, Chu P K, Tian X B, Wang SYand Wang X F 1999 J. Vac. Sci. Technol. B [11] Shamim M M, Sridharan K, Fetherston R P, Chen A and Conrad J R 1994 J. Vac. Sci. Technol. B [12] Schumacher R W, Matossian J N and Goebel D M 1994 High impedance plasma ion implantation method and apparatus US Patent Specification [13] Blawert C and Mordike B L 1997 Nucl. Instrum. Methods B [14] Khvesyuk V I and Tsygankov P A 1997 Surf. Coating Technol [15] Matossian J N and Wei R 1996 Surf. Coating Technol [16] Chu P K, Tang B Y, Cheng Y C and Ko P K 1997 Rev. Sci. Instrum [17] Tian X B, Tang B Y and Chu P K 1999 J. Appl. Phys [18] Seller J 1999 Pulse plasma why, what, where 2nd Int. Pulsed Plasma Surface Technologies Workshop (San Diego, CA, April 1999) [19] Zeng X C, Kwok DTK,Chan C and Chu P K 2000 IEEE Int. High Voltage Workshop 2000 (Newport Beach, CA, April 10 12, 2000) [20] Lieberman M A and Lichtenberg A J 1994 Principles of Plasma Discharges and Materials Processing (New York: Wiley) p 467 [21] Chapman B 1980 Glow Discharge Process (New York: Wiley) p 106 [22] Adjaottor A A, Ma E and Metetis E I 1997 Surf. Coating Technol [23] Budtz-jorgensen C V, Bottiger J and Kringhoj P 2000 Vacuum [24] Dhali S K 1989 IEEE Trans. Plasma Sci [25] Hugon R, Fabry M and Henrion G 1996 J. Phys. D: Appl. Phys [26] Dhali S K and Low L H 1988 J. Appl. Phys