V A Lisovskiy, V D Yegorenkov - Alpha-gamma transition in rf capacitive discharge in low-pressure oxygen - страница 1
Available online at www.sciencedirect.com
Vacuum 74 (2004) 19-28
Alpha-gamma transition in RF capacitive discharge in
V.A. Lisovskiy, V.D. Yegorenkov
Kharkov National University, Svobody Sq. 4, Kharkov 61077, Ukraine
We report the recorded current-voltage characteristics of a RF capacitive discharge in oxygen. Low-frequency oscillations of the plasma potential in a kilohertz frequency range are observed to accompany the transition of the discharge from a weak- (a-) to a strong-current (g-) regime in the low-pressure range. The weak-current regime of the RF capacitive discharge is observed within the pressure range limited not only from the medium pressure side but also from the lower-pressure one. Electron temperature and plasma density are registered with a probe technique. © 2003 Elsevier Ltd. All rights reserved.
Keywords: Radio-frequency capacitive discharge; Lowpressure; Probe; Negative ions
Radio-frequency capacitive discharge in oxygen is widely used for processing semiconductor and polymer materials (etching, deposition, oxidizing, cleaning, etc.) [1-3]. A large number of experimental and theoretical papers studying the characteristics of this type of the discharge were published (see, e.g. [4-15]). But up to nowthe data are actually absent on the regimes of RF-capacitive discharge in oxygen and on the transition of the discharge from a weak-current (a-) to a strong-current (g-) regime (a-g transition). Therefore, it is vital to study the inner parameters of and regimes for the RF discharge in oxygen.
As is known [16-18], a RF discharge can exist in two different regimes: weak- (a-) and strongcurrent (g-) ones. In a-regime the electrons acquire energy for ionising gas molecules in the RF field in the quasineutral plasma, the conductivity of the near-electrode sheaths is small, and the electron emission from the electrode surface does not play an essential role in the discharge sustainment. In g-regime the electron avalanches develop in the near-electrode sheaths. The main ionisation of gas molecules by electrons in g-regime occurs near the boundaries of the near-electrode sheaths, and the RF field in the quasineutral plasma cannot supply the electrons with sufficient energy for ionisation.
Ionisation instability can be observed in a gas discharge with negative ions [19,20].In DC discharge in electro-negative gases such as CO2, CO and O2, the attachment-induced ionisation instability was observed by the authors of papers [21-23]. The paper  presented the most detailed theoretical treatment of the mechanism of this instability. The authors of this paper derived the criterion for the onset of this instability in plasma
0042-207X/$ - see front matter © 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.vacuum.2003.11.003
containing negative ions. This criterion was used by the authors of papers [13,24] for elucidating the nature of the instability they had observed in RF-capacitive discharge in oxygen.
Our paper reports the measurements of the RF current amplitude Irf and of the phase shift j between RF current and voltage as well as of the dependence of the active RF current Irf cos j on RF voltage amplitude Urf (current-voltage characteristics of the RF discharge) for various values of oxygen pressure p. The extinguishing curve and the a—g transition curve for the RF discharge are registered. It is observed that within the oxygen-pressure range p& 26.6-306 Pa the a-g transition is accompanied by low-frequency oscillations of the plasma potential in the kilohertz frequency range. Electron temperature was measured with a single probe, and the axial profiles of the plasma density were also obtained.
Fig. 1 depicts the scheme of the experimental device used for registering the RF discharge characteristics. The section of a fused silica cylindrical tube (not shown in Fig. 1) with inner diameter of 100 mm vacuum-sealed on both ends with stainless-steel planar round electrodes 1 served as a discharge chamber. The puffing system 2 delivered oxygen into the chamber through small orifices in a grounded electrode. The chamber was
Fig. 1. Scheme of the experimental device.
evacuated through the orifices 3 in the same electrode. We used a thermoelectric gauge 4 for registering gas pressure from 0.133 Pa to the atmospheric one. The evacuation was accomplished with pre-vacuum and turbo molecular pumps permitting to achieve the limiting pressure of 2.7 x 10—4 Pa. One of the electrodes of the discharge chamber was grounded. The generator 5 delivered the RF voltage to another electrode via the matching box 6. The choke of Lc = 4mH inductance switched across the electrodes outside the chamber removed the self-bias voltage.
The experiments were performed within the gas pressure range p = 3.3-666 Pa in the range of RF voltage amplitudes Urf P700V and the RF field frequency f = 13.56 MHz. The distance between the electrodes was L = 33 mm. The amplitude of the RF current Irf was measured with a Rogowski coil 7 located on the bus, which grounded one of the electrodes. The signal from the Rogowski coil was fed to the FK2-12 device 8. The signal from the capacitive divider 9 connected to the RF electrode was fed to another input of the FK2-12 device. The device for measuring the phase difference FK2-12 is capable of registering the amplitude of the RF signal having the frequency in the range of f1 = 1 MHz-1 GHz and phase shift j between two signals (in our case between RF voltage and RF current). In the absence of the discharge the registered phase shift value is j = 90° = ft/2, i.e., we register only the displacement current. On igniting the RF discharge the active RF current appeared, the phase shift angle j becomes less than p/2. The accuracy of measuring the phase shift angle with FK2—12 was 0.1°.
Electron temperature, plasma potential and density were registered with a single cylinder nichrome probe 10 (5.5 mm in length and 0.18 mm in diameter). We employed the resonance filter 11 consisting of a high-frequency choke and a variable capacitor tuned to the frequency fof the RF generator to prevent the RF current from penetrating the probe circuit. We also included an additional choke of L/?=10mH inductance to suppress the main frequency signal and its harmonics in the probe circuit. All three possible regimes of probe operation (collisionless, transitional and collisional) can be observed within the
pressure range studied. That is, at different values of gas pressure ions may traverse the near-probe sheath without collisions (this is observed under lowgas pressure, p < 5 Pa); they may experience several collisions (intermediate pressure) or multiple collisions (p > 100 Pa). The plasma density (density of positive ions) ni was calculated from the ion branch of the probe current Ipr and electron temperature Te was measured according to the technique described in papers [25-27], permitting accurate determination of the density of positive ions from the probe current-voltage characteristic at arbitrary value of the gas pressure. Electron temperature Te was determined from the linear sections of the probe current-voltage characteristics drawn to a semi-logarithmic scale, as well as from the second derivative of the probe current with respect to the DC voltage on the probe. The electron temperature Te values measured with both techniques differed not more than by 10-20%. The presence of negative ions does not affect these techniques of measuring the electron temperature [28,29], what ensures the correct determination of Te. To measure the second derivative d2Ipr/dUpr, the second harmonic technique was applied, i.e., the probe current was modulated with a low frequency voltage (with the frequency ff e 1-3 kHz), but the signal was registered at the frequency 2 flf. The set-up for measuring d2Ipr/dUp2r was described in paper . Regretfully, this set-up did not permit to make correct measurements of the second derivative d2Ipr/dUp2r near the plasma potential, therefore we could not employ the technique for measuring the temperature Tn and density nn of negative ions suggested in [28,29]. The same set-up based on the SK-54 spectrum analyser was used for recording the spectrum of low-frequency oscillations generated in the RF gas discharge.
Fig. 2 shows the RF current amplitude Irf (Fig. 2a), the phase shift j (Fig. 2b,c) and the active RF current Irfcos(j) (Fig. 2d) against the RF voltage at different oxygen pressure values. The RF generator supplied the RF voltage up to
30 100 300 467
30 100 300
Fig. 2. RF current amplitude (a), phase shift angle (b, c) and active RF current (d) against RF voltage at different oxygen pressure values.
Urf P700V, thus enabling us to study the characteristics of RF discharge in the weak-current regime, the a—g transition and within a small section of the strong-current regime. Therefore, we observe a feebly expressed dogleg feature of the current-voltage characteristics (Fig. 2a) occurring when the discharge is passing to a strong-current regime. With the oxygen pressure fixed the curves in Fig. 2a possess an almost constant tilt which experiences strong changes only at the RF voltage value when the discharge is extinguished. However, we did not observe the normal current density feature  within the oxygen pressure range studied. It is clear from Fig. 2b,c that at low oxygen pressure (p< 13 Pa) on increasing the RF voltage the phase shift angle increases, reaches a maximum and then decreases slowly in the
g-regime. At higher oxygen pressures and low voltages Urf the phase shift angle first decreases, passes through the minimum and then increases, reaches a maximum at the end of the a—g transition, and then decreases in the strong-current regime of RF discharge. Fig. 2c gives a clear indication of the maximum through which the dependence of the phase shift j on the RF voltage is passing.
The current-voltage characteristics of the discharge (Fig. 2d) possess a region with a negative differential conductivity. Similar effect was earlier observed in the RF capacitive discharge in argon [18,31]. In the weak-current regime of discharge on increasing the RF voltage the active RF current first grows, it decreases under the a—g transition, reaches a minimum and grows quickly in the strong-current regime.
Let us realize exactly what change of the RF discharge parameters is expedient to take as a criterion of the a—g transition. People often regard the value of the RF voltage Urf at which a discontinuity in the derivative dIrf/dUrf is observed as the voltage value of the a—g transition of the RF discharge [33,34]. In this case the electron avalanches develop in the near-electrode sheaths, and with the RF voltage increasing a fast growth of the plasma density within the total discharge gap is observed. According to the opinion of the author of , at lowpressure the a—g transition curve coincides with such section of the ignition curve of the RF discharge (Paschen branch ), that after the breakdown the RF discharge burns at once in g-regime. However, at low pressure these two criteria do not agree between themselves. Again, visual observations showthat at lowand intermediate gas pressure the structure of the RF discharge becomes very similar to that of the DC glowdischarge when the RF voltage across the electrodes is clearly insufficient to cause the breakdown of near-electrode sheaths. From the current-voltage characteristics of the RF discharge reported in  one can also drawa conclusion that lesser values of the RF voltage are required for the discharge to pass from a-to g- regime with gas pressure decreasing. Therefore, our paper adopted the following change of the RF discharge parameters as a criterion for the passage from a-tog-regime. Within almost all oxygen pressure range the a—g transition is accompanied by a decrease in the active Irf cos(j), i.e., the RF voltage value at which the active RF current reaches maximum can be used as a criterion of the start of the a— g transition of the RF discharge. Simultaneously, the a—g transition is accompanied by the rebuilding of the discharge structure: the uniform glowof the positive column characterizing a-regime is transformed into two negative glows and two Faraday spaces overlapping at the centre of the discharge gap at lowoxygen pressure. Visual observations clearly register the start of this rebuilding of the discharge structure and the appearance of the dark region at the discharge centre. The discharge glownear the boundaries of near-electrode sheaths also experiences changes: the white glow in a-regime acquires a violet tint during the a— g transition and in g-regime indicating the appearance of fast electrons in the discharge.
Fig. 3 shows the smallest RF voltages for discharge burning (extinguishing curve) 1 and the largest RF voltages for the weak-current regime to exist (start curve of the a—g transition) 2 against the oxygen pressure. It is clear from the figure that at the oxygen pressures p >700 Pa the a—g transition curve approaches the extinguishing curve of the RF discharge. As is known [16,32], the region of stable existence of weak-current regime of the RF discharge is limited from the moderate
30 I-■ ■ ■ "m.i-.........-.........
1 10 100 1000
Fig. 3. Extinguishing curve of the RF discharge (1), the alpha-gamma transition curve (2) and the region where the RF oscillations exist (shaded region bounded by solid triangles) 3.
pressure side (рв 1000 Pa), i.e., for a fixed inter-electrode distance there exists a pressure value pcr, such that for the pressures pXpcr the RF discharge can exist only in the strong-current regime. On decreasing the pressure, the RF voltage across the electrodes Ua-g, at which the a—g transition is observed first decreases and reaches a minimum (similar to a minimum at the ignition curve of the DC glowdischarge). Then Ua-g, increases a little and at the oxygen pressure p&90—100 Pa reaches a maximum. Then similar to the RF discharge in argon , the further decrease in the oxygen pressure leads to a fast decrease of Ua-g, and in the pressure range pp 5Pa the a—g transition curve coincides with the extinguishing curve of the RF discharge. Thus the range where the weak-current RF capacitive discharge exists is limited not only from the medium pressure side but also from the low-pressure one. At pp 5 Pa the RF discharge can exist only in the strong-current regime.
Fig. 3 also shows the region of existence of low-frequency oscillations (shaded region 3, limited with triangles). Within the pressure range p&27—133 Pa the lower boundary of this region practically coincides with the a—g transition curve, i.e., the oscillations in the RF discharge appear just belowthe a—g transition curve. In the weak-current regime the RF discharge appears visually as a column of uniform glowoccupying the total cross section of the discharge tube between the boundaries of darker near-electrode sheaths. Under the a—g transition the structure of the RF discharge experiences changes: on growing the RF voltage the glowbecomes brighter at the boundaries of near-electrode sheaths, and in the central region the glowfades gradually. At the same time, the probe records the low-frequency oscillations of the plasma potential (see Fig. 4).
Within the oxygen pressure range pE 133—307 Pa with the increase of the RF voltage at the start of the a—g transition the uniform glow is splitting into three distinctly expressed separate glowing regions: two of them are near the boundaries of near-electrode regions, and the third one is in the central part of the discharge. The central glowis similar to the positive column of the DC discharge, and the darker regions between it and the glows located near the boundaries of
200 300 400 500 600
Fig. 4. Plasma potential oscillation frequency at the discharge centre against RF voltage at different oxygen pressure values.
sheaths are similar to the dark Faraday spaces. The low-frequency oscillations do not appear in this range of RF voltages. On increasing the RF voltage further, the intensity of the glowin the central part decreases, and then this glowdis-appears. The low-frequency oscillations in the central part of the discharge are observed just in this range of RF voltages. Therefore within the pressure rangepE 133—307 Pa the lower boundary of the shaded region (Fig. 3) of the existence of oscillations is above the a—g transition curve. After the RF discharge has experienced the transition into the strong-current regime, the low-frequency oscillations disappear.
Fig. 5 shows the spectra of low-frequency oscillations generated at the RF discharge centre. It is clearly seen from the picture that an increase in the RF voltage leads to the decrease of the oscillation frequency. The oscillation amplitude was in the range of 1-5 V. The oscilloscope pictures of the probe signal also indicate that these oscillations possess a relaxation-like and not a harmonic pattern, their wavelength exceeding several times the inter-electrode distance. Therefore we cannot construct a dispersion law for these oscillations.
Let us consider the processes taking place when the RF discharge is passing from a-tog-regime at various oxygen pressure values. At lowpressure p< 5 Pa secondary electrons, having emerged from the electrode surface under bombardment by ions,
Fig. 5. LF oscillation spectra of plasma potential at the RF discharge centre with the oxygen pressure /7=133 Pa and different RF voltage values.
metastable atoms and photons, traverse the near-electrode sheaths actually without collisions acquiring over its width the energy up to ee eeUsh (where Ush is the RF voltage drop across the sheath). When UshXUi (Ui is the ionisation potential of the oxygen molecule via electron impact), a beam of fast electrons emerges from the sheath and penetrates the plasma, these electrons being capable to ionise gas molecules along their track. The RF discharge experiences transition to the g-regime, the characteristics of the quasineutral plasma become similar to two negative glows of the DC discharge overlapping at the discharge centre. When the oxygen pressure is sufficiently small, RF discharge can burn only in
the g-regime, because the voltage drop across the near-electrode sheath exceeds the ionisation potential of gas molecules in the whole region of discharge existence.
At higher pressure secondary electrons traversing the near-electrode sheath have an opportunity to experience one or several elastic or inelastic collisions what results in lesser energy acquired by them over the sheath width compared with the case without collisions. Therefore, with pressure increasing the a—g transition occurs at higher RF voltage values. At pressure values p>100 Pa the characteristics of the a—g transition are similar to the ignition of the DC glowdischarge: the near-electrode sheath is in the pre-breakdown state, the a—g transition curve is similar to a Paschen curve
Let us check whether the breakdown criterion for the near-electrode sheath is met. We will assume the RF electric field E to drop linearly on the way from the electrode surface to the sheath boundary
(x) = ^r(
where d is the thickness of the near-electrode sheath, x = 0 at the surface of the electrode. As the electric field is not constant within the sheath, we should write the breakdown criterion for the sheath in the following form:
where g is the coefficient of the secondary ion-electron emission from the electrode surface, a is the first Townsend's coefficient. We can employ for a the following approximate formula:
a = Ap exp
where in the case of strong fields E/p >150V/ (cmTorr) for oxygen A = 22.5 Torr— 1 cm— 1 and B = 340 V/(cmTorr). Let us first evaluate the coefficient g. At oxygen pressure of 1 Torr (133 Pa), measured thickness of the near-electrode sheath of 3 mm and the RF voltage value of
Uag = 385V for the start of the a—g transition the sheath breakdown criterion (2) is met if gE0.011. Then for same value of the coefficient g, the sheath thickness of 5 mm, the oxygen pressure of 0.1 Torr (13.3 Pa) and the RF voltage value Uag = 96V the coefficient m — 2.4 • 10—3 5 1; i.e., the criterion for breakdown of the near-electrode sheath (2) for the a—g transition of the RF discharge in low-pressure oxygen is not met. Generally, under these conditions we did not manage to find the value of the RF voltage in the range below10 kV, at which m — 1; even with the 10-fold increase of the quantity g, in any case we have m5 1. Breakdown curves of the DC discharge in oxygen possess a minimum at pdmin e 56 Pa cm (0.42 Torr cm) and UminE415V . The minimum of the a—g transition curve (Fig. 3) possesses the coordinates jpminE200Pa (1.5 Torr) and UagmnE374 V. For the measured sheath thickness de 0.25 cm the value of the product is pd=50.7 Pa cm (0.38 Torr cm), what is in good agreement with the data for the DC breakdown. Therefore we can conclude that at oxygen pressure of /?>100Pa the a— g transition is indeed accompanied by the breakdown of the near-electrode sheath. However at /?dmin = 0.05Torrcm (6.65 Pa cm) large voltage values Uagb 1000V are required for the breakdown of the near-electrode sheath . Consequently, at lowgas pressure /?<60Pa the near-electrode sheath of the RF discharge is not broken down, but is a source of beams of fast electrons.