Electron Density and Ion Flux in Diffusion Chamber of Low Pressure RF Helicon Reactor

WDS'06 Proceedings of Contributed Papers, Part II, 150 155, 2006. ISBN 80-86732-85-1 MATFYZPRESS Electron Density and Ion Flux in Diffusion Chamber of Low Pressure RF Helicon Reactor R. Šmíd Masaryk University, Faculty of Science, Department of Physical Electronics, Brno, Czech Republic. A. Granier, A. Bousquet, G. Cartry IMN-CNRS, Université de Nantes, Nantes, France. L. Zajíčková Masaryk University, Faculty of Science, Department of Physical Electronics, Brno, Czech Republic. Abstract. The electron density, ion flux and argon emission lines were measured in a radio frequency helicon reactor for different static magnetic field amplitude and low rf powers using cylindrical Langmuir probe, planar probe and optical emission spectroscopy (OES). The static magnetic field amplitude created around the plasma source varied from 0 to 10 mt in the helicon source and from 0 to 1.4 mt in the diffusion chamber. The discharge was created in argon and oxygen at 0.7 Pa with rf powers, between 50 and 600 W, applied to the helicon antenna. The variations of the electron density, ion flux taken from Langmuir probe measurement and from planar probe measurement and Ar 750 nm line intensity with the rf power were in good agreement. Hence, it is concluded that the maximum of electron density observed in the diffusion chamber for the highest values of magnetic field is likely to be due to some confinement of the plasma in the source. Measurements in oxygen discharge showed slightly different results. Introduction Helicon generation of plasmas was first employed by Boswell [Boswell, 1970]. The study of helicon propagations and applications in material processing was intensified in next years (e. g. [Perry, 1991]). Helicons are propagating whistler wave modes in a finite diameter, axially magnetized plasma column, excited by an rf driven antenna that couples to the transverse mode structure across an insulating source wall [Lieberman, 1994]. Beside the wave coupling, two other coupling modes, capacitive and inductive, can be ignited in the source for certain (low) input rf powers and for certain magnetic field [Degeling, 1996]. Helicon applications to material processing utilize a diffusion chamber downstream from the source. The present paper is focused on an investigation of the rf helicon reactor used for material processing, namely on an influence of dc magnetic field and low rf powers on the discharge parameters in the diffusion chamber. The reactor was previously used for the deposition of silicon oxides and organosilicon polymer films as well as plasma diagnostics [Granier, 1997; Aumaille, 2000a, 2000b; Borvon, 2002a]. It was, however, equipped with one magnetic coil around the source tube. In recent study, two coaxial coils producing more homogeneous axial magnetic field were used. This study resumes on early results on the effect of the magnetic field on the discharge generated by the one coil source [Borvon, 2002b] and deals with non depositing gases only because of difficulties met in deposition conditions. Since the RF input powers were very low, the helicon modes are not supposed in our experiments. Experimental setup The studied helicon reactor consisted of the helicon source and the diffusion chamber. The helicon source was made of a glass tube surrounded by two Helmholtz coils and a helicon antenna (Fig. 1). The 30 cm long tube, 15 cm in diameter, was attached to the stainless steel diffusion chamber with the length and diameter, both of 30 cm. The two coils were fed by a DC current 0 2 A. The rf input power (13.56 MHz) was applied to the helicon antenna through an L-type matching box. RF continuous wave (cw) input power was changed 150

Figure 1. Simplified scheme of the rf helicon reactor. between 50 and 600 W. Argon (Ar) or oxygen (O 2 ) were injected through a small tube above the source directly to the source tube. Operating gas pressure was in both cases 0.7 Pa. The reactor was pumped by a turbomolecular pump placed under the diffusion chamber. The magnetic field trajectories were parallel to the reactor axis in the source and diverged in the diffusion chamber [Granier, 1997]. Compared to the one-coil source [Granier, 1997; Aumaille, 2000a, 2000b; Borvon, 2002a, 2002b], the amplitude of the axial magnetic field created by the two Helmholtz coils was more uniform. Calculations of this amplitude (B) showed that variation of the current ( ) in the coils from 0 to 2 A, corresponded to B = 0 10 mt in the source tube and 0 1.4 mt in the diffusion chamber. Such a weak magnetic field was assumed to have no influence on the probe measurements in the diffusion chamber. Diagnostic methods Langmuir probe: The discharge in the diffusion chamber was investigated by a commercial cylindrical RF compensated Langmuir probe (SmartProbe, 8 mm long tungsten tip, 0.1 mm in diameter) and a planar probe described below. Both probes were placed perpendicularly to the axis of the chamber, 15 cm under the source tube orifice, i. e. 4 cm below the organosilicon reactant injection ring (used for the deposition) and 4 cm above the substrate holder. The Langmuir probe characteristics reveal that in both gases, Ar and O 2, the electron energy distribution could be approximated by Maxwellian distribution and electron density as well as temperature were evaluated from these measurements. Planar probe: The future experiments are planned to be performed in deposition mixtures (such as oxygen/hexamethyldisiloxane), the results from the Langmuir probe were compared with the rf driven planar probe that is not affected by the deposition of thin dielectric film [Braithwaite, 1996; Booth, 2000]. The planar probe is, on principle, RF biased through a capacitance (C m ) and a self bias appears on probe surface [Braithwaite, 1996; Booth, 2000]. When the probe RF voltage is switched off, only ions can reach the probe. Our probe consisted of a concentric central disc (7 mm in diameter) and the outer guard ring (annulus, 16 mm in outer diameter). Both, the disc and the ring, behaved as rf electrodes and were connected to variable capacitors and resistors and they were grounded according to Fig. 2. through rf generator. We could write for the current I p potential V during the discharge: I p = C m dv dt = A pγ i (1) where A p is the probe surface. and Γ i ion flux. This equation stay valid even if the thin dielectric film is deposited on the probe disc and the guard ring [Braithwaite, 1996; Booth, 2000]. 151

Plasma Disc sheath Annulus 7,00 mm 16,00 mm Cm Cd disc Annulus Figure 2. Simplified scheme of the rf planar probe used to measure ion flux. Optical emission spectroscopy: To confirm the results from both probes we used an optical fiber to detect optical emission through a small quartz window in the chamber wall and above the source and a monochromator (JY HR640) equippped with Hamamatsu R928S photomultiplier. This arrangement enabled us to compare variations of the emission intensity in the source and in the diffusion chamber. In our experiments we supposed that Ar atoms came to their radiative state by direct excitation collisions with electrons. The intensity of Ar emission line (750 nm) should be in direct proportion to electron density since the electron temperature is almost constant under our experimental conditions (see next section). To study Ar emission line in both, Ar and O 2 discharge, we added 5% of Ar to pure O 2 discharge. Results and Discussion Results Measured Langmuir probe IV characteristic were time-averaged over the RF period. Calculations of electron energy probability function (EEPF) showed Maxwellian distribution function. We could neither observe double-maxwellian distribution [Godyak, 1992] in argon discharge, which is assigned to Ramsauer effect, nor perturbance in distribution due to negative ions in oxygen discharge. Example of EEPF s in argon discharge at 250 W are presented on the Fig. 3. Absence of low-energetic part (at the very beginning) of double Maxwellian distribution fuction could be caused by probe-plasma interaction or due to differentiation method [Rohmann, 1993]. So our disributions were treated as Maxwellian and the result temperatures were supposed to refer to high-energetic part of double-maxwellian distribution. Moreover we do not really know the effect of electron diffusion from source but we suppose that were could be some collisional effect since electron-neutral collisions are supposed in the diffusion chamber (see below). The electron temperature in the diffusion chamber of Ar discharge was obtained as 3.2 ev and almost constant for all input powers for = 1.5 and 2 A. For weaker magnetic fields ( 1 A), the electron temperature decreased strongly from about 3.2 ev for 25 W to 2.7 ev for 150 W and stayed constant for higher input powers. The electron temperature in the diffusion chamber of O 2 discharge decreased continuously with the input power from 3.8 ev to 2.8 ev for < 1 A. For 1 A the electron temperature stayed constant around 4 ev. Argon: The electron density dependence on the rf input power in the diffusion chamber of Ar discharge for different current to the coils ( ) is shown in Fig. 4(a). The electron density increased with the input power when a weak magnetic field was applied ( < 1 A). For more intensive magnetic field ( 1 A), the electron density exhibited a maximum which position and magnitude changed with. It reached 4 10 16 m 3 at 350 W for = 2 A and 2 10 16 m 3 at 200 W for = 1.5 A. For = 1 A, a very weak maximum of 4 10 15 m 3 appeared at 100 W. If we concentrate on the ion flux and 1 A (Fig. 4(b)), we observe the same trends compared to electron density. Ion flux exhibit the same maxima - for the same input powers and. This indicates that the results aren t caused by any Langmuir probe disturbance and both methods produce comparable results in argon discharge. 152

EEPF [m -3 ev -3/2 ] 10 14 10 13 0 2 4 6 8 10 12 0 E [ev] 1 0.5 2 1.5 [A] Figure 3. Electron energy probability function in argon discharge at 250 W and for different current to the coils. (a) (b) electron density [10 16 m -3 ] 5 4 3 2 1 0 = 0 A = 0.5 A = 1 A = 1.5 A = 2 A 0 100 200 300 400 500 600 ion flux [ma/cm 2 ] 0.8 0.6 0.4 =1 A =1.5 A =2 A 0 100 200 300 400 500 600 Figure 4. Electron density (a) and ion flux (b) in the diffusion chamber for different rf input power and current to the coils in argon discharge. In addition, we compared the ion saturation flux taken from the Langmuir probe measurements and the ion flux measured by the planar probe. Example of this comparison is shown in the Fig. 5(a) for = 1.5 A in argon discharge. Ion saturation flux (ion flux, Langmuir) is taken from ion saturation current at -50 V according to last term in eq. (1). Probe surface is different from planar probe surface and we could aproximate it by a cylinder (8 mm long and 0.1 mm in diameter). Bohm sheath criterion was taken into account. The resulted ion flux, Langmuir shows more than two times higher values than the ion flux, planar. This inequality could be caused due to bad ion flux saturation on Langmuir probe. Nevertheless, both methods have the same trends with the maxima at the same powers for all magnetic fields ( ). The electron density and ion flux show the same trends which conclude more or less constant electron temperature [Braithwaite, 1996]. To confirm the results from the Langmuir probe and the planar probe we compared the electron density and ion flux with intensity of Ar emission line. Assuming a direct electron excitation and a weak dependence of the electron temperature on the rf power (see above), the intensity of Ar emission line at 750 nm is proportional to the electron density. The comparison for Ar discharge for = 1.5 A is given in Fig. 5(b). All three diagnostics methods indicate the same trends. 153

(a) (b) ion flux [ma/cm 2 ] 1.4 1.2 0.8 0.6 0.4 Γ i, planar n e, Langmuir Γ i, Langmuir 0 100 200 300 400 500 600 30 25 20 15 10 5 0 electron density [10 15 m -3 ] ion flux [ma/cm 2 ] 0.5 2.0 1.8 0.4 1.6 1.4 0.3 1.2 0.8 ion flux 0.6 0.1 argon line 0.4 electron density 0 100 200 300 400 500 600 700 electron density [10 16 m -3 ] 3.5 3.0 2.5 2.0 1.5 0.5 argon line intensity [10 4 cps] Figure 5. Comparision among: (a) electron density, ion flux (by Langmuir probe) and ion flux (planar probe), (b) electron density, ion flux (planar probe) and intensity of Ar emission line at 750 nm in the diffusion chamber in Ar discharge at 0.7 Pa and for = 1.5 A. Oxygen discharge and source confinement: Similar agreement comparing the methods explained above was achieved for O 2 discharges. The electron density in oxygen discharge has slightly different behavior from argon discharge in the same power range. The electron density increased strongly with the input power for all except 1 A where a plateau of about 7 10 15 m 3 was observed from 400 W. The maximal electron density was about 1.1 10 16 m 3 for = 1.5 A. The intensity of Ar emission line in the source in both gases, Ar and O 2, increased with input power for all. Discussion In order to propagate the helicon waves in the helicon reactor, the power exceeding 500 W [Lieberman, 1994] or more than 1.5 kw is needed [Boswell, 1970]. Therefore, the helicon mode was not coupled in the power range used in the present study. Only the capacitive or inductive coupling was possibly present [Degeling, 1996]. The behavior of electron density observed especially in the diffusion chamber of Ar discharge is more likely due to some confinement of the plasma source, maybe magnetic field gradients at the source exit prevent diffusion. The diffusion coefficient D parallel to the magnetic field vector (and parallel to the reactor axis) is not influenced by magnetic field. The diffusion coefficient perpedicular to the magnetic field can be calculated as [Chen, 1974] D D = 1 + ω c2 τ 2 (2) where ω c is electron cyclotron frequency, τ 1 is mean collision frequency and D standard diffusion coefficient. The diffusion coefficient D calculated from eq. (2) using a reference data of electron to Ar collision cross section [Frost, 1964], varies in the range from 1 to 005 D for = 0 2 A, assuming a homogeneous magnetic field and electron temperature around 3 ev. The electron mean free path in argon discharge was about 12 cm, which is quite comparable to the reactor dimensions. Similarly, the electron to O 2 collision cross section taken from [Vahedi, 1993] gives the diffusion coefficient D in the range from 1 to 007 D for = 0 2 A. The electron mean free path is around 8 cm for O 2 discharge. This means that magnetic field could affect the difussion coeficients in both, the Ar and O 2 discharges, and then the electron density in the difussion chamber. Conclusion The variation of the electron density, ion flux, ion saturation flux and emission of argon lines in the diffusion chamber of rf helicon reactor for different static magnetic field amplitude and rf powers were measured using different techniques. It was shown that these experiments were in good agreement. The results depend on the magnetic field amplitude. In the case of argon discharge, it was found that the 154

electron density in the diffusion chamber increased continuously with the rf power for the current to magnetic coils < 1 A while it exhibited some maxima for 1.5 A. Ion flux taken from Langmuir probe IV characteristic and using planar probe shows the same trends for all input powers and currents to the coils but it is more than two times higher in the case of Langmuir probe measurement. The argon emission lines in the diffusion chamber respect the same trends as the probe results but the argon emission lines in the source did not present any maxima and continuously increased with the rf power. This could be explained by a magnetic confinement of the plasma in the source. Measurements in oxygen did not show any maxima of electron density in the power range presented. Electron density increased with input power in both, the plasma source and the diffusion chamber, in oxygen discharge. Acknowledgments. This research has been supported by the Barrande project 2005 2006 023 and by the projects MSM 0021622411, 202/03/H162. The first author would like to thank the French government for sponsorship of his foreign student short stay in Nantes. References R. W. Boswell: Phys. Lett. 33A, 1970, 457. A. J. Perry, D. Vender, R. W. Boswell: J. Vac. Sci. Technol. B, 9(2), 1991, 310. M. A. Lieberman and A. J. Lichtenberg: Principles of Plasma Discharges and Materials Processing. John Willey& Sons, New York 1994, 434. A. W. Degeling, C. O. Jung, R. W. Boswell and A. R. Ellingboe: Phys. Plasmas, 3(7), 1996, 2788. A. Granier, F. Nicolazo, C. Vallée, A. Goullet, G. Turban and B. Grolleau: Plasma Sources Sci. Technol., 6, 1997, 147 156. K. Aumaille, C. Vallée, A. Granier, A. Goullet, F. Gaboriau, G. Turban: Thin Solid Films, 359, 2000a, 188-196. K. Aumaille, A. Granier, M. Schmidt, B. Grolleau, C. Vallée and G. Turban: Plasma Sources Sci. Technol., 9, 2000b, 331-339. G. Borvon, A. Goullet, A. Granier, G. Turban: Plasmas and Polymers, 7(4), 2002a, 341. G. Borvon: Dépôt PECVD de matériaux à faible constante diélectrique. Ph.D. thesis, Université de Nantes, 2002b, in French. N. St. J. Braithwaite, J.P. Booth and G. Cunge: Plasma Sources Sci. Technol., 5, 1996, 677 684. J. P. Booth, N. St. J. Braithwaite, A. Goodyear and P. Barroy: Review of Scientific Instruments, 71, 7, 2000, 2722. V. A. Godyak, R. B. Piejak and B.M Alexandrowich: Plasma Sources Sci. Technol., 1,1992, 36 58. J. Rohmann, S. Klagge: Contrib. Plasma Phys, 33, 1993, 111 123. R. W. Boswell, A study of waves in gaseous plasmas, Ph.D. thesis, Flinders University of South Australian, 1970. A. W. Degeling, C. O. Jung, R. W. Boswell and A. R. Ellingboe: Phys. Plasmas, 3, 7, 1996, 2788. F. F. Chen: Introduction to Plasma Physics. Plenum Press, New York 1974. L. S. Frost, A. V. Phelps: Phys. Rev., 136, 1964, A1538. V. Vahedi, C. K. Birdsall, M. A. Lieberman, G. DiPeso and T. D. Rognlien: Plasma Sources Sci. Technol., 2, 1993, 273. 155