Transcript Slide 1
Cylindrical probes are best high bias Plane probes have undefined collection area. low bias If the sheath area stayed the same, the Bohm current would give the ion density. A guard ring would help. A cylindrical probe needs only a centering spacer. A spherical probe is hard to make, though the theory is easier. UCLA In RF plasmas, the probe is more complicated! ChenB probe tip detail 3 .250 x .188 x 5/8 .001 Nickel .030 Tungsten 1 .094 x .063 x 1 1/8 .005 Tungsten rod 4 .060 OD vacuum slip joints 2 .050 x .020 x 3/8 Spot weld Solder 6 .125 x .040 x 1/4 5 double hole (1) (2) (3) (4) Coors #65682 Coors #65650 Coors #65658 Ceramaseal #11288-02-X Vacuum epoxy .998 Alumina Alumina or pyrex Copper Tungsten UCLA Parts of a probe’s I–V curve Vf = floating potential Vs = space (plasma) potential Isat = ion saturation current Iesat = electron saturation current I here is actually –I (the electron current collected by the probe UCLA The electron characteristic 0.10 0.08 Iesat Knee I (A) 0.06 0.04 Exponential 0.02 0.00 -0.02 0 5 10 15 V 20 25 30 UCLA The transition region (semilog plot) 100.0 Ie, w/o ions Ie (fit) Ie, raw I (mA) 10.0 1.0 0.1 0 5 Vp 10 15 20 UCLA The exponential plot gives KTe (if the electrons are Maxwellian) I e I es e e(V p Vs ) / KTe ) I es eAne v / 4 æöKTe ene A ç÷ èø2p m 1/ 2 UCLA Ion saturation current 0.020 0.015 This gives the plasma density the best. However, the theory to use is complicated. I (A) 0.010 0.005 0.000 -0.005 -100 -80 -60 -40 -20 0 20 V UCLA Thin sheaths: can use Bohm current Sheath edge Probe tip Ii = necs I B neAcs , cs ( KTe / M ) ½ 0.5, but is larger for thicker sheaths UCLA Floating potential (I = 0) ½ Isat I B ½neAcs , cs ( KTe / M ) Ie Ies exp[e(Vp Vs )/ KTe )] Set Isat = Ie. Then Vp = Vf, and KTe æ 2M ö V f Vs ln ç ÷ 2e è p m ø UCLA Electron saturation usually does not occur 0.10 Reasons: 0.08 sheath expansion collisions magnetic fields Iesat Knee I (A) 0.06 Ideal for plane 0.04 Exponential 0.02 0.00 -0.02 0 5 10 15 V 20 25 30 UCLA Extrapolation to find “knee” and Vs 4E-03 Vs? Electron current 3E-03 2E-03 1E-03 0E+00 -1E-03 -10 0 10 20 Probe voltage 30 40 50 60 UCLA Vs at end of exponential part 0.08 dIe/dVp = 0 or 2 0.06 2 d Ie/dVp = 0 Ie (A) 0.04 0.02 0.00 -0.02 -10 -5 0 5 Vp 10 15 20 25 UCLA Ideal minimum of dI/dV 2 0 -2 dI / dV -4 -6 -8 -10 -12 -14 0 5 10 V 15 20 25 UCLA Finally, we can get Vs from Vf KTe æ 2M ö V f Vs ln ç ÷ 2e è p m ø Vf and KTe are more easily measured. 6 5 However, for cylindrical probes, the normalized Vf is reduced when the sheath is thick. Vf /KT e 4 3 2 1 0 0.1 1 Rp / l D 10 100 1000 F.F. Chen and D. Arnush, Phys. Plasmas 8, 5051 (2001). UCLA Two ways to connect a probe v Cs PROBE Resistor on the ground side does not see high frequencies because of the stray capacitance of the power supply. R Floating milliammeter PROBE Resistor on the hot side requires a voltage detector that has low capacitance to ground. A small milliammeter can be used, or a optical coupling to a circuit at ground potential. R v UCLA RF Compensation: the problem B e e e e As electrons oscillate in the RF field, they hit the walls and cause the space potential to oscillate at the RF frequency. In a magnetic field, Vs is ~constant along field lines, so the potential oscillations extend throughout the discharge. UCLA E le ctro n current Effect of Vs osc. on the probe curve -20 -10 0 10 20 30 Vp -Vs UCLA The I-V curve is distorted by the RF 0.5 T e = 3 eV Helium 0.4 Electron C urrent V rf V) 0 0.3 5 10 0.2 15 0.1 0.0 -0.1 -20 -15 -10 -5 eV/KT e 0 5 10 UCLA Solution: RF compensation circuit* Epoxy seal 0.239" 0.094" coax compensation electrode 10 pF capacitor * V.A. Godyak, R.B. Piejak, and B.M. Alexandrovich, Plasma Sources Sci. Technol. 1, 36 (19920. I.D. Sudit and F.F. Chen, RF compensated probes for high-density discharges, Plasma Sources Sci. Technol. 3, 162 (1994) UCLA Effect of auxiliary electrode The chokes have enough impedance when the floating potential is as positive as it can get. The Chen B probe ChenB probe tip detail 3 .250 x .188 x 5/8 .001 Nickel .030 Tungsten 1 .094 x .063 x 1 1/8 .005 Tungsten rod 4 .060 OD vacuum slip joints 2 .050 x .020 x 3/8 Spot weld Solder 6 .125 x .040 x 1/4 5 double hole (1) (2) (3) (4) (5) (6) Coors #65682 Coors #65650 Coors #65658 Ceramaseal #11288-02-X Coors #65673 K.J. Lesker #KL-320K Vacuum epoxy .998 Alumina Alumina or pyrex Copper Tungsten Epoxy Nickel Auxiliary electrode: Nickel foil is wrapped around ceramic tube and spotwelded along length. Tabs are then cut, and the rest wrapped tightly to cylinder. A small tab is left for spotwelding to .030 tungsten rod. UCLA Sample probes (3) A commercial probe with replaceable tip UCLA Sample probes (1) A carbon probe tip has less secondary emission UCLA Example of choke impedance curve 1000 Probe 70103 900 2a = .010" = .0254 cm, L = 0.95 cm, R = 12.6 è 800 700 Z (k è ) 600 500 400 300 200 100 0 10 15 MHz 20 25 30 The self-resonant impedance should be above ~200KW at the RF frequency, depending on density. Chokes have to be individually selected. UCLA Equivalent circuit for RF compensation The dynamic sheath capacitance Csh has been calculated in F.F. Chen, Time-varying impedance of the sheath on a probe in an RF plasma, Plasma Sources Sci. Technol. 15, 773 (2006) UCLA Electron distribution functions If the velocity distribution is isotropic, it can be found by double differentiation of the I-V curve of any convex probe. n 4p c f (c)dc g (c)dc 0 2 0 d2 j 2e g (c) 8 2 , where (V p Vs ) m d (A one-dimensional distribution to a flat probe requires only one differentiation) This applies only to the transition region (before any saturation effects) and only if the ion current is carefully subtracted. UCLA Examples of non-Maxwellian distributions EEDF by Godyak A bi-Maxwellian distribution UCLA Example of a fast electron beam The raw data After subtracting the ion current After subtracting both the ions and the Maxwellian electrons UCLA Cautions about probe EEDFs • Commercial probes produce smooth EEDF curves by double differentiations after extensive filtering of the data. • In RF plasmas, the transition region is greatly altered by oscillations in the space potential, giving it the wrong shape. • If RF compensation is used, the I-V curve is shifted by changes in the floating potential. This cancels the detection of non-Maxwellian electrons! F.F. Chen, DC Probe Detection of Phased EEDFs in RF Discharges, Plasma Phys. Control. Fusion 39, 1533 (1997) UCLA Summary of ion collection theories (1) Langmuir’s Orbital Motion Limited (OML) theory Integrating over a Maxwellian distribution yields UCLA Langmuir’s Orbital Motion Limited (OML) theory (2) For s >> a and small Ti, the formula becomes very simple: I2 varies linearly with Vp (a parabola). This requires thin probes and low densities (large lD). UCLA Langmuir’s Orbital Motion Limited (OML) theory (3) 4 Ii squared Ii(OML) 2 2 I (mA) 2 3 1 0 -100 -80 -60 -40 Vp -20 0 20 UCLA Summary of ion collection theories (2) The Allen-Boyd-Reynolds (ABR) theory SHEATH PROBE The sheath is too thin for OML but too thick for vB method. Must solve for V(r). The easy way is to ignore angular momentum. UCLA UCLA Allen, Boyd, Reynolds theory: no orbital motion d æ d ö 1/ 2 J e 0 ç ÷ d è d ø This equation for cylinders was given by Chen (JNEC 7, 47 (65) with numerical solutions. Absorption radius Allen, J.E., Boyd, R.L.F., and Reynolds, P. 1957 Proc. Phys. Soc. (London) B70, 297 ABR curves for cylinders, Ti = 0 p = Rp/lD, p = Vp/KTe UCLA Summary of ion collection theories (3) The Bernstein-Rabinowitz-Laframboise (BRL) theory The problem is to solve Poisson’s equation for V(r) with the ion density depending on their orbits. Those that miss the probe contribute to ni twice. UCLA The Bernstein-Rabinowitz-Laframboise (BRL) theory (2) The ions have a monoenergetic, isotropic distribution at infinity. Here b is Ei/KTe. The integration over a Maxwellian ion distribution is extremely difficult but has been done by Laframboise. F.F. Chen, Electric Probes, in "Plasma Diagnostic Techniques", ed. by R.H. Huddlestone and S.L. Leonard (Academic Press, New York), Chap. 4, pp. 113-200 (1965) UCLA The Bernstein-Rabinowitz-Laframboise (BRL) theory (3) Example of Laframboise curves: ion current vs. voltage for various Rp/lD UCLA Summary of ion collection theories (4) Improvements to the Bohm-current method 0 .0 3 B R L the o ry p = 20 L ine a r fit Te = 3 e V , n = 4 x 1 0 B o hm c urre nt 12 cm -3 I (m A) Vf Vs 0 .0 0 I neApvB = 0 .7 4 -0 .0 3 -1 6 0 -1 4 0 -1 2 0 -1 0 0 -8 0 V -6 0 -4 0 -2 0 0 20 UCLA Variation of with p 2.0 Evaluated at Vf - 25KTe Extrapolated to Vf 1.5 1.0 0.5 0.0 0 10 20 p R p / lD , 30 40 50 2 lD 0 KTe / ne2 I B neAcs , cs ( KTe / M )½ , 0.5 Summary: how to measure density with Isat High density, large probe: use Bohm current as if plane probe. Ii does not really saturate, so must extrapolate to floating potential. Intermediate Rp / lD: Use BRL and ABR theories and take the geometric mean. Small probe, low density: Use OML theory and correct for collisions. Upshot: Design very thin probes so that OML applies. There will still be corrections needed for collisions. UCLA Parametrization of Laframboise curves 10 1 4 (A 1 1 B 4 ) (C Rp/lD D 4 ) 0 1 2 I / Io 2. 5 3 1 Fit 4 5 10 50 1 00 0 0 1 10 100 UCLA The fitting formulas BRL ABR F.F. Chen, Langmuir probe analysis for highdensity plasmas, Phys. Plasmas 8, 3029 (2001) UCLA Comparison of various theories (1) 6 BRL mWave BRL*ABR ABR 4 Density (10 11 -3 cm ) 5 3 2 1 0 200 300 400 500 600 Prf 700 800 900 1000 The geometric mean between BRL and ABR gives approximately the right density! UCLA Comparison of various theories (2) 4 Data OML, n = 3.45E11 ABR, n = 1.76E11 I2 (mA)2 3 BRL, n = 5.15E11 2 1 0 -100 -80 -60 -40 V -20 0 20 UCLA Comparison of various theories (3) Density increases from (a) to (d) ABR gives more current and lower computed density because orbiting is neglected. BRL gives too small a current and too high a density because of collisions. UCLA Reason: collisions destroy orbiting An orbiting ion loses its angular momentum in a chargeexchange collision and is accelerated directly to probe. Thus, the collected current is larger than predicted, and the apparent density is higher than it actually is. UCLA Including collisions makes the I - V curve linear and gives the right density Z. Sternovsky, S. Robertson, and M. Lampe, Phys. Plasmas 10, 300 (2003).\ However, this has to be computed case by case. UCLA UCLA The floating potential method for measuring ion density cs Vf d s 2 d (2 f )3 / 4 lD 1.018 f 3 / 4lD 3 (Child-Langmuir law) Unexpected success of the C-L (Vf) method 6 Ne(MW) Ni(CL) Ni(ABR) Ne,sat n (1011cm-3 ) 5 4 10 mTorr 3 2 1 0 300 450 600 Pr f (W) 750 900 Problems in partially ionized, RF plasmas • Ion currents are not as predicted • Electron currents are distorted by RF • The dc plasma potential is not fixed UCLA Peculiar I-V curves: not caused by RF 3.0E-04 0.0014 2.5E-04 0.0009 Amps 2.0E-04 1.5E-04 -I (A) Mk2, short tube, 100W Ideal OML curve 0.0004 -0.0001 -100 -80 -60 -40 -20 0 20 40 60 80 100 Vp 15 mTorr 3 mTorr 1.0E-04 5.0E-05 0.0E+00 -5.0E-05 -20 0 20 40 Volts 60 80 UCLA 100 Probe electron current can pull Vs up if the chamber is not grounded Antenna Not used 10.50 Port 2 21.00 Port 1 ` Grounding plate Permanent magnets on surface 36.00 Pump Port 2 EPN in Z-drive Grounding plate Port 1 1/4" probe in Wilson seal UCLA Direct verification of potential pulling 0.0003 Vp Amps 0.0002 0.0001 -0.0001 EPN I-V curves as Vp on ChenA is varied -50V 0V 10V 20V 30V 40V 50V -0.0002 -0.0003 -100 -75 -50 Volts -25 0 UCLA 25 Correcting for Vf shift gives better I-V curve 3.5E-03 3.0E-03 Manual, Vf corrected Manual, uncorrected Hiden MKIU 2.5E-03 Amps 2.0E-03 1.5E-03 1.0E-03 5.0E-04 0.0E+00 -5.0E-04 -50 -40 -30 -20 -10 0 Volts 10 20 30 40 UCLA 50 Hence, we must use a dc reference electrode ChenB probe with reference electrode 3 .250 x .188 x 1/4 .375 x .250 x 18 3 7 .250 x .188 x 5/8 .001 Nickel .030 Tungsten 1 .094 x .063 x 1 1/8 .219 x .156 x 22 4 .060 OD vacuum slip joints 2 .050 x .020 x 3/8 .005 Tungsten rod Compensation electrode Spot weld HERE Reference electrode Soft solder 6 Vacuum epoxy .998 Alumina ChenB probe with reference electrode Alumina or pyrex 3 .250 x .188 x 1/4 .375 x .250 x 18 3 Copper .001 Nickel .030 Tungsten 1 .094 x .063 x 1 1/8 .005 Tungsten rod 7 .250 x .188 x 5/8 Tungsten 4 .060 OD vacuum slip joints .219 x .156 x 22 8 2 .050 x .020 x 3/8 Epoxy Compensation electrode Nickel Spot weld Auxiliary electrode: Nickel foil is wrapped around ceramic tube and spotwelded along length. Tabs are then cut, and the rest wrapped tightly to cylinder. A small tab is left for spotwelding to .030 tungsten rod. (1) (2) (3) (4) (5) (6) (7) (8) .125 x .040 x 1/4 5 double hole Coors #65682 Coors #65650 Coors #65658 Ceramaseal #11288-02-X Coors #65673 K.J. Lesker #KL-320K Coors #65660 Coors 65657 or pyrex tube For large chokes, must use #65658, specially ordered to fit into #65660 Reference electrode Soft solder 6 .125 x .040 x 1/4 5 double hole (1) (2) (3) (4) (5) (6) Coors #65682 Coors #65650 Coors #65658 Ceramaseal #11288-02-X Coors #65673 K.J. Lesker #KL-320K Vacuum epoxy New, as of 1/1/2005 .998 Alumina Alumina or pyrex Copper Tungsten UCLA 8 Recent data in Medusa 2 (compact) Probe outside, near wall 0.30 10.0 Ii squared Ii(OML) 0.25 Ie Ie(fit) Ie (2) 1.0 I (mA) 0.15 2 I (mA) 2 0.20 0.1 0.10 0.0 0.05 0.00 -100 0.0 -80 -60 -40 V -20 n = 0.81 1011 cm-3 0 20 -6 -1 4 V Te = 1.37 eV UCLA 9 Recent data in Medusa 2 (compact) Probe under 1 tube, 7” below, 3kW, 15 mTorr 3.5 Ii squared Ii(OML) 3.0 2 1.5 I (mA) 2.0 2 2.5 1.0 0.5 n = 3.12 1011 cm-3 0.0 -100 -80 -60 -40 -20 0 V UCLA Recent data in Medusa 2 (compact) 100.0 Ie Ie(fit) Ie (2) I (mA) 10.0 1.0 Te (bulk) = 1.79 eV 0.1 0.0 -20 -15 -10 -5 V 0 5 10 Probe under 1 tube, 7” below, 3kW, 15 mTorr 10.00 Ie Ie(fit) Ie (2) I (mA) 1.00 Te (beam) = 5.65 eV 0.10 0.01 0.00 -20 -15 -10 -5 V 0 5 10 UCLA Other important probes not covered here • Double probes (for ungrounded plasmas) • Hot probes (to get space potential) • Insulated probes (to overcome probe coating) • Surface plasma wave probes (to overcome coating) Conclusion There are many difficulties in using this simple diagnostic in partially ionized RF plasmas, especially magnetized ones, but the problems are understood and can be overcome. One has to be careful in analyzing probe curves! UCLA Conclusion There are many difficulties in using this simple diagnostic in partially ionized RF plasmas, especially magnetized ones, but the problems are understood and can be overcome. One has to be careful in analyzing probe curves! UCLA