- Email: [email protected]
A study of plasma parameters in a BAI 730 M triode ion plating system by means of a Langmuir probe and plasma mass and energy spectroscopy
Surface and Coatings Technology 135 Ž2001. 208᎐220
A study of plasma parameters in a BAI 730 M triode ion plating system by means of a Langmuir probe and plasma mass and energy spectroscopy Marijan Macek ˇ a,b,U , Boris Navinsek ˇ b, Peter Panjan b, Stanislav Kadlec c a
Faculty of Electrical Engineering, Uni¨ ersity of Ljubljana, Trzaska ˇ ˇ 25, 1000 Ljubljana, Slo¨ enia b Jozef ˇ Stefan Institute, Jamo¨ a 39, 1000 Ljubljana, Slo¨ enia c Institute of Physics, Academy of Sciences, Na Slo¨ ance 2, 18040 Prague 8, Czech Republic Received 18 April 2000; accepted in revised form 25 August 2000
Abstract The plasma in a physical vapor deposition ŽPVD. system used for the deposition of hard coatings ŽTiN, CrN. was studied by means of a Langmuir probe and energy resolved spectroscopy ŽBalzers plasma process monitor PPM 421.. I᎐V measurements gave the plasma ŽUpl . and floating ŽUfl . potentials, as well as the electron temperature Te and plasma density n i . Upl deduced from I᎐V measurements agreed well with the peak of the positive ion energy distribution, as well as with the highest positive potential for the given operational mode. Energy spectra measured in deposition of TiN show a high degree of ionization of Ti, with Ti 2q as the prevalent ion. Te calculated from the Maxwellian distribution for the standard deposition of TiN is rather high ŽTe s 6᎐8 eV.. We believe that the oscillations of the plasma potential with the measured amplitude up to 15 V are most probably the reason. The electron energy distribution F Ž E . is better described by the Druyvesteyn distribution one than by a Maxwellian one. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Physical vapor deposition; Plasma; hard coatings; Langmuir probe; Energy resolved spectroscopy
1. Introduction Plasma-assisted processes are widely used in various areas of modern manufacturing, including the growing field of physical vapor deposited ŽPVD. hard coatings. These coatings can be deposited by several methods. One of them is triode ion plating which is widely used for deposition of films like TiN, CrN and TiŽC,N.. Triode ion plating systems use a filament-based ionization source to form a low voltage ŽLV. arc expanding into the vessel. Such a plasma is an efficient source of low energy electrons, used for enhancing ionization
U
Corresponding author. E-mail address: [email protected] ŽM. Macek ˇ ..
during deposition, as well as for substrate heating and plasma etching processes used prior to the deposition itself to prepare the surface of the substrate w1᎐3x. From the first attempts to implement plasma assisted processes there was a need for characterization of plasma parameters. Beside the conventional Langmuir probe technique, plasma spectroscopy w4x is being more and more widely used. From the I᎐V characteristics measured by the Langmuir probe, plasma parameters such as the plasma and floating potentials ŽUpl , Ufl ., electron temperature ŽTe . and the density of ions and electrons Ž n i , n e . in the plasma could be extracted, assuming a Maxwellian distribution of electrons. Without doubt, parameters obtained by this relatively simple method are very important ones, but not sufficient for a thorough characteri-
0257-8972r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 2 5 7 - 8 9 7 2 Ž 0 0 . 0 0 9 9 8 - 1
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
zation of the plasma, especially in the presence of a magnetic field when the interpretation of the I᎐V curve becomes less accurate. However, surface reactions during plasma assisted processes Žthin film growth, etching, cleaning. strongly depend on the mass and energy distribution of the principal species of reactants in the plasma. This is the reason why energy and mass resolved spectroscopy has nowadays become more and more important in studies of plasma-assisted processes. This paper compares the results obtained by the Langmuir plasma probe with the results of energy and mass resolved plasma spectroscopy during three modes of operation Žheating, etching, deposition. in a commercial triode ion plating system ŽBlazers BAI 730 M. used to deposit hard coatings like TiN, CrN and TiŽC,N.. Special attention was paid to the characterization of the TiN deposition process.
2. Experimental
209
during etching and deposition they are connected to the negative bias to attract the ions. The plasma, confined in the magnetic field formed by the top and bottom Helmholtz coils, is an efficient source of low energy electrons, used primarily for enhancing ionization during deposition, secondarily as an efficient heating source, and thirdly, as a source of electrons for melting of the evaporation material. Argon ions from the LV arc plasma are also used for physical cleaning of the substrate surfaces as needed to assure good adhesion of the hard coatings. Heating and etching processes performed by the LV arc plasma are used prior to the deposition itself to improve coating adhesion and microstructure. For TiN deposition, metals are evaporated in a mixture of argon and nitrogen gases. The Ar inlet is in the LV ionization source chamber and N2 comes directly into the main chamber through the top chamber wall as shown in Fig. 1. The standard parameters for heating, etching and for the deposition of the TiN layers, as well as voltages measured on the crucible and the auxiliary anode are given in Table 1.
2.1. Description of the system 2.2. Langmuir probe measurements Energy distributions of positive ions were measured in the commercial Balzers BAI 730 M triode ion plating systems. This system uses a filament-based ionization source, which forms a low voltage ŽLV. arc expanding into the reaction chamber. The power supply for the LV arc is electrically floating with respect to the chamber walls. Its negative pole is connected to the arc cathode, while the positive one can be connected, as shown in Fig. 1a᎐c, to different anodes, depending on the operation mode Žheating, etching and evaporation.. The ion plating system has axial symmetry of the overall vessel with respect to the vertical axis. The inner diameter of the vessel is 70 cm and the height 60 cm. Two Helmholtz coils producing an axial magnetic field are placed at the top and bottom of the vessel and are separated by 52 cm. The crucible is centrally located on the bottom and it can be moved upwards vertically up. All the measurements presented in this paper were made with the crucible at the bottom position. The LV arc source is mounted centrally on the top of the chamber. The LV beam is focused by the axial magnetic field. Thus the metal in the crucible is melted directly by the axial LV arc. Substrates are placed on 12 rotating holders symmetrically distributed along the chamber axis at a radius of 26 cm. The crucible is surrounded by the auxiliary anode in the form of a concentric ring with an outer diameter of 26 cm. This anode is used mainly for etching and is electrically floating during deposition. Cylindrical dummy substrates were located symmetrically, with the inner radii aligned with the spectrometer aperture. During heating they are floating, and
A cylindrical molybdenum Langmuir probe Žlength 1.5 mm, diameter 1 mm. with an area A p is located horizontally in the middle of the vessel height, similarly to the spectrometer. The radial position of the probe tip varies from 10 to 35 cm from the vessel axis. If not mentioned specifically, the radial position of the probe tip was 20 cm from the center, like the orifice of the spectrometer and was aligned with the inner surface of the dummy substrates. Data acquisition is governed by PC controlled electronics that enable accurate measurements to be made. I᎐V plots are measured in the range of at least 50 V below the floating potential and up to 50 V above the plasma potential. The region of special interest, i.e. between Ufl and Upl , is plotted with at least 30 equidistant points necessary for further digital smoothing of the signal as described in w5x, using the theory of non-recursive filters. Typically, a convolution over nine points and quadratic least squares fitting is used. The plasma parameters are derived from the measured I᎐V characteristic assuming a Maxwellian distribution of electrons, w6x. Of special interest is the region with probe voltages between the floating potential Žthe potential when Ie s Ii . and the plasma potential, Ufl - V - Upl . In this case, the measured current I is the sum of the saturated ion Ž IiU . and electron current Ž Ie ., I s yIUi q Ie Ž V .. The latter can be expressed as Ie s IeU exp Ž e Ž Upl y V . rkTe .
Ž1.
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
210
Fig. 1. BAI 730 in three different operating modes: Ža. heating, Žb. etching and Žc. ion plating mode.
Table 1 Parameters and potentials measured for standard operating conditions in all three operating modes Mode
pAr Žmbar.
pN 2 Žmbar.
Iarc ŽA.
Icoil ŽA.
B Žmtorr.
Usubstr. ŽV.
Uarc ŽV.
Uanode ŽV.
Ucruc. ŽV.
Heating ŽH. Etching ŽE. Deposition ŽD.
2.5 E-03 1.5 E-03 1.5 E-02
r r 5.0 E-03
150 130 200
8r8 8r8 15r15
3.7 3.7 7.0
12᎐17 y130 y125
49 55 56
y0.6 8.3 50
y12.5 y20.6 53
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
211
Obviously the electron temperature Te , can be determined from the slope of the electron current vs. probe voltage curve, Te s Ž erk .Ž ⭸ln Ier⭸V .
y1
Ž2.
This definition was used in our work if not explicitly noted. Ion density is estimated from the ion saturation current IiU . It is obtained by linear fitting of the I᎐V plot below Ufl and extrapolating to the plasma potential Upl . ni s
IiU 0.6 eA p
(
mi kTe
Ž3.
Similarly, the electron density is calculated from the portion of the I᎐V plot where the probe voltage exceeds the plasma potential, V ) Upl . In this case the saturated electron current IeU is the electron current at the plasma potential Upl . ne s
IeU eA p
(
2 me kTe
The plasma potential can be set to the point of the kink in the I᎐V curve, or to the point of the highest maximum of the first derivative of the electron current, ⭸ln Ier⭸V. The latter point is a little bit lower than the former one w4,6x and was used in the present work if not stated otherwise. From the second derivative of electron current the electron energy distribution function F Ž E . can be calculated: F Ž E. s
4 3 e Ap
(
m e E ⭸ 2 Ie 2 ⭸V 2
Ž5.
By definition the plasma potential can be set to the point where the second derivative equals zero. The plasma in the system is a cold plasma ŽTi < Te ., and, therefore, the difference between the plasma and floating potentials is proportional to the electron temperature w4,6x. For a typical argon plasma Ž Ms 40., the proportionality factor is 5.1. Upl y Ufl f w 3.3q 0.5ln Ž M .x ⭈
kTe kT s 5.1⭈ e e e
Ž6.
Knowing Te and n i the ratio ␣ between fluxes of ions Ž ji . and neutrals Ž jn . can be calculated: ␣ s jirjn
Fig. 2. PPM421 energy and mass analyzer.
Ž4.
Ž7.
where the flux of neutrals is equal to pr '2 MkT , the flux of ions equals ji s n i¨s and ¨s stands for the velocity of the ions accelerated by the pre-heat, ¨s
s kTerm i , respectively. For all calculations we assume the absolute gas temperature to be 600 K for the heating and etching step while for deposition 700 K was chosen.
'
2.3. Energy and mass spectrometry Energy distributions of positive ions were measured using the energy and mass analyzer ŽBalzers PPM421, Fig. 2., located at one side of the chamber at the mid-height. The analyzer axis is horizontal, crossing the chamber axis. The additional turbomolecular pump evacuates the spectrometer differentially. The floating orifice of the spectrometer is located 1 cm behind an aperture with a 5 mm hole which also serves as a shutter. This aperture is positioned 20 cm from the vertical axis of the vessel, and is aligned with the cylindrical dummy substrates. Both the aperture and the substrates are connected to the same potential. The orifice of the energy and mass analyzer is electrically floating. The ions from the plasma are focused by the ion optics, filtered by the cylindrical mirror energy analyzer and then pass through the quadrupole mass analyzer. The ionization chamber is a source of ions when working in the mode for detection of neutral particles. The secondary electron multiplier counts the ions, generating an avalanche of electrons to improve sensitivity. It can give readings from 10y1 to 10 7 counts per second ŽCPS.. More details are given in the literature w3,7x. The analyzer can perform a mass scan at selected ion energy or an energy scan at selected mass numbers.
212
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
The measured energy spectrum is actually a spectrum of the stopping potential; and, therefore, for multiply charged ions the real energy is the product of the stopping potential V and the ion charge q, Es qV. It is worth noting that ion transmission through the ion optics depends on the trajectory, so the probability of detecting ions from different directions is not the same. Moreover, this probability also depends on the energy. Therefore, the measured ion energy spectra may differ from the real spectra in the plasma.
3. Results 3.1. Heating Let us start with the first step of the normal deposition process, with the heating mode. As shown in Fig. 1a, the system is configured in such a way that the electrons from the plasma strike the surface of the substrate, which is at the positive floating potential. Heating depends on the LV arc current, Iarc , and on the substrate potential Žtypically 10᎐15 V.. In this way, the temperature can be raised to some 300᎐400⬚C to out-gas contaminated surfaces prior to deposition. 3.1.1. Langmuir probe measurements During this step, the plasma is a classical cold one, Ti < Te . Moreover, the mean free path for ions in the plasma is much greater than the probe dimensions itself rp and the Debye length D ; 4 rp 4 D . Therefore, the probe operates as a classical Langmuir probe. Typical results of the plasma probe I᎐V measurements at this step are given in Fig. 3. This shows the family of I᎐V plots for different arc currents in the range from 50 to 200 A. The highest current is used for heating of heavy process loads. As one can see, the plots differ mostly in the ion and electron saturation currents, while the other features do not differ significantly: i.e. the slopes between the floating and plasma potential, as well as the floating and plasma potentials. Typical plasma parameters Ž n i , Te , Upl , Ufl . are collected in Table 2 for a set of arc currents measured with the probe positioned 20 cm from the center. As is to be expected, the plasma potential Upl , the plasma density n i , the electron temperature Te , as well as the ratio between the fluxes of ions and neutrals jirjn increase with increasing arc current. In the same way as the plasma potential increases, the voltage measured on the dummy substrates connected to the arc power supply ŽPS. ŽFig. 1a. also increases. The results in Table 2 for the standard arc current Ž150 A. represent the typical range of measured values for all substrates, while the others are the values measured on a single dummy substrate opposite the spectrometer, and 90⬚ to the probe axis. Obviously the plasma potential roughly
Fig. 3. A family of I᎐V plots for heating configuration with different arc currents. The argon pressure is 2 = 10y3 mbar and coil current is 8 A.
follows this highest positive potential in the chamber. The correctness of the plasma probe measurements is verified by the fact that the proportionality factor which relates Te , Upl and Ufl in Eq. Ž6., calculated for the above plots, is 3.93" 0.30, reasonably close to the theoretical value of 5.1 for a cold Ar plasma. For lower arc currents this value is even closer to the theoretical one and vice versa. The plasma density n e determined from the electron part of the Langmuir probe characteristic is typically a factor two to four lower than the plasma density n i determined from the ion part of the Langmuir probe characteristic. This can be attributed to the magnetic field, which reduces electron saturation current confining them to magnetic field lines w6x. However, current can be also overestimated. Reasons are: secondary electron emission as a result of high probe temperature Ž) 700 K. and ion bombardment as well as to the fact that Eq. Ž3. is only the first approximation of the complex problem of ion transport. Parameters of the plasma vary along the radius of the chamber. In the middle of the chamber height, the radial profile of the main plasma parameters is as shown in Fig. 4a,b. While the plasma potential is almost constant for distances over 15 cm from the center, the density of ions and electrons, and in the same way the ratio between the ion flux and neutral flux, strongly depends on the radius. The electron temperature also shows a radial dependence up to the limits of the dummy substrates Ž r s 20 cm., and becomes constant for radii above 20 cm. 3.1.2. Energy spectra Energy spectra measured during the heating step reveal, besides the singly and doubly charged Ar ions, ŽArq, mrqs 40 and Ar 2q, mrqs 20., a high concen-
Machine parameters
Plasma parameters
Mode
Iarc ŽA.
Uarc ŽV.
H H H H
50 100 150 200
42 48 49 54
E
130
55
D D D D D D
50 100 150 175 200 220
30 41 48 51 54 56
Usub 噛1 ŽV.
DRŽ I .rDR Ž200 A. Ž%.
Ufl ŽV.
Upl ŽV.
Te ŽeV.
ni Žm-3.
ne Žm-3.
jirjn Ž%.
ŽUpl y Ufl . ŽkTe re.
r r r r
7.8 8.4 8.6 9.0
12.0 13.7 14.3 14.9
0.99 1.33 1.48 1.50
6.6 E q 16 1.2 E q 17 2.0 E q 17 2.8 E q 17
2.7 E q 16 4.2 E q 16 5.3 E q 16 8.4 E q 16
2 4 6 9
4.25 4.00 3.85 3.95
y130
r
8.7
13.8
1.36
1.7 E q 17
5.1 E q 16
4
3.72
y125 y125 y125 y125 y125 y125
0 0 8 56 100 139
29.1 38.4 42.8 34.3 23.6 16.1
32.2 41.8 47.1 50.4 52.4 53.3
0.62 0.70 0.85 3.25 6.01 7.50
9.7 E q 16 2.5 E q 17 3.0 E q 17 3.2 E q 17 4.1 E q 17 4.6 E q 17
8.9 E q 16 1.8 E q 17 3.1 E q 17 2.5 E q 17 3.9 E q 17 2.9 E q 17
1 6 8 14 25 33
4.98 4.82 5.14 4.95 4.80 4.96
13.2 13.6 12᎐17 14.6
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
Table 2 Summary of plasma parameters derived from I᎐V measurements 20 cm from the center axis for three operating modes. Operating conditions, except the arc current Ia are the same as in Table 1
213
214
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
tration of water ŽH 2 Oq, mrqs 18., oxygen ŽOq, mrq . s 16. and nitrogen ŽNq, mrqs 14, Nq 2 , mrqs 28 ions. In Fig. 5a,b the energy spectra of the main species, singly Ža. and doubly Žb. charged Ar ions are shown for the same set of arc currents Ž50᎐200 A. as in Fig. 3 Žplasma probe characteristics .. The energy spectra of Arq ions are affected mainly by the increased second peak at approximately 35 eV. While this peak is not present for low arc current Ž50 A., it is very pronounced at arc currents higher than 100 A. The height of the main peak of the energy distribution does not change with the arc current. The effect of increase in arc current, i.e. higher ionization, is to shift the peak position to values of potentials similar to those measured at the substrate holders. In Table 2 one can see that these potentials increase slightly with arc current from 13 to approximately 15 V.
Fig. 5. Energy spectra for heating under conditions given in Table 1: Ža. singly charged Ar Ž 40Arq .; Žb. doubly charged Ar Ž 40Ar 2q . ions.
The reason for the second peak at approximately 35 eV is not yet clearly understood. It could be related to some kind of discrimination in the spectrometer, or, more probably, it could represent the high energy ions coming from the central arc column, where the plasma potential is probably higher, Ž Varc s 49 V, Vcruc s y12.5 V, Varc-ground s 37.5 V., than outside the column. The energy distributions for doubly charged Ar ions in Fig. 5b do not show this second peak at 35 V up to the highest arc current. However, the height of the peak increases with arc current. The distribution is also narrower than the distribution of singly charged Ar ions under the same conditions, but the peak is centered around the same energy. During heating under the standard conditions for 1.5 h, the intensity of water, oxygen and nitrogen ions exponentially decreases by approximately an order of magnitude over a typical heating time of 45 min. This is related to out-gassing at the elevated temperature, but it is not sufficient for good cleaning of the surface. Fig. 4. Radial profile of the plasma parameters during the heating step. Ža. Plasma ŽUpl . and floating ŽUfl . potentials and electron temperature Te ; Žb. ion Ž n i ., electron Ž n e . density and the ratio between ion and neutral fluxes ji r jn .
3.2. Etching During the etching step the BAI730 M system is
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
215
configured in such a way that the Ar ions are accelerated towards the biased Žtypically Usub s y130 V. substrates Žsee Fig. 1b.. These accelerated ions clean the surface by sputtering. This effect depends on the flux of impinging ions and their energy wA ŽUpl y Usub .x. 3.2.1. Langmuir probe measurements The I᎐V measurements of the Langmuir probe reveal quite similar results to those in the heating mode. Plasma parameters are summarized in Table 2 for standard etching with Iarc s 130 A. Plasma parameters are somewhere between the results obtained during heating with arc currents of 100 and 150 A in spite of the differences in configuration of the electrodes. Upl is again close to the highest positive potential in the chamber. In this case, however, it is the auxiliary anode, which is connected directly to the arc PS and to ground via a resistor on the same way as substrates in the heating mode ŽFig. 1b.. They are now at fixed negative potential Usub but float at 12 to 17 V during heating. The radial profile of the plasma measured at midheight of the chamber is shown in Fig. 6a,b. The plasma potential is almost constant for distances over 15 cm from the center. The electron temperature increases toward the center, as well as the ratio jirjn between the ion and neutral fluxes reaching the surface of the probe, as well as the densities of ions and electrons. However, generally there is no substantial difference in the plasma radial distribution from the one measured during heating. The main observable difference to the heating mode is that the plasma potential is the highest at a radius of approximately 15 cm, roughly corresponding to the anode radius. During heating, the maximum of plasma potential is found close to the radius of the substrate holders, which function as the anode in that mode. 3.2.2. Energy spectra Energy spectra of the singly and doubly charged species in this mode are not shown. They are very similar to those measured during the heating mode. The main differences in energy spectra are the absence of the second peak located at approximately 35 eV for singly charged particles and a much narrower energy distribution. There is no measured signal beyond 35 eV for etching Žcomparing with 60 eV for heating. at the highest Iarc s 200 A. Because of the extreme conditions in the center of the chamber Žvery high plasma density resulting in probe melting. we do not have direct evidence for the plasma potential there. However, from measurements of the floating potential of the crucible it seems that the plasma potential can be negative during both heating and etching, some 20 V more negative during etching. This high potential barrier could prevent the
Fig. 6. Radial profile of plasma parameters during the etching step. Ža. plasma ŽUpl . and floating ŽUfl . potentials and electron temperature Te ; Žb. ion Ž n i ., electron Ž n e . density and the ratio between ion and neutral fluxes ji r jn .
high energy positive ions coming from the chamber axis from being detected at the mass spectrometer position. The position of the peak does not change significantly with arc current and is close to the measured potential on the auxiliary anode ŽUan s 7.5 V at Iarc s 50 A and 8.6 V at 200 A.. However, the position of both peaks can be influenced by the magnitude and shape of the magnetic field. Fig. 7a,b shows the energy spectra of Arq ions during etching. The curves differ in respect to the magnetic field, produced by the coil currents I bottom and Itop in the bottom and top coil, respectively. The central curve shows the result for the standard conditions Ž I bottom s Itop s 8 A.. The other curves were measured for different combinations of coil currents. Fig. 7a was measured for various magnetic field
216
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
plasma potential increases, reflected in the peak position of the measured curve. In energy spectra measured during the etching step there are also other ions, such as atomic ŽOq, mrqs . 16. and molecular oxygen ŽOq 2 , mrqs 32 , and water ŽH 2 Oq, mrqs 18.. While the intensity of molecular oxygen Oq 2 decreases at the same rate as during heating Žout-gassing at elevated temperatures., the intensity of atomic oxygen and water drops very quickly after the system is switched from heating to the etching. At the same time the intensity of iron ŽFe, mrqs 56. increases, indicating the beginning of surface etching via the sputtering process. The intensity of iron sputtered from the surface is stabilized after approximately 15 min Žalso the prescribed etching time.. 3.3. Deposition
Fig. 7. Energy spectra of Arq ions for etching with different magnetic fields Žcoil currents .. Other conditions are the same as in Table 1: Ža. symmetric magnetic field, I bottom s Itop ; Žb. asymmetric magnetic field with I bottom / Itop .
strengths, while I bottom s Itop. The measurements show that the maximum of the curve, corresponding to the plasma potential, shifts towards higher energy if the magnetic field increases. Also the width of the curve increases. This effect is related to the magnetic confinement of the plasma. This is confirmed even better by the measurements shown in Fig. 7b, measured for various combinations of coil currents I bottom and Itop . The total curve width can be quite small if the current I bottom is small Ž1 A. relative to the top coil current Ž Itop s 20 A.. This corresponds to a small positive anode voltage as shown by the vertical lines ŽUa s 1.6 V.. In this case the magnetic field lines are divergent from top to bottom. Thus the plasma column is guided from the top filament source to the ring anode at the chamber bottom. The other case is when the current I bottom is large Ž20 A. relative to the top coil current Ž Itop s 1 A.. Then the plasma has to cross the magnetic field lines, divergent from bottom to top. Therefore, the anode voltage increases ŽUa s 31.1 V. and similarly increases the
3.3.1. Langmuir probe measurements I᎐V characteristics measured during the deposition of TiN ŽFig. 8. with various arc currents Žother settings the same as in Table 1. show a strong influence of Iarc on the characteristics, and therefore the plasma parameters are very sensitive to variations in Iarc . Especially pronounced is the shift of Ufl , which changes from 42.8 V at Iarc s 150 A to 16.1 V at 220 A. However, Upl shifts only slightly from 47.1 to 53.3 V at the same arc currents. This shift would result in a huge increase of Te , from 0.85 to 7.9 eV, if the distribution were Maxwellian. For 50 - Iarc - 150 A, the situation is less clear Žsee Table 2.. While Upl still decreases with decreasing Iarc , as above 150 A, Ufl decreases, but the difference ŽUpl y Ufl ., which is proportional to Te , does not change significantly ŽTe changes only from 0.85 to 0.6 eV.. The radial distribution of plasma parameters under the standard deposition conditions is shown in Fig. 9.
Fig. 8. A family of I᎐V plots for deposition of TiN with different arc currents. Other parameters Ž pAr , p N 2 , Ic . were the same as in Table 1 for the standard process.
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
217
jirje increase rapidly for arc currents above 150 A, when the melting of Ti visually starts. However, the plasma densities during heating and etching increases gradually with increasing arc current. A very important parameter is the electron energy distribution function F Ž E .. If the electrons are in thermal equilibrium, the distribution is Maxwellian, normally used to describe cold plasma: F Ž E . A 'E exp y
ž
E kTe
/
Ž8.
If the electrons are not in thermal equilibrium, the distribution can be the Druyvesteyn one w8x F Ž E . A 'E exp y0.24
Fig. 9. Radial profile of plasma parameters during TiN deposition. Ža. plasma ŽUpl . and floating ŽUfl . potentials and electron temperature Te , Žb. ion Ž n i ., electron Ž n e . density and the ratio between ion and neutral fluxes ji r jn .
Plasma potential Upl is constant over all measured distances Ž50᎐53 V., even behind the dummy cylinders, shown as gray rectangles in the figure. This value is close to the measured potential on the crucible ŽTable 1.. The electron temperature is also almost constant, with a small peak at the radii of the substrates. However, ion and electron densities Ž n i , n e ., and the ratio between ion and neutral fluxes Ž jirjn . rapidly decrease from the inside Ž r s 12.5 cm. towards the radius of the substrates Ž25 cm.. Beyond this radius the densities drop very rapidly. The ratio between ion and neutral fluxes during TiN deposition is rather high. It is higher than 0.25 at the inner radius of the substrates Ž r s 20 cm.. For comparison, the ratio measured during the heating with the same current and total pressure is only 0.09. From Table 2 one can see that the effects of a plasma characterized by parameters Te , Ni , Ne and
E kTe
ž /
2
Ž9.
In Fig. 10 we compare electron energy distributions for standard deposition conditions and for heating with the same magnetic field as used for deposition, B s 7 mtorr. Both measured distributions are very similar up to 12 eV and can be fitted better with the Druyvesteyn distribution function given by Eq. Ž9. than by the Maxwellian one wEq. Ž8.x. The electron temperature Te is estimated as 2.0 eV for heating and 2.5 eV for deposition. The latter is much smaller than the electron temperature estimated from Eq. Ž6. ŽTe s 5.8 eV. or obtained from the best fit of the I᎐V plot ŽTe s 7.5 eV. in Fig. 8. It appears that during TiN deposition we are dealing with electrons coming from different plasma potentials. ŽThis effect is not seen in the electron distributions for heating or etching.. In fact, measurements of Ufl with a very fast sampling rate Ž318 s. did show large fluctua-
Fig. 10. Electron energy distribution function for standard deposition of TiN and for heating with the same magnetic field as during deposition Ž Ic s 15r15 A.. Obviously the electron energy distribution during the deposition is not Maxwellian Ždashed line..
218
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
tions up to 15 V. This makes the evaluation of the probe characteristic difficult, if not impossible. In measurement such as those in Fig. 8 the plasma potential most of the time remains approximately 50 V but for short periods it falls by 5᎐15 V, collecting mostly the electrons instead of ions. This can explain the shape of F Ž E . for energies above 12 eV as well as the large difference between the apparent Upl and Ufl and increasing Te . 3.3.2. Energy spectra Energy spectra of Ar and Ti ions for different Iarc are shown in Fig. 11a᎐d. In Fig. 11a,b spectra for doubly charged ions Ž 40Ar 2q, mrqs 20, 48 Ti 2q, mrqs 24. are represented, while in Fig. 11c,d spectra for singly charged ions Ž 40Arq, mrqs 40, 48 Tiq, mrqs 48. are shown. The main difference in Ar and Ti spectra is that the intensity of the doubly charged ŽTi 2q . ion is much stronger than that of the singly charged ŽTiq. ion. This is the opposite of that for argon and nitrogen ions. In ref. w9x it was shown that the ion with mrqs 16, that is triply charged 48 Ti 3q ion, Žnot oxygen., was detected with relatively strong intensity. In Fig. 11 the peaks have almost identical shape for all measured ions and arc currents. For the standard Iarc , the peak is centered at 57 " 0.5 eV. This is close to the plasma potential measured by the plasma probe and to the potential of the crucible. A very important fact is that the intensity of Ti ions ŽTiq, Ti 2q, Ti 3q . increases with Iarc , while that of Ar and N decreases. These effects are partially due to the
increased evaporation rate and partially due to the increased ionization as it is discussed in Section 4. There is also a difference in the shape of the distribution, since the Tiq ions exhibit almost no low energy tail Žbelow 40 eV. up to the Iarc over 200 A. This tail is relatively strong for other ions shown, as well as for the nitrogen. This low energy tail seems to correspond well with the temporal distribution of the plasma potential. This ion signal seems originate from the plasma potential in the short time when it is lower for some 5᎐15 V.
4. Discussion On the basis of the I᎐V measurements in the previous section it was shown how the plasma behaves during heating Žalmost the same as during etching., and during the deposition of TiN under different arc currents. With increasing current all plasma parameters except Te and Ufl more or less gradually increase in both modes of operation ŽTable 2.. These relationships are shown in Fig. 12. During deposition, the electron temperature sharply increases after the onset of Ti evaporation. The difference ŽUpl y Ufl . increases in a similar way. The explanation seems to be the onset of instability in the plasma potential. This is reflected also in the shape of the ion energy distribution, see Fig. 11. A difficult question to answer is how such plasma instability with large amplitude Ž; 15 V. starts and is sustained during Ti evaporation. The interaction
Fig. 11. Energy spectra of Ar Ža,c. and Ti Žb,d. ions for deposition of TiN with different Iarc .Ž150, 175, 210 A. Other conditions are the standard ones as given in Table 1: Ža. 40Arq2 Ž mrqs 20.; Žb. 48 Tiq2 Ž mrqs 24.; Žc. 40Arq Ž mrqs 40.; Žd. 48 Tiq Ž mrqs 48..
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
219
Consequently, the higher is the probability of Penning ionization according to the following relationship: ArU q Tiª Ar q Tiqq ey
Ž 10 .
Another process in the plasma is the electron-impact ionization of excited Ar atoms: U
eyq Ar ª Arqq 2 ey Eq y Impact y Ion
Fig. 12. Plasma parameters vs. Iarc for heating and TiN deposition. Conditions are the same as in Table 1: Ža. plasma potential Upl and floating potential Ufl . Žb. electron temperature Te ; ratio between ion and neutral fluxes ji r jn
between the electron beam in the magnetic field and the plasma containing Ti ions could be the cause. In the Langmuir probe measurements the electron temperature Te deduced from the fit of the I᎐V plots may be a little overestimated, since the ratio between ŽUpl y Ufl .rTe is even below 4 at higher currents for heating mode. According to theory it should be 5.1, and therefore either Upl is underestimated or Te is overestimated. The reason for underestimation of Upl could be small differences in potential of individual substrates during heating. The deposition rate Ž DR ., measured by the net consumption of nitrogen gas, correlate very well Ž r 2 ) 99%. with the arc current and with the electron temperature calculated from Eq. Ž6. for Iarc ) 150 A, as can be seen in Table 2, where the results for DR are normalized to the values for standard deposition of TiN. The higher is the arc current, the higher is the evaporation rate and the concentration of Ti atoms.
Ž 11.
Since the excitation energy of Ar, 11.5 eV, is large compared with the ionization energy of the Ti, 6.8 eV, the probability of Penning ionization of Ti should increase and for impact ionization of Ar should decrease as the number of Ti atoms increases with increasing arc current. Therefore, the source of additional electrons and ions increasing the ratio jirjn during TiN deposition ŽFig. 12c. could be Penning ionization. This fact is also confirmed by spectroscopic ŽPPM. measurements. Integration of the energy distribution in Fig. 11 shows that the intensity of the singly Ž mrqs 48. and to an even more pronounced degree of doubly Ž mrqs 24. and triply Ž mrqs 16. charged Ti ions follows this relationship Žincrease with Iarc .. On the other hand the intensity of singly Ž mrqs 40. and also doubly Ž mrqs 20. charged Ar ions decreases with increasing arc current during the deposition of TiN. The intensity of the ion flux relative to the flux of neutrals ŽFig. 12c. in deposition with Iarc over 150 A increases very strongly mainly due to the increase in Ti evaporation and its easy ionization to the doubly and singly charged states. Moreover Penning ionization may also leads to fragmentation of metal clusters. It was suggested in ref. w10x that metal evaporation at typical low pressure ion plating conditions predominantly occurs in the form of clusters with more than 10 atoms per unit charge. In our mass spectroscopic measurements up to 511 a.m.u. we never detected any particle with mrq larger than 50 Ž 50 Tiq .. If evaporated metal is really in the form of clusters, than Penning process can also cause their multiple ionization. From the facts that approximately 90% of ions striking the surface of substrates is Ti and that we never detected particles with mrq higher than 50 we may conclude that Penning ionization leads to the complete fragmentation of Žat least. clusters with less than 10 Ti atoms per unit charge. During the heating Žand etching. mode the situation is different. The main species in the vessel is Ar. Therefore the electron-impact ionization of Ar is the most important ionization mechanism. The intensity of singly and doubly charged ions increases gradually with Iarc in the range from 50 to 200 A, and there is no ‘anomaly’ with a decreasing floating potential as during the deposition. As the intensity of Ar ions detected by the spectrometer increases linearly with Iarc , in the same way the
220
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
ratio between the fluxes of Žmainly. Ar ions and neutrals also increases.
5. Conclusions On the basis of plasma probe measurements and of energy and mass spectroscopy we can conclude that the plasma during the heating and etching modes behaves like a classical cold plasma with electrons in thermal equilibrium with a mean energy below 1.5 eV for the standard operating conditions. The flux of ions is low compared with the flux of gas neutrals, only a few percent. The plasma potential is very uniform over the vessel radius. It corresponds to the highest potential in the chamber, i.e. the potential of the substrates during heating or the potential of the auxiliary anode during etching. During the deposition of TiN, as well as other metals like Cr, the plasma is very intense. The ratio between fluxes of ions and neutrals Žgas and metal. is high compared with heating and etching, jirjn s 25% at the standard conditions. Mass spectroscopy shows that the majority of the Ti ions are doubly charged ones. The apparent energy of electrons under the standard deposition conditions with a strong magnetic field, B s 7 mtorr is very high. Assuming thermal equilibrium and a quiescent plasma, the electron temperature Te should be approximately 6 eV. The energy distribution of the electrons is better described by the Druyvesteyn function than by the Maxwellian one. Moreover, a large plasma instability was observed, especially during TiN deposition. This affects the apparent Langmuir probe characteristics, as well as the ion energy distribution.
Acknowledgements This work was supported by the Ministry of the Science and Technology of the Republic of Slovenia and is also part of the bilateral scientific collaboration between the Czech Republic and the Republic of Slovenia. We also acknowledge the support of a NATO-Collaborative Research Grant, Project HTECH.CRG 970529, and the project ME110 of the Czech Ministry of Education. This work was also partly supported by the Grant Agency of the Czech Republic, Grant no. 106r96rK245. References w1x E. Bergman, Surf. Coat. Technol. 57 Ž1993. 133᎐137. w2x E. Mol, E. Bergman, Surf. Coat. Technol. 37 Ž1989. 453᎐509. w3x M. Nesladek, C. Quaeyhaegens, S. Wouters, L.M. Stals, E. ´ Bergmann, G. Rettinghaus, Surf. Coat. Technol. 68r69 Ž1994. 339. w4x A. Grill, Cold Plasma in Materials and Fabrication, IEEE Press, New York, 1994. w5x R.W. Hamming, Digital Filters, 3rd ed, Prentice-Hall, Engelwood Cliffs, NJ, 1989. w6x N. Herskowitz, Langmuir probe diagnostics, in: D.A. Glocker, S. Ismath Shah ŽEds.., Handbook of Thin Film Process Technology, Institute of Physics Publishing, Bristol, 1995, p. D3.0:1. w7x S. Wouters, S. Kadlec, M. Nesladek, C. Quaeyhaegens, L.M. ´ Stals, Surf. Coat. Technol. 76 Ž1995. 135. w8x M.J. Druyvesteyn, F.M. Penning, Rev. Mod. Phys. 12 Ž2. Ž1940. 88. w9x M. Macek, ˇ B. Navinsek, ˇ P. Panjan et al., Surf. Coat. Technol. 113 Ž1999. 149. w10x K.S. Fancey, A. Matthews, Appl. Phys. Lett. 55 Ž9. Ž1989. 834.
A study of plasma parameters in a BAI 730 M triode ion plating system by means of a Langmuir probe and plasma mass and energy spectroscopy Marijan Macek ˇ a,b,U , Boris Navinsek ˇ b, Peter Panjan b, Stanislav Kadlec c a
Faculty of Electrical Engineering, Uni¨ ersity of Ljubljana, Trzaska ˇ ˇ 25, 1000 Ljubljana, Slo¨ enia b Jozef ˇ Stefan Institute, Jamo¨ a 39, 1000 Ljubljana, Slo¨ enia c Institute of Physics, Academy of Sciences, Na Slo¨ ance 2, 18040 Prague 8, Czech Republic Received 18 April 2000; accepted in revised form 25 August 2000
Abstract The plasma in a physical vapor deposition ŽPVD. system used for the deposition of hard coatings ŽTiN, CrN. was studied by means of a Langmuir probe and energy resolved spectroscopy ŽBalzers plasma process monitor PPM 421.. I᎐V measurements gave the plasma ŽUpl . and floating ŽUfl . potentials, as well as the electron temperature Te and plasma density n i . Upl deduced from I᎐V measurements agreed well with the peak of the positive ion energy distribution, as well as with the highest positive potential for the given operational mode. Energy spectra measured in deposition of TiN show a high degree of ionization of Ti, with Ti 2q as the prevalent ion. Te calculated from the Maxwellian distribution for the standard deposition of TiN is rather high ŽTe s 6᎐8 eV.. We believe that the oscillations of the plasma potential with the measured amplitude up to 15 V are most probably the reason. The electron energy distribution F Ž E . is better described by the Druyvesteyn distribution one than by a Maxwellian one. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: Physical vapor deposition; Plasma; hard coatings; Langmuir probe; Energy resolved spectroscopy
1. Introduction Plasma-assisted processes are widely used in various areas of modern manufacturing, including the growing field of physical vapor deposited ŽPVD. hard coatings. These coatings can be deposited by several methods. One of them is triode ion plating which is widely used for deposition of films like TiN, CrN and TiŽC,N.. Triode ion plating systems use a filament-based ionization source to form a low voltage ŽLV. arc expanding into the vessel. Such a plasma is an efficient source of low energy electrons, used for enhancing ionization
U
Corresponding author. E-mail address: [email protected] ŽM. Macek ˇ ..
during deposition, as well as for substrate heating and plasma etching processes used prior to the deposition itself to prepare the surface of the substrate w1᎐3x. From the first attempts to implement plasma assisted processes there was a need for characterization of plasma parameters. Beside the conventional Langmuir probe technique, plasma spectroscopy w4x is being more and more widely used. From the I᎐V characteristics measured by the Langmuir probe, plasma parameters such as the plasma and floating potentials ŽUpl , Ufl ., electron temperature ŽTe . and the density of ions and electrons Ž n i , n e . in the plasma could be extracted, assuming a Maxwellian distribution of electrons. Without doubt, parameters obtained by this relatively simple method are very important ones, but not sufficient for a thorough characteri-
0257-8972r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 2 5 7 - 8 9 7 2 Ž 0 0 . 0 0 9 9 8 - 1
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
zation of the plasma, especially in the presence of a magnetic field when the interpretation of the I᎐V curve becomes less accurate. However, surface reactions during plasma assisted processes Žthin film growth, etching, cleaning. strongly depend on the mass and energy distribution of the principal species of reactants in the plasma. This is the reason why energy and mass resolved spectroscopy has nowadays become more and more important in studies of plasma-assisted processes. This paper compares the results obtained by the Langmuir plasma probe with the results of energy and mass resolved plasma spectroscopy during three modes of operation Žheating, etching, deposition. in a commercial triode ion plating system ŽBlazers BAI 730 M. used to deposit hard coatings like TiN, CrN and TiŽC,N.. Special attention was paid to the characterization of the TiN deposition process.
2. Experimental
209
during etching and deposition they are connected to the negative bias to attract the ions. The plasma, confined in the magnetic field formed by the top and bottom Helmholtz coils, is an efficient source of low energy electrons, used primarily for enhancing ionization during deposition, secondarily as an efficient heating source, and thirdly, as a source of electrons for melting of the evaporation material. Argon ions from the LV arc plasma are also used for physical cleaning of the substrate surfaces as needed to assure good adhesion of the hard coatings. Heating and etching processes performed by the LV arc plasma are used prior to the deposition itself to improve coating adhesion and microstructure. For TiN deposition, metals are evaporated in a mixture of argon and nitrogen gases. The Ar inlet is in the LV ionization source chamber and N2 comes directly into the main chamber through the top chamber wall as shown in Fig. 1. The standard parameters for heating, etching and for the deposition of the TiN layers, as well as voltages measured on the crucible and the auxiliary anode are given in Table 1.
2.1. Description of the system 2.2. Langmuir probe measurements Energy distributions of positive ions were measured in the commercial Balzers BAI 730 M triode ion plating systems. This system uses a filament-based ionization source, which forms a low voltage ŽLV. arc expanding into the reaction chamber. The power supply for the LV arc is electrically floating with respect to the chamber walls. Its negative pole is connected to the arc cathode, while the positive one can be connected, as shown in Fig. 1a᎐c, to different anodes, depending on the operation mode Žheating, etching and evaporation.. The ion plating system has axial symmetry of the overall vessel with respect to the vertical axis. The inner diameter of the vessel is 70 cm and the height 60 cm. Two Helmholtz coils producing an axial magnetic field are placed at the top and bottom of the vessel and are separated by 52 cm. The crucible is centrally located on the bottom and it can be moved upwards vertically up. All the measurements presented in this paper were made with the crucible at the bottom position. The LV arc source is mounted centrally on the top of the chamber. The LV beam is focused by the axial magnetic field. Thus the metal in the crucible is melted directly by the axial LV arc. Substrates are placed on 12 rotating holders symmetrically distributed along the chamber axis at a radius of 26 cm. The crucible is surrounded by the auxiliary anode in the form of a concentric ring with an outer diameter of 26 cm. This anode is used mainly for etching and is electrically floating during deposition. Cylindrical dummy substrates were located symmetrically, with the inner radii aligned with the spectrometer aperture. During heating they are floating, and
A cylindrical molybdenum Langmuir probe Žlength 1.5 mm, diameter 1 mm. with an area A p is located horizontally in the middle of the vessel height, similarly to the spectrometer. The radial position of the probe tip varies from 10 to 35 cm from the vessel axis. If not mentioned specifically, the radial position of the probe tip was 20 cm from the center, like the orifice of the spectrometer and was aligned with the inner surface of the dummy substrates. Data acquisition is governed by PC controlled electronics that enable accurate measurements to be made. I᎐V plots are measured in the range of at least 50 V below the floating potential and up to 50 V above the plasma potential. The region of special interest, i.e. between Ufl and Upl , is plotted with at least 30 equidistant points necessary for further digital smoothing of the signal as described in w5x, using the theory of non-recursive filters. Typically, a convolution over nine points and quadratic least squares fitting is used. The plasma parameters are derived from the measured I᎐V characteristic assuming a Maxwellian distribution of electrons, w6x. Of special interest is the region with probe voltages between the floating potential Žthe potential when Ie s Ii . and the plasma potential, Ufl - V - Upl . In this case, the measured current I is the sum of the saturated ion Ž IiU . and electron current Ž Ie ., I s yIUi q Ie Ž V .. The latter can be expressed as Ie s IeU exp Ž e Ž Upl y V . rkTe .
Ž1.
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
210
Fig. 1. BAI 730 in three different operating modes: Ža. heating, Žb. etching and Žc. ion plating mode.
Table 1 Parameters and potentials measured for standard operating conditions in all three operating modes Mode
pAr Žmbar.
pN 2 Žmbar.
Iarc ŽA.
Icoil ŽA.
B Žmtorr.
Usubstr. ŽV.
Uarc ŽV.
Uanode ŽV.
Ucruc. ŽV.
Heating ŽH. Etching ŽE. Deposition ŽD.
2.5 E-03 1.5 E-03 1.5 E-02
r r 5.0 E-03
150 130 200
8r8 8r8 15r15
3.7 3.7 7.0
12᎐17 y130 y125
49 55 56
y0.6 8.3 50
y12.5 y20.6 53
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
211
Obviously the electron temperature Te , can be determined from the slope of the electron current vs. probe voltage curve, Te s Ž erk .Ž ⭸ln Ier⭸V .
y1
Ž2.
This definition was used in our work if not explicitly noted. Ion density is estimated from the ion saturation current IiU . It is obtained by linear fitting of the I᎐V plot below Ufl and extrapolating to the plasma potential Upl . ni s
IiU 0.6 eA p
(
mi kTe
Ž3.
Similarly, the electron density is calculated from the portion of the I᎐V plot where the probe voltage exceeds the plasma potential, V ) Upl . In this case the saturated electron current IeU is the electron current at the plasma potential Upl . ne s
IeU eA p
(
2 me kTe
The plasma potential can be set to the point of the kink in the I᎐V curve, or to the point of the highest maximum of the first derivative of the electron current, ⭸ln Ier⭸V. The latter point is a little bit lower than the former one w4,6x and was used in the present work if not stated otherwise. From the second derivative of electron current the electron energy distribution function F Ž E . can be calculated: F Ž E. s
4 3 e Ap
(
m e E ⭸ 2 Ie 2 ⭸V 2
Ž5.
By definition the plasma potential can be set to the point where the second derivative equals zero. The plasma in the system is a cold plasma ŽTi < Te ., and, therefore, the difference between the plasma and floating potentials is proportional to the electron temperature w4,6x. For a typical argon plasma Ž Ms 40., the proportionality factor is 5.1. Upl y Ufl f w 3.3q 0.5ln Ž M .x ⭈
kTe kT s 5.1⭈ e e e
Ž6.
Knowing Te and n i the ratio ␣ between fluxes of ions Ž ji . and neutrals Ž jn . can be calculated: ␣ s jirjn
Fig. 2. PPM421 energy and mass analyzer.
Ž4.
Ž7.
where the flux of neutrals is equal to pr '2 MkT , the flux of ions equals ji s n i¨s and ¨s stands for the velocity of the ions accelerated by the pre-heat, ¨s
s kTerm i , respectively. For all calculations we assume the absolute gas temperature to be 600 K for the heating and etching step while for deposition 700 K was chosen.
'
2.3. Energy and mass spectrometry Energy distributions of positive ions were measured using the energy and mass analyzer ŽBalzers PPM421, Fig. 2., located at one side of the chamber at the mid-height. The analyzer axis is horizontal, crossing the chamber axis. The additional turbomolecular pump evacuates the spectrometer differentially. The floating orifice of the spectrometer is located 1 cm behind an aperture with a 5 mm hole which also serves as a shutter. This aperture is positioned 20 cm from the vertical axis of the vessel, and is aligned with the cylindrical dummy substrates. Both the aperture and the substrates are connected to the same potential. The orifice of the energy and mass analyzer is electrically floating. The ions from the plasma are focused by the ion optics, filtered by the cylindrical mirror energy analyzer and then pass through the quadrupole mass analyzer. The ionization chamber is a source of ions when working in the mode for detection of neutral particles. The secondary electron multiplier counts the ions, generating an avalanche of electrons to improve sensitivity. It can give readings from 10y1 to 10 7 counts per second ŽCPS.. More details are given in the literature w3,7x. The analyzer can perform a mass scan at selected ion energy or an energy scan at selected mass numbers.
212
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
The measured energy spectrum is actually a spectrum of the stopping potential; and, therefore, for multiply charged ions the real energy is the product of the stopping potential V and the ion charge q, Es qV. It is worth noting that ion transmission through the ion optics depends on the trajectory, so the probability of detecting ions from different directions is not the same. Moreover, this probability also depends on the energy. Therefore, the measured ion energy spectra may differ from the real spectra in the plasma.
3. Results 3.1. Heating Let us start with the first step of the normal deposition process, with the heating mode. As shown in Fig. 1a, the system is configured in such a way that the electrons from the plasma strike the surface of the substrate, which is at the positive floating potential. Heating depends on the LV arc current, Iarc , and on the substrate potential Žtypically 10᎐15 V.. In this way, the temperature can be raised to some 300᎐400⬚C to out-gas contaminated surfaces prior to deposition. 3.1.1. Langmuir probe measurements During this step, the plasma is a classical cold one, Ti < Te . Moreover, the mean free path for ions in the plasma is much greater than the probe dimensions itself rp and the Debye length D ; 4 rp 4 D . Therefore, the probe operates as a classical Langmuir probe. Typical results of the plasma probe I᎐V measurements at this step are given in Fig. 3. This shows the family of I᎐V plots for different arc currents in the range from 50 to 200 A. The highest current is used for heating of heavy process loads. As one can see, the plots differ mostly in the ion and electron saturation currents, while the other features do not differ significantly: i.e. the slopes between the floating and plasma potential, as well as the floating and plasma potentials. Typical plasma parameters Ž n i , Te , Upl , Ufl . are collected in Table 2 for a set of arc currents measured with the probe positioned 20 cm from the center. As is to be expected, the plasma potential Upl , the plasma density n i , the electron temperature Te , as well as the ratio between the fluxes of ions and neutrals jirjn increase with increasing arc current. In the same way as the plasma potential increases, the voltage measured on the dummy substrates connected to the arc power supply ŽPS. ŽFig. 1a. also increases. The results in Table 2 for the standard arc current Ž150 A. represent the typical range of measured values for all substrates, while the others are the values measured on a single dummy substrate opposite the spectrometer, and 90⬚ to the probe axis. Obviously the plasma potential roughly
Fig. 3. A family of I᎐V plots for heating configuration with different arc currents. The argon pressure is 2 = 10y3 mbar and coil current is 8 A.
follows this highest positive potential in the chamber. The correctness of the plasma probe measurements is verified by the fact that the proportionality factor which relates Te , Upl and Ufl in Eq. Ž6., calculated for the above plots, is 3.93" 0.30, reasonably close to the theoretical value of 5.1 for a cold Ar plasma. For lower arc currents this value is even closer to the theoretical one and vice versa. The plasma density n e determined from the electron part of the Langmuir probe characteristic is typically a factor two to four lower than the plasma density n i determined from the ion part of the Langmuir probe characteristic. This can be attributed to the magnetic field, which reduces electron saturation current confining them to magnetic field lines w6x. However, current can be also overestimated. Reasons are: secondary electron emission as a result of high probe temperature Ž) 700 K. and ion bombardment as well as to the fact that Eq. Ž3. is only the first approximation of the complex problem of ion transport. Parameters of the plasma vary along the radius of the chamber. In the middle of the chamber height, the radial profile of the main plasma parameters is as shown in Fig. 4a,b. While the plasma potential is almost constant for distances over 15 cm from the center, the density of ions and electrons, and in the same way the ratio between the ion flux and neutral flux, strongly depends on the radius. The electron temperature also shows a radial dependence up to the limits of the dummy substrates Ž r s 20 cm., and becomes constant for radii above 20 cm. 3.1.2. Energy spectra Energy spectra measured during the heating step reveal, besides the singly and doubly charged Ar ions, ŽArq, mrqs 40 and Ar 2q, mrqs 20., a high concen-
Machine parameters
Plasma parameters
Mode
Iarc ŽA.
Uarc ŽV.
H H H H
50 100 150 200
42 48 49 54
E
130
55
D D D D D D
50 100 150 175 200 220
30 41 48 51 54 56
Usub 噛1 ŽV.
DRŽ I .rDR Ž200 A. Ž%.
Ufl ŽV.
Upl ŽV.
Te ŽeV.
ni Žm-3.
ne Žm-3.
jirjn Ž%.
ŽUpl y Ufl . ŽkTe re.
r r r r
7.8 8.4 8.6 9.0
12.0 13.7 14.3 14.9
0.99 1.33 1.48 1.50
6.6 E q 16 1.2 E q 17 2.0 E q 17 2.8 E q 17
2.7 E q 16 4.2 E q 16 5.3 E q 16 8.4 E q 16
2 4 6 9
4.25 4.00 3.85 3.95
y130
r
8.7
13.8
1.36
1.7 E q 17
5.1 E q 16
4
3.72
y125 y125 y125 y125 y125 y125
0 0 8 56 100 139
29.1 38.4 42.8 34.3 23.6 16.1
32.2 41.8 47.1 50.4 52.4 53.3
0.62 0.70 0.85 3.25 6.01 7.50
9.7 E q 16 2.5 E q 17 3.0 E q 17 3.2 E q 17 4.1 E q 17 4.6 E q 17
8.9 E q 16 1.8 E q 17 3.1 E q 17 2.5 E q 17 3.9 E q 17 2.9 E q 17
1 6 8 14 25 33
4.98 4.82 5.14 4.95 4.80 4.96
13.2 13.6 12᎐17 14.6
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
Table 2 Summary of plasma parameters derived from I᎐V measurements 20 cm from the center axis for three operating modes. Operating conditions, except the arc current Ia are the same as in Table 1
213
214
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
tration of water ŽH 2 Oq, mrqs 18., oxygen ŽOq, mrq . s 16. and nitrogen ŽNq, mrqs 14, Nq 2 , mrqs 28 ions. In Fig. 5a,b the energy spectra of the main species, singly Ža. and doubly Žb. charged Ar ions are shown for the same set of arc currents Ž50᎐200 A. as in Fig. 3 Žplasma probe characteristics .. The energy spectra of Arq ions are affected mainly by the increased second peak at approximately 35 eV. While this peak is not present for low arc current Ž50 A., it is very pronounced at arc currents higher than 100 A. The height of the main peak of the energy distribution does not change with the arc current. The effect of increase in arc current, i.e. higher ionization, is to shift the peak position to values of potentials similar to those measured at the substrate holders. In Table 2 one can see that these potentials increase slightly with arc current from 13 to approximately 15 V.
Fig. 5. Energy spectra for heating under conditions given in Table 1: Ža. singly charged Ar Ž 40Arq .; Žb. doubly charged Ar Ž 40Ar 2q . ions.
The reason for the second peak at approximately 35 eV is not yet clearly understood. It could be related to some kind of discrimination in the spectrometer, or, more probably, it could represent the high energy ions coming from the central arc column, where the plasma potential is probably higher, Ž Varc s 49 V, Vcruc s y12.5 V, Varc-ground s 37.5 V., than outside the column. The energy distributions for doubly charged Ar ions in Fig. 5b do not show this second peak at 35 V up to the highest arc current. However, the height of the peak increases with arc current. The distribution is also narrower than the distribution of singly charged Ar ions under the same conditions, but the peak is centered around the same energy. During heating under the standard conditions for 1.5 h, the intensity of water, oxygen and nitrogen ions exponentially decreases by approximately an order of magnitude over a typical heating time of 45 min. This is related to out-gassing at the elevated temperature, but it is not sufficient for good cleaning of the surface. Fig. 4. Radial profile of the plasma parameters during the heating step. Ža. Plasma ŽUpl . and floating ŽUfl . potentials and electron temperature Te ; Žb. ion Ž n i ., electron Ž n e . density and the ratio between ion and neutral fluxes ji r jn .
3.2. Etching During the etching step the BAI730 M system is
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
215
configured in such a way that the Ar ions are accelerated towards the biased Žtypically Usub s y130 V. substrates Žsee Fig. 1b.. These accelerated ions clean the surface by sputtering. This effect depends on the flux of impinging ions and their energy wA ŽUpl y Usub .x. 3.2.1. Langmuir probe measurements The I᎐V measurements of the Langmuir probe reveal quite similar results to those in the heating mode. Plasma parameters are summarized in Table 2 for standard etching with Iarc s 130 A. Plasma parameters are somewhere between the results obtained during heating with arc currents of 100 and 150 A in spite of the differences in configuration of the electrodes. Upl is again close to the highest positive potential in the chamber. In this case, however, it is the auxiliary anode, which is connected directly to the arc PS and to ground via a resistor on the same way as substrates in the heating mode ŽFig. 1b.. They are now at fixed negative potential Usub but float at 12 to 17 V during heating. The radial profile of the plasma measured at midheight of the chamber is shown in Fig. 6a,b. The plasma potential is almost constant for distances over 15 cm from the center. The electron temperature increases toward the center, as well as the ratio jirjn between the ion and neutral fluxes reaching the surface of the probe, as well as the densities of ions and electrons. However, generally there is no substantial difference in the plasma radial distribution from the one measured during heating. The main observable difference to the heating mode is that the plasma potential is the highest at a radius of approximately 15 cm, roughly corresponding to the anode radius. During heating, the maximum of plasma potential is found close to the radius of the substrate holders, which function as the anode in that mode. 3.2.2. Energy spectra Energy spectra of the singly and doubly charged species in this mode are not shown. They are very similar to those measured during the heating mode. The main differences in energy spectra are the absence of the second peak located at approximately 35 eV for singly charged particles and a much narrower energy distribution. There is no measured signal beyond 35 eV for etching Žcomparing with 60 eV for heating. at the highest Iarc s 200 A. Because of the extreme conditions in the center of the chamber Žvery high plasma density resulting in probe melting. we do not have direct evidence for the plasma potential there. However, from measurements of the floating potential of the crucible it seems that the plasma potential can be negative during both heating and etching, some 20 V more negative during etching. This high potential barrier could prevent the
Fig. 6. Radial profile of plasma parameters during the etching step. Ža. plasma ŽUpl . and floating ŽUfl . potentials and electron temperature Te ; Žb. ion Ž n i ., electron Ž n e . density and the ratio between ion and neutral fluxes ji r jn .
high energy positive ions coming from the chamber axis from being detected at the mass spectrometer position. The position of the peak does not change significantly with arc current and is close to the measured potential on the auxiliary anode ŽUan s 7.5 V at Iarc s 50 A and 8.6 V at 200 A.. However, the position of both peaks can be influenced by the magnitude and shape of the magnetic field. Fig. 7a,b shows the energy spectra of Arq ions during etching. The curves differ in respect to the magnetic field, produced by the coil currents I bottom and Itop in the bottom and top coil, respectively. The central curve shows the result for the standard conditions Ž I bottom s Itop s 8 A.. The other curves were measured for different combinations of coil currents. Fig. 7a was measured for various magnetic field
216
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
plasma potential increases, reflected in the peak position of the measured curve. In energy spectra measured during the etching step there are also other ions, such as atomic ŽOq, mrqs . 16. and molecular oxygen ŽOq 2 , mrqs 32 , and water ŽH 2 Oq, mrqs 18.. While the intensity of molecular oxygen Oq 2 decreases at the same rate as during heating Žout-gassing at elevated temperatures., the intensity of atomic oxygen and water drops very quickly after the system is switched from heating to the etching. At the same time the intensity of iron ŽFe, mrqs 56. increases, indicating the beginning of surface etching via the sputtering process. The intensity of iron sputtered from the surface is stabilized after approximately 15 min Žalso the prescribed etching time.. 3.3. Deposition
Fig. 7. Energy spectra of Arq ions for etching with different magnetic fields Žcoil currents .. Other conditions are the same as in Table 1: Ža. symmetric magnetic field, I bottom s Itop ; Žb. asymmetric magnetic field with I bottom / Itop .
strengths, while I bottom s Itop. The measurements show that the maximum of the curve, corresponding to the plasma potential, shifts towards higher energy if the magnetic field increases. Also the width of the curve increases. This effect is related to the magnetic confinement of the plasma. This is confirmed even better by the measurements shown in Fig. 7b, measured for various combinations of coil currents I bottom and Itop . The total curve width can be quite small if the current I bottom is small Ž1 A. relative to the top coil current Ž Itop s 20 A.. This corresponds to a small positive anode voltage as shown by the vertical lines ŽUa s 1.6 V.. In this case the magnetic field lines are divergent from top to bottom. Thus the plasma column is guided from the top filament source to the ring anode at the chamber bottom. The other case is when the current I bottom is large Ž20 A. relative to the top coil current Ž Itop s 1 A.. Then the plasma has to cross the magnetic field lines, divergent from bottom to top. Therefore, the anode voltage increases ŽUa s 31.1 V. and similarly increases the
3.3.1. Langmuir probe measurements I᎐V characteristics measured during the deposition of TiN ŽFig. 8. with various arc currents Žother settings the same as in Table 1. show a strong influence of Iarc on the characteristics, and therefore the plasma parameters are very sensitive to variations in Iarc . Especially pronounced is the shift of Ufl , which changes from 42.8 V at Iarc s 150 A to 16.1 V at 220 A. However, Upl shifts only slightly from 47.1 to 53.3 V at the same arc currents. This shift would result in a huge increase of Te , from 0.85 to 7.9 eV, if the distribution were Maxwellian. For 50 - Iarc - 150 A, the situation is less clear Žsee Table 2.. While Upl still decreases with decreasing Iarc , as above 150 A, Ufl decreases, but the difference ŽUpl y Ufl ., which is proportional to Te , does not change significantly ŽTe changes only from 0.85 to 0.6 eV.. The radial distribution of plasma parameters under the standard deposition conditions is shown in Fig. 9.
Fig. 8. A family of I᎐V plots for deposition of TiN with different arc currents. Other parameters Ž pAr , p N 2 , Ic . were the same as in Table 1 for the standard process.
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
217
jirje increase rapidly for arc currents above 150 A, when the melting of Ti visually starts. However, the plasma densities during heating and etching increases gradually with increasing arc current. A very important parameter is the electron energy distribution function F Ž E .. If the electrons are in thermal equilibrium, the distribution is Maxwellian, normally used to describe cold plasma: F Ž E . A 'E exp y
ž
E kTe
/
Ž8.
If the electrons are not in thermal equilibrium, the distribution can be the Druyvesteyn one w8x F Ž E . A 'E exp y0.24
Fig. 9. Radial profile of plasma parameters during TiN deposition. Ža. plasma ŽUpl . and floating ŽUfl . potentials and electron temperature Te , Žb. ion Ž n i ., electron Ž n e . density and the ratio between ion and neutral fluxes ji r jn .
Plasma potential Upl is constant over all measured distances Ž50᎐53 V., even behind the dummy cylinders, shown as gray rectangles in the figure. This value is close to the measured potential on the crucible ŽTable 1.. The electron temperature is also almost constant, with a small peak at the radii of the substrates. However, ion and electron densities Ž n i , n e ., and the ratio between ion and neutral fluxes Ž jirjn . rapidly decrease from the inside Ž r s 12.5 cm. towards the radius of the substrates Ž25 cm.. Beyond this radius the densities drop very rapidly. The ratio between ion and neutral fluxes during TiN deposition is rather high. It is higher than 0.25 at the inner radius of the substrates Ž r s 20 cm.. For comparison, the ratio measured during the heating with the same current and total pressure is only 0.09. From Table 2 one can see that the effects of a plasma characterized by parameters Te , Ni , Ne and
E kTe
ž /
2
Ž9.
In Fig. 10 we compare electron energy distributions for standard deposition conditions and for heating with the same magnetic field as used for deposition, B s 7 mtorr. Both measured distributions are very similar up to 12 eV and can be fitted better with the Druyvesteyn distribution function given by Eq. Ž9. than by the Maxwellian one wEq. Ž8.x. The electron temperature Te is estimated as 2.0 eV for heating and 2.5 eV for deposition. The latter is much smaller than the electron temperature estimated from Eq. Ž6. ŽTe s 5.8 eV. or obtained from the best fit of the I᎐V plot ŽTe s 7.5 eV. in Fig. 8. It appears that during TiN deposition we are dealing with electrons coming from different plasma potentials. ŽThis effect is not seen in the electron distributions for heating or etching.. In fact, measurements of Ufl with a very fast sampling rate Ž318 s. did show large fluctua-
Fig. 10. Electron energy distribution function for standard deposition of TiN and for heating with the same magnetic field as during deposition Ž Ic s 15r15 A.. Obviously the electron energy distribution during the deposition is not Maxwellian Ždashed line..
218
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
tions up to 15 V. This makes the evaluation of the probe characteristic difficult, if not impossible. In measurement such as those in Fig. 8 the plasma potential most of the time remains approximately 50 V but for short periods it falls by 5᎐15 V, collecting mostly the electrons instead of ions. This can explain the shape of F Ž E . for energies above 12 eV as well as the large difference between the apparent Upl and Ufl and increasing Te . 3.3.2. Energy spectra Energy spectra of Ar and Ti ions for different Iarc are shown in Fig. 11a᎐d. In Fig. 11a,b spectra for doubly charged ions Ž 40Ar 2q, mrqs 20, 48 Ti 2q, mrqs 24. are represented, while in Fig. 11c,d spectra for singly charged ions Ž 40Arq, mrqs 40, 48 Tiq, mrqs 48. are shown. The main difference in Ar and Ti spectra is that the intensity of the doubly charged ŽTi 2q . ion is much stronger than that of the singly charged ŽTiq. ion. This is the opposite of that for argon and nitrogen ions. In ref. w9x it was shown that the ion with mrqs 16, that is triply charged 48 Ti 3q ion, Žnot oxygen., was detected with relatively strong intensity. In Fig. 11 the peaks have almost identical shape for all measured ions and arc currents. For the standard Iarc , the peak is centered at 57 " 0.5 eV. This is close to the plasma potential measured by the plasma probe and to the potential of the crucible. A very important fact is that the intensity of Ti ions ŽTiq, Ti 2q, Ti 3q . increases with Iarc , while that of Ar and N decreases. These effects are partially due to the
increased evaporation rate and partially due to the increased ionization as it is discussed in Section 4. There is also a difference in the shape of the distribution, since the Tiq ions exhibit almost no low energy tail Žbelow 40 eV. up to the Iarc over 200 A. This tail is relatively strong for other ions shown, as well as for the nitrogen. This low energy tail seems to correspond well with the temporal distribution of the plasma potential. This ion signal seems originate from the plasma potential in the short time when it is lower for some 5᎐15 V.
4. Discussion On the basis of the I᎐V measurements in the previous section it was shown how the plasma behaves during heating Žalmost the same as during etching., and during the deposition of TiN under different arc currents. With increasing current all plasma parameters except Te and Ufl more or less gradually increase in both modes of operation ŽTable 2.. These relationships are shown in Fig. 12. During deposition, the electron temperature sharply increases after the onset of Ti evaporation. The difference ŽUpl y Ufl . increases in a similar way. The explanation seems to be the onset of instability in the plasma potential. This is reflected also in the shape of the ion energy distribution, see Fig. 11. A difficult question to answer is how such plasma instability with large amplitude Ž; 15 V. starts and is sustained during Ti evaporation. The interaction
Fig. 11. Energy spectra of Ar Ža,c. and Ti Žb,d. ions for deposition of TiN with different Iarc .Ž150, 175, 210 A. Other conditions are the standard ones as given in Table 1: Ža. 40Arq2 Ž mrqs 20.; Žb. 48 Tiq2 Ž mrqs 24.; Žc. 40Arq Ž mrqs 40.; Žd. 48 Tiq Ž mrqs 48..
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
219
Consequently, the higher is the probability of Penning ionization according to the following relationship: ArU q Tiª Ar q Tiqq ey
Ž 10 .
Another process in the plasma is the electron-impact ionization of excited Ar atoms: U
eyq Ar ª Arqq 2 ey Eq y Impact y Ion
Fig. 12. Plasma parameters vs. Iarc for heating and TiN deposition. Conditions are the same as in Table 1: Ža. plasma potential Upl and floating potential Ufl . Žb. electron temperature Te ; ratio between ion and neutral fluxes ji r jn
between the electron beam in the magnetic field and the plasma containing Ti ions could be the cause. In the Langmuir probe measurements the electron temperature Te deduced from the fit of the I᎐V plots may be a little overestimated, since the ratio between ŽUpl y Ufl .rTe is even below 4 at higher currents for heating mode. According to theory it should be 5.1, and therefore either Upl is underestimated or Te is overestimated. The reason for underestimation of Upl could be small differences in potential of individual substrates during heating. The deposition rate Ž DR ., measured by the net consumption of nitrogen gas, correlate very well Ž r 2 ) 99%. with the arc current and with the electron temperature calculated from Eq. Ž6. for Iarc ) 150 A, as can be seen in Table 2, where the results for DR are normalized to the values for standard deposition of TiN. The higher is the arc current, the higher is the evaporation rate and the concentration of Ti atoms.
Ž 11.
Since the excitation energy of Ar, 11.5 eV, is large compared with the ionization energy of the Ti, 6.8 eV, the probability of Penning ionization of Ti should increase and for impact ionization of Ar should decrease as the number of Ti atoms increases with increasing arc current. Therefore, the source of additional electrons and ions increasing the ratio jirjn during TiN deposition ŽFig. 12c. could be Penning ionization. This fact is also confirmed by spectroscopic ŽPPM. measurements. Integration of the energy distribution in Fig. 11 shows that the intensity of the singly Ž mrqs 48. and to an even more pronounced degree of doubly Ž mrqs 24. and triply Ž mrqs 16. charged Ti ions follows this relationship Žincrease with Iarc .. On the other hand the intensity of singly Ž mrqs 40. and also doubly Ž mrqs 20. charged Ar ions decreases with increasing arc current during the deposition of TiN. The intensity of the ion flux relative to the flux of neutrals ŽFig. 12c. in deposition with Iarc over 150 A increases very strongly mainly due to the increase in Ti evaporation and its easy ionization to the doubly and singly charged states. Moreover Penning ionization may also leads to fragmentation of metal clusters. It was suggested in ref. w10x that metal evaporation at typical low pressure ion plating conditions predominantly occurs in the form of clusters with more than 10 atoms per unit charge. In our mass spectroscopic measurements up to 511 a.m.u. we never detected any particle with mrq larger than 50 Ž 50 Tiq .. If evaporated metal is really in the form of clusters, than Penning process can also cause their multiple ionization. From the facts that approximately 90% of ions striking the surface of substrates is Ti and that we never detected particles with mrq higher than 50 we may conclude that Penning ionization leads to the complete fragmentation of Žat least. clusters with less than 10 Ti atoms per unit charge. During the heating Žand etching. mode the situation is different. The main species in the vessel is Ar. Therefore the electron-impact ionization of Ar is the most important ionization mechanism. The intensity of singly and doubly charged ions increases gradually with Iarc in the range from 50 to 200 A, and there is no ‘anomaly’ with a decreasing floating potential as during the deposition. As the intensity of Ar ions detected by the spectrometer increases linearly with Iarc , in the same way the
220
M. Macek ˇ et al. r Surface and Coatings Technology 135 (2001) 208᎐220
ratio between the fluxes of Žmainly. Ar ions and neutrals also increases.
5. Conclusions On the basis of plasma probe measurements and of energy and mass spectroscopy we can conclude that the plasma during the heating and etching modes behaves like a classical cold plasma with electrons in thermal equilibrium with a mean energy below 1.5 eV for the standard operating conditions. The flux of ions is low compared with the flux of gas neutrals, only a few percent. The plasma potential is very uniform over the vessel radius. It corresponds to the highest potential in the chamber, i.e. the potential of the substrates during heating or the potential of the auxiliary anode during etching. During the deposition of TiN, as well as other metals like Cr, the plasma is very intense. The ratio between fluxes of ions and neutrals Žgas and metal. is high compared with heating and etching, jirjn s 25% at the standard conditions. Mass spectroscopy shows that the majority of the Ti ions are doubly charged ones. The apparent energy of electrons under the standard deposition conditions with a strong magnetic field, B s 7 mtorr is very high. Assuming thermal equilibrium and a quiescent plasma, the electron temperature Te should be approximately 6 eV. The energy distribution of the electrons is better described by the Druyvesteyn function than by the Maxwellian one. Moreover, a large plasma instability was observed, especially during TiN deposition. This affects the apparent Langmuir probe characteristics, as well as the ion energy distribution.
Acknowledgements This work was supported by the Ministry of the Science and Technology of the Republic of Slovenia and is also part of the bilateral scientific collaboration between the Czech Republic and the Republic of Slovenia. We also acknowledge the support of a NATO-Collaborative Research Grant, Project HTECH.CRG 970529, and the project ME110 of the Czech Ministry of Education. This work was also partly supported by the Grant Agency of the Czech Republic, Grant no. 106r96rK245. References w1x E. Bergman, Surf. Coat. Technol. 57 Ž1993. 133᎐137. w2x E. Mol, E. Bergman, Surf. Coat. Technol. 37 Ž1989. 453᎐509. w3x M. Nesladek, C. Quaeyhaegens, S. Wouters, L.M. Stals, E. ´ Bergmann, G. Rettinghaus, Surf. Coat. Technol. 68r69 Ž1994. 339. w4x A. Grill, Cold Plasma in Materials and Fabrication, IEEE Press, New York, 1994. w5x R.W. Hamming, Digital Filters, 3rd ed, Prentice-Hall, Engelwood Cliffs, NJ, 1989. w6x N. Herskowitz, Langmuir probe diagnostics, in: D.A. Glocker, S. Ismath Shah ŽEds.., Handbook of Thin Film Process Technology, Institute of Physics Publishing, Bristol, 1995, p. D3.0:1. w7x S. Wouters, S. Kadlec, M. Nesladek, C. Quaeyhaegens, L.M. ´ Stals, Surf. Coat. Technol. 76 Ž1995. 135. w8x M.J. Druyvesteyn, F.M. Penning, Rev. Mod. Phys. 12 Ž2. Ž1940. 88. w9x M. Macek, ˇ B. Navinsek, ˇ P. Panjan et al., Surf. Coat. Technol. 113 Ž1999. 149. w10x K.S. Fancey, A. Matthews, Appl. Phys. Lett. 55 Ž9. Ž1989. 834.