Optical and Langmuir probe diagnostics of an Argon plasma in an RF reactive magnetron sputtering system Shinsuke Mori1, Houssam Fakhouri2, Jerome Pulpytel2, Farzaneh Arefi-Khonsari2 1. Department of Chemical Engineering, Tokyo Institute of Technology, 152-8552 Tokyo, Japan 2. LGPPTS, Université Pierre et Marie Curie, ENSCP, 75231 Paris, France Abstract: Plasma diagnostics of an argon RF magnetron reactive sputtering system, used for the deposition of titanium oxide or titanium oxide doped with nitrogen were performed using optical emission spectroscopy and Langmuir probe systems. In the optical method, we have used a modified corona model to estimate the electron temperature, which has been recently developed by Boffard et al. [1] The influence of electron energy distribution function (EEDF) on the estimation of electron temperature by the modified corona model is addressed. Electron temperature estimated by the optical method assuming MaxwellianEEDF is much lower than that measured by the Langmuir probe system in the experimental conditions tested. However, when non-Maxwellian EEDF calculated by Boltzmann solver including electron-electron collisions was applied to the modified corona model, the discrepancy between the optical method and the Langmuir probe system was improved and the estimated electron temperatures are almost identical with those of the Langmuir probe systems. Keywords: Reactive Magnetron Sputtering, Langmuir Probe, Optical Emission Spectroscopy, Corona Model, Argon plasma 1. Introduction 2. Experimental Optical emission spectroscopy (OES) has been most widely employed as a non-intrusive diagnostic tool to characterize plasma alternative to Langmuir probe system. In order to determine the electron temperature by the OES method, one usually applies the so-called line-ratio technique. The determination of electron temperature in low-temperature plasma by the OES method, however, is still unreliable and needs to be developed. In this study, we have used a modified corona model to estimate the electron temperature, which has been recently developed by Boffard et al. and adopted to an argon ICP system showing good agreement with Langmuir probe results [1]. This method was applied to a magnetron discharge system in this study which allowed to examine the agreement between the OES and Langmuir probe results under a variety of plasma conditions. We used the SPT120-H model which is a Sputter Deposition System (PLASMIONIQUE), comprised of a MAGNION-02 magnetron sputtering gun, with an appropriate pumping system. The RF power generator delivered a maximum power of 300 W at a frequency of 13.56 MHz. A metallic Ti target (50 mm diameter) having a purity of 99.95%, was sputtered in a reactive gas atmosphere containing Ar, O2 and N2. Different nitrogen to oxygen ratios were used. The initial base pressure was between 3 to 7x10-6 Torr, while the total working pressure was between 3 to 20 mTorr. The Langmuir probe system is an RF compensated Smartprobe from Scientific Systems. A detailed description of the Langmuir probe system is explained elsewhere [2]. The target to Langmuir probe distance was about 10 mm. The radial position of the probe tip is 15 mm away from the center of the target. 3. Langmuir probe results Figures 1 and 2 show the experimental results of the Langmuir probe measurements. Electron density increases and electron temperature decreases with an increase in a pressure, monotonously. When the N2 is added into the feed gas, electron temperature decreases at higher pressures, although the influence of O2 mixture on the electron temperature is negligible. On the other hand, the influence of O2 mixture on the electron density is more significant as compared with N2 admixtures. In general, the poisoning effect on the sputtering target is more p prominent in the case of O2 admixtures and the electron energy loss due to the inelastic collisions is more significant in the case of N2 admixtures because Electron density (1010 cm-3) 5 4. OES model In order to estimate the electron temperature by the OES data, we utilized modified corona model developed by Boffard et al [1]. Here, we explain the model briefly. In the simplest corona model, only two processes are considered: the electron-impact excitation from the ground-state species and the spontaneous radiations from the excited species. Then, the rate balance equation and emission intensity Iij become as follows: ne n0 k0 i = ni Σj Aij 4 (1) Æ Iij = hvij Aij ni = hvij Γij ne [n0 k0 i] Æ 3 2 ● : Pure-Ar: 20sccm △ : O2/Ar = 2sccm/20sccm ▽ : N2/Ar = 2sccm/20sccm 1 0 0 5 10 15 20 Pressure (mTorr) Figure 1. Electron number density measured by Langmuir probe at a radial position of R = 15mm in a 200 W Ar plasma. 5 Electron temperature (eV) because of the resonant vibrational excitations [3]. Therefore, it would be reasonable to consider that the reduction in the electron density and electron temperature may be mostly related to the poisoning effect and the inelastic collisions, respectively. 4 2 ● : Pure-Ar: 20sccm △ : O2/Ar = 2sccm/20sccm ▽ : N2/Ar = 2sccm/20sccm 5 10 15 Æ (1) If the apparent excitation cross sections are used instead of direct excitation cross sections, the radiative cascade from the higher-lying levels can be considered. (3) Electron impact excitations from the long-lived lower levels have to be considered. 0 0 where ne is electron density, n0 is the number density of the atoms in the ground level, ni is the number density of atoms in the level-i, hvij is the energy gap between levels i and j, k0 i is the electron impact rate constants out of the ground level-0, Aij is the Einstein coefficient for the i Æ j transition, Γij (= Aij / Σj Aij ) is the branching fraction. The corona model can be more generalized after the proper modifications [1]: (2) In order to consider the radiation trapping due to the reabsorption of line radiation by the lower level, the effective branching fraction Γijeff (= Rij Γij , where Rij is a reabsorption factor) should be utilized instead of Γij. 3 1 (2) 20 Pressure (mTorr) Figure 2. Electron temperature measured by Langmuir probe at a radial position of R = 15mm in a 200 W Ar plasma. After having considered the modifications listed above (1-3), eq. (2) can be generalized as follows: Iij = hvij Γijeff ne [Σl nl kl app i ] Æ (3) Æ = I pq / v pq ∑ nk ∑ nk Γijeff l eff Γpq l app l l →i app l l→ p (4) Two lines in Ar(3p54p Æ 3p54s) transition array (2px Æ 1sy in paschen’s notation) are chosen for the line intensity ratio and Γijeff can be simplified as unity for these two lines [4]: 750.39 nm (2p1 Æ 1s2) and 811.53 nm (2p9 Æ 1s5). In this study, the model assumes that the 2p1 and 2p9 levels were populated by electron-impact excitation from the ground state and from the 1s3 and 1s5 metastable levels. The density of the Ar(3p54s) levels (1s3 and 1s5 levels) were estimated considering the reabsorption of 3p54p to 3p54s emissions by the Ar(3p54s) levels using the least square method developed in the paper reported by Schulze et al. [5] When the common upper level is chosen, the line intensity ratio is simplified and determined by the escape factors γ. I ij / vij I ik / vik = Γijeff Γikeff = γ ij Aij γ ik Aik 11.83 eV 11.72 eV 11.62 eV 11.55 eV 0 5 10 15 20 Pressure (mTorr) Figure 4. Number densities of Ar(3p54s) levels in a 200 W Ar plasma derived from the OES branching fraction method of Schulze et al. [5]. Key: 1s2 level (▽), 1s3 level (X), 1s4 level (▲), 1s5 level (●). 100 10-1 3.0 eV 4.0 eV 5.0 eV -2 10 10-3 2.0 eV 10-4 10-5 2p1(J=0) 2p2(J=1) 10-6 2p3(J=2) 2p4(J=1) 0 5 10 15 20 25 30 Electron energy (eV) 811.53 nm 852.14 nm 794.82 nm 840.82 nm 738.40 nm 706.72 nm 826.45 nm 727.29 nm 696.54 nm 2p9(J=3) 750.39 nm 109 (5) Because the escape factor is a function of the lower level density, we can estimate the latter by comparing the experimental line intensity ratio with the calculated one using eq (5). Schulze et al. estimated the density of the lower levels, Ar(3p54s), using 5 intensity ratios with the least square method. Figure 3 shows the Ar lines utilized in this study. 13.48 eV 13.33 eV 13.30 eV 13.28 eV 13.08 eV 1010 EEDF (eV-3/2) I ij / vij Our calculation results are shown in Figure 4 and seem to be reasonable as compared with other experimental measurements [4,5]: the number density of Ar(3p54s) levels are around 109 to 1010 1/cm3 and the density of 1s5 metastable is much higher than that of 1s3 level and the densities of 1s2 and 1s4 are almost identical. Number density (1/cm3) where nl is the number density of atoms in the initial level-l, kl iapp is the apparent electron impact rate constants out of the initial level-l. Then the line intensity ratio becomes as follows: 1s2(J=1) 1s3(J=0) 1s4(J=1) 1s5(J=2) Figure 3. Partial energy level diagram for Ar(3p54pÆ 3p54s). Line intensity ratios of 696.54nm: 826.45nm, 727.29nm: 826.45nm, 706.72nm: 840.82nm, 738.40nm: 840.82nm, and 794.82nm: 852.14nm are used for the estimation of Ar(3p54s) levels and 750.39 nm: 811.53 nm is used for the corona model. Figure 5. EEDF used in the calculation of rate constants for the electron impact excitation of Ar. Key: Maxwellian-EEDF ( … ), EEDF calculated including electron-electron collisions (―), EEDF calculated excluding electron-electron collisions (- - -). Finally, we calculated the line intensity ratio using eq. (4) and from the best fit with the experimental line intensity ratio, the electron temperature has been determined. In order to evaluate the influence of electron energy distribution function (EEDF) on the estimation of electron temperature by the corona model, we utilized three different EEDFs for the calculation of rate constants kl iapp in eq. (4): Maxwellian-EEDF, EEDFcalculated by the Boltzmann solver (Bolsig+ [6,7]) excluding the electron-electron collisions, and EEDF-calculated including the electron-electron collisions. As shown in the Figure 5, when the electron-electron collisions are considered in the calculation of EEDF, the high energy electrons are increased and the shape of EEDF becomes similar to the Maxwellian-EEDF. Æ 5. Comparison of Langmuir probe results and those of OES method Electron temperatures estimated by the modified corona model are shown in Figure 6. For the comparison, Langmuir probe results are also plotted in this figure. Electron temperature estimated by the modified corona model assuming Maxwellian-EEDF is much lower than that measured by the Langmuir probe system in the experimental conditions tested. On the other hand, the modified corona model using the non-Maxwellian EEDF calculated by the Boltzmann solver without electron-electron collisions gives higher electron temperatures. However, when non-Maxwellian EEDF calculated by Boltzmann solver including electron-electron collisions was applied to the modified corona model, the estimated electron temperatures are almost identical with those of the Langmuir probe systems. 6. Conclusions Plasma diagnostics of an argon RF magnetron reactive sputtering system, used for the deposition of titanium oxide or titanium oxide doped with nitrogen were performed using optical emission spectroscopy and Langmuir probe systems. Electron temperatures are 2.2-4.5 eV and electron densities are 1.6-4.5 x 1010 cm-3 by the Lanmuir probe measurements. Electron temperature estimated by the optical method assuming Maxwellian-EEDF is much lower than that measured by the Langmuir probe system in the experimental conditions tested. On the other hand, the modified corona model using the nonMaxwellian EEDF calculated by the Boltzmann solver without electron-electron collisions gives higher electron temperatures. However, when nonMaxwellian EEDF calculated by Boltzmann solver including electron-electron collisions was applied to the modified corona model, the estimated electron temperatures are almost identical with those of the Langmuir probe systems. 5 Electron temperature (eV) References 4 [1] J. B. Boffard et al, Plasma Sources Sci. Technol. 19, 065001 (2010). 3 [2] J. Pulpytel, F. Arefi-Khonsari and W. Morscheidt, J. Phys. D: Appl. Phys. 38, 1390-1395 (2005). 2 [3] J. Pulpytel, W. Morscheidt and F. Arefi-Khonsari, J. Appl. Phys. 101, 073308 (2007). ● : Langmuir probe (R = 15 mm) △ : OES (calculated-EEDF w/o e-e collisions) ▽ : OES (calculated-EEDF with e-e collisions) × : OES (Maxwellian-EEDF) 1 0 0 5 10 15 20 Pressure (mTorr) Figure 6. Comparison of electron temperatures estimated by different methods in a 200 W Ar plasma. [4] J. B. Boffard et al, Plasma Sources Sci. Technol. 18, 035017 (2009). [5] M. Schulze et al, J. Phys. D: Appl. Phys. 41, 065206 (2008). [6] http://www.bolsig.laplace.univ-tlse.fr/ [7] G.J.M. Hagelaar and L.C. Pitchford, Plasma Sources Sci. Technol. 14 722-733 (2005).
© Copyright 2026 Paperzz