AFM Study of Hydrocarbon Thin Films

Transcription

1 WDS'05 Proceedings of Contributed Papers, Part II, , ISBN MATFYZPRESS AFM Study of Hydrocarbon Thin Films M. Valtr, I. Ohlídal Masaryk University in Brno, Faculty of Science, Brno, Czech Republic. P. Klapetek Czech Metrology Institute, Brno, Czech Republic. Abstract. In this paper atomic force microscope (AFM) study on hydrocarbon thin films was performed. Thin films were prepared using plasma enhanced vapour deposition (PECVD) in pulsed regime. The aim was to see whether the surface roughness of the films changes with the duty cycle. The AFM is a direct measuring technique for scanning surfaces of samples. From obtained images one can compute many quantities characterizing the surface. It is shown that the root mean square (RMS) value of the heights seems to be independent on duty cycle except for region from 10 to 15 %. Outside of this region the RMS values of the heights vary in the range from 6,3 to 14,8 nm, while for 10 % it is 96,3 nm and for 15 % it s 61,7 nm. Unfortunately also the deposition time, which corresponds to different thicknesses of the films, was varied. This could be the reason for many times higher roughness inside the region. Further study of this problem would be needed. Introduction Plasma polymerization is widely used technique having many applications [Janča and Sodomka, 1998; Denes and Manolache, 2004]. Pulsing the discharge is one possibility to control the chemistry during the deposition [Choukourov et al., 2003]. In this paper we concentrated on the study of the surface morfology of deposited thin films. Probably most powerfull technique to deal with this problem is the atomic force microscopy (AFM). It enables us to determine quantities like root mean square (RMS) value of the heights, autocorrelation length, power spectral density function (PSDF), etc. These quantities have then significantly influence on optical properties [Ohlídal and Vižďa, 1999]. AFM Statistical Characterization of Thin Films One of the most important parameters of the statistically random rough surfaces is RMS value of the heights σ. It is defined by the following equation [Klapetek, 2003] σ 2 = z 2 w(z)dz, (1) where z denotes value of the height of the irregularity and w(z) is distribution of the probability density of the heights. RMS value of the heights is only first moment describing statistical properties in individual points. In order to perform more complete statistics, higher order moments should emerge. Usually second order moments, which describe mutual interaction of two points on the surface are sufficient. One of these quantities is the two-dimensional power spectral density function. It can be written in terms of Fourier transform of the autocorrelation function [Klapetek, 2003] G(τ x, τ y ) W 2 (K x, K y ) = 1 G(τ x, τ y )e i(kxτx+kyτy) dτ x dτ y, (2) 4π where K x and K y are components of the wave vector of the harmonic component of a certain spatial frequency of the boundary roughness and τ = τx 2 + τy 2 is the distance between the points 391

2 z 1 (x 1, y 1 ) and z 2 (x 2, y 2 ). Within the AFM measurements we usually evaluate one-dimensional PSDF, which is given by the following equation W 1 (K x ) = W 2 (K x, K y )dk y. (3) If we suppose autocorrelation function to be Gaussian[Klapetek, 2003], we obtain for onedimensional PSDF W 1 (K x ) = σ2 T 2 π exp( K2 xt 2 /4), (4) where T is the autocorrelation lenght. It should be emphasized that above definition formulae are valid for surfaces made by continuous stationary isotropic and ergodic stochastic processes. Moreover we assume the mean value of the heights to be zero. An AFM scan consists of a dataset having N M data points. In order to calculate the above defined parameters, we have to use corresponding discrete formulae: σ 2 = 1 NM M N zij, 2 (5) j=0 i=0 The values of W 1 (K x ) can be calculated by the fast Fourier transform, i.e. W 1 (K x ) = 2π NMh N ˆP j (K x ) 2, (6) k=0 where ˆP j (K x ) is the Fourier coefficient of the jth row, i.e. ˆP j (K x ) = h 2π N z kj exp( ik x kh). (7) k=0 Function w(z) must be calculated using following formula: N (z, δz) w(z) = (8) NMδz where function N (z, δz) represents the number of the values of z ij lying within the interval z δz/2, z + δz/2. Experimental Thin films were prepared by plasma enhanced chemical vapor deposition (PECVD) technique in tubular type reactor. A schematic view of the apparatus is in the figure 1. The horizontally mounted SIMAX c glass tube was closed by two aluminium electrodes. Distance of the electrodes was 186 mm and inner diameter of the tube was 77 mm. The apparatus was pumped out by a rotary pump. Total pressure was measured using Leybold CERAVAC diaphragm gauge. Gas flows were controlled by Hastings flowmeters. Gases used for deposition were argon and acetylene. Flow rate of Ar and/or C 2 H 2 was kept constant at 4 sccm and/or 1 sccm. Corresponding total presure in the apparatus was about 39 Pa. Power was delivered via matching unit from Dressler CESAR 133 generator operating at MHz. Delivered peak power during deposition was 10 W. The depositions were performed in pulsed regime, where frequency of the cycles was always 1 Hz and duty cycle varied from 5 to 70 %. Silicon substrates were located on a glass holder in the middle of the reactor. The deposition time changed from 10 to 40 minutes. 392

3 Prior to every deposition surface treatment in Ar was performed. Cleaning runned for 5 minutes at 5 W in continuous wave regime. After that deposition was started immediately. AFM measurements were made with the help of Topometrix Accurex II L device in contact mode. Scanned area was µm Figure 1. Schematic view of the apparatus. 1 - power source, 2 - matching unit, 3 - samples, 4 - SIMAX tube, 5 - rotary pump, 6 - pressure gauge, 7 - glass holders, 8 - Al electrodes, 9 - flowmeters, 10 - control unit, 11 - Ar cylinder, 12 - acetylene cylinder. Results and Discussion The AFM measurements were processed using Gwyddion analysis tool (downloadable from It should be mentioned that the Gwyddion program uses in the fitting procedure well known Marquart-Leveberg algoritm. The results on RMS rougness of the heights σ and autocorrelation lenghts T are summarized in the table 1. The quantity σ AFM was computed directly from AFM data according to formula (5). The rest two quantities were acquired by fitting experimental data by the least squares method. Experimental PSDF function was calculated using formula (6) and the fitting function used was Gaussian PSDF function (4) (σ and T were free parameters). Best fit values are then in the table 1 denoted as σ PSDF and T PSDF. If we compare σ AFM and σ PSDF we could say that they are in good agreement. Table 1. The results on RMS roughness of the heights σ and autocorrelation lenghts T. duty cycle [%] σ AFM [nm] σ PSDF [nm] T PSDF [nm] 5 13,9 13,2±0,2 202±6 7 14,8 13,8±0,2 75± ,3 90,6±0,2 262± ,7 56,9±0,1 304± ,8 10,39±0,09 132± ,3 14,6±0,1 124±3 50 7,0 6,2±0,2 125± ,3 4,97±0,04 103±2 393

4 σ AFM σ PSDF 60 σ [nm] duty cycle [%] Figure 2. Comparison of RMS rougness of the heights computed directly from AFM measurement - σ AFM and values obtained by the fitting procedure - σ PSDF. In the figure 2 is the graphical representation of results of RMS roughness of the heights written in table 1. At first sight we can see that the roughness shows constant dependency except for duty cycles 10 and 15 %. This discrepancy could have two explanations. The roughness could be strongly dependent on the duty cycle within this region or there is some correlation between roughness and film thickness. From ellipsometry, the estimates of film thicknesses were done and it was seen that this two films were about 350 µm thick, while the thickness of the rest of the films was about 150 µm. It seems that the film thickness has some influence on the rougness, but to be really sure, more depositions should have been done. In order to do more detailed study of the film surface, another quantities have been calculated. Namely it was the distribution of heights and PSDF function. In the figures 3 and 4 are plots of these functions for single sample, where duty cycle was 15 %. From the figure 3 one can see that the mean height of the peaks on the surface is about 219 nm. It can be also easily seen that the w(z) function is asymetrical towards to higher values, which means that there are some very high peaks present. Comparison of a measured PSDF with its best fit is ploted in the figure 4. Furthermore on picture 5 is a 3D image of the sample surface. Conclusion The AFM technique is very powerfull tool to characterize surfaces of materials. It enables us to estabilish quantities like RMS roughness of the heights, autocorrelation length, PSDF, etc. These parameters strongly influence e.g. optical properties. Hydrocarbon thin films were examined with this measuring technique and their surface parameters were evaluated. It has been shown that the RMS roughness of the heights seems to be almost independent on duty cycle. Values of σ vary in the range from 6,3 to 14,8 nm. However in the region from 10 to 15 % there is a discrepancy. Further depositions would be needed to perform more detailed examination of the influence of the film thickness on the roughness. 394

5 VALTR ET AL.: AFM STUDY OF HYDROCARBON THIN FILMS w(z) z [nm] Figure 3. Plot of the distribution of probability density of heights. Note that the mean value is 219 nm. PSDF [x10 5 nm 3 ] AFM data fit K x [nm -1 ] Figure 4. Plot of the power spectral density function computed from AFM data together with the best fit with Gaussian PSDF. 395

6 Figure 5. 3D image of an sample surface. Hydrocarbon thin film grew for 25 minutes with these deposition conditions: P = 10 W, Q Ar = 4 sccm, Q C2 H 2 = 1 sccm, f = 1 Hz, duty cycle was 15 %. Note that σ AFM = 61, 7 nm. The present work was supported by the Ministry of Education under con- Acknowledgements. tract FRVŠ 2005/2181. References Choukourov, A., Biederman, H., Slavinska, D., Trchova, M., Hollander, A., The influence of pulse parameters on film composition during pulsed plasma polymerization of diaminocyclohexane, Surface and Coating Technology, , p , 2003 Denes, F., S., Manolache, S., Macromolecular plasma-chemistry: an emerging field of polymer science, Prog. Polym. Sci., 29, , 2004 Janča, J., Sodomka, L., Plasma-polymerized organosiloxane thin films as selective gas sensors, Surface and Coatings Technology, 98, , 1998 Klapetek, P., Characterization of randomly rough surfaces in nanometric scale using methods of modern metrology, PhD Thesis, Masaryk University in Brno, Faculty of Science, Brno, Czech Republic, 2003 Ohlídal, I., Vižďa, F., Optical quantities of multilayer systems with correlated randomly rough boundaries, Journal of Modern Physics, 46, 14, p ,