An Investigation of Metal Thin Films Using X-ray Reflectivity and Atomic Force Microscopy D.M.SOLINA, R.W.CHEARY, Microstructural F.A.LUPSCHA, and P.D.SWIFT Analysis Unit, Department of Applied Physics University of Technology, Sydney P.O.Box 123, Broadway, 2007. Australia. ABSTRACT Magnetron sputtered thin films of tungsten, 100 silicon were investigated as part of a systematic of the interfaces. titanium and molybdenum using X-ray Reflectivity results are presented atomic number, and gas pressure during fabrication. the surface film roughness obtained Microscopy are noted. Roughness on and Atomic Force Microscopy study on the effects of deposition Preliminary deposited conditions on the roughness on the effects of film thickness, Significant using X-ray Reflectivity differences between and Atomic Force was found to increase with increasing gas pressure, and with lower atomic number materials. Little effect on roughness was observed with thickness. INTRODUCTION Multilayers, can be used consisting of alternating layers of material such as Tungsten and Carbon, as mirrors in x-ray astronomy, soft x-ray lithography, and x-ray microscopy[‘l. In these devices one material acts as the diffracting plane for the x-rays and the other a spacer. The requirements of a high reflectivity mirror are that the step in the refractive index of the two bilayer materials be as large as possible, and that the interface between the two materials be well defined and very smootht2]. X-ray mirrors are used widely in the soft x-ray region (clkev), but their use in the hard x-ray region (-SkeV) has only recently been exploited13]. Hard x-ray mirrors require ultra-smooth surfaces and interfaces to maintain high reflectivity and ultra-thin layers to give reflection at angles higher than grazing incidence (i.e. >0.5”8). For example a spacing of approximately 15w is required to reflect at 28 = 6 ’ using 1.54A (8 keV) radiation. In order to identify optimum conditions for fabricating smooth interfaces a systematic study is being carried out to determine the relationship between the deposition parameters and the smoothness of single layer metal films. In this paper, the variation of roughness with film thickness, and gas pressure during magnetron sputtering has been investigated using tungsten, titanium and molybdenum deposited on 100 silicon. This is part of a larger study which encompasses other deposition techniques including cathodic arc deposition and e-beam evaporation. Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 This document was presented at the Denver X-ray Conference (DXC) on Applications of X-ray Analysis. Sponsored by the International Centre for Diffraction Data (ICDD). This document is provided by ICDD in cooperation with the authors and presenters of the DXC for the express purpose of educating the scientific community. All copyrights for the document are retained by ICDD. Usage is restricted for the purposes of education and scientific research. DXC Website – www.dxcicdd.com ICDD Website - www.icdd.com EXPERIMENTAL CONDITIONS All samples were deposited cut from highly polished, on polished substrates, of dimensions 18mm by 30mm, 100 silicon supplied by Infra-Red Products. Substrates were cleaned ultrasonically, first in ANALAR grade acetone and then PET ether, followed by drying with dry nitrogen. Prior to deposition the samples were further dried with a heat gun to remove any condensed water. The films were produced by magnetron Parameters investigated COOLING sputtering using the set-up shown in Figure 1. are listed in Table 1. w*TER?yy- WORKIN GAS VACUUM I I CHAMBER I ’ VACUUM PUMPS I Figure 1. Set-up for DC Magnetron Sputtering. Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 The X-ray Reflectivity (XRR) measurements for this work were carried out on a SIEMENS D5000 8-28 X-ray diffractometer. to measurement the equipment The set-up is shown in Figure 2. Prior was carefully aligned to obtain accurate zero positions for the detector angle and the specimen angle. The tilt and vertical displacement of the specimen were adjusted to an accuracy of +O.O05mm with three micrometers fixed to the specimen stage as shown in Figure 3. The XRR scans were done in two parts. In the vicinity of the critical angle, an attenuator was placed in the beam owing to high count rates. At higher angles the attenuator was removed. All scans were made by collecting rocking curves at equal intervals in 28 of 0.02” along the specular ridge. For a typical sample the rocking curves were recorded over the range 8+0.1” in steps of 0.005’ with a 1 second dwell time at each point. For a range of 0.6-6.0” 28 the collection time is approximately 3.5 hours. The reflectivity value at each 28 was obtained by first subtracting the background and diffuse scatter from the rocking curve and then integrating intensity. the specular reflected higher angle data sets were merged reflectivity and normalised The attenuated with I,, low angle and corresponding = 1. An example of data obtained from a smooth high atomic number (W) film of nominallv 2OOAcan be seen in Figure 4. Source to Divergence slit : 1lOmm Sample to Receiving Figure 2. Schematic diagram of X-ray Reflectivity Slit : 215mm set-up. I Three Pin Micrometer Height djustment 4 Three Pin Micrometer Height Adjustment I Side View I to a Sample Holder I I I -lF Spring loaded sample support Figure 3. Schematic diagram of sample holder. Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 Tar, View le.6 le-7 I 0 1 Figure 4. Measured 2 3 reflected silicon. Atomic Force Microscopy intensity 4 5 6 for a smooth nominal 7 8 28 9 2OOA W film on 100 (AFM) was also used to measure the rms roughness of the top surface and the film thickness. AFM scans were made with an AutoProbe Atomic Force Microscope from Park Scientific. The tip was cone shaped and crafted from silicon with a radius of curvature lOO& Scans were made over areas of 20ym by 20um to lum by lym. Six or more different areas on the sample were scanned to give an average value for the rms roughness of the top surface. The rms roughness was calculated with the manufacturer’s software, Proscan. The AFM was also used to measure the thickness of the film. To do this, steps were produced on the film with a mark of ink on a of silicon substrate which was sputtered at the same time as the XRR samples. These marks were then dissolved with high purity acetone and the thickness subsequently determined by scanning across the revealed step. MODELING THE XRR DATA The C++ software written to analyse and fit the reflectivity data implements Rouard’s treatment [41 for multilay ers and incorporates a Debye-Waller factort5’ to represent interface roughness (o,.&. The parameters obtained from fitting were the interfacial roughness ((T,,), film thickness (t) and density (p) of each layer of the film. All the XRR data in this study were fitted by synthesising XRR curves over a range of conditions and comparing the log values to obtain the best fit using the following expression to represent the goodness of fit : Goodness of Fit = i “2 C i=n (lnRpbS - lnRfa1c)2 1 Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 N where ni and n2 define the angular region being fitted and N the number of points fitted. Various models were examined to represent the structure of the film. The first model examined (Model Ai) consisted densitiest6-‘I of a layer of tungsten but allowing the thickness and roughness on silicon using the bulk of the tungsten and silicon to refine. This model was unable to fit the data (see Figure S).The fit in Al can be improved by allowing the tungsten density to refine also (Model AZ) but the value obtained for the density of the tungsten film was unrealistically low and the thickness obtained was much less than that measured by AFM. Various other layers were added to this model to obtain a better fit. In the model that we have adopted, an oxide layer has been added on both the silicon parameters and tungsten surfaces (Model B). The film obtained for models Ai, AZ, and model B are drawn up in Table 2. It is worth noting that in all cases the thickness of the tungsten layer was constant to within f 1A for each model. Errors with such fittings are hard to define but from the above we can give an estimation of the uncertainties to be c 2% for thickness, ~5% for density and c 10% for rms roughness. The results for the oxide film model are consistent in a number of ways: 1. The inclusion of an oxide layer on both the tungsten and silicon surfaces gives a density for the tungsten layer only 6% less than that for the bulk material, 2. X-ray Photoelectron Spectroscopy on the films has shown the presence of high concentrations of oxygen on the surface of the tungsten film and at the tungsten-silicon interface, 3. AFM measurements to determine the film thickness have been found to be consistently greater than the measured XRR thickness for the tungsten layer (Table 3). The AFM thickness is comparable to the combined thickness of the tungsten oxide and pure tungsten layers found by XRR, 4. Ellipsometry measurements have also shown the presence of a native layer of SiO2 on the surface of the silicon with a thickness in the vicinity of 10-20,&t91, 5. The densities obtained for the amorphous WO, and SiO2 layers from the XRR results are comparable to the densities of the bulk materials WO3 (6.3-7.2 gcme3) and amorphous silica (-2.1 gem-3). Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 le+O ____** -B . Measured le- 1 h le-2 P ;? s $ le-3 0 B ’ le-4 le-5 le-6 / 0 Figure 5. I I 1 2 3 4 Fits obtained using models Ai and B for a nominally Table 2. XRR parameters ANALYSIS 29 6 lOOA W film on silicon. obtained for the W films on 100 Si using models Ai, AZ, OF RESULTS The AFM results and the fitted XRR fitting parameters groups listed in Table 1 are given in Tables 3,4, and 5. The general conclusions follows. 1. 5 obtained for the different drawn from the three groups of results are summarised Increasing the thickness of the W layer has little or no effect on the roughness or densities of the layers. Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 as Table 3. Values obtained for fitting parameters whilst varying thickness of tungsten deposited on silicon 100 at 3mTorr. Nominal Thickness -1oots -2OOA -300w AFM t (A) (TInIS (8) 113+4 1.8kO.4 218f6 1.8fO.l 3Olk1.5 1.6kO.2 wo, t (fi) %ls (8) p (gem”) 16.5 2.1 5.2 16.6 2.8 4.4 15.2 2.7 4.8 w t (A) d Ems (A) p (gem”) 97.2 1.1 18.1 189.8 1.5 18.1 280.0 1.1 18.1 Si02 t (-Q %I, (A) p (gcme3) 24.6 2.3 1.7 24.6 1.5 1.7 24.6 2.3 1.7 Si %Is (A) p (gem”) 1.4 2.33 1.6 2.33 1.4 2.33 XRR Table 4. Values obtained for fitting parameters at varying gas pressures for tungsten deposited on silicon 100. Table 5. Values obtained for fitting parameters titanium deposited on silicon 100 at 3mTorr. AFM 1 Metal- for tungsten, molybdenum Metal W MO Ti 0 mls (A) 1.820.4 1.9kO.6 4.2k1.9 16.6 15.0 16.0 t (A) Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 and 2. With increasing gas pressure during fabrication there is marked increase in the roughness at the W/WO, and the WOJAir interfaces. The XRR fitting results also indicate that the density of the tungsten layer increases with increasing gas pressure. 3. Roughness at the metal/oxide interface does not appear to display any systematic changes with atomic number. The oxide on the titanium film was found to be the roughest with a thickness approximating that of the measured roughness. Further analysis using a FEGSEM will be carried out examining the nature of these surfaces. Table 3 also shows that the film thickness obtained using AFM is invariably greater than the W thickness thickness measured by XRR, but are comparable of the tungsten and oxide films. In this table the roughness interface values given by AFM and XRR are comparable. to the combined of the WOJAir In other films the agreement is not so good. I le-6 i 0 1 2 3 4 5 ze 6 Figure 6. XRR data for 8 mTorr tungsten on silicon. Figure 6 shows the rapid damping of the interference fringes observed in the XRR data which arises when the rms roughness becomes large (-Sfi). Although the XRR data and XRR fitting parameters (Table 4 and 5) show high roughness, it is not reflected in the AFM results. The low levels of roughness measured using AFM are probably due to the large size of the tip used to probe the surface. The tip is approximately lOOA in diameter, therefore cannot probe into the deep, narrow crevices on the surface, As a consequence the AFM will always lateral correlation distance is small [lo, ‘ll. The results given in Table 4 and 5 also show that with the roughness of the metal-oxide surface is comparable metal/metal-oxide interface. This indicates that the upper Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997 give a low roughness when exception of the Ti film, the with the roughness of the oxide/air interface conforms with the metal/metal-oxide roughness interface. In the case of the Ti film, the very large suggests some breaking up of the oxide which will need to be investigated further for confirmation. FUTURE WORK Further investigations are needed to confirm and quantify the preliminary conclusions in the present work. The work will be extended to include other film materials and deposition Diffuse procedures including electron beam evaporation X-ray and X-ray Photoelectron characterise the roughness concentrations and conformity Spectroscopy and filtered cathodic arc. will also be investigated of the interfaces to as well as the elemental with depth. ACKNOWLEDGMENTS I would like to thank the Australian Research Council for providing funding to carry out this work. Also Maree Anast, Richard Wuhrer and Geoff McCredie for helping to make this work possible. REFERENCES 1. 2. Barbee, T.W. (1990) Opt. Eng. 29.711 Puik, E.J. vander Wied, M.J. Zeijlemaker, H. and Verhoeven, (1991) J.Appl. Surf. Sci., 47. 63-67 3. 4. 5. 6. Schuster, M.and Gobel, H. (1995) J. Phys. D : Appl. Phys., 28. A270-A275 Heavens, “Optical Properties of Thin Solid Films” Butterworth Scientific Publications. (1955) Chason, E. Falco, C.M. Ourmazd, A. Schubert, E.F. Slaughter, J.M.and Williams, R.S. (1993) MRS Symp. Proc., 280.203-238 Edited by Wilson, A. J.C., IUCr., “International Tables for Crystallography. Volume C.“Kluwer Academic Publishers. (1992) 7. 8. 9. 10. 11. Cullity, B.D.,“Elements of X-ray Diffraction - 2nd Ed.” Addison Wesley Publishing Company, Inc. (1978) Lide, D.R, Editor in Chief.“Handbook of Chemistry and Physics- 74th Ed” CRC Press. (1993-1994) Ghez, R. “A Primer of Diffusion Problems” John Wiley & Sons. (1988) Westra K.L., and Thomson D.J., (1995) J.Vac. Sci. Technol. B 13 (2). 344-349 Lengeler B., and Huppauff M., (1993) Mat. Res. Sot. Symp. Proc. 280.245250 Copyright (C) JCPDS-International Centre for Diffraction Data 1997 Copyright 0 JCPDS-International Centre for Diffraction Data 1997
© Copyright 2026 Paperzz