KINETICS AND THERMODYNAMICS OF THE FORMATION OF THIN FILM TITANIIUM DISILICIDE R.J. Kasica, E.J. Cotts and R.G. Ahrens Department of Physics, State University of New York at Binghamton, Binghamton, NY 13902 ABSTRACT Multilayered diffusion couples consisting of alternating layers of titanium (Ti) and amorphous silicon (a-si) have been fabricated using sputter deposition with a range of modulation lengths corresponding to an average composition of Ti 33. We have used differential scanning calorimetry to measure the enthalpy evolved during the solid state reaction a-si + Ti C49-TiSi 2 and have characterized the phases formed using x-ray diffraction analysis. An average measured enthalpy of formation, H was found to be H = -58 + 9 kj/g atom for thin film samples. Using scanning and isothermal calorimetry measurements, we have also characterized the kinetics involved during the initial intermixing stage. INTRODUCTION Silicides have been well received as a choice material for microelectronic device applications based on their low resistivities and thermal stability 1,2. Titanium disilicide has been extensively applied because it has one of the lowest electrical resistivities of the suicides. However, as microelectronic devices continue to scale in line widths and film thicknesses, their production requires more stringent design and manufacturing tolerances. Physical vapor deposition methods are limited to depositing material in a line of sight, making their use for small complex device geometries tenuous. The use of chemical vapor deposition (CVD) in the proccessing of these devices offers several important advantages over current vapor phase technology, which includes the ability to improve step coverage and coat complex three dimensional surfaces 3. Its use however, has been hampered by inaccuracies in calculating the phases formed during a CVD reaction 4. These inaccuracies are primarily due to widely varying thermodynamic data which are incorporated into calculations of CVD phase diagrams. In this paper we report thermodynamic results of the study of free-standing Ti/a-Si composites. Using differential scanning calorimetry (DSC), we have been able to monitor the total heat evolved during the solid state reaction of elemental layers while simultaneously monitoring the reaction kinetics. We show that DSC proves to be a useful method for enthalpy of formation measurements for thin solid films and that these measurements can be used to approximate bulk enthalpy values as well. EXPEREMENTAL Titanium-amorphous silicon structures were prepared by sputter depositing silicon (99.9% pure) and titanium (99.99% pure) sequentially onto cleaved NaCl crystals and/or thick ( > 50 pm) oxidized Si wafer substrates. Samples with overall modulation lengths of 10 to 200 nm were fabricated. Once free standing films were removed from the substrate, samples were cut up and hermetically sealed in aluminum DSC pans under an inert atmosphere before loading into DSC sample cups. Heating scans were programmed from just above room temperature to above 800 K at a constant rate, generally 20 K/min. Samples scanned to temperatures above 873 K required the use of other pans which included C, Pt, or Cu. Before and after samples were reacted in the DSC, the films were analyzed using a standard Θ-2Θ X-ray diffractometer with a CuK α radiation source to identify both the initial and final phases formed. 1
DISCUSSION A. Thermodynamics Measurements of heat flow versus temperature upon heating Ti/a-Si composites reveal two disti nct exothermic signals as shown in Figure 1. This form is similar to previous observations 5 which correlated the low temperature exotherm (T < 750 K) with the formation of an amorphous phase. The second or high temperature exotherm has previ ously 5 been identified with the formation of the C49-TiSi 2 phase. We show two plots of heat flow versus temperature for composites with modula tion lengths of 18 nm and 37 nm fabricated with average compositions of Ti 33. Our x-ray dif fraction analysis (XRD) is consistent with these previous studies which used transmission electron microscopy as well as XRD. In Figure 2 we show an XRD profile of an Ti/a-Si composite heated to a temperature above 800 K where the observed Bragg peaks with Miller indices (131) and (060) correspond to the C49 titanium disilicide phase. Integration of the heat flow data pro vides the total heat, L evolved during the formation of C49-TiSi 2 from Ti/a-Si composites as Figure 1. Heat flow versus temperature as measured by differential scanning calorimetry for Ti-a-Si multilayered composites of average composition Ti 33 with modulation lengths as labeled in the figure. Samples were heated at a constant rate of 20K/min Figure 2. X-ray diffraction pattern of a Ti/a-Si composite heated to 8l0 K at 20 K/min. CuKα radiation was used in a standard Θ - 2Θ geometry. proportional to the amount of heat flow, dh/dt, as 7 H = I dh/dt dt. (1) From these measurements and those from additional composites with modulations between 10-200 nmn, we have calculated an average enthalpy, H = -58 ± 9 kj/mol. Previous investigation into the Ti/a-Si system has revealed the formation of an amor phous interlayer during the sputter deposition pro cess 6. We can account for this initial interlayer by assuming planar growth and that the thickness, λ ο, of such a layer is independent of the compos ite modulation, λ, for the sample. Since the en thalpy of formation, H, of this initial interlayer is clearly not measured during our DSC scans, we can examine the dependence of our measured en thalpy values, H on λ by writing Hα λ λ o (2) The growth rate, dx/dt of the reaction product is 2
Figure 3. Measured enthalpies of formation, H plotted as a function of the inverse modulation length, λ -1 for composite film structures of λ = 10 to 200 nm. DH/dt = β dx/dt. (3) The proportionality constant β = ρa H f /M. (4) with H, the enthalpy of formation, A the total interfacial area, ρ the density and M the molar mass of the reaction product, We may then write H = H f cλ o /λ (5) where c is a constant proportional to H o. If we plot our measured values, H versus I/λ, (shown in Figure 2), we can arrive at an estimate for the amount of heat evolved during the thin film reaction a-si + Ti C49-TiSi 2 having taken into account the initial intermixed layer. From a least squares regression analysis fit to the data of Figure 3, we find for the heat of formation of C49-TiSi 2 from Ti and a-si a value of H f = -61 ± 9 kj/mol. B. Kinetics of Amorphization It has been established previously that the Ti/a-Si system proceeds through a solid state amorphization reaction (SSAR) prior to forming a stable titanium disilicide compound phase 8,9. Our x- ray diffraction analysis of samples heated to temperatures between the exothermic peaks in Figure 1 is consistent with these observations. We can analyze the amorphization reaction using DSC to obtain kinetic data on the growth of the amorphous product. In Figure 4, we have plotted the heat flow and growth rate as a function of the temperature for the initial amorphization reaction. By assuming planar diffusion-controlled growth then, we can write x dx/dt = k/2 (6) 3
Figure 4. Growth rate, dx/dt, and heat flow, dh/ dt, the amorphous product phase plotted as a function of the temperature. Figure 5. Isothermal plot of the enthalpy release (squared) as a function of time for the growing amorphous interlayer. Temperature = 713 K. where the interlayer thickness, x and growth rate, dx/dt provide the reaction constant, k 2. Using the relationship of Equation 3, we then find that H dh/dt α k 2. (7) In table 1, we have calculated this reaction constants at a number of temperatures for the Table I. Reaction constants calculated for scan shown in scan shown in Figure 5. Fig. 4. The heat flow as a function of time was measured isothermally at several temperatures during amorphization of Ti/a-Si composites. Plots of the integrated enthalpy squared, H 2, versus time were examined [cf Figure 5, a plot for an isothermal temperature of 713K]. Equation 3 indicates that H 2 is directly proportional to x 2. For diffusion-limited growth kinetics, where x 2 = k 2 t, one expects a straight line plot in Figure 5. The deviations from a straight line (the slope decreases by a factor of two over 180 seconds) are consistent with previous observations of the formation of Kirkendall voids at the interface 8. Studies have shown that planar growth occurs during amorphization reactions 10,11. If we assume planar diffusion-limited growth during the course of the isotherm in Figure 5, than the calculated value of k 2 decreases from 3.0x10-17 m 2 /sec. during the anneal. Estimates of the averaged interdiffusion coefficient, D, in the growing amorphous phase can be calculated from our values of k 2 by D = ck 2 (8) Where c depends on the concentration across the diffusion couple 12. Using an estimate of c based on microscopy studies 13, we calculate a value of D = 7x10-17 m 2 /sec upon heating at a constant rate of 20 K/ min to a temperature of 650K. This value is higher than that reported by De Avinez et al. 5 however this difference is consistent with the indication that D is decreasing as a function of time [cf Fig.5]. 4
CONCLUSIONS Using differential scanning calorimetry, we have calculated the enthalpies of formation of C49- TiSi 2 from free standing modulated thin films of titanium and amorphous silicon. We have shown that DSC provides an accurate means of measuring the enthalpy involved in a solid state reaction. In addition, we have also shown that DSC provides a very useful means (once the rate controlling growth modes have been established) of monitoring the kinetics involved during a solid state reaction. ACI(NOWLEDGEMENTS We gratefully acknowledge the support of the National Science Foundation, DMR-9202595, and DUE- 9452604. REFERENCES I. S.P. Murarka, Silicides for VLSI Applications, (Academic Press, New York, 1983), pp. 1-28. 2. C.M. Osburn, in Rapid Thermal Processing Science and Technology, edited by R-B. Fair (Academic Press, New York, 1993) pp. 227-3 09. 3. H.O. Pierson, Handbook of Chemical Vapor Deposition, (Noyes Publications, Park Ridge, NJ, 1992), pp. I- 16. 4. J. Engqvist, PhD Thesis, Uppsala University, 1992. 5. KR- De Avillez, L.A. Clevenger, C.V Thompson and KN. Tu, J. Mater. Res. 5, 593 (1990). 6. K. Holloway, PhD Thesis, Stanford University, 1988. 7. E.J. Cotts in Thermal Analysis of Metallurgical Systems, edited by R-D. Shull and A. Joshi (TMS, Warrendale, PA 1992), pp. 299-328. 8. I.J.M.M. Raaijmakers, A.H. Reader, PH. Oosting, J. Appl. Phys. 63,2790 (1988). 9. K. Holloway and R- Sinclair, J. Appl. Phys. 61, 1359 (1987). IO. E.J. Cotts, WJ. Meng, and W.L. Johnson, Phys. Rev. Lett. 57, 2295 (1986). 11. H. Schroeder, K Samwer, and U. K6ster, Phys. Rev. Lett. 54, 197 (1985). 12. U. Goesele and KN. Tu, J. Appl. 53, 3252 (1982). 13. I.J.M.M. Raaijmakers, PH. Oosting, A.H. Reader in Multilayers: Synthesis, Properties, and Nonelectronic Applications, edited by T.W Barabee, E Spaepen, and L. Greer (Mater. Res. Soc. Proc. 103, Boston, MA, 1988) pp. 229-233. 5