Surface modified spin-on xerogel films as interlayer dielectrics

Surface modified spin-on xerogel films as interlayer dielectrics S. V. Nitta, V. Pisupatti, A. Jain, P. C. Wayner, Jr., W. N. Gill, and J. L. Plawskya) Department of Chemical Engineering, Rensselaer Polytechnic Institute, Troy, New York 12180 ~Received 17 June 1998; accepted 30 October 1998! SiO2-based xerogels are highly porous materials that may enhance the performance of microelectronic devices due to their extremely low dielectric constants ( e 51.36– 2.2). Conventional xerogel and aerogel manufacturing techniques include an expensive and hazardous supercritical drying step to deposit crack free, high porosity films. Ambient drying techniques have recently been developed and in this article, we discuss how the process parameters in the ambient drying process affect the properties of a spin-coated film. Successful spin-on deposition of highly porous ~.70%!, thick ~.1 mm!, crack-free, xerogel films was accomplished using a solvent saturated atmosphere during spinning and aging. The saturated atmosphere allowed for the isolation of each processing step and a better understanding of the effects of process variable changes. The film porosity was controlled by varying the extent of silylation ~surface modification!, the aging time, or the initial water/silane ratio. Fourier transform infrared spectra demonstrated that silylation of xerogel films helps eliminate bound moisture in these films and renders them hydrophobic. Finally, the dielectric constants extrapolated from refractive index measurements were in good agreement with those obtained from our conventional electrical measurements. © 1999 American Vacuum Society. @S0734-211X~99!01701-1# I. INTRODUCTION ~C2H5O!3Si–OC2H51HO–Si~OC2H5!3 With ultralarge scale integration ~ULSI! and deep submicron device dimensions of perhaps 0.1 mm in the near future, the performance of microelectronic devices is seriously inhibited by the RC delay associated with the resistance of the metal ~R! and capacitance of the dielectric ~C!, used as interconnect materials. The RC delay is directly proportional to the resistivity, r, of the metal and the dielectric constant, e, of the dielectric. However, as device dimensions shrink, the relatively high dielectric constant of SiO2 ( e 53.9) will preclude its use as an interlayer dielectric due to an increased RC delay, as well as increased cross talk, noise, and power dissipation. Hence materials with a lower intrinsic dielectric constant need to be developed and evaluated as potential replacements for SiO2. Due to their high thermal stability and their ability to be rendered hydrophobic, silica xerogels ~prepared by the ambient drying process and deposited using a spin-coating procedure! have the potential to emerge as an attractive alternative to conventional SiO2. Previous work: Silica xerogels are prepared by the hydrolysis and condensation of alkoxysilanes.1 Tetraethylorthosilicate ~TEOS! is first hydrolyzed by reacting with water in the presence of a mutual solvent and a catalyst as shown in the following: hydrolysis: Si~OC2H5!41H2O→~C2H5O!3Si–OH1C2H5OH; ~1! water condensation: ~C2H5O!3Si–OH1HO–Si~OC2H5!3 →~C2H5O!3Si–O–Si~OC2H5!31H2O; alcohol condensation: a! Electronic mail: plawsky@rpi.edu 205 J. Vac. Sci. Technol. B 17„1…, Jan/Feb 1999 ~2! →~C2H5O!3Si–O–Si~OC2H5!31C2H5OH. ~3! Once hydrolysis has started, condensation reactions occur, liberating either a molecule of water or alcohol upon siloxane bond formation ~–Si–O–Si–!. As these reactions proceed, larger siloxane oligomers are formed.1 Since TEOS is tetrafunctional, the monomer undergoes a nonlinear polymerization and forms a three dimensional structure. Bond formation leads to fractal aggregates that grow until they begin to impinge on one another and form clusters. At the gel point, a single, spanning cluster appears that coexists with a sol phase containing many smaller clusters. These smaller clusters gradually become attached to the network in a process called aging.1 The rates of hydrolysis and condensation reactions, as well as the structure of the final gel, vary depending on factors such as the type of catalyst employed and the hydrolysis ratio ~ratio R, of H2O to Si in the initial mixture!. In general, acid catalysis with low hydrolysis ratios (R,2) produces long chained, less highly branched polymers, whereas base catalysis with higher hydrolysis ratios (R.2) produces more compact, highly branched polymers.1 In this work, we use a two-step, acid-base catalysis scheme to prepare the sol for the deposition of xerogel films. The two-step procedure allows us to control the average pore size so that it is much less than the ~;0.1 mm! device dimensions planned in future integrated circuits. As bond formation continues beyond gelation, the network contracts and expels pore fluid. The gel shrinks. Shrinkage happens most dramatically during drying as the fluid evaporates from the pores. To obtain highly porous aerogels or xerogels, it is important to prevent shrinkage during any part of the processing, particularly drying. This can 0734-211X/99/17„1…/205/8/$15.00 ©1999 American Vacuum Society 205 206 Nitta et al.: Surface modified spin-on xerogel films be accomplished by a number of techniques including supercritical drying and ambient drying with surface modification.2 Conventional aerogel preparation uses a supercritical drying process that eliminates the capillary stress exerted by the solvent on the gel.1 While this helps avoid shrinkage during drying, it is an expensive and hazardous high pressure and high temperature process. An alternative to supercritical drying is the ambient pressure drying technique, which involves a predrying surface modification step known as silylation. This technique has been developed within the last few years2,3 and involves replacing the hydroxyl groups on the surface with inert, trimethylsilyl groups by reacting the wet gel either with trimethylchlorosilane ~TMCS! or hexamethyldisilazane ~HMDS!. Unlike TMCS, HMDS does not corrode metallization layers.4 However, the structural effects of using HMDS are indistinguishable from adding TMCS. Prakash, Brinker, and Hurd2,3 reported one of the first successful fabrications of dip coated, thin film xerogels by this technique. Unfortunately dip coating is incompatible with semiconductor processing because both surfaces of the substrate are coated by this method and the process is harder to control than spin coating. Several authors5–8 have reported the successful deposition of spin-coated silica sol-gel films. Hrubesh and Poco5 used supercritical drying to form aerogel films, while the others6–8 formed densified sol-gel films that had extremely low porosities. Smith and co-workers4,9,10 successfully deposited spincoated xerogel films under ambient drying conditions. Process improvements made by them have cut down the total process time for film coating to the order of minutes, thus making these materials commercially attractive. Integration studies of their films with aluminum as well as copper as the metal layers have also been reported.11,12 Jo and co-workers, have reported the successful deposition of supercritically dried films13,14 as well as ambient dried15 films using a process flow similar to that of Smith et al.9 However, while they achieved high porosity using supercritically dried films, the porosity they achieved with their ambient dried films is quite low ~,70%!. While the successful deposition of ambient dried xerogel films via spin coating has been reported, the relationship between the processing parameters and the final properties of the film has not been reported in detail in the open literature. In this article we report on the effects of varying the extent of surface modification, the aging time, the drying temperature, and the water–silane ratio, on the thickness, porosity, and dielectric constant of ambient dried silica xerogel films. II. EXPERIMENTAL PROCEDURES An overview of the fabrication scheme for depositing high porosity xerogels is given in Fig. 1. This scheme is modified from the one developed by Prakash, Brinker, and Hurd2,3 for preparing dip-coated xerogel films and results in fairly long, overall process times. Our approach enabled us to isolate and control all aspects of film formation during J. Vac. Sci. Technol. B, Vol. 17, No. 1, Jan/Feb 1999 206 FIG. 1. Xerogel fabrication process schematic ~Ref. 5!. each stage of the process. Thus we had greater control over, and sensitivity to, changes in the process variables. Step 1 of the sol preparation was carried out by reacting a mixture of TEOS, water, and HCl at 60 °C for 90 min using ethanol as the solvent. The following molar ratio of reactants was used (TEOS:H2O:EtOH:HCl):1:2:3.8:7.331024 . Stoichiometrically, a water/TEOS ratio, R, of 2 is necessary for complete conversion. Step 2 of the sol preparation–base catalysis, was carried out by mixing 10 ml of the solution from step 1 with 1 ml of 0.05 M NH4OH. Immediately following base catalysis, 2 ml of the acid-base catalyzed mixture was filtered through a 0.2 mm filter and deposited on a precleaned silicon wafer in a Laurell 4SNPP spin coater. During the deposition stage, rapid evaporation forces the precursors into close proximity, thus increasing the reaction rate. An undesirable side effect of this rapid evaporation is that films shrink during and immediately following deposition and thus it is difficult to obtain highly porous coatings. In thin films, fluid flow due to either gravitational ~in dipcoating systems! or centrifugal forces ~in spin-coating systems!, along with attachment of the precursor species to the substrate, impose a shear stress within the film during deposition. Rapid shrinkage due to removal of solvent and continued condensation creates a tensile stress within the film and changes the rheological properties of the film. Both effects generally lead to cracking or delamination of the film.1 Thus it is crucial to control the evaporation of solvent from the film during and following deposition to be able to obtain gels with high porosities. 207 Nitta et al.: Surface modified spin-on xerogel films In this work, the sol was spun at 2000 rpm for approximately 12 s. A solvent saturated atmosphere during spin coating was insured by sealing off all the ports of the spin coater and flooding the bottom of the spin coater with a mixture of ethanol, water, and ammonium hydroxide. The saturated atmosphere allowed us to control the evaporation rate from the film and was used by Hrubesh and Poco5 as well. Following spin coating, wafer rotation was stopped and the spun-on film was allowed to gel in the spin coater for about 45 min ~gel formation takes between 10 and 20 min at room temperature and the catalyst concentration used!. At the end of the gelation stage, the wafer was transferred to a petri dish containing an aging solution that consisted of a mixture of ammonium hydroxide in ethanol. The ammonium hydroxide was added to enhance gelation during aging. Before removing the film from the solvent saturated atmosphere, 2 ml of the aging solution was deposited onto the gelled film through the lid of the spin coater to prevent premature drying of the film during the transfer. The petri dish containing the film and the aging solution was then placed in an oven held at 40 °C. Aging times were varied between 30 min and 24 h. After the aging step, a surface modification ~silylation! procedure was performed to replace the unconverted surface silanol groups with unreactive methyl groups. The pore solvent, ethanol, was also exchanged with hexane during this step. Hexane has a lower surface tension than ethanol and exerts lower capillary pressure during drying, leading to reduced shrinkage and cracking. The silylation was effected by immersing the aged film in a solution of hexane and TMCS for 60 min. The amount of TMCS in hexane was varied from 1% to 12%. The surface modified films were then dried in a furnace with either air or dried nitrogen as the environment. The drying temperature was set between 50 and 450 °C. The furnace ramp rate was held constant at 1 °C per minute. Film characterization: The thickness and refractive index ~RI! of the films were measured by ellipsometry at a wavelength of 632.8 nm and an angle of 70°. Since ellipsometry reports a cyclic film thickness, an interferometer and a stylus profilometer were used to obtain an estimate of the film thickness and set the ellipsometric period. Infrared spectroscopy ~ATR FTIR! was performed on these films to check for bound and adsorbed water. The measurements were performed under atmospheric conditions. Transmission infrared spectra of the films also enabled important information to be obtained on the structure and composition of the films before and after drying. The sample preparation for the transmission Fourier transform infrared ~FTIR! measurements involved scraping a film off the substrate, mixing the scraped off powder with potassium bromide, and preparing a transparent pellet. Since potassium bromide is hygroscopic, its spectra was subtracted from that of the combined spectra to yield the bare film spectra. Field emission scanning electron microscopy ~SEM! cross-sectional images of the films were obtained to provide JVST B - Microelectronics and Nanometer Structures 207 an estimate of the average pore size, the pore size distribution, and the size of the largest pore in the sample observed. Xerogels films were prepared on highly doped, p type, ^100& silicon wafers ~resistivity 0.005–0.02 V cm! to measure the dielectric constant electrically. One hundred nanometers of thermal oxide was grown on the surface of these wafers prior to xerogel deposition. After deposition of the xerogel films, the wafers were capped with 100 nm of PECVD TEOS oxide. Metal dots ~0.2–0.7 cm diam! were evaporated onto the films in an e-beam evaporator through a shadow mask. A blanket layer of Al was also deposited on the backside of the wafer to act as a second contact. The capacitance of the sandwich structure was measured using an impedance analyzer and the capacitance of the xerogel film was calculated based on the capacitors-in-series model. The value of the capacitance of the oxide layer, was measured separately using an identical structure. III. RESULTS AND DISCUSSION A. Refractive index, xerogel porosity, and xerogel dielectric constant Several correlations exist which relate measured values of the refractive index to the porosity and dielectric constant of porous bodies. Henning and Svensson ~HS!16 use volume average definitions for the density and refractive index of xerogels to relate the two: h gelV tot5 h SiO2V SiO21 h air@ V tot2V SiO2# , ~4! r gelV tot5 r SiO2V SiO21 r airV air . ~5! Assuming r air'0, h air'1 and with h SiO251.45 and r SiO2 52.19 g/cm3, the refractive index, h gel , and porosity, p gel , are given by h gel5120.21r gel , p gel512 ~6! r gel 1.452 h gel . 5 r SiO2 0.45 ~7! Other relations exist for correlating the refractive index to the porosity of a porous body.17 The Maxwell–Garnett ~MG! approximation for the xerogel is given by N 2gel21 N 2gel12 2 N SiO 2N 2air 2 , 5 f SiO2 2 N SiO 12N 2air ~8! 2 where N i are the complex refractive indices of the xerogel, SiO2 and air. f SiO2 is the volume fraction of inclusions which we take as being SiO2. The effective medium approximation ~EMA! is given by the following equation: 2 N SiO 2N 2gel N 2air2N 2gel 2 f SiO2 2 1 f 50. air 2 N SiO 12N 2gel N air12N 2gel ~9! 2 For a film of refractive index 1.1, the Maxwell–Garnet, effective medium, and empirical correlation of HS yield a porosity of 0.76, 0.77, and 0.78, respectively. All of these values are in good agreement so, in principle, any correlation 208 Nitta et al.: Surface modified spin-on xerogel films could be used. In this work we use the correlation given by Henning and Svensson16 since this was developed specifically for the case of xerogel films and shown to apply experimentally. We should mention that none of these correlations include the effect of surface water or trimethylsilyl groups. These groups are present in such small quantities that they would not affect our refractive index measurements and hence our estimates of porosity. Their presence does affect the dielectric constant measurements at 1 MHz as we will show in Fig. 3. Hrubesh, Keene, and Lattorre18 also obtained empirical correlations for relating the dielectric properties of silica aerogels to their density and hence, the refractive index. They measured the complex relative permittivity ~dielectric constant! of bulk aerogel samples with and without ‘‘significant moisture content.’’ The presence of residual water is reported to contribute to a 7% increase in the measured dielectric constant of these bulk samples when compared to samples that were calcined to remove the bound water. The dielectric constant measurements were made using a ‘‘cavity perturbation’’ technique, which relates the shift in the resonant frequency of an air filled microwave cavity upon introduction of an aerogel sample, to its dielectric constant. Based on their experiments, Hrubesh, Keene, and Lattorre,18 obtain the following correlation for bulk aerogels: e gel5117.1~ h gel21 ! calcined aerogels, no moisture, ~10! e gel5117.7~ h gel21 ! as prepared aerogels, ~11! applicable for aerogels with a porosity in the range between 73% and 99.5%. The results obtained from our direct measurements of the dielectric constant using the metal–insulator–semiconductor sandwich structure are shown in Fig. 2 and in Table I. The dielectric constant ~measured at 1 MHz! for the film is obtained from the slope of the best fit line. Figure 3 shows how the dielectric constant varies with porosity. The porosity was calculated from measured values of the refractive index using Eqs. ~6! and ~7!. The measured values of the dielectric constants were 2.0 or lower as expected for ‘‘dry’’ films of porosity greater than 70%. The porosity of these films was controlled by the extent of surface modification and aging. The higher porosity films were treated with higher concentrations of TMCS. This may explain why the dielectric constant data are better represented by Eq. ~11! at low porosity and by Eq. ~10! at higher porosity. At higher porosity, we have less bound water in the films. Chow et al.19 also report values for the dielectric constant of a cured film that ranged from 2.5 when measured in an inert atmosphere to 4.5 when measured in the ambient air. They explain this as a result of moisture absorption. Since all our films were tested under ambient conditions ~;40% relative humidity! and we observe low dielectric constants, we must have a negligible amount of moisture in the films and this is confirmed by our FTIR spectra. J. Vac. Sci. Technol. B, Vol. 17, No. 1, Jan/Feb 1999 208 FIG. 2. Capacitance measurement of xerogel film. B. Xerogel IR features Figure 4 shows a typical transmission FTIR spectra of a silica xerogel film. The main peak at 1080 cm21 is associated with the transverse optical vibration mode corresponding to the asymmetric stretching of the intertetrahedral oxygen atoms in the Si–O–Si linkage.20,21 The relative intensity of this peak compared to densified glasses or sol-gels, suggests that the compressive stress in the sample is low, and that the silica network is comparatively stiff.20 The peak centered around 1200 cm21 that appears as a shoulder at the high frequency end of the 1080 cm21 peak has been attributed to the corresponding longitudinal optical ~LO! vibration mode21 of the Si–O–Si linkage. Though light was incident normal to the sample, Almeida and Pantano21 explain the appearance of the LO shoulder as scattering due to the porous nature of the samples. The peaks between 875 and 950 cm21 in Fig. 4 are attributed to the stretching vibrations of Si–OH or Si–O2 groups. These absorptions are very strong and since these peak are relatively small, there is little or no moisture in the sample. This conclusion is supported by the absence of a Si–OH shoulder on the low frequency end of the 1080 cm21 peak, and the absence of any OH stretching peaks around 3400– 3600 cm21. The relative absence of water is confirmed by the electrical, dielectric constant measurements shown in Fig. 3. TABLE I. Xerogel film properties used in dielectric constant measurements. Process variables Sample No. Aging time ~h! %TMCS Thickness ~mm! Porosity ~%! Dielectric constant ~k! 1 2 3 12 12 4 8 4 2 1.35 1.84 1.32 80 76 70 1.66 1.74 2.06 209 Nitta et al.: Surface modified spin-on xerogel films FIG. 3. Dielectric constant vs porosity. The peaks centered around 2980 and 1295 cm21 are associated with the C–H and the Si–CH3 stretching vibrations, respectively. The C–H vibrations can arise either from residual ethoxy groups due to incomplete hydrolysis, or from the methyl groups introduced during silylation. However, the presence of the peak is mainly due to the methyl groups present on the surface after silylation. Figures 5 and 6 show comparative FTIRs of xerogels formed using 1% TMCS in hexane and 10% TMCS in hexane during the surface modification step. Figure 6 highlights the C–H stretching region. As the extent of silylation is increased, the C–H stretching peak at 2980 cm21 intensifies proportionally indicating that the height of this peak is due to the presence of the methyl groups on the surface, not unreacted silane. Figure 5 shows a similar increase in the Si–CH3 stretching peak at 1295 cm21. 209 FIG. 5. Transmission FTIR of xerogel films with different TMCS loadings. 10% TMCS processed film is more porous. shape. Analyzing the SEM pictures shows that the average pore size is in a range between 10 and 20 nm and that the average particle size is also about 20 nm. The films also appear isotropic and have no ‘‘large’’ pores ~.100 nm!. Our pore size estimates reflect only those pores that are visible in the images. Unlike BET measurements, we have no access to the micropores that may exist within our individual particles. A full adsorption study would be required to determine the actual pore size distribution. C. SEM analysis of xerogel films Figure 7 shows high resolution SEM fracture images of a xerogel film that is 78% porous. The film is composed of a collection of bonded particles that are largely spherical in FIG. 4. ATR-FTIR spectrum of a typical xerogel film. JVST B - Microelectronics and Nanometer Structures FIG. 6. Transmission FTIR of xerogel films with different TMCS loadings. 210 Nitta et al.: Surface modified spin-on xerogel films 210 FIG. 8. Film thickness and refractive index of xerogel films as a function of aging time. FIG. 7. Cross-sectional, fracture SEMs of 78% porous xerogel film; ~a! magnification: 25 0003; ~b! magnification: 100 0003. D. Processing results Aging the film helps reduce shrinkage and cracking because it strengthens and stiffens the network and enables it to withstand the capillary pressures better during the drying stage.1 At the pH of 10, used in these experiments, and with all other processing parameters held constant, the refractive index of the gels decreases with an increase in aging time as shown in Fig. 8. This implies that as the aging time is increased, the gels retain more of the porosity developed during gelation. Strengthening of the network is also assisted by coarsening during the drying stage. Coarsening helps to reduce the interfacial area and causes the growth of necks between particles due to dissolution and reprecipitation.1 Neck growth stiffens and strengthens the gel network and the reduction in interfacial area reduces the capillary pressures developed during drying.1 Thus, during aging, both continued polymerization as well as coarsening contribute towards minimizing shrinkage. Corresponding to the increase in porosity, we also expect a decrease in shrinkage of the final gel films. As shown in Fig. 8, the thickness of the films increases with an increase in aging time. Each data point in the figure corresponds to a separate film. These results are consistent with those observed by Prakash, Brinker, and Hurd,2 using dip coated films. However, in dip-coated films, aging is controlled by J. Vac. Sci. Technol. B, Vol. 17, No. 1, Jan/Feb 1999 the withdrawal speed and the chemical reactions leading to coarsening occur predominantly in the extended meniscus that exists above the sol pool. Since dissolution, reprecipitation, and solvent evaporation/condensation are all functions of interfacial curvature, the high curvature gradients, and evaporation/condensation processes in the extended meniscus make the process difficult to control. By contrast, the spin coating technique separates deposition from aging. No bulk curvature gradients exist in the film and aging takes place uniformly over the wafer. The effects of increasing the extent of silylation on the refractive index and the film thickness are shown in Fig. 9. The role of the organosilane is to cap off the reactive terminal hydroxyl groups with organic ligands that are unable to participate in further condensation reactions. As the gels are treated with higher concentrations of silylating agent, they become thicker and more porous. This trend continuous until an optimum is reached at around 8% TMCS in hexane for the contact time used here. We observe an increase in refractive index beyond a TMCS concentration of 8%. One possible explanation for this observation is that the methyl molecules begin filling up some of the pore space. The FTIR spectra shown in Figs. 5 and 6 indicate that as the films are treated with higher concentrations of TMCS, more trimethylsilyl groups are introduced. Prakash, Brinker, and Hurd observe a similar relationship between porosity and silylation in dip coated, ambient dried films.2,3 For their processing conditions, the optimum concentration of TMCS was 6%. In Fig. 10, the effect of drying temperature on the refractive index and film thickness is shown. Two films were processed under identical conditions. One was dried in dry N2 and the other in ambient air. The refractive index and film thickness are independent of temperature until about 450 °C. 211 Nitta et al.: Surface modified spin-on xerogel films 211 FIG. 11. Transmission FTIR of film dried to 450 °C. FIG. 9. Film thickness and refractive index of xerogel films as a function TMCS concentration used in the silylation step. At that temperature there is a slight increase in porosity as evidenced by the decrease in refractive index. We also observe from Fig. 10 that the effect of annealing on film thickness is very slight. Figure 11 is a FTIR spectra of the film dried at 450 °C. The – CH3 peak is effectively absent indicating that the methyl groups have decomposed during the 450 °C anneal. This could also explain the decrease in refractive index at this temperature. FIG. 10. Film thickness and refractive index of xerogel films as a function of drying temperature. Two samples processed under identical conditions were used. One was dried in air and the other in dried N2. JVST B - Microelectronics and Nanometer Structures The effect of increasing the water to silane ratio of the initial sol is shown in Fig. 12. As the water to silane ratio of the sol increases, the refractive index of the film produced from that sol increases. As we increase the water to silane ratio in the starting sol, the overall rates of the hydrolysis and condensation reactions increase. This increased reaction rate leads to denser initial clusters, increased condensation during aging and drying, and an overall denser film. The leveling off in refractive index as R increases coincides with the fact that stochiometrically, only 2 moles of water per mole of TEOS are required for complete hydrolysis. It also demonstrates that there are no mass transfer limitations resulting FIG. 12. Refractive index of the xerogel films as a function of the initial water to silane ratio used in the sol formulation. 212 Nitta et al.: Surface modified spin-on xerogel films from incomplete mixing. Using excess water in the sol decreases the process time. However, it makes control of the thickness and uniformity of the films more difficult and also requires the use of extra capping agent to remove the added bound water locked in the Si–OH groups. IV. CONCLUSIONS Xerogel films can be fabricated with extremely low dielectric constants ~,2.0! and can also be manufactured to be thick ~.1.0 mm!, hydrophobic, and thermally stable. Hence, they are attractive candidates as interlayer dielectrics. The spin-coating process decouples the deposition and drying steps making uniform xerogel films with controlled properties possible. Saturating the atmosphere of the spin coater with the solvent prevents premature drying and shrinkage of the xerogel films and is critical to this decoupling process. Aging strengthens the gel network so that it can withstand the capillary forces during drying. The longer a gel is allowed to age, the more of its original porosity it retains upon drying and the lower its dielectric constant. Silylation of the xerogel films prior to drying inhibits the condensation reactions during the drying stage by replacing many of the free surface hydroxyl groups with inert methyl groups. This procedure ensures that the films contain minimal bound water and the procedure also renders the films much more resistant to moisture absorption from the environment. An optimum extent of silylation exists. Too much surface modification results in lower porosities and higher dielectric constants as the pores get filled with organic moieties. FTIR analysis of the xerogel films demonstrates that there is very little water present and gives important clues to their composition and structure. This also is corroborated by electrical measurements of the dielectric constant. Both the electrical and optical dielectric constant measurements are in good agreement. ACKNOWLEDGMENTS Some of the authors ~S.V.N., J.L.P., and P.C.W., Jr.! were supported by the Department of Energy under Grant No. DE-FG02-89ER14045.A000. Any opinions, finding, and conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the view of the DOE. Some of the authors ~W.N.G., V.P., and A.J.! were supported by the Semiconductor Research Corporation under Contract No. SRC-450-96. The authors also wish to thank Motorola for obtaining the FTIR spectra of our films shown in Fig. 4 and the Dow Chemical Company for their assistance in obtaining key SEM photographs of our films shown in Fig. 7. J. Vac. Sci. Technol. B, Vol. 17, No. 1, Jan/Feb 1999 212 NOMENCLATURE C f air f SiO2 r R R V SiO2 V tot e e gel h air h gel h SiO2 p gel r air r gel r SiO2 1 capacitance of dielectric ~farad! volume fraction of air volume fraction of SiO2 resistivity of metal ~V m! hydrolysis ratio resistance of metal ~V! volume of SiO2 ~m3! total volume ~m3! dielectric constant dielectric constant of the xerogel refractive index of air refractive index of the xerogel refractive index of SiO2 porosity of the xerogel density of air ~kg/m3! density of the xerogel ~kg/m3! density of SiO2 ~kg/m3! C. J. Brinker and G. W. Scherer, Sol-Gel Science ~Academic, New York, 1990!. 2 S. S. Prakash, C. J. Brinker, and A. J. Hurd, J. Non-Cryst. Solids 190, 264 ~1995!. 3 S. S. Prakash, C. J. Brinker, A. J. Hurd, and S. M. Rao, Nature ~London! 374, 439 ~1995!. 4 C. Jin, J. D. Luttmer, D. M. Smith, and A. T. Ramos, MRS Bull. 39 ~1997!. 5 L. H. Hrubesh and J. F. Poco, J. Non-Cryst. Solids 188, 46 ~1995!. 6 S. Maekawa, K. Okude, and T. Ohishi, Electron. Commun. Jpn., Part 2: Electron. 77, 86 ~1994!. 7 X. Zhongfu and P. Gunther, IEEE Trans. Electr. Insul. 1, 31 ~1994!. 8 K. Vorotilov et al., J. Sol-Gel Sci. Technol. 5, 173 ~1995!. 9 D. M. Smith, J. Anderson, C.-C. Cho, and B. E. Gnade, Mater. Res. Soc. Symp. Proc. 371, 261 ~1995!. 10 C. C. Cho, D. M. Smith, and J. Anderson, Mater. Chem. Phys. 42, 91 ~1995!. 11 R. S. List, C. Jin, S. W. Russell, S. Yamanaka, L. Olsen, L. Le, L. M. Ting, and R. H. Havemann, Symposium on VLSI Technology Digest of Technical Papers, 1997, pp. 77–79. 12 E. M. Zielinski, S. W. Russell, R. S. List, A. M. Wilson, C. Jin, K. J. Newton, J. P. Lu, T. Hurd, W. Y. Hsu, V. Cordasco, M. Gopikanth, V. Korthuis, W. Lee, G. Cerny, N. M. Russell, P. B. Smith, S. O’Brien, and R. H. Havemann, Proc. IEDM 1997, 31.7.1 ~1997!. 13 M.-H. Jo, J.-K. Hong, H.-H. Park, J.-J. Kim, and S.-H. Hyun, Microelectron. Eng. 33, 343 ~1997!. 14 M.-H. Jo, H.-H. Park, D.-J. Kim, S.-H. Hyun, S.-Y. Choi, and J.-T. Paik, J. Appl. Phys. 82, 1299 ~1997!. 15 M.-H. Jo, J.-K. Hong, H.-H. Park, J.-J. Kim, S.-H. Hyun, and S.-Y. Choi, Thin Solid Films 308, 490 ~1997!. 16 S. Henning and L. Svensson, Phys. Scr. 23, 697 ~1981!. 17 H. G. Tompkins, A Users Guide to Ellipsometry ~Academic, New York, 1993!. 18 L. W. Hrubesh, L. E. Keene, and V. R. Lattorre, J. Mater. Res. 8, 1736 ~1993!. 19 L. Chow, T. Yu, B. S. Dunn, K. N. Tu, and C. Chiang, Mater. Res. Soc. Symp. Proc. 476, 105 ~1997!. 20 T. M. Parrill, J. Mater. Res. 7, 2230 ~1992!. 21 R. M. Almeida and C. G. Pantano, J. Appl. Phys. 68, 4225 ~1990!.