SiO2 aerogel film as a novel intermetal dielectric Moon-Ho Jo, Hyung-Ho Park,a) Dong-Joon Kim, Sang-Hoon Hyun, Se-Young Choi, and Jong-Tae Paikb) Department of Ceramic Engineering, Yonsei University, Seodaemun-ku, Seoul 120-749, Korea
~Received 2 January 1997; accepted for publication 28 April 1997! A low dielectric constant material for an intermetal dielectric ~IMD! is imperative to reduce power dissipation, cross talk, and interconnection delay in the deep submicron device regime. SiO2 aerogel is one of the possible candidate with an inherent low dielectric constant. This article reports on the results of the successful fabrication of a SiO2 aerogel film as well as its material properties and electrical properties. Fundamental physical, chemical, and electrical material properties were evaluated for a SiO2 aerogel film before and after thermal treatment. An inherent low dielectric constant of 2.0 was realized for about 70% porosity of the SiO2 aerogel film and the leakage current density held at a level of 1027 A/cm2. Preliminary results of the SiO2 aerogel film investigated in our study represent a very positive prospective to IMD applications. © 1997 American Institute of Physics. @S0021-8979~97!02615-7#
I. INTRODUCTION
II. EXPERIMENT
The basic technological trend in ultralarge scaled integration is the realization of a higher device speed with closer packing density, which results in multilevel interconnection structure. Interconnection delay, generally termed resistance–capacitance ~RC! time delay, which is mainly dominated by parasitic capacitance between metal interconnections, has received a great deal of attention over the basic gate delay in the deep submicron devices.1–3 The implementation of conventional SiO2 as an intermetal dielectric ~IMD! with a new material of lower dielectric constant can be a hopeful solution to reducing parasitic capacitance. SiO2 aerogel film may be a promising one because it has a very low dielectric constant due to its inherent high porosity which is controllable in the fabrication process. SiO2 aerogel has been well known as a sol-gel derived solid with nanoscaled particles and pores that can be tailored in the chemical solution stage. With respect to the dielectric constant, it has been reported that a dielectric constant below two, even very close to one, can be easily realized for a monolith SiO2 aerogel.4,5 With this in mind, we proposed a SiO2 aerogel film for application in a new IMD material with very low dielectric constant in an earlier article.6 Dielectric properties of a solgel derived SiO2 film are greatly influenced by their chemical composition and nanostructure such as low density and high porosity. Therefore, quantitative analysis on the film density and chemical composition of SiO2 aerogel films is most important in order to evaluate the films’ material properties from a microelectronics application point of view. In this work, its basic material characteristics, classified into physical, chemical, and electrical properties, were investigated for use in possible device applications. a!
Author to whom correspondence should be addressed. Electronic mail:
[email protected] b! Current address: Semiconductor Technology Division, ETRI, Yusung, Taejon, 305-600, Korea. J. Appl. Phys. 82 (3), 1 August 1997
A. Film synthesis
SiO2 sol was prepared by a two-step process involving acid and base catalysts with tetraethoxysilane ~TEOS! as a precursor. We used iso-propanol ~IPA!, in particular, as the solvent because it has a relatively low critical temperature and pressure in the supercritical drying process.7 The composition of the sol was as follows: TEOS:H2O:NH4OH:HCl:IPA5 1:4:8.23 1023 :1.83 1024 :3 in a molar ratio. The sol was spin deposited on silicon substrates by a commercial photoresist spinner with parameters such as viscosity and spin rate optimized under the solvent atmosphere to minimize solvent extraction during the spin deposition. Then each spun-on film was immersed in IPA and subsequently placed in an autoclaving apparatus. The solvent could be extracted from internal pores between the skeletal solid fraction without any shrinkage using the supercritical drying method. Subsequently, each dried film was thermally treated up to 450 °C under vacuum which is the maximum temperature for the Al metallization process and the resultant properties were compared to as-deposited films.
B. Material characterization
The surface morphology of the films was observed by scanning electron microscopy ~SEM!. A quantitative interpretation of the surface particle size and its distribution was attempted using atomic force microscopy ~AFM! ~Autoprobe CP, Park Scientific Instruments! in noncontact mode. The noncontact mode was able to overcome many drawbacks by interaction between the tip and soft sample surface that occurred in contact mode and therefore presented a better nanoscale image. The film density/porosity and chemical composition were measured by Rutherford backscattering spectrometry ~RBS! with collimated 2 MeV He11 ions. Energy loss of the injected He11 ions to the sample depends only on the amount of material traversed regardless of the physical thickness which can be measured by cross-sectional SEM
0021-8979/97/82(3)/1299/6/$10.00
© 1997 American Institute of Physics
1299
observation.8 From the spectra of each element, the areal density value, which is the film density multiplied by the thickness, can be measured by counting the energy width between the front and rear edges of each element, that is, the energy difference between the detected He11 ions that are backscattered from the sample surface and from the SiO2 aerogel film/Si interface. Then, for the film density measurement it is coupled by SEM cross-sectional measurement of the film thickness. The chemical structure of the SiO2 aerogel film was complemented by x-ray photoelectron spectroscopy ~XPS! and Fourier transformed infrared ~FTIR! spectroscopy ~Bomem DA 3.16 spectrophotometer! analyses. We operated XPS ~VG Scientific ESCALAB 220i! with Mg K a ~1253.6 eV! radiation operating at 250 W. A wide scan spectrum was recorded with 100 eV pass energy to observe all the surface elements. Then narrow scan spectra of all regions of interest were recorded with 20 eV pass energy in order to identify the chemical bonding states of the elements and their variation according to in situ thermal heating under 1.0310 210 Torr at a heating rate of 10 °C/min. FTIR was also used to identify the chemical species on the internal surface. The refractive indices of the SiO2 aerogel film were also measured by ellipsometer ~Gaertner Scientific Corporation, L117C! at its most shallow incident angle of 30°. Dielectric properties and leakage current behaviors were evaluated by C – V characteristics with a HP 4194A impedance/ gain-phase analyzer and a HP 4155A semiconductor parameter analyzer in the metal/insulator/semiconductor ~MIS! structure which was fabricated by depositing an Al electrode at room temperature onto the SiO2 aerogel film. The back side of the Si wafer was also metallized after etching off the native oxide with a diluted HF solution in order to form a better back contact. III. RESULTS AND DISCUSSION
FIG. 1. Typical RBS spectra of SiO2 aerogel films: ~a! as-deposited and ~b! thermally treated at 450 °C.
work is located at 1060– 1080 cm21 due to the asymmetric stretching mode. Various organic species are observed in the IR spectrum of the as-deposited film. The C–H stretching band at 2936– 2978 cm21 and the juxtaposed deformation band at 1380– 1467 cm21 correspond to the Si–OR bond.11 As was seen in a spectrum of a thermally treated film at 450 °C, there is a loss of organic groups which is considered to be the conversion from Si–OR to Si–O–Si bonds that forms a more complete SiO2 network. The sharp peak at
A. Chemical aspects of the SiO2 aerogel film
In Fig. 1, typical RBS spectra obtained from both asdeposited and thermally treated at 450 °C films are shown. The composition of films, e.g., ratios of O/Si and C/Si, could be deduced by comparing the yield heights of the elements in spectra.9 The atomic ratio of Si:O:C was measured to be 1:2.55:0.9 for the as-deposited film and 1:2.20:0.3 for the thermally treated film. The carbon content in the films obviously decreased to less than half for the thermally treated film. These carbon species are thought to come from the partially hydrolyzed TEOS and residual IPA used as the precursor and the solvent for film preparation. Also, the O/Si ratio was reduced, closer to 2, for the thermally treated film and is considered to be a result of the reduction of the organic portion. The O/Si and C/Si ratios can be a useful parameter by which to judge the films stability to moisture absorption because additional oxygen ~the stoichiometry of O/Si is two for ideal SiO2! constitutes Si–OH and Si–OR bonds on the internal pore surface other than the skeletal SiO2 network. And these surface chemical species can result in the degradation of the film reliability.10 The IR spectra of the films are shown in Fig. 2. The most intense absorption peak of the Si–O bond that constitutes the skeletal SiO2 net1300
J. Appl. Phys., Vol. 82, No. 3, 1 August 1997
FIG. 2. FTIR absorbance spectra of SiO2 aerogel films: ~a! as deposited and ~b! thermally treated at 450 °C. Jo et al.
FIG. 4. XPS narrow scan spectra of the SiO2 aerogel film with the function of heating temperature: ~a! O 1s and ~b! Si 2 p narrow scan spectra.
FIG. 3. XPS spectra of the SiO2 aerogel film with the function of heating temperature: ~a! wide scan spectra showing all elements present and ~b! C 1s narrow scan spectra.
940 cm21 and the broad region between 3200 and 3800 cm21 correspond to a stretching of Si–OH and physisorbed moisture on the surface in several modes, respectively.12 The evolution of single Si–OH by thermal treatment is also shown. Considering that one of the main by-products created by thermal decomposition to the Si–OH bond is C2H4~Si–OC2H5→Si–OH1C2H4!, and that most of physisorbed moisture is bridged by the hydrogen bond to the internal surface, the Si–OH bond is thought to be established by pyrolysis of partially hydrolyzed TEOS and the removal of physisorbed moisture. Figure 3~a! shows typical wide scan spectra that survey all the elements present on the film. There was no Cl or N species that was used as the catalyst in the sol-gel solution stage. Thermal behavior of the film was monitored during the in situ anneal treatment. Vacuum drop down was observed particularly during the anneals at 150–200 and 330– 380 °C, and corresponds to the evaporation of physisorbed moisture and the pyrolysis of residual carbon, respectively.13 The drastic decrease of carbon is easily seen in the wide scan spectrum obtained after the anneal at 450 °C. But, with the evaporation of physisorbed moisture, we could not observe any intensity difference of oxygen in between the spectra. This may be explained as follows. During the introduction of the sample into the ultrahigh vacuum ~UHV! analysis chamber, the physisorbed moisture on the uppermost surface is evaporated. But by annealing of the sample, any remaining physisorbed moisture at the inner surface of the sample could be evaporated and cause the vacuum drop down. Due to the surface sensitive characteristic of XPS, we could not observe any oxygen intensity difference. A similar phenomenon may J. Appl. Phys., Vol. 82, No. 3, 1 August 1997
have been observed for carbon. We could not observe carbon on the surface of the film by XPS analysis after the anneal at 450 °C, but it was still detected by RBS and IR analyses. In Figs. 3~b! and 4, there are typical narrow scan spectra of each elements. The resultant peak positions of the elements were corrected with the Au 4 f 7/2 peak as an external standard. In general, all the spectra show broad and asymmetric lines, indicating overlaps of more than one bonding state. In the C 1s spectra seen in Fig. 3~b!, at least two bonding states are observed, one is the typical hydrocarbon ~C–C/H! and the other with higher binding energy is carbonyl species ~C– O!. Comparing these to crystalline SiO2 such as pure quartz,14 the SiO2 aerogel film exhibited a particularly lower binding energy of Si 2p by 0.6 eV while the position of O 1s is relatively comparable. Considering the chemistry of the film synthesis and the IR analysis discussed above, Si 2p and O 1s peaks would exist both as Si–O–Si for the skeletal network and as Si–OR or Si–OH for the internal surface coverage. Specific deconvolution into each component in Si 2 p and O 1s was not attempted yet because the peak positions of these constituents are slightly split and they also did not show any detectable spectral change during thermal treatment. Instead, the Si 2 p and O 1s peaks were qualitatively evaluated in terms of their centroid peak positions. The peak position of Si 2 p is lower than that of quartz which is composed of regular @ SiO4# tetrahedra indicating a somewhat different siloxane ~Si–O–Si! environment in the SiO2 network structure of the sample. The Si 2 p core line is known to be more sensitive to this siloxane environment in various O/Si stoichiometries15 than the O 1s core line because it is greatly influenced by the more electropositive environment. The oxygen content ~O/Si 5 2.55 and 2.20, for the as-deposited and thermally treated films, respectively, as discussed above! was high enough to establish the complete SiO2 network. However, a considerable amount of oxygen contributes to the internal surface species, such as Si–OR and Si–OH, and this surface contribution causes a different siloxane environment with different bonding angles and disJo et al.
1301
FIG. 7. Particle size distribution on the SiO2 aerogel film surface.
B. Physical aspects of the SiO2 aerogel film
FIG. 5. Cross-sectional SEM photographs of the SiO2 aerogel films: ~a! as deposited and ~b! thermally treated at 450 °C.
tances in the SiO2 aerogel. This assumption is consistent with the behavior of the Si 2 p line during thermal treatment above 450 °C and further treatment at 650 °C. Below 450 °C, with the drastic reduction of residual carbon as C–C/H and C–O associated with Si in the SiO2 network forms, the binding energy of Si 2 p was shifted to a higher binding energy of the more complete SiO2 network. A further peak shift after a subsequent thermal treatment at 650 °C was observed. The overall chemical structure of the SiO2 aerogel film can be depicted as SiO2 aerogel is composed of an inorganic SiO2 network and internal surface coverages such as Si–OH or organic Si–OR. During the thermal treatment, organic Si–OR was effectively eliminated, whereas Si–OH is evolved, resulting in a more ideal SiO2 network formation.
In Fig. 5, cross-sectional SEM photographs of asdeposited and thermally treated SiO2 aerogel films at 450 °C are shown. The thickness of the as-deposited film was measured at 1.2 mm but after thermal treatment at 450 °C, film thickness was decreased by 10%. The surface morphology of the films is shown in Fig. 6, a highly randomly crosslinked structure of nanoscale particles is illustrated along with three-dimensional projected AFM images. The increase in surface particle size after the thermal treatment can be observed on 1 mm film surface area. The convoluted surface particle size distribution for more than 200 particles with line profiling is presented in Fig. 7. The actual particle size with AFM measurements was determined by considering the relationship between the pyramid shaped tip with a 10 nm diameter and soft surface particles of the SiO2 aerogel and is illustrated elsewhere.16,17 Their average size range was 10–30 nm. After thermal treatment, the particle size increased and its distribution became more uniform which resulted in more contact between the particles. This induced a decrease of root mean square roughness from 4 to 3.2 nm. Considering the combined chemical evolution during ther-
FIG. 6. AFM images of the SiO2 aerogel films with the corresponding SEM surface images: ~a! as deposited and ~b! thermally treated at 450 °C. 1302
J. Appl. Phys., Vol. 82, No. 3, 1 August 1997
Jo et al.
FIG. 9. Leakage current density (J) with the function of applied electric field (E) of the SiO2 aerogel films. The inset displays the log J vs log E plot.
FIG. 8. Dielectric constant of the SiO2 aerogel films in a MIS structure. The square of the refractive indice is shown for two extreme pore distribution models.
mal treatment as discussed above, i.e., the removal of organic species ~Si–OR! and moisture (H2O) on the particle surface followed by a polycondensation reaction forming the siloxane bond ~Si–O–Si!, the increase of particle size and smoother surface of the film can be explained. At the same time, the decrease in thickness with thermal treatment can be explained by the same reason. The density of the film was determined by RBS analysis, coupled with the measured film thickness by cross-sectional SEM observation. Stopping cross-section factor, @ e 0 # of each element was calculated by normalizing its value to the concentration of each element in the film for Si–O–C system, considering the significant amount of residual carbon content. The resultant density was in the range of 0.59 and 0.68 g/cm3 for both the as-deposited and the thermally treated films. The corresponding film porosities were then calculated to be 74% and 70%, respectively. For the porosity calculation, we have referenced thermally grown SiO2 which has 2.27 g/cm3 density as a measure for the degree of densification of the film. Even though the film thickness decreased during thermal treatment, the film density remains almost constant. And from this phenomenon we realized that the elimination of residual carbon and moisture was directly associated with the formation of a siloxane bond in the network as much as it was to their reduced amounts. C. Electrical aspects of the SiO2 aerogel film
The measured average dielectric constants in the 1 MHz region were 2.1 and 2.0 for the as-deposited and thermally treated films, respectively, in the MIS structure ~Fig. 8!. The polarization of the SiO2 aerogel film is made up of electronic, ionic, and dipolar constituents. The measured dielectric constant was compared to the square value of the refractive indices which reflect only electronic polarization because the optical refractive indice is measured in the visible light range with a 632.8 nm He–Ne laser. It was also J. Appl. Phys., Vol. 82, No. 3, 1 August 1997
compared to the calculated value considered from porosity. A model for an ideal mixture of more than two dielectrics can be expressed as k n 5 ( V i • k i n , 2 1 < n < 1; one is that the main orientation of pore distribution is parallel to the applied field when n 5 1 8 , the other is perpendicular to the applied field when n 5 21. For both the as-deposited and thermally treated films, the calculated dielectric constants from the refractive indices were in between the two extreme cases indicating a complete random pore distribution in the SiO2 aerogel film. Also, any deviation of measured dielectric constant from either that of the ideal mixture model or the calculated value from the refractive indices can be qualitatively understood as a contribution by additional dipolar polarization which is caused by surface chemical species on the internal pore surface. As can be seen in XPS and IR analyses results, the internal surface is covered mainly by hydroxyl and residual organics that result in additional polarization besides the Si–O skeletal contribution because these are highly polarizable species, especially in the 1 MHz regime. The decrease of the dielectric constant during thermal treatment can be also explained by this effect. Therefore, we concluded that an inherent low dielectric constant of the SiO2 aerogel is guaranteed by its high porosity; however within a certain range of porosity it was sensitive to the chemical species of the internal surface coverage. Leakage current characteristics through the SiO2 aerogel film at up to 1 MV/cm field is shown in Fig. 9. Both the as-deposited and thermally treated films were dehydrated in Ar atmosphere at 150 °C before the measurements to minimize the effect of physisorbed moisture in the film. The overall characteristic of leakage current density (J) can be depicted in a similar fashion as that as the applied field (E) is increased, the leakage current increased parabolically up to a saturation point. In the inset, plot of log E vs log J shows linear behavior in the low field region and nonlinear behavior in the high field region. This linear behavior can be explained in terms of charge carrier traps, indicating that there is a small amount of charge carrier traps in spite of the very porous microstructure of the dielectrics. The onset of nonlinear conduction behavior was shifted to a higher electrical field for the thermally treated film. Nonlinear behavior Jo et al.
1303
that the film porosity and density remain at almost the same value, we could conclude that the leakage current of the SiO2 aerogel film is strongly dependent on the amount of residual organics. IV. SUMMARY
FIG. 10. Leakage current density distribution at 1.5 V of the SiO2 aerogel films.
at higher field can be interpreted by several mechanisms, namely, space-charge-limited conduction, field-enhanced Schottky emission, or Poole–Frenkel electron emission inside the dielectric.18,19 More specific studies will be devoted to clarifying this point. Instead, I–V characteristics were evaluated in terms of leakage current value according to the amount of residual carbon content. In Fig. 10, the leakage current distribution at 1.5 V that would be the operational voltage in the subquarter-micron regime is shown. Better leakage behavior was achieved by a thermal treatment. This improvement seems to result from a reduction of residual carbon content. To consolidate this relation, films were analyzed with 4.26 MeV He11 ion resonance RBS to quantify the total carbon content in the films. Compared to 2.0 MeV He11 ion RBS which represents a relatively weak carbon peak because of the low mass magnitude, with resonant RBS the scattering cross section of carbon is greater and we can observe the dramatic change of its content with the thermal treatment. From a comparison of Figs. 10 and 11, a close relationship between leakage current and the amount of residual carbon contained in the film is strongly suggested. Two major chemical species of the surface coverage in SiO2 aerogel films, such as Si–OH and Si–OR, are known to influence the leakage current behavior.20,21 Considering that during the thermal treatment free Si–OH is evolved and also
FIG. 11. The residual carbon content in resonant RBS spectra of the SiO2 aerogel films: ~a! C/Si 5 0.9 for as deposited and ~b! C/Si 5 0.3 for those thermally treated at 450 °C.
1304
J. Appl. Phys., Vol. 82, No. 3, 1 August 1997
SiO2 aerogel film was successfully fabricated and showed a very low dielectric constant which is inherent to the material. Since SiO2 aerogel film is fabricated from an organic precursor ~TEOS! in an alcohol solvent ~IPA!, the residual carbon content, which covered a large internal surface as Si–OR, was not negligible and it influenced the electrical properties. During thermal treatment, most residual organics were eliminated, leaving the siloxane bond, and in the meantime more Si–OH developed. This subsequent polycondensation reaction directly resulted in the shrinkage of film thickness and the growth of particle size. Within the finite density range, dielectric properties were very sensitive to the organic surface coverage due to the additional dipolar polarization and were improved by thermal treatment. The dielectric constant of as-deposited SiO2 aerogel film was 2.1 and it decreased to 2.0 for the thermally treated film with the same tendency of leakage current characteristic. ACKNOWLEDGMENT
The authors wish to express their thanks for the financial support of this research by the Electronics and Telecommunications Research Institute ~ETRI! in Korea under Contract No. 96398. 1
S. P. Jeng, R. H. Havemann, and M. C. Chang, Mater. Res. Soc. Symp. Proc. 25, 337 ~1994!. 2 C. B. Case, A. Kornblit, K. S. Y. Lau, and N. H. Hendricks, Proc. ISMIC Conf. 104, 116 ~1995!. 3 S. P. Murarka, Solid State Technol. 39, 83 ~1996!. 4 L. W. Hrubesh and J. F. Poco, J. Non-Cryst. Solids 188, 46 ~1995!. 5 L. W. Hrubesh, L. E. Keene, and V. R. Latorre, J. Mater. Res. 8, 1736 ~1993!. 6 M. H. Jo, J. K. Hong, H. H. Park, J. J. Kim, and S. H. Hyun, Microelectron. Eng. 33, 343 ~1997!. 7 D. J. Kim and S. H. Hyun, J. Korean Ceram. Soc. 33, 485 ~1996!. 8 W. K. Chu, J. W. Mayer, and M.-A. Nicolet, Backscattering Spectrometry ~Academic, Orlando, 1978!, p. 86. 9 L. R. Doolittle, Nucl. Instrum. Methods Phys. Res. B 9, 334 ~1985!; 15, 227 ~1986!. 10 B. K. Hwang, J. H. Choi, H. J. Shin, U. I. Chung, S. I. Lee, and M. Y. LEE, Proc. ISMIC Conf. 104, 113 ~1995!. 11 T. Homma and Y. Murao, Proc. ISMIC Conf. 101, 65 ~1992!. 12 P. L. Pai, A. C. Chetty, R. Root, N. Cox, and C. Ting, J. Electrochem. Soc. 134, 2829 ~1987!. 13 T. Woignier and J. Phalippou, J. Mater. Sci. 25, 3118 ~1990!. 14 J. Chastain, Handbook of X-ray Photoelectron Spectroscopy ~PerkinElmer, Eden Prairie, MN, 1992!, p. 231. 15 P. Kelfve, B. Blomster, H. Siegbahn, and K. Siegbahn, Phys. Scr. 21, 75 ~1980!. 16 P. D. Dietz, P. K. Hansma, O. Inacker, H. D. Lehmann, and K. Heinz, J. Membrane Sci. 65, 101 ~1992!. 17 W. R. Bowen, N. Hilal, R. W. Lovitt, and P. M. Wiliam, J. Membrane Sci. 110, 233 ~1996!. 18 C. S. Hwang, S. O. Park, H. J. Cho, C. S. Kang, H. K. Kang, S. I. Lee, and M. Y. Lee, Appl. Phys. Lett. 67, 2819 ~1995!. 19 H. Y. Lee, K. C. Lee, J. N. Scunke, and L. C. Burton, IEEE Trans. Compon. Hybrids, Manuf. Technol. CHMT-7, 443 ~1984!. 20 M. Matsuura, Y. Hayashide, H. Kotani, and H. Abe, Jpn. J. Appl. Phys. 1 30, 1530 ~1991!. 21 T. Homma and Y. Murao, J. Electrochem. Soc. 140, 2046 ~1993!.
Jo et al.
