This article is one of five papers to be presented exclusively on the web as part of the October 2000 JOM-e the electronic supplement to JOM.
	The following article appears as part of JOM-e, 52 (10) (2000), http://www.tms.org/pubs/journals/JOM/0010/Rosenthal/Rosenthal-0010.html JOM is a publication of The Minerals, Metals & Materials Society

Infrared Analysis of Advanced Thin Film Materials

TABLE OF CONTENTS
INTRODUCTION EXPERIMENTAL ULTRATHIN OXIDES PHOTORESIST MEASUREMENTS CONCLUSIONS ACKNOWLEDGEMENTS References

The technique of Fourier transform infrared (FTIR) reflectometry has been advanced to characterize the thickness and optical properties of thin films commonly used in advanced integrated circuits (I.C.s). This article describes the demonstration of a new, high accuracy reflectometer to characterize the reflectance of ultrathin gate oxides and chemically amplified deep ultraviolet (UV) photoresist thin films. The gate oxide reflectance data were related to the deposition time to model the thermal oxidation growth kinetics. Model based analysis was employed to extract the dielectric function and thickness of the photoresist layers. Changes in the absorption features in the dielectric function were related to the exposure dose.

INTRODUCTION

As the semiconductor industry moves to develop integrated circuits (I.C.s) with smaller feature sizes, faster switching speeds, and lower power consumption, the materials employed in the basic wiring, dielectric, and photolithographic layers are changing dramatically. In particular, the paradigm of aluminum/silicon dioxide interconnect technology, diffused, implanted, or epitaxial active silicon layers, and mercury line photoresists (a paradigm that has lasted for several decades with little change) is giving way to a new I.C. architecture employing copper/low-k interconnects, silicon-germanium and silicon on insulators based transistor structures, chemically amplified deep ultraviolet (UV) and x-ray lithography, and new metal-silicide ohmic contact materials. One unifying theme shared by these new materials is the greater complexity and importance of composition in controlling their properties.

The goals of timely and cost effective integration of these new materials into mainstream manufacturing has motivated the introduction of new metrology and process control practices focused on meeting new specs on layer composition as well as thickness. Furthermore, to meet the industry wide goals articulated by National Technology Roadmap for Semiconductors (NTRS), the roles of both old and new metrology are expected to shift from interactive off-line and stand-alone measurement configurations to automated in-line and in situ configurations. This shift is needed to enable overall increased automation of process control tasks such as run-to-run control, fault detection and classification, as well as to decrease the reliance on non-product wafers used for process control and quality assurance. Ultimately, the goal is to employ non-destructive measurements on every product wafer as a means to gather data to control the process.

Ultraviolet and visible (UV-VIS) reflectometry and ellipsometry has emerged over the last decade as the most widely accepted method for production monitoring of transparent thin films. Their excellent accuracy, sensitivity, and ability to extract multiple parameters from a film stack have made them widely applicable for many film processes. Nevertheless, UV-VIS techniques are insensitive to compositional effects, and often suffer from complications associated with roughness induced scattering, sample-to-sample variations in crystallinity and optical properties, and correlations of parameters such as thickness and index of refraction for the thinnest samples. In light of the increased interest in controlling layer composition, in structures that often exhibit significant microroughness and variations in microstructure, these issues are becoming increasingly important.

Infrared spectroscopy offers a metrology approach, complementary to UV-VIS techniques, that provides excellent sensitivity to layer composition, including chemical bond densities (through their vibrational mode intensities), and free carriers, with enhanced immunity to roughness induced scattering. Because it can be implemented as a reflectance sensor, infrared spectroscopy shares many of the inherent advantages of UV-VIS spectroscopy as a non-destructive process control tool.

We have recently developed a new FTIR reflectometry tool optimized for the characterization of films with complex chemistry. This new metrology tool employs two recent inventions: a high sensitivity optical reflectometer design that can measure thin films on transparent substrates while suppressing backside reflected light that would otherwise interfere with the analysis of the front surface reflections; and model-based fitting to extract the dielectric function of a layer from the reflectance spectrum, thereby separating out the compositional information from interference fringes and substrate related artifacts. The mid infrared dielectric function, typically expressed in terms of the complex refractive index n +ik, encodes a great deal of chemical information related to chemical bond related vibrational modes, porosity, stress, and doping, in the form of free-carrier absorption.

EXPERIMENTAL

Samples of ultrathin thermal oxides and deep UV photoresist were studied using the Fourier transform infrared (FTIR) system. Reflectance spectra were acquired using a system incorporating the FTIR reflectometer equipped with a linearized liquid nitrogen detector. Data were collected at an angle of incidence of 47°, with an s polarization fraction of 0.43, in the spectral range between 600 cm^-1 and 5,000 cm^-1. Reflectance data were referenced to lightly doped silicon wafer samples. Overall, the instrument drift was observed to be less than 1% over a 24-hour period, and with frequent references, reflectance data were reproducible in the short term to better than 0.1%. As inputs to the model-based analysis software, the angle of incidence and polarization of the probing beam were carefully calibrated using samples of silicon and silicon dioxide.

After the spectra were collected, model-based (curve fitting) thin film algorithms were applied to analyze the data to determine layer thicknesses and optical constants.^1-5 The data of optical constants were then related to the layer composition or process parameters. Employing a model for the film stack and the DF (optical constants) of each of the layers in the film stack, the software performs a transfer matrix calculation of the reflectance.^6,7 The optimization or fitting algorithm then iteratively varies the model parameters such as layer thicknesses, carrier concentrations, index of refraction, and other dielectric function (DF) parameters, recomputing the simulated reflectance until the weighted mean square difference is minimized between the measured and simulated spectra.

The film stack model provides a parameterization of the structure that contains a list of layers and compositions of the materials in each layer. The model may include regions with abrupt and graded composition profiles. If present, graded layers are approximated as segmented stacks of uniform layers, each slightly different than its neighbors. The reflectance model computes a simulated reflectance from the segmented layer stack model.

During fitting, the dielectric function models compute the frequency dependent optical constants n and k from the layer compositions prescribed in the film stack model. For particularly fast and accurate fitting with only a few free parameters, constrained dielectric function models are tailored to the optical behavior of the materials. For reflectance analysis, silicon, for example, is nearly perfectly modeled in the infrared by a Drude term for the free carriers, and a smooth weakly varying background dielectric function. While most dielectrics, such as silicon dioxide, silicon nitride and photoresist have relatively simple, compositionally insensitive Cauchy dielectric functions in the visible,⁸ in the infrared these materials generally exhibit significant chemical variability, encoded in the vibrational absorption band structure. To analyze these compositionally complex materials in the infrared, a more general approach is required to model and extract dielectric functions of materials from reflectance spectra. In this method, the dielectric function of the layer is modeled with a basis set of damped harmonic oscillators, i.e. as a sum of Lorentzian terms (SOL), closely spaced in frequency, with equal damping constants and spacing. The arrays of oscillators are located in the spectral regions, where absorption is expected in the film. The width and spacing of the Lorentzians set describing these oscillators are fixed at values that depend on the desired spectral resolution. During the fit, the amplitudes of the oscillators, high frequency dielectric constant, and layer thickness are varied to fit the model to the measured data. This procedure provides a robust, well-conditioned extraction of a dielectric function of a layer, independent of the layer thickness, using only the reflectance spectrum. By virtue of the properties of the Lorentzian oscillator basis set, the solutions are implicitly Kramers-Kronig consistent, and can model a DF of arbitrary shape or complexity. This combination of powerful algorithms and accurate reflectometry data provides the basis for analyzing a wide variety of thin film materials.

ULTRATHIN OXIDES

Figure 1. Measured and fitted reflectance of 4.9 nm oxide film.		Figure 2. Magnified view of the Si-O stretch spectral region.		Figure 3. SiO2 thickness measured by FTIR for a series of ultrathin thermal oxides on heavily doped substrates as a function of deposition time, overlaid with a power-law fit. The open square symbol is a suspected misprocessed wafer.

A series of gate oxide films were acquired from the New Jersey Institute of Technology (NJIT). The films were fabricated in a rapid thermal processor with deposition times of 100, 200, 300, 400, 500, and 600 seconds; all processed at the same temperature. The substrates were heavily doped p+ silicon. Figure 1 shows the measured and fitted reflectance of the 100-second sample. The spectra were fit to a multiparameter model in which the substrate carrier concentration and scattering rate, as well as the oxide thickness, were allowed to vary during the analysis. Literature values for the oxide DF were employed for these fits.⁹ The oxide layer thickness was encoded mostly in the optical vibrational modes in the band around 1,100 wavenumbers, while the substrate carrier concentration affected the spectrum broadly, with a dramatic plasma edge visible around 700 wavenumbers. Figure 1 shows that the fitted spectrum is virtually identical to the measured data. The fits were similarly excellent for all the wafers, indicating that the films were high quality oxide layers, and that the analysis model employed an accurate description of the optical properties of both the substrate and film.

Figure 2 shows an expanded view of the absorption region of the spectrum around 1,000-1,300 wavenumbers. The SiO₂ film has a transverse vibrational mode at 1,070 wavenumbers whose amplitude is indicative of the surface density of silicon-oxygen bonds. An associated longitudinal mode is also visible at 1,250 wavenumbers. The longitudinal mode position is much more dependent on film stress and density than the transverse mode and can be considered an indicator of film quality. Figure 2 shows that the measured LO mode is shifted to slightly higher frequency than that observed in the fitted reflectance, which was calculated using bulk thermal oxide optical constants. This indicates the possible presence of stress in the film perturbing the index of refraction.

The extracted SiO₂ layer thicknesses, summarized in Table I, ranged from approximately 5 nm to 12 nm. A repeatibility study on several of the samples established the precision (1 sigma) of the measurement to be around 0.7 Å. The SiO₂ thickness vs. deposition time is plotted in Figure 3, along with a fit to a power law. This fit resulted in an exponent of 0.498, a nearly perfect square root law. This square root dependence on time is commonly observed in diffusion limited processes such as thermal oxidation. All but wafer #5 fell within 0.2 nm of the trend line. This sample, labeled as a 500-second growth, was nearly identical to the 400-second sample.

PHOTORESIST MEASUREMENTS

Figure 4. Extracted index of refraction and extinction coefficient for the UV5 resist as a function of exposure level.

A chemically amplified Deep UV (DUV) photoresist sample was analyzed to explore the ability to characterize an organic thin film with extremely complicated photochemistry and optical constants. A wafer was obtained from National Semiconductor with a layer of Shipley UV5 photoresist deposited, exposed and post-baked according to the manufacturer's recommendations. A series of 1-cm fields on the wafer was exposed with different doses to investigate the effect of exposure dose on the infrared properties of the resist. Each of the fields was measured using the infrared reflectometer, and analyzed for the resist dielectric function using the SOL based dielectric function extraction algorithm. The DF data in Figure 4a, Figure 4b, Figure 5a, Figure 5b, Figure 5c, and Figure 5d are labeled with increasing label numbers for decreasing doses. Figure 4a and Figure 4b show the extracted index of refraction and extinction coefficient, while Figure 5a, Figure 5b, Figure 5c, and Figure 5d show the expanded spectral regions that exhibited the most dramatic chemistry-related changes. The figures show clear systematic variations with exposure dose, both in the detailed absorption bands as well as in the high frequency index of refraction. The reduction of the band around 1,150 cm^-1 corresponds to -CO groups reacting and producing -OH groups, for which a corresponding increase is observed in the 3,100-3,500 cm^-1 region. Changes in the 1,700 cm^-1 region have been attributed to the deprotection reaction involving ester bonds reacting with the photo-generated acid to yield carboxylic acids.¹⁰

Figure 5. Variations in chemistry for UV5 resist.

CONCLUSIONS

By combining model-based infrared spectral analysis with high performance reflectometry hardware , it is possible to extract quantitative data on multiple parameters related to film properties. The technique has a unique sensitivity to film composition, which is broadly applicable to a wide range of films including ultrathin oxides, doped semiconductors, and complex materials such as photoresists and low-k dielectrics. With wider application, FTIR could significantly enhance the development, introduction, and production of the new materials currently being considered for future generations of high performance I.C.s. More work is clearly needed to develop a deeper understanding of the relationship between the infrared optical properties of thin films and their process and performance parameters.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge the support of the NSF SBIR grant No. DMI-9631216 and No. DMI-9860798, and Army SBIR Grant No. DAAH04-C-0090 for the generous support for the R&D programs that generated these results.

References

1. S. Charpenay et al. "Model-based Analysis for Precise and Accurate Epitaxial Silicon Measurements," Solid, State Technology (July 1998).
2. B.W. Fowler et al., "The Measurement of Sub-micron Epitaxial Layer Thickness and Free Carrier Concentration by Infrared Reflectance Spectroscopy," Proc. Electro-Chemical Soc., 94-33 (1994), p. 254.
3. S. Liu et al., "FT-IR Spectroscopy as a Process Monitor for Integrated Circuit Manufacturing," SPIE (January 1993).
4. Peter A. Rosenthal, "Integrated FTIR Reflectometer Controls Semiconductor Fabrication Process," Laser Focus World-Design and Applications, 34 (April 1998), pp. 173-176.
5. T. Buffeteau and B. Desbat, Applied Spectroscopy, 43 (6) (1989), pp. 1027-1032.
6. F. Abeles, Advanced Optical Techniques (Amsterdam: North-Holland, 1967).
7. K. Yamaoto and H. Ishida, Applied Spectroscopy, 48 (7) (1994), p. 775.
8. E. Hecht and A. Zajac, Optics, 2nd ed. (Reading, MA: Addison-Wesley Publishing Co., 1987).
9. E. Palik, Handbook of Optical Constants of Solids (New York: Academic Press, Inc., 1985).
10. N. Jakatdar et al., "Novel Metrology for the DUV Photolithographic Sequence," Characterization and Metrology for ULSI Technology, Seiler et al. (College Park, MD: The American Institute of Physics, 1998).

P.A. Rosenthal, J. Xu, and S. Charpenay are with On-Line Technologies, Inc.; J.E. Cosgrove is with Advanced Fuel Research, Inc.; and N.M. Ravindra is with the New Jersey Institute of Technology.

For more information, contact Peter Rosenthal, On-Line Technologies, Inc., 87 Church Street, East Hartford, Connecticut 06108; (860) 291-0719; fax (860) 289-7975, e-mail prosenth@online-FTIR.com or e-mail N.M. Ravindra at nmravindra@home.com.