Characterization and in-line monitoring of low-k porogen formation
12/01/2007
EXECUTIVE OVERVIEW
The competitive requirement for increased device performance has compelled the IC industry to introduce lower dielectric constant (k) materials. The move away from conventional oxide deposition techniques has significantly increased the complexity of the manufacturing process. There are a number of methods that can be used to achieve low-k films, including the controlled introduction of porosity. Here the dielectric constant is lowered in relation to the film’s decrease in bulk density. Mass analysis can quantitatively monitor the UV cure process for the production of porous low-k materials.
As ULSI design rules shrink below 90nm, the manufacturing challenges related to processing low-k materials becomes significant. One such challenge is to accurately understand the transition from a dense as-deposited composite to a stable porous low dielectric constant material with the desired mechanical and electrical properties.
Experiment: Pre-film cure analysis
The low-k materials used for this study were deposited in PECVD mode using Air Products and Chemicals’ DEMS low-k precursor as the structure former in conjunction with an organic porogen precursor [1]. The relative concentration of precursors, along with the chosen deposition conditions, determines the level of porosity incorporation and final film properties.
All depositions were performed on an Applied Materials Precision 5000 in a 200mm DxZ chamber fitted with a TEOS-oxide process kit. The process conditions used were 275-300°C, 8-10Torr, 600-750W RF (single frequency, 13.56MHz), 0.35-in. gap between the substrate and showerhead/electrode, 200sccm CO2 carrier, 0-25sccm oxygen, and total chemical flow rate of 700-1100 sccm.
The dense composite films were treated by exposure to a Fusion H+ ultraviolet lamp (model I-600) in order to render them porous. Exposures were performed with the Fusion lamp system retrofitted onto an Applied P-5000 DxL chamber. The chamber was modified by replacing the standard lid with a synthetic silica window to allow for transmission of radiation. The bulb was positioned in the reflector to provide collimated UV light with an approximate exposure region of 2 in. × 10 in. The UV bulb was cycled horizontally across the surface of the film at a rate of ~10 sec/cycle with the wafer held at ~400°C and vacuum ambient.
Process monitoring is relevant to both the deposition stage and the treatment stage (i.e., the generation of porosity) of porous dielectric film production. After deposition, the film exists as a composite of silicate and organic material. This composite must be optimized and monitored for reproducibility. Here we evaluate a 25-wafer lot with a nominal composite film thickness of 6000Å (Fig. 1a). Under this set of processing conditions, the film thickness stability is excellent with a process capability well within the range of ±5% for the targeted value.
The corresponding mass stability for the same 25-wafer lot (Fig. 1b) is excellent, with a process capability well within the ±5% mass limits. By relating the film mass to the corresponding film thickness we can directly determine the film density for the porogen-containing film without the need for complex modelling [4]. The density measurements are in excellent agreement with those determined by x-ray reflectivity.
The stability of the film density as derived from ellipsometry and Metryx mass measurement provides a rapid and effective means to monitor the stability of the porogen-containing film. This can be used to ensure the porogen content of the film is consistent pre-cure. The film stability for thickness, mass and density are comparable with a 3σ value of approximately 1% of mean (Fig. 1c). Characterization of the composite film as deposited allows for a meaningful study of the subsequent UV curing process.
Experiment: UV exposure and porogen removal
After the composite film is deposited, the second stage of processing involves removal of the porogen species through exposure to an appropriate energetic source such as UV radiation. Multiple areas of interest can be studied by monitoring mass change during the UV exposure step. Optimization of film properties requires knowledge of the effects of UV exposure time and wavelength range. Figure 2a shows the mass reduction as a function of UV exposure. As the porogen is removed from the film, the mass is reduced. The rate and quantity of porogen removed is directly determined with mass.
In this section, the cure process used the Axcelis RapidCure 320FC UV tool. The process uses microwave driven electrode-less bulbs emitting UV radiation in the 100-400nm range under inert atmosphere, with the wafer temperature maintained at 400°C. The bulbs can be easily changed to allow use of different wavelength distributions. Previous studies have shown that UV exposure has a simultaneous strengthening effect on the silicate lattice, through selective removal of terminal groups [2].
The first experimental set was designed to quantify the mass of porogen removed from the film with increasing exposure time to UV source ‘B’ (Fig. 2a). Asymptotic mass loss with respect to UV exposure time is observed. The reduction in overall mass of the film is significant, with approximately 20% of the film mass lost as a result of UV curing after 13 minutes of exposure. The film reaches a near steady state with respect to mass loss after ~3 min. The use of mass loss measurements in this way can effectively show the ‘end-point’ of porogen loss and maximum porosity under the cure conditions used. While other metrology techniques such as FT-IR, XRR, and ellipsometric porosimmetry (EP) can be used to quantify porogen loss, the ability to measure mass accurately and quickly can provide unique information.
Figure 2b compares the shrinkage with mass loss, where continual film shrinkage with UV cure time gives no indication of the extent of porogen removal. Techniques such as FTIR can be used to follow porogen removal via -CH3 or -CH loss. However these signals are dependant on porogen type used, and can often be confounded with desirable Si-R lattice moiety.
Figure 3 shows changes in the C-H/Si-O peak area ratio, which in this case, is made up of porogen C-H and lattice SiC-H3 signals. This ratio initially drops rapidly due to porogen removal, but shows a more gradual constant loss due to UV-induced Si-R network cleavage with continued exposure.
The rate and final magnitude of porogen loss is dependant on the particular wavelength chosen. In the second part of this study, four different UV broadband sources were evaluated to determine the magnitude of mass reduction of the porogen-containing film. The exposure time was fixed at 13 min for all four UV sources (see table).
 |
Clearly, different wavelengths have a significant impact on the total mass of porogen removed from the film. In this case, we see that Wavelength ‘A’ removes just over 17% of the mass of the original film, while both C and D remove approximately 23% of the original porogen-containing film.
The table compares this bulb dependant mass loss with the dielectric constant obtained post cure. The clear correlation between mass loss and k values demonstrates the need for UV source optimization for the low-k cure process. The choice of wavelength results in significantly different porogen removed from the film.
 Figure 4. Dielectric constant (k) post-curing correlates well to the mass loss for porous low-k films. |
A general relationship between k value and mass removed from the film can be observed across all collected data. There is the strong correlation between mass removed from the film and post-cure k value (Fig. 4).
Although the mass removed from the film is strongly correlated to the desirable reduction in k value, it is not correlated to the nano-indentation modulus. Therefore the goal of achieving a low-k value and a film with the desired mechanical properties are not mutually exclusive. Figure 5 shows that a variety of modulus values are achieved at a k value of 2.55. Therefore there is some variability in k value and modulus, known to be controlled by the final chemistry and structure of the film [3], which may also be altered by the wavelength of UV exposure during the cure step.
Optimization of the intensity and wavelength distribution of the UV light source may lead to significantly improved cure performance.
Such UV optimization is needed due to the complex nature of the cure process, which involves simultaneous porogen removal, chemical changes to the network, and film densification. Each of these factors is intimately related to the mechanical, electrical, and integration properties of the final film.
Clearly porogen removal is a function of UV exposure intensity, wavelength, and time exposed. Remaining porogen can result in higher film k values, poorer film stability and degraded electrical properties. In-line monitoring on product is critical to ensure consistent low-k properties. A repeatable mass reduction is expected for a nominal deposition thickness on product wafers. This mass reduction can be normalized for different film thickness by expressing the mass change as a percentage of the deposited porogen-containing film. The development of reliable dielectric materials can be enhanced through the use of highly accurate mass measurement techniques.
Conclusion
The advantages of using mass change to monitor porogen-containing films and porogen removal are clear. Mass change provides an absolute, direct method to monitor porogen formation and avoids the complexity associated with other techniques. These advantages extend to both the process development phase along with high sample rate in-line monitoring of the product [5]. This method is applicable to a wide range of low-k materials, which achieve porosity through a curing process that removes mass from the intermediate state of the process.
Acknowledgments
The authors would like to thank Dr. Adrian Kiermasz and Barry Tomkins of Metryx Ltd. for their support and contributions to this article. DEMS is a trademark of Air Products and Chemicals Inc.
References
- R.N. Vrtis, M.L. O’Neill, J.L. Vincent, A.S. Lukas, M. Xiao, J.A.T. Norman, US Patent 6,846,515, January 25, 2005
- A.S. Lukas, M.L. O’Neill, J.L. Vincent, R.N. Vrtis, M.D. Bitner, E.J. Karwacki, US Patent 7,098,149, August 29, 2006.
- M.L. O’Neill, M.K. Haas, B.K. Peterson, R.N. Vrtis, S.J. Weigel, D.J. Wu, M.D. Bitner, E.J. Karwacki, MRS Symposium Proceedings, 2006.
- Micron Technologies: Determining Dielectric Constant Variation of SiOC Low-k Film Using Density Measurement, AMC 2003.
- Liam Cunnane, Adrian Kiermasz, Y. Travaly, “Cost Effective SPC/APC Monitoring of a Film’s Dielectric Constant on Product,” EuroAsia Semiconductor, Oct. 2006.
Liam Cunnane received his BS in experimental physics from UCD. He is the worldwide technology director for Metryx Ltd., Bristol UK; ph 44 0/127-585-9988, e-mail [email protected].
Darren Moore received his BSc and PhD in chemistry from Bristol U., UK. He is a principal process engineer for Axcelis Technologies.
Carlo Waldfried received his PhD in physics from the University of Nebraska. Waldfried is a staff scientist at Axcelis Technologies.
Mary K. Haas received her PhD in chemistry from Princeton U. She is a principal research scientist for Air Products and Chemicals Inc.
Mark L. O’Neill received his PhD in physical chemistry from Carleton U., Canada. Mark is the team leader for thin film technology in the Electronics Technology Division at Air Products and Chemicals Inc.