Steam-based RTP for advanced processes
10/01/1998
Steam-based RTP for advanced processes
Rahul Sharangpani, R.P.S. Thakur, Nitin Shah, S.P. Tay, AG Associates Inc., San Jose, California
Rapid thermal processing in steam-containing ambients is a new method that is strongly aligned with current and anticipated future device needs. There are numerous advantages of steam-based rapid thermal processing technology in conjunction with the fast ambient switching capability of a single-wafer rapid thermal processing system. Our discussions will include some of the key issues involved in designing integrated steam generator
apid thermal processing systems that can offer the desired process control and capability, throughput, and flexibility. The studies presented here focus on the physical and electrical properties of the oxides grown in the new integrated steam system; correlate the oxide growth curves with the model that explains the oxidation behavior; and briefly show the overall advantages and key areas of integrated circuit processing where this technique may be used.
The next generation of IC devices is providing renewed opportunities for rapid thermal processing (RTP). In addition to conventional advantages of RTP, such as reduced thermal budget, contact anneals, implant damage repair, and dopant activation, advanced features like fast ambient switching and multiple ambient processing are becoming equally important. Lately, steam ambients in RTP have drawn considerable attention. Several processes such as dielectric growth and metal reflow provide opportunities for reduced thermal budgets, and superior thin films like gate and cell dielectrics can be formed with steam ambients.
For a long time, the semiconductor industry did not consider steam-based RTP technology as a viable option for single-wafer systems due to the problems associated with steam condensation on cold walls and subsequent nonuniformity and film contamination. Recent advances in equipment design and process innovation, however, allow steam as one of the process gases in state-of-the-art RTP systems, consolidating the benefits of RTP with steam ambients (see figure top right). The integrated systems thus provide considerable technical leverage by addressing the need for a very wide spectrum of advanced applications.
 |
Due to the wide range of steam concentrations, temperature requirements, and the growing emphasis on contamination-free processing, both the steam generator and the RTP system should be carefully selected. We will present some results that can guide the selection of steam-RTP systems for different applications. The basic characterization of the integrated system uses rapid thermal oxidation as a process and performance criterion. Our discussions will also include some recent trends in various methods of steam generation and the benefits of RTP in steam ambients.
Advantages of RTP-based steam oxidation
RTP-grown dry oxides are superior to furnace oxides in bulk and interface properties [1]. Due to its short processing time capability at higher temperatures and faster ramp rates, steam-based RTP is ideal for growing thin oxides. This is significant because steam oxidation offers additional benefits, some of which are summarized below [2-4]:
1. Steam oxides have lower interfacial stress and fewer defects (pinholes, keyholes, micropores, etc.) than dry oxides.
2. As a consequence, steam oxides have improved reliability, as measured by breakdown voltage (Ebd) and charge to breakdown (Qbd). Studies show that steam oxides give smoother interfaces, leading to higher effective electron mobility than dry oxides.
3. Steam oxides are denser and more impervious to impurity diffusion.
4. Steam oxidation allows higher productivity due to the faster growth rate at a given temperature and steam concentration.
5. As a consequence, a much broader range (100 nm and higher) of oxides can be easily grown using steam oxidation.
Steam generation schemes for RTP systems
Steam generation schemes for RTP applications exist in the form of external techniques, where steam generation occurs outside of the RTP system using either the bubbler, pyrogenic, or catalytic methods; and internal techniques, where steam is generated by combusting hydrogen and oxygen inside the RTP chamber. While steam generation with bubblers is cost-effective, it limits the flexibility of steam generation over a wider process window and may not be suitable for advanced RTP applications.
Pyrogenic steam generator. In a pyrogenic steam system, igniting a controlled mixture of hydrogen and oxygen generates ultrapure steam, which then travels through heated lines to the RTP chamber. Figure 1a shows the schematic of a pyrogenic steam generator. Pyrogenic generators have been in existence for several decades and have been extensively used and characterized over the years. However, this technique of steam generation has serious limitations. For instance, low steam concentrations (which are required for controllable growth of thin oxides at elevated temperatures) cannot be generated using pyrogenic generators. Table 1 summarizes some of the other advantages and disadvantages of this technique.
 |
Figure 1. a) Pyrogenic steam generator, b) in situ steam generator, and c) catalytic steam generator.
 |
Internal steam generation. Steam generation inside the RTP system is based on a similar concept (Fig. 1b). The principal difference is that the wafer acts as a heating source to bring the temperature of the reactants to a level high enough to initiate spontaneous combustion. The steam concentration depends on many factors such as the composition of reactants and the temperature of the reactant species. This method of steam generation requires very even mixing of the reactant gases and consistent gas temperature control to ensure uniform heating and oxide growth (Table 2).
Catalytic steam generation. A new steam generator that produces steam externally has shown promising performance for advanced applications (Fig. 1c). This new technique generates steam in a catalytic chamber external to the RTP system. The reaction in a catalytic environment proceeds by adsorption of hydrogen and oxygen on the catalytic surface, followed by reaction of the adsorbed species. The catalyst lowers the activation energy of the reaction of hydrogen and oxygen, allowing steam generation at a reduced temperature. This provides a suitable environment to accommodate a wide range of flow rate and steam concentration. Table 3 compares the advantages and disadvantages of catalytic steam generators.
 |
Steam generation and application trends
Steam generation using bubblers or internal techniques has safety hazards, limited controllability, and poor uniformity in thin-film growth. It is thus easier to eliminate or substantially limit bubblers and internal steam generation. In light of the growing importance of contamination-free processing and the use of very low steam concentrations for thin oxide growth, catalytic steam generation may be the method of choice for the future. The main concern with using metal catalysts is that the catalyst material itself could introduce metal contamination in the gas stream. Thorough investigations, however, have reported nonexistent metal contamination for wafers oxidized using catalytic steam systems [5]. Catalytic systems capable of generating 2 slm and higher steam flow rates are expected to be available shortly. This would increase the versatility of the integrated steam system for both low (<1%) and high (>50%) steam concentrations.
Table 4 shows process steps that can potentially benefit from catalytic-based RTP to meet the requirements for current device generations.
Oxide growth using catalytic steam-based RTP systems
1. Temperature and concentration sensitivity curves. Figure 2 shows growth rate curves using steam RTO at various temperatures. The thickness represents the average thickness of 49 points measured using single wavelength ellipsometry (l = 632.8 nm), with 3-mm edge exclusion on 200-mm wafers.We have derived a model to predict the oxide thickness under a given set of processing conditions (temperature, time, steam concentration, etc.). This model also calculates one of the growth parameters required to achieve a target thickness if the other two parameters as well as the initial oxide thickness are specified. (For more details of the model and a comparison of experimental and theoretical growth curves, see [6].) For purposes of illustration, we compared the model against experimentally observed growth rates for two different conditions (Fig. 3).
 |
Figure 2. Experimental growth-rate curves.
 |
Figure 3. Experimental vs. theoretical growth curves.
2. Uniformity. We used spectroscopic ellipsometry (KLA-Tencor UV 1250SE) to measure the thickness uniformity on 200-mm dia. p(100) wafers on 121 points, with 3-mm edge exclusion. These measurements were taken to find out whether the presence of steam degrades the uniformity compared to dry oxides grown under similar conditions. Oxides for uniformity measurement were grown at 1000?C using 30% steam in an oxygen ambient for 40 sec. Figure 4 shows an illustrative thickness contour map in which the direction of gas flow is from the top to the bottom. The measured uniformity (0.949% 1 s) is similar to that obtained with dry oxides on the same tool.
 |
Figure 4. Thickness contour map.
3. Electrical data. The cumulative plots of time to breakdown (Tbd) demonstrate how steam oxidation improves oxide reliability (Fig. 5). They were obtained under constant current density (100 mA/cm2) gate injection on 100 ? 100 ?m in situ-doped polysilicon capacitors formed on 4-nm thermal oxides grown at 800?C. The plots indicate an order of magnitude improvement of 50% steam oxides over dry oxides. We used FTIR to characterize the interfacial stress, which, measured on the samples used in Fig. 5 from the position of the absorption frequency peaks of the Si-O bonds [7], showed that the samples with higher steam concentration impose less stress on the Si-O bond.
 |
Figure 5. Time to breakdown (Tbd) of 4-nm, 800?C grown dry andwet oxides.
4. Metal contamination. Figure 6 shows bidirectional C-V plots of 20-nm oxides grown using 50% steam at 1100?C before and after bias temperature stressing (250?C, 1 MV/cm stress). These plots show characteristic absence of any shift in the flat band voltage, thereby confirming that mobile ions were not present in our oxides. The small amount of hysteresis in the curves for each voltage direction is significant as it indicates a low interface trap density. Total reflection x-ray fluorescence spectroscopy (TXRF) spectra of the oxides were taken to detect contamination in oxides due to heavy metals (Table 5). Results showed that almost all the metals were below the practical detection limits of the tool at all three measured points.
 |
Figure 6. C-V plots of wet oxides before and after bias temperature stressing.
 |
Conclusion
We have shown results of steam ambient processing in an RTP system, with comparisons of various methods of steam generation. Our studies showed that the performance of the catalytic method was most compatible with future needs. The steam oxide films we grew using the integrated steam-RTP system showed superior uniformity, minimal contamination, high throughput, and improved electrical performance.n
Acknowledgment
The authors wish to thank Y. Tanabe of Hitachi, Japan, and Krishna Saraswat and T.C. Yang of Stanford University for their help in electrical characterization of oxide samples and several technical discussions related to the subject matter.
References
1. J. Nulman, J.P. Krusius, A. Gat, IEEE Electron Device Letters EDL-6, Vol. 205, 1986.
2. M. Gluck et al., "Homogeneity of Wet Oxidation by RTP," MRS Symp. Proc., Vol. 242, p. 215, MRS, Pittsburgh, PA, 1994.
3. E. A. Irene, R. Ghez, "Silicon Oxidation Studies: The Role of H20,"J. Electrochemical Society, Vol. 124, No. 11, p. 1757, November 1977.
4. F. Bryant, F. Liou, "Thin Gate Oxides Grown in Argon-diluted Oxygen with Steam Ambient and HCl Treatment," Proc. Symp. on Silicon Nitride and Silicon Dioxide Thin Insulating Films, Vol. 89-7, p. 220, eds. S.B. Bibyk, V.J. Kapoor, N.S. Alvi, The Electrochem. Soc., 1989.
5. Y. Tanabe et al., "Diluted Wet Oxidation: A Novel Technique for Ultra-thin Gate Oxide Formation," Proc. 1997 IEEE International Symposium on Semiconductor Manufacturing, Meeting Plus, p. p-49.
6. R. Sharangpani et al., "Growth and Characterization of Thin Wet Oxides Grown by Rapid Thermal Processing," in MRS RTP Symposium Proceedings, in press, 1998.
7. T.C. Yang, N. Bhat, K.C. Saraswat, "Effect of Interface Stress on Reliability of Gate Oxide," 4th Symp. on Silicon Nitride and Silicon Oxide Thin Insulating Films, 191st Mtg. of the Electrochem. Soc., Montreal, Canada, May 1997.
RAHUL SHARANGPANI received his PhD in electrical engineering from Clemson University in August 1997. He holds two patents and is the recipient of the 1996 DuPont Plunkett student award, the 1996 Harriss outstanding researcher award, and was a MRS 1995 graduate student award finalist. Sharangpani develops steam and RTP processes for sub-0.18-?m technology at AG Associates.
R.P.S. THAKUR received his BS in electronics and communication engineering, and MS and PhD in electrical engineering. He is AG Associates` VP of technology and R&D. Thakur holds more than 35 patents and has authored nearly 130 papers in journals and conference proceedings.
NITIN SHAH received his MS in chemical engineering from the University of Pittsburgh. He has 14 years of experience in process development and integration, and has authored more than 20 publications on RTP. Shah is director of marketing and technical sales at AG Associates Inc., 4425 Fortran Drive, San Jose, CA 95134-2300; ph 408/935-2189, fax 408/935-2740, e-mail [email protected].
S.P. TAY received his PhD from the University of Salford, UK. He has been manager of the process development department at AG Associates since 1995, specializing in front-end silicon processing with an emphasis on oxidation and diffusion. Tay has authored more than 50 technical papers and has been granted 12 patents.