Rapid photothermal processing
10/01/1997
R. Singh, R. Sharangpani, Dept. of Electrical & Computer Engineering, Clemson University, South Carolina
The differences between the radiation spectra emitted by the heating elements in furnaces and rapid thermal processing (RTP) systems result in fundamental differences in the operating mechanisms of these two methods. Quantum effects can be initiated in RTP by a suitable choice of lamps and lamp orientation. This technique, called rapid photothermal processing (RPP), meets many current and long-term semiconductor processing challenges. This article presents a simple quantum mechanical description of some of the key operating mechanisms in RPP, along with supportive experimental results.
Until recently, furnace batch-processing was one of the most popular methods for manufacturing the entire spectrum of semiconductor devices. As a result of shrinking device geometries and increasing wafer sizes, however, trends point in the direction of single wafer processing. There is also a focus on long-term requirements for reduced microscopic defects, lower processing temperature, reduced cost of ownership, reduced cycle times, smaller feature sizes and environmentally friendly processing of sub-0.18-µm feature size ICs.
The significantly higher heating and cooling rates of RTP systems with lamp heaters, along with RTP's compatibility with single-wafer technology, make it one of the most promising alternatives to furnaces for many applications [1]. Recently introduced minifurnaces and RTP systems using resistive heaters [2, 3] provide comparable heating and cooling rates, but are not suitable for the photon-assisted processing discussed in this article.
By examining the role of photoeffects in RTP [4–8], we have demonstrated how ultraviolet (UV) and vacuum ultra violet (VUV) light sources, used in conjunction with conventional (lamp-based) RTP, can optimize any thermal step in device fabrication. RPP uses a different photon spectrum from that obtained in furnace processing or conventional RTP to optimize device properties. This paper presents the operating principles and technology of RPP in light of anticipated manufacturing requirements.
Operating principles
Radiation spectra. According to the laws of black body radiation, a heated body (such as the filament of a lamp) emits electromagnetic radiation. The Planck distribution, shown in Fig. 1 for six different temperatures, gives the precise dependence of the energy output on wavelength (λ).
Figure 1. Radiation spectra for bodies heated at different temperatures. |
Figure 1 has two striking features. First, the peak shifts to smaller wavelengths as the temperature of the body increases. Second, at high temperatures, an appreciable portion of the emitted radiation lies in the visible (0.4–0.8 µm), UV (0.2–0.4 µm), and VUV (0.1–0.2 µm) regions of the spectrum. In a furnace process step at 1000°C, the chamber walls are in thermal equilibrium with the wafer. The photons emitted at this temperature are in the infrared (IR) region (λ > 1 µm). However, the filaments of the tungsten halogen lamps used in an RTP system must be maintained at a higher temperature than the wafers to overcome radiation losses. Since, for the same 1000°C process, the lamps are at a higher temperature than the furnace walls, they emit a different spectrum of radiation and many more photons (of all wavelengths). More significant, lamps emit substantially more photons/sec in the UV/VUV and visible range. These photons have energy in the range of 1–10 eV, much higher than the energy of IR photons emitted by the furnace walls (<1 eV). This difference in the amount of UV/VUV and visible photons emitted per second causes differences in properties between RPP and furnace-processed samples.
Matter-light interaction. The equation, Photons + Matter = Thermal Effects + Quantum Effects, describes the photon-matter interaction. Thermal effects dominate for photons with wavelength greater than about 0.8 µm. Photons in this wavelength range increase the vibrational amplitude of the atoms, raising the temperature of the film and the underlying substrate so that the desired thermal effects can take place. Thermal excitation is the basic mechanism in furnace processing, where all of the photons emitted have wavelengths > 0.8 µm.
Figure 2. Dependence of activation energy, process temperature, and time on processing technique; a) purely thermal process, b) photothermal process. |
With photons of wavelength < 0.8 µm, quantum effects occur in addition to thermal effects. Atoms and molecules transition from ground state to electronically excited states when bombarded with photons of energy of a few eV. As shown in Fig. 2, quantum excitation of species changes the process in several ways:
- The process occurs at a lower temperature, reducing thermal stress and energy consumption/processed wafer. Excited states also reduce the intrinsic stress. Intrinsic and extrinsic stress reduction improves performance and reliability.
- Reduced microscopic defects in the bulk and interface yield higher minority carrier lifetimes, higher surface and bulk mobilities, etc., improving device performance.
- Reduction of the processing time further reduces stress and increases the material usage efficiency. Costs and ecological threats are reduced as a result.
All of these effects lower the cost of ownership of the equipment [9].
Quantum effects can be exploited only where UV/VUV photons dominate. However, mere emission of these photons does not guarantee that they will be incident on the active side of the wafer. The relative orientation between the lamps and the wafers determines whether or not maximum use is made of the UV/VUV photons.
Design considerations in RPP
Several factors need to be considered to design the most efficient RPP systems.
Directionality of incident photons. The performance of each RPP step depends on the direction of photons falling on the semiconductor. If the backside of the wafer faces the lamps, then most of the UV/VUV photons are absorbed in the substrate before they can reach the front (active) side. The only photons arriving at the front surface are high wavelength (low energy) photons in the IR range. (The absorption coefficient of silicon is much smaller for high wavelength photons at process temperatures [10].) If the front side of the wafer directly faces the lamps, however, the UV and VUV photons are directly incident on the active side. There is an appreciable difference in the photon flux (especially for photons < 0.8 µm) between processes with front and back heating [11].
Figure 3. Lamp configurations. |
Four lamp configurations were studied to establish the dependence of high and low dielectric constant (K) material properties on the photon spectrum [11]. One of these configurations (Fig. 3b), used both front and back heating, providing a photon flux on the active side that was intermediate between the flux provided by front (Fig. 3c) or back (Fig. 3a) heating alone. The other configuration (Fig. 3d) used a deuterium lamp (wavelengths in the range of 0.1 to 0.19 µm) in conjunction with the tungsten halogen heater lamps to maximize quantum effects. Each of the four configurations gave a different photon spectrum on the front side. The order of lowest to highest photon flux is the same as the order of the figures (a–d). The UV/VUV photon flux directly correlated with the overall quality of the film.
|
Lamp selection. Once the correct directionality is obtained, it is important to select lamps that emit a high number of photons in the UV/VUV range. Argon, krypton, fluorine, and xenon lamps give a spectrum in the 0.1–0.2-µm wavelength range. Arc lamps can give an even richer UV/VUV photon spectrum. A recent study compared the film stress for hydrogen silsequioxane films processed in a furnace, a tungsten halogen lamp-based RTP system, and an arc lamp-based RTP system [12]. The arc lamp system, which has the highest number of UV and visible photons, produced films with the lowest stress, followed by tungsten halogen lamp-based RTP and furnace-processed samples. Since this was also the order of highest to lowest UV/VUV photon flux, the experiment established a direct correlation between the UV/VUV photon flux and stress. Table 1 summarizes the different processing methods.
Experimental results
Rapid photothermal diffusion. We diffused phosphorus in p-type Si wafers that were spin-coated with high-purity phosphosilicate glass using configurations a), c), and d). More detailed experimental data can be found [13]. Table 2 summarizes the results.
|
The number of photons on the diffusion side of the wafer was the only difference in the samples. There is a direct correlation between the flux and the diffusion coefficient, and therefore between the flux and the time required to obtain a certain junction depth. Looked at another way, increased UV/VUV photon flux lowers the temperature required to realize a specific junction depth if the time is kept constant. Configuration d) can give cycle times and/or processing temperatures lower than conventional methods. Since the minority carrier lifetimes in a semiconductor depend on the defect densities, the correlation of lifetimes with the UV/VUV photon flux indicates a reduction of microscopic defects.
Ohmic contact formation. Any semiconductor device requires ohmic contacts. We studied the role of photoeffects in the formation of ohmic contacts by screen printing back contacts on two p-type Si wafers, using Ag-Al paste. In case i), we annealed the wafer with the contact facing the bottom bank of lamps (configuration c). The temperature was ramped up to 720°C at 10°C/min and immediately ramped down to 225°C in 50 sec, after which the wafer was allowed to cool to room temperature. The total cycle time was 122 sec. In case ii), we annealed the wafer using lamp configuration a), with the contact not facing the top bank. The same ramp rates were used, but the temperature was held at 720°C for 30 sec to ensure identical sheet resistivity in both cases. The total cycle time was therefore 152 sec. Configuration a) delivered fewer UV/VUV photons to the contact side than configuration c). The shorter cycle time (122 vs. 152 sec) achieved with higher UV/VUV photon-assisted processing translates into a higher throughput and less energy waste.
|
To compare contact characteristics, we evaporated Al contacts on the front side of the wafers to form Schottky diodes. Table 3 compares the ideality factor (n), series resistance (Rs), minority carrier lifetime (t) of the bulk silicon, surface roughness (measured by atomic force microscopy), and stress. The device characteristics improve with quantum effects. The higher minority current lifetime obtained with configuration c) illustrates the lower microscopic defects possible with UV and VUV photons. The lower stress obtained with configuration c) suggests that UV and VUV photons can also enhance device reliability.
Figure 4. Stress vs. lamp configuration for cured polyimide samples. |
Low k dielectric processing. Portable low-voltage microsystems have stimulated great interest in low k dielectrics as possible replacements for SiO2 in future interconnect schemes. Quantum photoeffects can reduce the annealing temperature for a spin-coated low k dielectric material [11] (DuPont company's polyimide 2570). Polyimide annealing induces crosslinking between oxydianiline and phenylenediamine chains. If all other conditions are unchanged, crosslinking increases at higher temperatures, giving dielectric films with better insulating properties. In spite of this, polyimide samples annealed using configuration d) at 300°C have higher resistivity (2 × 1016 Ω cm) than obtained using configuration a) at 400°C (2.4 × 1015 Ω cm) at 5 V. An additional sample was processed in a furnace for comparison. The stress values for all films (Fig. 4) follow the same trend as the leakage current.
Conclusion
The results reported in this paper indicate that RPP will be important in fabrication of transistors with feature sizes as low as 25 nm. New equipment using photothermal effects should be designed to exploit the contribution of photons in the UV/VUV range. Since this technique is equally applicable to a wide variety of low thermal mass systems, we expect new applications outside conventional semiconductor manufacturing.
References
- R. Singh, "Rapid Isothermal Processing," J. Appl. Phys., 63, R59-R114, 1988.
- K.G. Reid, A.R. Sitaram, "RTP for ULSI Applications: An overview," Solid State Technology, 63, Feb. 1996.
- A. Dip, "Fast thermal processing: Batch comes back," Solid State Technology, p. 113, June 1996.
- R. Singh, "A Better Way to Form III-V Junctions," Electronics, 58, 19, Dec. 1985.
- R. Singh, "New Processes for Junction Formation in Compound Semiconductors by Rapid Isothermal Processing," Semiconductor International, 9 (1), 28, 1986.
- R. Singh, et al., "Oxidation of Tin on Silicon Substrate by Rapid Isothermal Processing," J. Appl. Phys., 66, 2381, 1989.
- R. Singh, et al., "Role of Photoeffects in Rapid Isothermal Processing," Appl. Phys. Lett., 1991.
- R. Singh, "Rapid Isothermal Processing," Handbook of Compound Semiconductors, eds. P. Holloway, G. McGuire, (Noyce Publications, Park Ridge, N.J.), 442, 1995.
- R. Singh, et al., "How Rapid Isothermal Processing can be a Dominant Semiconductor Processing Technology in the 21st Century," Mat. Res. Soc. Symp. Proc., eds. J.C. Gelpey, et al., Material Research Society, Pittsburgh, PA, Vol. 429, p. 81, 1996.
- G.E. Jellison Jr., F. A. Modine, "Optical Functions of Silicon at Elevated Temperatures," J. Appl. Physics, 39, 3758, 1994.
- R. Sharangpani, K.C. Cherukuri, R. Singh, "Importance of High Energy Photons in the Curing of Spin-on Low Dielectric Constant Interconnect Materials," J. Electrochem. Soc., 144 (2), 669, 1997.
- T.E. Gentle, "Rapid Thermal and Related Processing Techniques," eds. M.M. Moslehi, R. Singh, D. Kwong, Int. Soc. Opt. Eng., Vol. 133, p. 146, 1991.
- R. Singh, et al., "Low Temperature Shallow Junction Formation Using Vacuum Ultraviolet Photons In Rapid Isothermal Processing," Appl. Ph. Lett., Vol. 70, p. 1700, 1997.
|
Rajendra Singh is D. Houser Banks Professor of electrical and computer engineering and director of the materials science and engineering program at Clemson University. His main research contributions are in the field of rapid photothermal processing. He is the author of more than 200 publications and editor or co-editor of eight books. Dept. of Electrical & Computer Engineering and Materials Science and Engineering Program, Clemson University, Clemson, SC 29634-0915; ph 864/656-0919, e-mail [email protected].
|
Rahul Sharangpani received his PhD in electrical engineering from Clemson University. He was the recipient of the 1996 DuPont Plunkett Student Award for Innovation with Teflon and the 1996 Harriss Outstanding Researcher Award, and he was a 1995 MRS Graduate Student Award finalist. He has authored or co-authored more than 25 papers related to rapid photothermal processing. Clemson University, ph 864/656-1183, e-mail [email protected].