III-nitride epitaxial growth with real-time access to wafer temp, ternary composition
02/01/2005
In all epitaxial growth processes of III-Vs and related materials (MOCVD, MBE, CVD), an accurate control of the true wafer temperature is essential, primarily because the composition of key materials (e.g., InGaN, AlGaN, InGaAsP) depends on it to a very large extent. On the other hand, high-precision, quantitative real-time analysis of film composition can only be performed if the wafer temperature is accurately known (to select the correct substrate refractive index out of a high-temperature database). Thus, direct real-time access to the true wafer temperature throughout the complete epitaxial growth process under all growth conditions is highly desirable.
When using the III-nitrides, one has to grow certain layers and structures in growth modes that in conventional III-V growth would not be considered ideal. Two examples are highly rough and 3D nucleation layers for defect reduction in the subsequent device structure, and highly strained structures causing substrate bowing (at growth temperatures). The latter is used to attain flat substrates after the cool-down phase in preparation for further processing (thermal expansion coefficients often differ significantly for the individual materials). The challenge for in situ monitoring is that surface morphology effects have to be strictly distinguished from the film composition effects, and the effect of wafer curvature on wafer temperature measurements must be taken into account.
Transparent substrates
For IR-absorbing substrates such as Si, GaAs, and InP, emissivity-corrected pyrometry of the wafer temperature is a familiar and extremely accurate technology. For the case of IR transparent substrates (e.g., sapphire and SiC), which are widely used for mass-producing blue LEDs, specific calibration routines also were recently introduced to monitor the wafer surface temperature - not the temperature of the carrier underneath, which is rarely the same. So, for a given type of IR-transparent substrate (regarding backside polishing, thickness, and doping), the wafer temperature is available during growth due to proper calibration of the pyrometer. To discover what happens when the substrate is “changing” (e.g., by strain-induced substrate bowing), systematic studies were performed in a project with the U. of Magdeburg by measuring the real wafer temperature during a full III-nitride LED growth run on a heavily bowing Si substrate (Fig. 1).
 Figure 1. Wafer temperature (red) and wafer curvature (blue) during growth of a III-nitride LED on silicon. The nominal process temperature (black) is added for comparison. |
During the MOCVD device growth process, the silicon wafer is bowing in a convex manner (i.e., the vertical distance z between susceptor and the center of the wafer is increasing while the outer edge of the wafer maintains contact). Interestingly, due to the way in which the low-temperature AlN strain-engineering layers in this structure are designed, the LED wafer is completely flat again, and therefore ready to be processed after cooling down to room temperature.
During epitaxial growth, however, the wafer temperature in the center is reduced by 45°C. Even if this factor is taken into account by respective process-temperature compensation in the growth recipe, it is disastrous to the resulting yield because only the center part of this wafer can be used. The importance of this issue increases dramatically with wafer diameter, as a simple calculation demonstrates: The maximum wafer curvature in Fig. 1 relates to a curving radius of ~2m. A 4" wafer of the same curving radius would have a gap of z = 0.5mm between the wafer center and the susceptor. Even if the related effects for growth on sapphire and SiC are smaller than for silicon, we expect that advanced in situ sensors combining wafer-curvature measurements both with wafer-temperature measurements and multiple-wavelength reflectance will be mandatory for III-nitride device process development on large substrate diameters.
Measuring true film composition
The actual film composition typically is assessed through real-time measurement of the refractive index in the film as it’s grown. The growth rate and the film composition are extracted from the shape, average level, and amplitude of reflectance oscillations (due to thin-film interference effects). The problem, however, is that this only works for ideal, 2D film growth! Figure 2 shows an example of an ideally grown AlN film on a GaN buffer, which gives exactly the same reflectance signal as a rough (~24nm RMS) GaN layer grown with a slightly reduced growth rate on the same GaN buffer. The film composition effect on the reflectance signal is “perfectly” mimicked by a surface morphology effect. A research team at LayTec analyzed the different contributions of surface morphology (roughness, waviness) and film composition to the reflectance measurement and is preparing new analysis features allowing composition measurements even for less than ideal growth.
Conclusion
Two main challenges for shortening process development cycles and increasing the yield of epitaxial processes by applying advanced in situ sensors have been investigated: III-nitride ternary composition in situ measurements and related real-time sensing of wafer bowing and nonuniform wafer temperatures.
A first conclusion is that single-wavelength in situ reflectance is not adequate for sensing InGaN or AlGaN ternary composition because of a well understood physical limitation in case of growth scenarios where the 3D surface morphology matters for device performance. For this reason, LayTec has focused from the very beginning on multiple-wavelength sensors, avoiding this limitation [e.g., the double-wavelength sensor (EpiTT) uses 633nm and 950nm]. The highest sensitivity to film composition (>±1%) is reached with a fully spectroscopic sensor (EpiR TT) that covers all wavelengths between 300nm and 750nm.
A second conclusion is that, as the wafer size increases in III-V technologies , especially in III-nitride LED and solar cell applications, stress effects causing strong wafer bowing during epitaxial growth will considerably affect the wafer-temperature uniformity. This has a strong effect on the resulting composition uniformity of ternary and quaternary layers and, thus, on the maximum yield of a production process. Therefore, even more advanced in situ sensors combining wafer-temperature sensing and multiple-wavelength reflectance with an additional real-time wafer-curvature measurement are needed.
Acknowledgments
The authors appreciate the collaboration with A. Dadgar (U. of Magdeburg), H. Hardtdegen (Research Center Jülich), A. Strittmatter (Berlin U. of Technology), and their co-workers who contributed to the research.
For more information, contact Elisabeth Steimetz at LayTec GmbH; ph 49/30-3980-0800, e-mail [email protected].
Thomas Zettler received his diploma and PhD in physics at Humboldt U. in Berlin, and his “habilitation” in physics at Berlin U. of Technology. He co-founded LayTec in 1999 and is head of R&D.
Elisabeth Steimetz received her diploma in physics at the RWTH in Aachen and PhD in physics at Berlin U. of Technology. She is head of strategic marketing at LayTec GmbH.