Technical Insight
Pyrometry and reflectance techniques provide in situ control for GaN growth
Emissivity-corrected pyrometry in conjunction with spectroscopic reflectance can improve the reproducibility and efficiency of GaN epitaxy, according to Kolja Haberland and colleagues.
Even more so than conventional III-V epitaxy, the growth of GaN remains challenging. Several different techniques have been developed to cope with the massive lattice mismatch between the substrate - usually sapphire, SiC or silicon - and the epilayer; these include nitridation and nucleation, epitaxial lateral overgrowth, and masking with SiO2.
In no other growth process does the quality of the final layers depend so much on the initial epitaxial steps. In addition, the nucleation step turns out to be extremely temperature-sensitive. In contrast to other material systems, optical in situ studies of GaN growth have been indispensable in improving material quality, optimizing surface roughness and obtaining run-to-run reproducibility.
One of the parameters most critical to control for all epitaxial growth processes is the temperature. However, measuring the susceptor temperature using only a thermocouple or light pipe from the backside is not sufficient, since there is usually a significant offset between wafer and susceptor temperature. Furthermore, the size of this offset is not a constant value but depends, for example, on wall or ceiling depositions, type of substrate wafer and susceptor coating. Thus, there is a strong need for tight control of the wafer temperature, which can only be done by emissivity-corrected pyrometry of the wafer.
In addition to pyrometry, single-wavelength or fully spectroscopic reflectance measurements are widely used for GaN growth in order to characterize the growing layers in real time (Compound Semiconductor June 2002). The reflectance data also allows the determination of growth rates or layer thicknesses together with roughness and composition information.
Emissivity-corrected pyrometry
Direct pyrometry of the wafer from the top is the method of choice in determining the true wafer temperature. This can easily be applied when, for instance, GaN is grown on an opaque substrate such as silicon. For transparent substrates such as sapphire or SiC, at least the temperature of the susceptor surface directly beneath the wafer can be measured.
Calibrated sensors such as LayTec s EpiTT (figure 1) or EpiR TT systems can measure the temperature with an accuracy of better than 1 K and an r.m.s. noise of less than 0.2 K. With a special (patent pending) calibration procedure, the true wafer temperature can also be determined on transparent substrates.
Things get more complicated during heteroepitaxy when the emissivity of the material changes with time during growth. In consequence, the measured pyrometry signal starts to oscillate with the same period as the Fabry-Perot oscillations of the simultaneous reflectance measurement (figure 2). Maintaining high accuracy of the temperature measurement during heteroepitaxy requires compensation for changes in reflectivity, using emissivity-corrected pyrometry. EpiTT or comparable systems can perform a simultaneous reflectance measurement at the pyrometry wavelength. By detecting the changing reflectance, the emissivity of the growing material can be determined in real time and thus the temperature measurement can be corrected to its true value. In figure 2, the offset between true temperature and nominal temperature at 1200 ºC is around 100 K. Depending on the viewport geometry of the MOCVD reactor, the temperature range of 400-1400 ºC can be assessed with a typical accuracy of better than 1 K.
Wafer-selective measurements
In multiwafer systems the true wafer temperature is measured for each wafer separately. This allows wafer-to-wafer reproducibility checks and can also be used to optimize susceptor temperature uniformity. Monitoring pocket-to-pocket temperature differences is also essential for assessing compositional variations in quantum well (QW) growth.
Much more can be learned about the growth if, in addition to infrared pyrometry, reflectance measurements are performed at a shorter wavelength, for example 635 nm. Shorter wavelengths are far more suitable with respect to the band edge of the material and the penetration depth of the light. A shorter wavelength also leads to a shorter period of the Fabry-Perot oscillations during growth, which enhances the accuracy of any growth-rate fit.
Figure 3 shows data recorded during the growth of a GaN buffer and five subsequent InGaN QWs on a set of six wafers in a Thomas Swan Scientific Equipment CCS reactor. The wafer-to-wafer temperature variations (bottom graph) are very small (< 1K) and only small wafer-to-wafer differences in the reflectance signal (top graph) can be seen. These variations are due to slightly different nucleation processes.
The measurements shown in figure 3 are perfectly suited for easy run-to-run and wafer-to-wafer comparison in device production. The reflectance trace allows precise determination of the growth rate. More advanced analysis of the growing layers is possible in multiwafer reactors, when the optical tool allows detection of highly time-resolved data as performed by EpiTT. These data can be used to characterize the reflectance and temperature uniformity of the wafers during all stages of the growth.
In no other growth process does the quality of the final layers depend so much on the initial epitaxial steps. In addition, the nucleation step turns out to be extremely temperature-sensitive. In contrast to other material systems, optical in situ studies of GaN growth have been indispensable in improving material quality, optimizing surface roughness and obtaining run-to-run reproducibility.
One of the parameters most critical to control for all epitaxial growth processes is the temperature. However, measuring the susceptor temperature using only a thermocouple or light pipe from the backside is not sufficient, since there is usually a significant offset between wafer and susceptor temperature. Furthermore, the size of this offset is not a constant value but depends, for example, on wall or ceiling depositions, type of substrate wafer and susceptor coating. Thus, there is a strong need for tight control of the wafer temperature, which can only be done by emissivity-corrected pyrometry of the wafer.
In addition to pyrometry, single-wavelength or fully spectroscopic reflectance measurements are widely used for GaN growth in order to characterize the growing layers in real time (Compound Semiconductor June 2002). The reflectance data also allows the determination of growth rates or layer thicknesses together with roughness and composition information.
Emissivity-corrected pyrometry
Direct pyrometry of the wafer from the top is the method of choice in determining the true wafer temperature. This can easily be applied when, for instance, GaN is grown on an opaque substrate such as silicon. For transparent substrates such as sapphire or SiC, at least the temperature of the susceptor surface directly beneath the wafer can be measured.
Calibrated sensors such as LayTec s EpiTT (figure 1) or EpiR TT systems can measure the temperature with an accuracy of better than 1 K and an r.m.s. noise of less than 0.2 K. With a special (patent pending) calibration procedure, the true wafer temperature can also be determined on transparent substrates.
Things get more complicated during heteroepitaxy when the emissivity of the material changes with time during growth. In consequence, the measured pyrometry signal starts to oscillate with the same period as the Fabry-Perot oscillations of the simultaneous reflectance measurement (figure 2). Maintaining high accuracy of the temperature measurement during heteroepitaxy requires compensation for changes in reflectivity, using emissivity-corrected pyrometry. EpiTT or comparable systems can perform a simultaneous reflectance measurement at the pyrometry wavelength. By detecting the changing reflectance, the emissivity of the growing material can be determined in real time and thus the temperature measurement can be corrected to its true value. In figure 2, the offset between true temperature and nominal temperature at 1200 ºC is around 100 K. Depending on the viewport geometry of the MOCVD reactor, the temperature range of 400-1400 ºC can be assessed with a typical accuracy of better than 1 K.
Wafer-selective measurements
In multiwafer systems the true wafer temperature is measured for each wafer separately. This allows wafer-to-wafer reproducibility checks and can also be used to optimize susceptor temperature uniformity. Monitoring pocket-to-pocket temperature differences is also essential for assessing compositional variations in quantum well (QW) growth.
Much more can be learned about the growth if, in addition to infrared pyrometry, reflectance measurements are performed at a shorter wavelength, for example 635 nm. Shorter wavelengths are far more suitable with respect to the band edge of the material and the penetration depth of the light. A shorter wavelength also leads to a shorter period of the Fabry-Perot oscillations during growth, which enhances the accuracy of any growth-rate fit.
Figure 3 shows data recorded during the growth of a GaN buffer and five subsequent InGaN QWs on a set of six wafers in a Thomas Swan Scientific Equipment CCS reactor. The wafer-to-wafer temperature variations (bottom graph) are very small (< 1K) and only small wafer-to-wafer differences in the reflectance signal (top graph) can be seen. These variations are due to slightly different nucleation processes.
The measurements shown in figure 3 are perfectly suited for easy run-to-run and wafer-to-wafer comparison in device production. The reflectance trace allows precise determination of the growth rate. More advanced analysis of the growing layers is possible in multiwafer reactors, when the optical tool allows detection of highly time-resolved data as performed by EpiTT. These data can be used to characterize the reflectance and temperature uniformity of the wafers during all stages of the growth.
