Technical Insight
Photoreflectance technique reduces need for destructive testing of VCSEL epiwafers
Martin Murtagh, Pat Kelly and Roy Blunt report on a novel approach to photoreflectance spectroscopy that measures both cavity-mode and quantum well transition energies, promising to improve the quality assurance of VCSEL epiwafers and reduce the need for destructive testing.
Vertical-cavity surface-emitting lasers (VCSELs) consist of one or more quantum wells (QWs) embedded within a microcavity with both top and bottom distributed Bragg reflector (DBR) stacks designed for optimal lasing output. VCSELs are efficient devices that consume only moderate power, but nevertheless they run hot, especially within their active areas. Both QW and cavity-mode energies require careful control, because their energies are offset at room temperature and are designed to be matched at the operating temperature. Extreme accuracy and reproducibility are required in the epitaxial growth of these devices to ensure satisfactory operation and yield, and ideally a technique that can characterize the QW and cavity-mode energies is needed.
At the heart of the VCSEL lies an active region composed of a small number of ultra-thin QWs, typically AlGaAs/GaAs epilayers. It is this region that presents the greatest challenge to non-destructive characterization techniques.
Access problemVCSEL researchers and manufacturers have struggled for years with the problem of how to access the active-layer transition energies of a VCSEL by non-destructive means. The main problem is that the top mirror is specifically designed to reflect the luminescence of the active region at very high efficiency (very close to 100%), precluding the use of methods such as photoluminescence (PL) spectroscopy. Any PL signal generated by the active layer would be totally reflected before it reaches the surface. Moreover, such PL would be influenced by the cavity modes present, as well as requiring sufficiently thick layers to achieve sufficient signal-to-noise levels. Characterization by PL thus requires specially grown multiple QW structures (without resonant cavity mirrors) to allow VCSEL wafer QW calibration.
Currently, VCSEL epitaxial growth production engineers resort to the growth of numerous calibration runs before actual production runs take place. They also engage in destructive testing of a significant fraction of total production runs after growth. Both of these approaches result in a relatively low effective total yield, in terms of the ratio of production wafers actually shipped to the total number of wafers grown. While such an approach affords a degree of quality control, it can never achieve total quality assurance. As a result, there is a limited confidence level for witness wafer testing, especially in cases where wafer rotation anomalies can occur in a large reactor.
Even a moderate reduction in either the number of calibration runs required and/or the frequency of destructive testing of VCSEL wafers, based on non-destructive photoreflectance (PR) spectroscopy for process analysis, could make major savings. For example, one study has shown that 450,000 ($580,000) per year could be saved if the frequency of destructive testing could be dropped from 1 wafer in 8 to 1 wafer in 16. This assumes a monthly production volume of 1200 wafers and a conservative cost of 500 per destructive test (including loss of wafer).
Photoreflectance yields more dataIn contrast to other techniques such as PL, ellipsometry and Raman, PR spectroscopy relies upon electric field modulation, exhibiting a derivative-like dependence yielding highly resolved signal responses such as transition energies, strain information and also electric field data. PR affords unrivaled spectral resolution, typically 2 meV even at room temperature, and with much sharper resolution observed in many practical examples. For example, reproducibility of AlGaAs bandgap energies to 0.4 meV has been observed in sample studies.
An advanced form of the technique - variable angle of incidence PR spectroscopy - holds the unique promise of truly non-destructive test and characterization of VCSEL epiwafers, revealing the technologically important cavity-resonance mode and ground-state QW exciton structures.
In simple reflectance spectroscopy, the Fabry-Perot dip characteristic of the VCSEL etalon, or the cavity mode, is evident. At normal incidence, this cavity-mode peak energy is designed to lie below the active-layer transition energy at room temperature. Varying its angle of incidence will increase the peak energy upwards and closer to the active-layer transition energy, until it eventually coincides with the energy region of the QW transition. PR spectroscopy is capable of detecting this effect, and can use it to determine the active-layer transition energy non-destructively.
At these angles of incidence, a significant change occurs in the PR signal due to either a resonant or an antiresonant interaction of the PR lineshapes, arising from both the Fabry-Perot dip and active region. This interaction forms the basis for a non-destructive measurement of the active-layer energy, and - most importantly - for determining the offset of the cavity-mode and active-layer energies at room temperature.
Figure 1 shows a typical suite of variable angle of incidence PR spectra for an 850 nm VCSEL epiwafer. The emergence of two PR signal resonances at angles near 40 and 55º can clearly be seen, corresponding to adjacent QW interactions or even arising from heavy/light hole transitions within the active region.
An understanding of the analysis of these complex lineshapes requires some background explanation of PR spectroscopy lineshape fitting. The PR lineshape in this case is a correlated low electric field lineshape, as represented by derivatives of a dielectric function. These derivative functional forms are used to represent the different contributions to the PR response. The cavity mode itself is reproduced using a Lorentzian function, which is then convoluted with a low field lineshape representing excitonic transitions within the active region. Apart from a (fixed) dimensionality parameter, four other parameters define each lineshape from its functional form, namely transition energy, broadening and amplitude parameters, as well as lineshape phase.
Also shown in figure 1 are the corresponding PR results recorded from a similar VCSEL structure, grown without the top DBR mirror. The results confirm the lineshapes and analysis method employed for the full VCSEL structure.
Etch studies of similar VCSEL structures, involving repeated top DBR mirror-thinning using wet chemical etches until angle invariant spectra were observed (thus yielding unambiguous QW transition energy levels), further verified the effectiveness of variable-angle PR spectroscopy.
At the heart of the VCSEL lies an active region composed of a small number of ultra-thin QWs, typically AlGaAs/GaAs epilayers. It is this region that presents the greatest challenge to non-destructive characterization techniques.
Access problemVCSEL researchers and manufacturers have struggled for years with the problem of how to access the active-layer transition energies of a VCSEL by non-destructive means. The main problem is that the top mirror is specifically designed to reflect the luminescence of the active region at very high efficiency (very close to 100%), precluding the use of methods such as photoluminescence (PL) spectroscopy. Any PL signal generated by the active layer would be totally reflected before it reaches the surface. Moreover, such PL would be influenced by the cavity modes present, as well as requiring sufficiently thick layers to achieve sufficient signal-to-noise levels. Characterization by PL thus requires specially grown multiple QW structures (without resonant cavity mirrors) to allow VCSEL wafer QW calibration.
Currently, VCSEL epitaxial growth production engineers resort to the growth of numerous calibration runs before actual production runs take place. They also engage in destructive testing of a significant fraction of total production runs after growth. Both of these approaches result in a relatively low effective total yield, in terms of the ratio of production wafers actually shipped to the total number of wafers grown. While such an approach affords a degree of quality control, it can never achieve total quality assurance. As a result, there is a limited confidence level for witness wafer testing, especially in cases where wafer rotation anomalies can occur in a large reactor.
Even a moderate reduction in either the number of calibration runs required and/or the frequency of destructive testing of VCSEL wafers, based on non-destructive photoreflectance (PR) spectroscopy for process analysis, could make major savings. For example, one study has shown that 450,000 ($580,000) per year could be saved if the frequency of destructive testing could be dropped from 1 wafer in 8 to 1 wafer in 16. This assumes a monthly production volume of 1200 wafers and a conservative cost of 500 per destructive test (including loss of wafer).
Photoreflectance yields more dataIn contrast to other techniques such as PL, ellipsometry and Raman, PR spectroscopy relies upon electric field modulation, exhibiting a derivative-like dependence yielding highly resolved signal responses such as transition energies, strain information and also electric field data. PR affords unrivaled spectral resolution, typically 2 meV even at room temperature, and with much sharper resolution observed in many practical examples. For example, reproducibility of AlGaAs bandgap energies to 0.4 meV has been observed in sample studies.
An advanced form of the technique - variable angle of incidence PR spectroscopy - holds the unique promise of truly non-destructive test and characterization of VCSEL epiwafers, revealing the technologically important cavity-resonance mode and ground-state QW exciton structures.
In simple reflectance spectroscopy, the Fabry-Perot dip characteristic of the VCSEL etalon, or the cavity mode, is evident. At normal incidence, this cavity-mode peak energy is designed to lie below the active-layer transition energy at room temperature. Varying its angle of incidence will increase the peak energy upwards and closer to the active-layer transition energy, until it eventually coincides with the energy region of the QW transition. PR spectroscopy is capable of detecting this effect, and can use it to determine the active-layer transition energy non-destructively.
At these angles of incidence, a significant change occurs in the PR signal due to either a resonant or an antiresonant interaction of the PR lineshapes, arising from both the Fabry-Perot dip and active region. This interaction forms the basis for a non-destructive measurement of the active-layer energy, and - most importantly - for determining the offset of the cavity-mode and active-layer energies at room temperature.
Figure 1 shows a typical suite of variable angle of incidence PR spectra for an 850 nm VCSEL epiwafer. The emergence of two PR signal resonances at angles near 40 and 55º can clearly be seen, corresponding to adjacent QW interactions or even arising from heavy/light hole transitions within the active region.
An understanding of the analysis of these complex lineshapes requires some background explanation of PR spectroscopy lineshape fitting. The PR lineshape in this case is a correlated low electric field lineshape, as represented by derivatives of a dielectric function. These derivative functional forms are used to represent the different contributions to the PR response. The cavity mode itself is reproduced using a Lorentzian function, which is then convoluted with a low field lineshape representing excitonic transitions within the active region. Apart from a (fixed) dimensionality parameter, four other parameters define each lineshape from its functional form, namely transition energy, broadening and amplitude parameters, as well as lineshape phase.
Also shown in figure 1 are the corresponding PR results recorded from a similar VCSEL structure, grown without the top DBR mirror. The results confirm the lineshapes and analysis method employed for the full VCSEL structure.
Etch studies of similar VCSEL structures, involving repeated top DBR mirror-thinning using wet chemical etches until angle invariant spectra were observed (thus yielding unambiguous QW transition energy levels), further verified the effectiveness of variable-angle PR spectroscopy.
