The enormous demand for fiber optic and wireless communication components has been the main driving force behind the development and high volume production of compound semiconductors. Since the majority of optoelectronic device structures are manufactured using epitaxial growth methods, rapid, non-destructive and non-contact characterization techniques are required for quality control. Photoluminescence and photo-luminescence mapping have become the most important process control tools for this early stage of production. First Principles When light of sufficient energy is incident on a semiconductor material, photons are absorbed and electronic excitations are created. Eventually, these excitations relax and the electrons return to the ground state. If radiative relaxation occurs, the emitted light is called photoluminescence (PL). This light can be collected and analyzed to yield a wealth of information about the material. The color (or wavelength) of the light provides information on transition energies, and the intensity is a measure of the relative rate of radiative and nonradiative recombination. Excitation Wavelength Since most optoelectronic and microelectronic devices are based on multi-layer structures with different doping types and energy gaps, it is very important to choose a suitable excitation wavelength and power density to maximize the information obtained from PL. The energy of the excitation photon should be greater than the band gap of the layer of interest, and because the absorption coefficient of most materials increases with photon energy, the penetration depth of incident light reduces with increasing photon energy. In direct band gap semiconductors, the above band gap energy photons have penetration depths less then 1 m. shows the absorption coefficient of GaAs at the wavelengths of the most commonly used excitation lasers and transmissions at thicknesses of 100 and 200 nm. Note that the transmission is reduced from 74% down to only 6% when the excitation wavelength is changed from 780 to 488 nm. This does not take into account the 2030% excitation power reflected at the surface. The diffusion of photo-generated carriers is strongly controlled by the energy band profile, and carriers can be confined within a thin layer or move up to 10 m in bulk material. Thus, by carefully selecting the energy of excitation, the PL signal can be enhanced and buried layers within the device structures can be investigated. In a typical AlGaAs/GaAs separate confinement QW laser structure such as shown in , a narrower band gap GaAs cap layer is used in order to the improve the ohmic contact. The electron-hole pairs generated in the cap layer can not reach the active region due to the high potential barrier present at the cap/cladding interface. Therefore, only the light absorbed in the cladding region would be capable of generating electron-holes pairs that could then reach the active region by diffusion. The PL signal from a top cladding layer can be eliminated by using an excitation source of photons with an energy between that of the band gap energies of cladding and waveguide layers. However, the PL signals are not normally proportional to the efficiency of electroluminescence in LEDs and laser diodes, which is due to quenching of the emission of the QWs by the built-in electric field. The electric field also effectively stops diffusion of either electrons or holes generated near the surface into the active region. Excitation Intensity The excitation intensity controls the density of photoexcited electrons and holes, which governs the behavior of these carriers. Each electron-hole recombination mechanism has a distinct functional dependence on the carrier density. For example, the number of interface and impurity states is finite, and recombination at these states will saturate at high excitation photon densities. When the excited carrier density is low, the recombination process is dominated by discrete defect and impurity sites at the interfaces and within the bulk of the material. For this condition the excitation power density is typically less than 10 W/cm2, and repeatable peak wavelength measurements are achieved at higher excitation power densities. For example, good correlation between the PL intensity and leakage current has been reported for InGaAs-based PIN photodetector structures at very low excitation level. The less critical surface state effects can be minimized by direct excitation of the InGaAs region. PL Mapping of Optoelectronic Devices The optical transition energy obtained from PL analysis can be used to determine the key specifications for optoelectronic devices, the color of LEDs and the wavelength of laser diodes. Wafer level testing benefits manufacturers by detecting defective wafers at an early stage of device production, thus reducing production costs. An example of a PL analysis system is the Accent Rapid Photoluminescence Mapping (RPM) system. This system has been developed to meet such requirements and can provide image quality PL maps within minutes. Four key parameters define PL peaks: the peak wavelength, the peak intensity, the full width at half maximum (FWHM) and the area of the peaks (integrated PL intensity). The simplest algorithm to determine the peak position is to choose the wavelength at which the intensity is a maximum. However, this algorithm is very noise sensitive over broad peaks. An alternative method is to use intersections. The wavelength can be chosen in the center for a symmetrical peak or at a fixed weighting ratio for an asymmetrical peak. Although both peak intensity and the area of the peaks can be used to measure the PL strength, the physics behind each can be different. For a multiple QW sample, broader peaks are usually associated with the uniformity of well width or well composition. This would not be a problem for LEDs, providing the color changes were not recognizable to the human eye. But it could increase the threshold current dramatically for a laser diode, due to a reduction in peak gain. In practice, it is necessary to develop a reliable analysis method to produce the useful statistical results from thousands of spectra in the mapping data. Using the Accent RPM system, a single recipe file has been used to set up all the variable parameters required, from data collection, storage statistical analysis and print out. Great care must be taken for all the measurement conditions, excitation wavelength, power density as well as the sample temperature. The PL peak parameters are all excitation power dependent. Integrated PL Intensity Mapping Integrated intensity maps are widely used as a quantitative measure of the quality of epitaxial wafers in production. shows an integrated PL map of the GaAs peaks at the interface of AlxGa1-xAs/GaAs QWs grown on the GaAs substrate. The feature near the prime flat is probably due to the handling damage. The red "blobs" at the wafer s center and in the four-fold distribution were seen in all the results from wafers from an individual growth run and were caused by temperature hot spots. Some of the features seen on the top of the wafer map were not seen on the map of QW emission, indicating that these defects were restricted to the substrate and did not penetrate through to the epilayer. Spectral Mapping The most important parameter for the optoelectronic device is the emission spectrum, especially from the active region. The tolerance of devices such as Nd:YAG or Er-doped fiber pumping lasers is much tighter than that for LED wafers. The PL mapping results shown in were obtained from an wafer containing red LEDs that was rejected. Although the wavelength variation is very smallless than 0.6 nmthe average wavelength is too short compared to the specification of 625 5 nm. The large variation in the PL intensity from this wafer was mainly due to non-uniformity in the thick window layers grown on top of the surface. All the statistical data shown in Figure 3 excluded data points less than 3 mm away from the edge of the wafer (indicated by the white line on the maps). For a ternary material, PL measurements can be used to calculate the composition. In the case of AlGaAs, the resolution can be higher than that commonly achieved using x-ray diffraction, and PL has proved particularly useful for electronic devices that require better than 1% repeatability in the Al composition. The Accent RPM system also allows white light reflectivity measurement. Reflectivity is employed in a production environment to control the composition and is useful in the analysis of Bragg mirror structures in VCSELs, and in LEDs. Epitaxial structures are increasingly being used to manufacture optoelectronic devices. Characterization of these structures at wafer level using non-contact high-speed photoluminescence mapping increases yields and benefits QA in today s compound semiconductor production environments. * Contact: Shouyin Wang. Tel: +44 20 8328 2290; E-mail: swang@accentopto.com or Gyles Webster. Tel: +1 650 934 8800; E-mail: gwebster@accentopto.com.