November_EDFA_Digital

edfas.org 31 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 4 hole injection efficiency into the device. [20] Having a precisemeasurement of the size (which can vary from20 to 500nm), density of V-pits across LED (or micro-LED) and spatial distribution across wafers wouldallowsignificant advances in the prediction of device efficiencies and reliability. Figure 4 shows an example of data taken on a 4-in. green LEDon sapphirewafer. These data can directly be used for statistical process control. The same type of wafer-level visual can be obtained with hyperspectral measurements to generate equivalent maps to those obtained from photo- luminescence mappers. Hotspot analysis and model testing. With the development of automated CL solutions, new possibilities arise in statistical analysis of wafer level properties of epitaxial wafers, for example the testing of physical models using wafer scale information to identify hotspots. Figure 5 shows anexampleofwafer level testingof anexpectation model with subsequent residual analysis. Here the expectation model is tested, a simple linear relationship between CL emission energy and CL intensity, and the wafer-level hyperspectral data (recorded on the same LED wafer as shown in Fig. 4) is evaluated with respect to the given expectation model. In the residual wafer map, rather than a random distribution of residuals, two clusters of data points are seen on the wafer Fig. 5 The leftside shows adistributionplot of CL intensityagainst CLemissionenergy forwafer level hyperspectral data recorded on a 4-in. green LED on sapphire. The p-value of 0.841 shown in the inset represents the Pearson correlation coefficient for a linear relationship between the two variables. The right side shows thewafer-level residual map obtained by fitting the distribution of the data shown in the left distribution plot with a linear expectation model and locally resolving the residual from the fit. This allows the local deviations from expectation model to be resolved. Fig. 4 Example of panchromatic images taken on a 4-in. green LED on sapphirewafer. Local images are takenover fieldof views of at least 20µm(typicallyas shownon the left-handside) tohave statistically significant numbers, andcanbepost processed toyieldwafer-wide statistical information such as the sizes and density of V-shaped pits, two factors known to influence the device efficiency. Here, the constant light collection efficiency over a large area achieved by in-column, aberration-corrected collection optics is crucial for efficient feature recognition at post processing.