With LED manufacturers narrowing device quality requirements, improvements to structural and optical properties of epiwafer surfaces have become critical. Doru Florescu from Veeco Instruments describes GaN-based LED layer optimization using atomic force microscopy.

As GaN-based LEDs expand into large-screen backlighting, outdoor signage and general illumination applications, manufacturers are seeking to improve the optical output efficiency of devices. Enhancements to packaging designs are one way of addressing performance thresholds to improve the device quality. Optimizing the surface morphology of LED constituent layers has become another key approach. Atomic force microscopy (AFM) has shown promising results in characterizing and helping to optimize GaN-based layers grown by MOCVD. In conjunction with finely tuned process parameters, AFM helps to significantly reduce surface defects, and ultimately contributes to brighter and more efficient devices. Surface topographyImproving GaN epilayers via AFM involves analyzing epitaxial surfaces measured at the nanoscale level. To obtain high-resolution images (typically with sub-Angstrom vertical and nanometer-scale lateral resolution), AFM uses small probes to monitor surface topography through direct, non-destructive contact with material samples. Unlike methods that can damage sample surfaces due to poorly controlled frictional forces, Veeco s "tapping mode" technique incorporates an intermittent contact process that avoids dragging probes across surfaces during scanning. Tapping mode AFM uses a probe with an end radius of only a few nanometers attached to a flexible cantilever. To provide a high-resolution image, the probe tip intermittently contacts the sample surface. During raster-scanning, the cantilever is oscillated at or near its resonant frequency, typically 100-400 kHz, depending on the type of cantilever, using a piezoelectric crystal. Changes in epilayer surface features modify the oscillation amplitude during the scan, generating a precise topographical image in a non-destructive manner. Since physical contact between the probe and sample is limited to short "taps", shear is effectively eliminated. This leads to long-lasting probes and higher-resolution images. Using nanoscale AFM surface morphology readings, growers can implement multistage optimization processes to attain smooth surfaces and control doping levels throughout device structures. Surface features such as "V-groove" defects that negatively affect performance can be minimized and controlled. In the case of MOCVD-grown GaN devices, 470 nm LEDs with an unpackaged chip-level output power in the 4-4.5 mW range and forward voltage less than 3.2 V at 20 mA have been consistently achieved. Fine-tuning LED structuresWhen used properly, AFM can be an effective tool for reducing the roughness of GaN-based LED surfaces at each layer in the structure, monitoring dopant levels in certain layers, and helping to determine substrate quality. Because substrate surface morphology can dramatically impact the quality of subsequently deposited films, selecting a substrate with the appropriate physical properties can be pivotal. For example, although widely used for GaN growth, sapphire substrates have lattice parameters that are not an exact match to GaN. The high density of crystalline defects caused by this mismatch amplifies the importance of the surface finishing of sapphire substrates. Using the detailed view provided by AFM, potential problems such as surface contamination and scratching due to aggressive polishing are easily observed and avoided. The major benefits of using AFM to optimize growth are realized in the n-type, active and p-type layers of GaN structures. Using material samples grown on an MOCVD platform, AFM proved an effective method for improving yield quality via its monitoring of the effects of growth condition adjustments. For bipolar, vertical GaN-based optoelectronic devices, silicon-doped GaN is the leading material for n-type layers, with a smooth surface essential for achieving sharper interfaces in LEDs. AFM provides a clear indication of surface quality, with a smooth surface showing atomic terrace lines and few dislocations, and a rough surface showing evidence of grain boundaries and numerous mixed dislocations (figure 1). Since the n-type layer is used as a template for the active region, subtle adjustments in the MOCVD process based on AFM observations of this layer can greatly improve overall device performance. Due to the tunable bandgap energy of InGaN, multiple quantum wells (MQWs) are commonly used as active layers for blue and green LEDs and laser diodes. Similar to the n-type layer, smooth surfaces are critical for high-power applications. In this case, AFM can screen for cracked surfaces and inclusions that are detrimental to thermal stability. Surface morphology of p-type layers also affects device performance. Since the p-type contact is fabricated directly above this layer, it is especially important to maintain smoothness to enhance light and heat extraction from the device. Optimized growth conditions for magnesium-doped GaN p-type layers result in few surface features, while non-ideal conditions display pyramidal features and a high degree of cracking (figure 2). In all of the layers described, as well as the substrate material, AFM can be used as both a sampling tool to monitor and calibrate manufacturing processes, and as a developmental tool for engineering new structures. Optimizing the active layerTo demonstrate the contributions of AFM optimization phases to LED performance, a four-period InGaN/GaN MQW active region of a blue GaN LED structure was monitored. By improving the quality of quantum heterostructures using AFM feedback as a reference, positive results are clearly visible on surface morphology images and these translate into increased device output power. In the initial stage of development, a highly defective surface is easily observable (figure 3a), with dZ/RMS values of 38.12 nm/4.69 nm, where RMS is the root mean square of height deviations. The image surface area differential (ISAD), another indication of surface roughness, is 5.29%. (The ISAD is the difference between the image s 3D surface area and its 2D footprint area, and for a perfect surface would be 0%.) Severe cracking leading to a high density of voids governs this active region surface. After adjusting growth conditions, a smoother surface with atomic terraces, typical V-shaped pits and inclusions is formed (figure 3b). The dZ/RMS values are now 65.72 nm/3.00 nm with an ISAD of 2.18%, with the higher dZ number possibly attributed to the presence of inclusions effectively elevated above the surface plane. To evaluate the effect of the device optimization between these two phases of AFM-based development, full LEDs were fabricated using the samples described. The unpackaged, chip-level output power improved from the 1.2-2.0 mW range to the 2.5-3.0 mW range. Using the same technique, further improvements to the optical and electrical properties of the structure were achieved by reducing the density of V-defects associated with indium-rich regions and completely suppressing the occurrence of inclusions. Using a typical morphology sample from the second stage of development (figure 4a), dZ/RMS values of 63.41 nm/5.87 nm were found with an ISAD of 4.8%. By precisely tuning the growth process to account for the second-stage findings, a surface with a high degree of smoothness, reduced density of V-pits and no inclusions was achieved (figure 4b). The new dZ/RMS values were 15.30 nm/0.52 nm with an ISAD of only 0.047%. As with the first set of improvements, full LEDs were fabricated and integrating sphere measurements were conducted. The unpackaged, chip-level output power improved from the 2.5-3.0 mW range of the second stage of development to the 4.0-4.5 mW range for the third stage of development. In addition, the third-stage sample exhibited improved thermal stability as revealed by photoluminescence measurements before and after the growth of p-contact GaN layers. These atomic-level improvements promise to help LED manufacturers to produce the brighter, more efficient GaN-based devices that are required for future solid-state lighting applications. Further readingTing et al. 2003 Morphological evolution of InGaN/GaN quantum well heterostructures grown by MOCVD. J. Appl. Phys. 94 1461.
Florescu et al. 2003 Investigation of V-defects and embedded inclusions in InGaN/GaN multiple quantum wells grown by metal organic vapor deposition on (0001) sapphire Appl. Phys. Letts 83 33.
Florescu et al. 2004 Using AFM to control and enhance structural and optical properties of GaN-based LEDs grown by MOCVD Veeco Applications Note AN80.