Optical monitoring comes of age at IC-MOVPE 2002

Metalorganic supplier Mochem was the center of much attention. As the conference opened, the company announced an agreement to join with Akzo Nobel to create a new metalorganics supplier combining Mochem s expertise in group V sources with Akzo s larger group III operation (Compound Semiconductor July 2002 p23). Mochem also ensured that crowds regularly gathered at its stand by installing a large-screen TV to show live World Cup soccer matches.

MOVPE is a relatively mature technology that is now widely used in mass-production environments. As such, much of the technical program focused on its application to the growth of particular structures rather than the MOVPE process itself. These applications included lasers, LEDs and HBTs, covering material systems as diverse as II-VIs, antimonides, dilute nitrides, InP and GaAs. With MOVPE being applied to the production of blue LEDs and lasers, and the development of high-power HEMTs, the III-nitrides were particularly well covered.

The conference may be best remembered as the one at which in situ optical monitoring came of age. A large number of talks were given by those implementing optical techniques to monitor and control the MOVPE process (Compound Semiconductor June 2002 p41). ORS of St Asaph, UK, and LayTec of Berlin gave presentations on their respective growth monitoring technologies, and many talks not specifically about growth monitoring made reference to its use. This indicates that such monitoring equipment is now being seen as an essential accessory to an MOVPE reactor.

GaN on Si

An invited talk from Armin Dadgar of Otto-von-Guericke University gave a good overview of the driving forces behind worldwide efforts to grow GaN on Si, and also the difficulties in achieving device-quality GaN on this substrate (Compound Semiconductor July 2002 p45). Dadgar pointed out that the main problem was the large difference between the thermal expansion coefficients of GaN and Si, and not the large lattice mismatch. The difference in thermal properties causes cracking as the wafer cools after growth or is cycled to different temperatures during deposition. This prevents large-area, thick, high-quality GaN layers being successfully grown. Dadgar therefore focused his talk on the strain engineering methods his group has adopted to grow thick, crack-free layers for LEDs and transistors.

The first of these methods was selective-area deposition to give islands of thick GaN small enough to prevent cracking. Crack-free LED structures containing AlGaN/GaN multilayers 2.5 µm thick were produced over an area of 300 x 300 µm2. Crack-free layers 3.6 µm thick were possible over a 100 x 100 µm2 area.

The second approach involved the use of low-temperature AlN interlayers 10-15 nm thick. Using these it was possible to grow crack-free device structures over the whole wafer. The low-temperature AlN layers introduce compressive stress to compensate the tensile thermal stress. Depositing the AlN interlayers at lower temperatures reduces the layer stress and dislocation density. By using this approach Dadgar s group was able to fabricate transistors and blue LEDs.

GaN ELO on Si

Several talks focused on epitaxial lateral overgrowth (ELO) as a route to achieving high-quality GaN on Si. Hiroyuki Naoi of Mie University in Japan showed how his group had successfully produced areas of GaN free from cracks by using grooved Si substrates. A 0.3 µm GaN layer was deposited on a Si wafer, followed by SiO2 masking stripes 5 µm wide and 100 µm long. The GaN and Si was etched to a depth of 1 µm and the SiO2 mask removed to leave behind GaN seed stripes for subsequent ELO. The stripe areas covered the Si wafer in a grid pattern to enable crack-free growth. Each 100 x 100 µm2 area of stripes was separated by 20 µm from its nearest neighbor.

Prior to GaN growth, the Si surface was nitrided in the MOCVD reactor. GaN growth was initiated on the seed stripes without deposition on the nitrided Si. At a growth temperature of 1000 °C the GaN layer successfully coalesced. The growth temperature was then increased to achieve a smooth surface and the dislocation density in the GaN islands was 1 x 108/cm2 (figure 1).

Paul Hageman of Nijmegen University, the Netherlands, reported an alternative ELO method using overgrowth of holes etched into Si. The holes were 4 µm deep, 1.5 µm in diameter and separated by 2.5 µm. A 100 nm thick AlN nucleation layer was deposited at 850 °C, and GaN growth was at 1170 °C. Figure 2 shows an SEM of a 2 µm thick GaN layer as it begins to overgrow the holes. Deposition starts on the AlN between the holes and then begins to coalesce over the holes. Growth in the vertical direction is faster then the lateral direction leading to inverted pyramid-shaped holes in the GaN layer. Complete coalescence was seen after 3 µm.

TEM of the GaN layer showed vertically propagating dislocations over the AlN. Dislocations in the overgrown GaN were only present in the basal plane and resulted from the bending of vertical dislocations in the GaN adjacent to the holes. On a larger scale, the GaN layer began to crack at a thickness of 2 µm.