Germanium takes the Strain out of GaN-on-silicon epi
If you want to grow n-doped GaN layers on silicon with minimal tensile strain, then consider using germanium, rather than silicon as the n-type dopant.
That’s the key finding of a study by researchers from the Otto-von-Guericke-University Magdeburg, Germany, that have compared the tensile strain of GaN layers doped with both of these group IV elements.
Figure 1 : In-situ curvature measurement, demonstrating the benefit of germanium-doping in comparison to silicon-doping in GaN-on-silicon
The researchers say that until now, good quality thick n-type doping of GaN-on-silicon has not been possible due to edge-dislocation climb. This results in an increasing tensile strain during epitaxial growth. In fact, previous studies have shown that doping GaN-on-silicon with silicon actually enhances edge-dislocation climb.
The scientists compared samples grown by MOVPE using standard precursors for AlN and
GaN growth processes. These were diluted silane (100ppm) and germane (10%) in H2for the silicon and germanium doping processes, respectively.
Germanium doping has already been tested by Nakamura in the early 1990s. However, with a lower doping efficiency and as a more costly option to silicon doping, germanium doping has not been established. The recent results obtained by the team headed by Alois Krost, on the other hand, show the absence of edge-dislocation climb in the case of germanium-doped GaN-on-silicon.
Krost points out that using germanium as a dopant opens up the possibilities of realizing a wide range of GaN-on-silicon devices, where thick highly n-conducting layers are required.
The researchers also found evidence that the formation of SiN at the edge dislocation core is the initiator of an enhanced dislocation climb, even for already tensely strained layers. This is in contrast to previous assumptions that the surface roughness of highly silicon-doped GaN-on-silicon leads to edge-dislocation climb.
"Our idea to dope GaN-on-silicon with an alternative dopant is quite old and reaches back to the early 2000s when we realized that silicon doping is unfavorable in regard of strain development, but we had no resources to investigate alternatives in detail," says Armin Dadgar.
"Thus having waited such a long time to test it, we are glad that our assumption was right and will give an additional impulse for GaN-on-silicon device applications".
The researchers expect not only an impact for GaN-based LEDs grown on silicon substrates, but also in power electronics. For example in vertical Schottky diodes which could now be grown with a highly conducting n-type layer and a thick, undoped layer on top without any impact on strain development.
-200111.jpg)
Figure 2: Image of LEDs grown on silicon-doped GaN-on-silicon ; an illustration for one application
Furthermore, n-doped HVPE layers for GaN pseudosubstrates would benefit from a reduced strain in the lower, more defective region of the layer.
This research is further described in the paper “Crack-Free, Highly Conducting GaN Layers on Si Substrates by Ge Doping,” by Armin Dadgar, Jürgen Bläsing, Annette Diez, and Alois Krost in Applied Physics Express 4, (2011) 011001 (DOI:10.1143/APEX.4.011001)
Further details of III-Vs on Silicon can be obtained from the publication “III–V Compound Semiconductors: Integration with Silicon-Based Microelectronics”, Editors: Tingkai Li, Michael Mastro, and Armin Dadgar, CRC-Press (2010), ISBN 9781439815229.
