News Article
Polarisation-free GaN shows promise for visible photonics
A new development at IBM culminates in a stress-free gallium nitride material ideally-suited for enabling polarisation-free visible light emitters
GaN-on-silicon LEDs promise to cut the cost of solid-state lighting and accelerate its adoption, but they suffer from a high dislocation density that stems from significant lattice and thermal mismatches between the substrates and epilayers.
However, scientists at IBM have shown that these issues can be addressed by growing a unique nanopattern on a CMOS-compatible silicon substrate –a cubic phase of stress-free GaN can result in a dislocation density of less than 108 /cm2, which is three orders of magnitude below that of conventional GaN-on-silicon.
What's more, turning to the cubic phase can improve LED performance. Conventional GaN materials are grown on three- or six-fold base-symmetric substrates such as sapphire, SiC, and silicon (111).
As such, conventional GaN devices are hexagonal (wurtzite) phase structures that are plagued by very high polarisation fields along the common growth direction of < 0001 >.
These large polarisation fields produce shifts in wavelength and current, and are thus detrimental to the performance of both photonic and vertical transport electronic devices. So, if these polarisation fields could be eliminated, it is possible that the performance of GaN devices could hit a new high.
The GaN formed by Dr. Can Bayram and co-workers from IBM could enable this, because it is polarisation-free along the common < 001 > growth direction. However, until now, this phase was thermodynamically unstable and required low temperature deposition conditions and unconventional substrates, such as GaAs.
Schematic sketch and TEM images of Ge, GaAs, and GaN grown on patterned Si (100) substrates. The threading dislocations inside the materials are shown as black lines in the sketches. Very different growth and dislocation behaviour is observed for GaN with respect to gerrmanium and GaAs. In the GaN case, high quality (defect density < 108 /cm2), planar, and cubic phase material is enabled
In order to overcome this thermal instability, Bayram developed a novel nano-groove patterning and MOCVD-compatible maskless selective area epitaxy processes.
After the integration of thermodynamically-stable, stress-free, and low-dislocation density-GaN on CMOS-compatible on-axis silicon, InGaN/GaN multi-quantum-well structures were grown on top. The material showed strong room temperature luminescence in the visible spectrum - promising polarisation-free visible emitter applications for this technology.
SEM images of (left) conventional and (right) selective MOCVD growth processes on the same groove structures
Furthermore, cubic phase GaN-on-silicon (100) has a cleavage plane suitable for mirror formation and shows promise for laser diodes.
Another strength of the IBM process is its compatibility with CMOS processing. Conventional GaN-on-silicon technologies employ silicon (111) substrates, a platform that renders them useless for co-integration with CMOS-devices.
That’s because (111)- oriented silicon substrates have a similar three-fold crystal symmetry to that of six-fold GaN in the (0001) plane. However, it is essential to grow GaN-based devices on on-axis silicon (100) substrates to integrate GaN devices with already-established silicon-based CMOS technology.
The performance of GaN devices is ultimately determined by the quality of the GaN. For many applications it would be ideal if one could achieve low dislocation densities (< 106 /cm2) and stress-free GaN over wafer-sized areas. However, according to Bayram, no such material has been demonstrated on CMOS compatible on-axis silicon (100).
Broader growth demonstrations of selective GaN regrowth on Silicon (100) substrate. This approach could be useful for interconnects as well as novel device architectures. Oxide pattern can be used to locate GaN layers along the periphery
For example, GaN-on-silicon (111) suffers from dislocation densities of greater than 1011/cm2 and wafer bow. Although single-crystal GaN can be epitaxially grown on off-cut on axis silicon (100) substrates, such substrates are not CMOS-compatible and GaN re-growth suffers from similar issues.
In comparison, the efforts at IBM have resulted in material growth compatible with conventional CMOS processes, and have provided a novel roadmap for making GaN-based visible emitters cost-competitive.
The work has been described in detail in the paper, "Cubic Phase GaN on Nano-grooved Si (100) via Maskless Selective Area Epitaxy," by Can Bayram et al in Advanced Functional Materials, 2014.
DOI: 10.1002/adfm.201304062
However, scientists at IBM have shown that these issues can be addressed by growing a unique nanopattern on a CMOS-compatible silicon substrate –a cubic phase of stress-free GaN can result in a dislocation density of less than 108 /cm2, which is three orders of magnitude below that of conventional GaN-on-silicon.
What's more, turning to the cubic phase can improve LED performance. Conventional GaN materials are grown on three- or six-fold base-symmetric substrates such as sapphire, SiC, and silicon (111).
As such, conventional GaN devices are hexagonal (wurtzite) phase structures that are plagued by very high polarisation fields along the common growth direction of < 0001 >.
These large polarisation fields produce shifts in wavelength and current, and are thus detrimental to the performance of both photonic and vertical transport electronic devices. So, if these polarisation fields could be eliminated, it is possible that the performance of GaN devices could hit a new high.
The GaN formed by Dr. Can Bayram and co-workers from IBM could enable this, because it is polarisation-free along the common < 001 > growth direction. However, until now, this phase was thermodynamically unstable and required low temperature deposition conditions and unconventional substrates, such as GaAs.
Schematic sketch and TEM images of Ge, GaAs, and GaN grown on patterned Si (100) substrates. The threading dislocations inside the materials are shown as black lines in the sketches. Very different growth and dislocation behaviour is observed for GaN with respect to gerrmanium and GaAs. In the GaN case, high quality (defect density < 108 /cm2), planar, and cubic phase material is enabled
In order to overcome this thermal instability, Bayram developed a novel nano-groove patterning and MOCVD-compatible maskless selective area epitaxy processes.
After the integration of thermodynamically-stable, stress-free, and low-dislocation density-GaN on CMOS-compatible on-axis silicon, InGaN/GaN multi-quantum-well structures were grown on top. The material showed strong room temperature luminescence in the visible spectrum - promising polarisation-free visible emitter applications for this technology.
SEM images of (left) conventional and (right) selective MOCVD growth processes on the same groove structures
Furthermore, cubic phase GaN-on-silicon (100) has a cleavage plane suitable for mirror formation and shows promise for laser diodes.
Another strength of the IBM process is its compatibility with CMOS processing. Conventional GaN-on-silicon technologies employ silicon (111) substrates, a platform that renders them useless for co-integration with CMOS-devices.
That’s because (111)- oriented silicon substrates have a similar three-fold crystal symmetry to that of six-fold GaN in the (0001) plane. However, it is essential to grow GaN-based devices on on-axis silicon (100) substrates to integrate GaN devices with already-established silicon-based CMOS technology.
The performance of GaN devices is ultimately determined by the quality of the GaN. For many applications it would be ideal if one could achieve low dislocation densities (< 106 /cm2) and stress-free GaN over wafer-sized areas. However, according to Bayram, no such material has been demonstrated on CMOS compatible on-axis silicon (100).
Broader growth demonstrations of selective GaN regrowth on Silicon (100) substrate. This approach could be useful for interconnects as well as novel device architectures. Oxide pattern can be used to locate GaN layers along the periphery
For example, GaN-on-silicon (111) suffers from dislocation densities of greater than 1011/cm2 and wafer bow. Although single-crystal GaN can be epitaxially grown on off-cut on axis silicon (100) substrates, such substrates are not CMOS-compatible and GaN re-growth suffers from similar issues.
In comparison, the efforts at IBM have resulted in material growth compatible with conventional CMOS processes, and have provided a novel roadmap for making GaN-based visible emitters cost-competitive.
The work has been described in detail in the paper, "Cubic Phase GaN on Nano-grooved Si (100) via Maskless Selective Area Epitaxy," by Can Bayram et al in Advanced Functional Materials, 2014.
DOI: 10.1002/adfm.201304062
