Gallium nitride has turn into the de facto materials in third-generation semiconductors. Nevertheless, making GaN epi wafers within the high quality you want and the thermal resistance you want are challenges that fabs are nonetheless making an attempt to beat.
The mismatch of lattice fixed and thermal enlargement coefficient between GaN epi layers and substrates akin to silicon, sapphire, and silicon carbide result in the dislocation and cracking of epi layers.
A standard methodology for thermal administration is utilizing substrates with excessive thermal conductivity, akin to SiC or diamond, because the heatsink. Nevertheless, each the lattice mismatch and the coefficient of thermal enlargement mismatch between GaN and SiC/diamond make the heteroepitaxy very difficult. Moreover, the standard nucleation layer reveals low thermal conductivity because of the defects and poor crystallinity. The thick buffer with low thermal conductivity provides important thermal resistance to the warmth dissipation path from the machine to the substrate, as many of the warmth is generated throughout the energetic layer on the prime. Defect and boundary scatterings throughout the transition layer, on the interface between the substrate and transition layer, and by near-interfacial dysfunction contribute collectively to giant thermal resistance.
Although there are decisions of substrates that can be utilized for rising GaN epi, some aren’t foundry-friendly, whereby CMOS processes are used. Another excuse is that the lithography instruments and different instruments for making CMOS gadgets which are cutting-edge can be found solely on larger-scale wafers. Therefore, GaN-on-Si with wafer sizes of as much as 12 inches has benefits. GaN-on-sapphire at 6 inches is comparatively cheap; nonetheless, many foundries don’t settle for sapphire, and its thermal conductivity is poor.
To develop high-quality GaN, costly substrates akin to bulk GaN and SiC are required. Due to this fact, the manufacturing value for the machine manufacturing is considerably larger than Si-based electronics. To attain cost-effective state-of-the-art GaN energy machine efficiency whereas effectively managing the generated warmth, the epi layer may very well be faraway from the substrate, enabling substrate reuse, and instantly bonded to a heatsink to enhance machine thermal efficiency. Nevertheless, current removing processes akin to these involving photoelectrochemical etching, mechanical spalling, and laser interface decomposition undergo from sluggish processing pace and/or important floor roughening/cracking, limiting the method yield and practicality of substrate reuse. Due to this fact, the method value of those standard strategies sometimes exceeds GaN substrate value, limiting manufacturing.
When the machine wants higher high quality, when it comes to dislocation density, thermal properties, and better frequencies which are wanted for high-voltage gadgets in energy for automotive, RF, and data-center purposes, GaN-on-SiC tends to be the best way to go.
Nevertheless, GaN-on-SiC is an costly resolution. As soon as the good-quality GaN epi layer is grown on the SiC substrate, you’re going to get a greater GaN machine for energy and RF purposes. The disadvantage is that SiC substrates are very costly. The SiC substrate is not wanted after the GaN epi layer is grown on prime of it.
To summarize:
- Massive GaN epi wafers of present know-how have larger dislocation density (poor crystallinity).
- GaN-on-Si wafers have a tendency to make use of very thick buffers and interlayers to handle the stress, making it tough to handle thermal conductivity.
- Most different substrates are very costly, and scaling to bigger wafers isn’t an choice.
References
Please go to the eBook for the whole article