Microwave Power Performance of Buffer-Free AlGaN/GaN MISHEMT With MOCVD Grown Ex Situ SiN

This work investigates the effect of ex situ MOCVD-grown SiN<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</i> gate dielectric and PECVD SiN<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</i> passivation layer on the microwave power performance of buffer-free AlGaN/GaN MISHEMT. We fabricated devices on a series of four samples diced from a 4 inch epi: first two samples had no gate dielectric while the latter two had upto 3 nm of ex situ SiN<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</i> as the gate dielectric. In each of the categories, one sample had baseline 100 nm SiN<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</i> passivation deposited under high frequency plasma conditions, while the other sample had 100 nm bilayer SiN<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">x</i> passivation. Devices had gate length of 0.22 μm and width of 2 × 50 μm. Samples with no gate dielectric exhibited low 2DEG density ~7 × 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>12</roman></sup> cm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-<roman>2</roman></sup> possibly due to significant charge depletion by the gate Fermi pinning owing to a thin barrier (10 nm, Al<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>0.3</roman></sub>Ga<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>0.7</roman></sub>N). Moreover, these HEMTs exhibited high current collapse (CC) <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\ge$</tex-math></inline-formula>38% irrespective of the passivation scheme when subjected to 1 μs pulse. The MISHEMTs on the other hand, exhibited high 2DEG density ~10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><roman>13</roman></sup> cm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-<roman>2</roman></sup> which also manifested as increased threshold voltage. Gate lag reduced to a mere 4%–6% in MISHEMTs, while under gate–drain double pulsing, the collapse was found to be 27% and 11% for different passivation schemes. Drain current transient (DCT) studies revealed suppression of surface states upon ex situ SiN deposition, and hinted at localized defects. Finally, under 28 V Class B operation at 6 and 10 GHz, output power of 8.1 W/mm power added efficiency (PAE 68%) and 5.2 W/mm (PAE 47%) respectively were measured under pulsed loadpull conditions.