Epitaxial Optimization of InP High Electron Mobility Transistors in Low-Noise Amplifiers for Qubit Readout

Doctoral thesis, 2024

The indium phosphide high electron mobility transistor (InP HEMT) is used in cryogenic low-noise amplifiers (LNAs) at C-band (4-8 GHz) for qubit readout in superconducting quantum computing. Improving the LNA performance with respect to noise, gain and dc power consumption is essential to meet future demands when upscaling quantum processors to handle thousands of qubits. Still, the noise temperature of today’s best C-band LNAs remains nearly an order of magnitude higher than that of a quantum-noise-limited amplifier. This motivates further studies into the noise reduction mechanisms of the InP HEMT. This thesis provides evidence of the importance of epitaxial optimization in reducing noise in cryogenic InP HEMTs for LNAs in quantum computing. A 100-nm gate length InP HEMT process was refined to a recess-first approach to mitigate electrochemical etching deterioration. The correlation between the subthreshold swing and noise temperature of the LNA observed at 4 K implied that the subthreshold swing served as an indicator of the amount of carrier fluctuations in the InP HEMT channel giving rise to noise. The noise of a C-band three-stage hybrid LNA was investigated by varying the spacer thickness (1 to 7 nm) and channel indium content (53% to 70%) in the InAlAs-InGaAs heterostructure of the 100-nm InP HEMT at 4 K. At the optimum noise bias, the 5 nm spacer InP HEMT LNA exhibited a 1.4 K average noise temperature with 6.1 mW dc power consumption. The best performance achieved with 60% indium content, yielding an average noise temperature and an average gain of 3.3 K and 21 dB, respectively, at 108 μW dc power, represents a new state of the art. Channel noise, expressed as a sum of thermal and excess noise measured from 4 K to 300 K, showed excess noise dominance at 4 K, independent of temperature. Excess noise was observed to vary with the channel indium content even with the same drain current and gate length, suggesting that channel noise cannot be described solely by suppressed shot noise. Additional channel noise contributions were proposed to originate from real-space transfer and impact ionization.