Investigation of Real-Space Transfer Noise in InP Quantum Wells

TL;DR

The paper addresses the origin of drain noise in InP HEMTs by testing a real-space transfer (RST) noise theory using gate-less transfer-length method structures with two barrier compositions to alter confinement without changing channel transport. A two-component model is developed, separating thermal noise $T_e$ and RST noise $T_{RST}$, and experiment measures microwave noise temperature $T_{N||}$ across varying electric fields and recess lengths; fitting yields electron-energy relaxation time $\overline{\tau}_E$ and RST relaxation time $\overline{\tau}_{RST}$, demonstrating that $T_{RST}$ can dominate at higher fields and that barrier composition and recess length modulate this contribution. The results qualitatively agree with the model, supporting RST as a contributor to drain noise in InP HEMTs and suggesting that enhanced quantum confinement via barrier engineering could mitigate noise. This work provides a pathway to understanding and reducing drain noise in state-of-the-art InP HEMTs, with implications for quantum computing and high-frequency applications.

Abstract

Indium phosphide (InP) high electron-mobility transistors (HEMTs) are widely used in many fields such as quantum computing because of their unparalleled microwave noise performance. Achieving improved noise performance requires a physical understanding of the noise mechanisms. Here, we experimentally test a theoretical proposal for drain (output) noise as originating in part from real-space transfer (RST) by characterizing the microwave noise temperature of transfer-length method structures with the same channel composition but two different barrier compositions. This choice was made to alter the confining potential of electrons in the channel, thereby affecting the RST mechanism, while avoiding changes to the channel transport properties. We observe trends of noise temperature with physical temperature and source-drain voltage which are compatible with the predictions of RST noise theory. This finding supports the hypothesis that RST contributes to drain noise in HEMTs.

Figures (6)

• Figure 1: (a) Cross section schematic of the ungated HEMT device. The epitaxial layers are grown on an InP substrate. From bottom to top, the epitaxial layers include a 500 nm In$_\text{0.52}$Al$_\text{0.48}$As buffer, a 15 nm In$_\text{0.53}$Ga$_\text{0.47}$As channel, a 9 nm In$_\text{x}$Al$_\text{1-x}$As barrier/spacer, a 4 nm InP etch stop, and a 20 nm In$_\text{0.53}$Ga$_\text{0.47}$As cap. The $\delta$ doping is located at 4 nm above the channel. A recess with controlled recess length $L_\text{re}$ is etched on the cap to the etch stop. (b) Top view SEM image of the ungated HEMT device. The cross section shown in (a) is located at the center of the device.
• Figure 2: (a) Room-temperature I-V characteristics of the LM device. Inset: drift velocity versus electric field. (b) Representative $S_{11}$ data on a LM device with $L_\text{re}$ = 13 from 3.5 - 8.5 GHz.
• Figure 3: (a) Representative raw noise power data versus time for a LM sample with recess length $L_\text{re}$ = 13 at 5 GHz and room temperature. The pulsed bias $V_\text{ds,Hot}$ (black line) and the diode detector hot voltage $V_\text{N,Hot}$ (orange line) are shown. (b) Zoom-in view of (a) showing a single bias pulse and the noise power data respectively. The offset between the bias pulse and noise measurement is due to a slight difference in trigger timing between the noise voltage and bias voltage oscilloscopes, which is accounted for in software.
• Figure 4: (a) Noise temperature parallel to the channel ($T_{N}||$) of LM sample vs electric field. All data are taken at 5 GHz at room temperature. (b) Analogous plot for TB sample.
• Figure 5: Noise temperature parallel to the channel ($T_{N}||$) of LM and TB samples versus electric field. The black (orange) data are taken on $L_\text{re}$ = 13 devices on the LM (TB) sample at 5 GHz at room temperature. The dashed lines are the predicted electron temperature ($T_{e}$) from \ref{['eq:Te']}. The dash-dotted lines are the calculated real-space transfer noise ($T_{RST}$) from the model described in \ref{['eq:TN']}. The solid lines are the combined noise temperature predicted by the electron temperature and the real-space transfer model. Inset: low field region, showing that both fitted model and the measured data are about 300 K at zero field.