Date of Award
Fall 2022
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Engineering and Applied Science
First Advisor
Xia, Fengnian
Abstract
Sensing is essential in every corner of our world nowadays, from on-chip cameras to medical tests, autonomous vehicles, etc. In conventional applications of sensing, physical properties, such as the radiant power of light and the amplitude of magnetic field, are directly measured by the sensors. However, only limited types of physical properties can be directly read out from those sensors in measurements, and the accuracy of these measurements is strictly limited by sensors themselves. Compared to the direct measurements of these physical properties, by utilizing post-measurement processing or more specifically machine learning techniques, intelligent sensing can help to infer properties that are hard to be directly measured or to produce more accurate results than direct measurements. Generally, the advances of intelligent sensing are enabled by three factors, the new designs of sensors, the ever-growing processing power of modern computers, and the developments in algorithms. Among these three factors, we explore different designs of tunable sensors and pinpoint appropriate algorithms to interpret sensor outputs and infer corresponding physical properties. The information to be sensed includes the spectrum and polarization of light, and the magnitude of magnetic field. We first explore the photoresponse and its tunability in optical materials, then encode the information to be sensed into sensor outputs using these materials. Finally, by using machine learning techniques, we can decode the sensor outputs and infer the information of interest. In this thesis, we first explore the tunability in black phosphorus (BP) and its isoelectronic materials. Two different approaches, the alloying and the electrical tuning, are studied to realize wide tunability in BP. We probe the electrical and optical properties of BP using these two approaches, and the band gap of BP is shown to be dramatically reduced. A mid- infrared radiation up to a wavelength of 7.7 μm can be detected in either approach, while intrinsic BP originally shows little response to radiations beyond 3.7 μm. Most importantly, in the electrical tuning approach, the vertical electrical displacement field can extend cut- off wavelength continuously. By utilizing this continuously tunable BP photodetector, we demonstrate a mid-infrared spectrometer in the 2–9 μm spectral range. Then, the photoresponse and its tunability in graphene and its moiré superlattice structures are investigated. The photocurrent generation mechanism in pristine graphene is first clarified. Both the transport and photoresponse show unique features in the transition between the metallic and insulating behaviors in graphene. Electron-phonon scattering channels are revealed to play different roles depending on the doping level of graphene. Furthermore, we show that high-quality graphene is an ideal high-speed bolometric material for the mid-infrared photodetection considering its broadband absorption, small heat capacity, and large temperature coefficient of resistance (TCR). These understandings enable a further improvement in twisted bilayer graphene (TBG). The enhancement of photoresponse in TBG benefits from the higher density of states and the reduced phonon coupling at superlattice gaps compare to monolayer graphene. Besides the enhanced photoresponse, the twist also leads to symmetry breaking in crystal structures. The symmetry of twist double bilayer graphene (TDBG) is dramatically reduced compared to monolayer graphene. This symmetry breaking gives rise to the tunable bulk photovoltaic effect (BPVE) in TDBG. Due to the BPVE, the tunable quantum geometric properties of TDBG generate two-dimensional photovoltage mappings, which are fully governed by the polarization states and wavelength of incident light. We then show that this tunable BPVE in TDBG and convolutional neural network (CNN) can enable a sub-wavelength-scale mid-infrared polarimeter. Beyond the demonstration of the miniaturized spectrometer and polarimeter, we explore an intelligent sensing scheme for magnetometers. We show that the high carrier mobility in graphene is highly desired in Hall sensor applications, and the noise in graphene magnetometers can be dramatically reduced using a real-time total variation denoising algorithm.
Recommended Citation
Yuan, Shaofan, "Intelligent Sensing Enabled by Tunable Optical Materials" (2022). Yale Graduate School of Arts and Sciences Dissertations. 794.
https://elischolar.library.yale.edu/gsas_dissertations/794
