UC Berkeley

Tunneling two-level systems (TLS) are considered by the glass community to be a hallmark characteristic of amorphous materials and even the disorder that is inherent to these systems. They are observed in all glassy materials and have thus historically been deemed “universal,” yet their mutability under prescriptive growth conditions emphasizes the plasticity of the structure-function relationship also inherent to disordered systems. Despite their theoretical and empirical identification almost fifty years ago, a structural characteristic of the amorphous phase that is both vastly prevalent and occasionally reducible which explains trends seen in TLS density has yet to be identified. This is in part due to how difficult it is to characterize amorphous structures relative to their crystalline counterparts. In crystalline materials, characterization of structure at one length scale is often generalizable to other length scales, whereas in amorphous materials, structural traits at, say, subnanometer length scales do not have a reliable relationship to structural traits occurring at tens of nanometers.

The monatomic nature and narrow distribution of hybridization types in amorphous silicon (a-Si) offers it as a relatively straightforward system by which to study covalent random networks. Low-TLS a-Si and a-Si:H are candidate dielectrics for phase resonant qubits; the presence of TLS is a limiting factor in their coherence time. Low-TLS a-Si is also a candidate for LIGO’s interferometer mirror coating which would increase precision detection of gravitational waves beyond its current means. In this work, we utilize an array of structural characterization techniques to develop a well-rounded description of a-Si thin film structure from $\sim$ 0.1-10 nm length scales as a function of growth conditions. We then relate these findings to trends seen in TLS in a-Si films grown under the same conditions.

After an introductory chapter, Chapter 2 describes findings from atomic density measurements, Raman spectroscopy, electron diffraction, high resolution transmission electron microscopy, fluctuation electron microscopy, electron energy loss microscopy, doppler broadening spectroscopy, electron spin resonance, sound velocity, and atomic force microscopy. Additionally, we etch samples and remeasure their atomic density and Raman spectra in order to understand better the depth dependence of those properties. We find that a-Si thin films develop bilayers within $\sim$ 20 - 30 nm from the substrate interface. These bilayers likely result from accumulated compressive strain in the bottom layer which, when saturated, causes an interface to form which initiates growth of tensile columnar structures. We find evidence that nanovoids present throughout the film coalesce as films grow thicker.

In Chapter 3, we compile new and old findings in TLS density as measured by nanocalorimetry and internal friction; additionally, we include a new analysis of previously unpublished a-Si:H TLS data. We find that TLS may be present in two structural traits present in a-Si thin films: at the surfaces of voids or nanovoids, as well as in the highly strained regions at or near the bilayer interface. An observed decoupling at low film densities of TLS measured by nanocalorimetry from those measured by internal friction can be explained by a theory put forth by Coppersmith in 1991. This theory posits that TLS which exist in frustrated regions are decoupled from phonons, whereas TLS which remain coupled to phonons have their tunneling suppressed by an effective interaction field and are thus less likely to be observed. Thus, phonon-decoupled TLS which exist in geometrically frustrated regions are observable via nanocalorimetry, but not via internal friction. TLS can be attributed to nanovoid surfaces for films with densities between 4.5 - 5 $\times$ 10$^{22}$ atoms/cm$^{2}$; a-Si film density shows highly regular trends as a function of growth conditions (especially growth temperature) and displays reliable tunability. TLS density in films with atomic densities less than 4.5 - 5 $\times$ 10$^{22}$ atoms/cm$^{2}$ show the decoupling between nanocalorimetry and internal friction measurement techniques; TLS in strained regions continue to cause an increase in nanocalorimetric TLS density as atomic density decreases, while the TLS density as measured by internal friction plateaus.

a-Si:H results support that TLS presence depends on both film density and hydrogen content, both of which are nontrivially related to growth temperature. As with a-Si, we find that TLS density in a-Si:H decreases as atomic density increases. Chapter 4 concludes this work by reviewing its main findings.