Investigation of the Dominant Noise Mechanism in MIM SETs at Low Frequencies

Single-Electron Transistors (SETs) are the most sensitive electrometers known to-date (sensitivities of 0.9 μe/Hz^1/2) [1, 2]. Advancement and optimization of SETs are important to fields such as quantum-based metrological standards [3-5], electron-spin based quantum computing and spintronics [6, 7], and scanning probe technology. Noise, the random fluctuations in voltage, current and/or resistance intrinsic to the device that obscure a signal, limits the sensitivity of these devices. To push the sensitivity of these devices to their ultimate limits, the noise sources in SETs need to be identified and studied to determine if and how they can be removed or reduced. This thesis will present the findings and conclusions on flicker noise from measurements made on Al/AlOx/Al SETs.

Flicker noise, or 1/f noise, is a significant component of the noise power spectral density in SETs at low frequencies (<1kHz). The flicker noise found in the SETs presented in this thesis originates from a superposition of the noise power from numerous two or more level fluctuators. In the limit of one, two-level fluctuator, the time domain noise signal resembles that of telegraph signals as a result of the fluctuator changing between its two possible states. Since these “signals” are random, this type of noise is called random telegraph signals (RTS). In order to avoid flicker noise, the radio-frequency single-electron transistor (rf-SET) [1, 8, 9] was developed, which moves the measurement frequency into higher frequencies where the 1/f noise is negligible compared to thermal and shot noise. While this engineered solution provides a work around for current SET applications, this thesis uses the study of flicker noise in SETs as an opportunity to characterize material imperfections. Identification of these imperfections guided the proposal of future work that will minimize their effect on the performance of the device across the frequency spectrum. Of special interest is the improvement of the low-frequency SET measurements. Improvement of this spectral region would allow for improved sensitivities without needing to implement the more complex radio-frequency setup.

As discussed in detail in this thesis, a substantial contribution to the flicker noise in Al/AlOx/Al SETs stems from atomic reconfigurations within the tunnel barriers that make up the SET. During the formation of aluminum oxide, according the model presented by Cabrera and Mott [10], aluminum ions diffuse through interstitial sites in the oxide as they move toward the surface to combine with oxygen adsorbed to the surface. If this process is abruptly stopped, as done in this thesis work through the removal of the oxygen source and sealing the surface of the oxide by the deposition of more aluminum on the oxide surface, then the aluminum oxide will be non-stoichiometric and highly disordered. In addition to non-stoichiometric aluminum oxide, another non-ideality of thermally grown aluminum oxide is the presence of hydroxide ions in the oxide. This is a well-known phenomenon which results from the dissociation of water at the oxide surface during its formation. The disorder of the constituent ions and presence of hydroxide ions are two possible physical sources that give rise to the two-level systems that produce the flicker noise presented in this thesis. Since this noise is a result from imperfect ion coordination within the oxide and impurity ions, it is proposed that thermal annealing in high vacuum will provide the oxide with sufficient energy to achieve a more energetically stable configuration. Results from the literature confirming this hypothesis are briefly discussed.