Impact of Charge Transitions at Atomic Defect Sites on Electronic Device Performance

7.3 Outlook and future investigations

With the still ongoing downscaling of electronic devices and new emerging technologies such as quantum computer or 2D FETs, investigations of the charge transfer dynamics and the localization of charge will remain relevant to improve reliability of future electronic devices. In this thesis, the stable charge states of different defects with respect to the Fermi level and their structural relaxations upon charge trapping were analyzed. While this is sufficient to model the effective phonon mode in 1D and qualitatively analyze the probability of a charge transition at a certain defect site, quantifying the exact transition rates captured by Fermi’s golden rule also requires the calculation of the electron matrix element. The evaluation of this matrix element would allow to directly link charge noise as detected in electronic devices, such as MOSFETs or spin qubits, to specific defect charge transitions based solely on first principle calculations.

In previous investigations, this matrix element was often approximated due to the high computational costs of generating and analyzing realistic interface structures. In such cases, the matrix element between a delocalized bulk wave function and a highly localized defect wave function has been estimated by a tunneling factor so far [32, 2, 125]. If the charge transitions parameters were directly extracted from first-principles calculations, this could significantly improve the physical modeling of charge transitions in real devices with TCAD. The recently published MLIP for the oxidation of Si [207] makes the generation of Si/SiO $_{2}$ computationally feasible also for statistical analysis. Such investigations could be relevant for linking experimentally detected RTN in MOSFETs as well as Si spin qubits. Besides the oxide defects already investigated in this thesis, suspected defects types which might be the cause for the localization of charge also include for example dangling bonds or Si–H and O–H sites at the interface. Since qubits states are very sensitive already to small fluctuations, such calculations could also be helpful to better understand decoherence phenomena and improve the device design.

Similar to the modeling approach for amorphous silicon nitride, the GAP method could be used to train an MLIP to accurately model amorphous silicon oxynitride (SiO $_{x}$ N $_{y}$ ) thin films as employed in ultra-thin MOSFET devices. While this high- $κ$ insulator is of great importance for electronics, theoretical investigations of charge trapping in silicon oxynitride thin films are rare, presumably due to the complicated composition and high defect density in melt-and-quench generated samples. As demonstrated in this thesis, the MLIP approach is promising to model realistic amorphous structures and an extension of this force field to also include O could be helpful for further investigations. SiO $_{x}$ N $_{y}$ structures could also be generated by modeling the nitridation of Si/SiO $_{2}$ interfaces using ab initio MD or by training an MLIP specifically for the nitridation process, similar to what has been done for the thermal oxidation of Si [24, 207]. Subsequently, charge trapping sites for different Si, O and N concentrations can be investigated, including vacancies, over- and undercoordinated atoms and H-related defects.

The analysis of optical broadening can be extended from vacancies in $α$ -Al $_{2}$ O $_{3}$ to other defects studied in this thesis. While the effective phonon mode can account for the vibrational broadening of optically stimulated electron transfer from defect sites to the band edges, internal electronic defect transitions can not be evaluated within this framework. More sophisticated techniques, such as time-dependent DFT (TDDFT) or constraint DFT (CDFT), can be employed to analyze the optical transitions to higher defect orbitals. In particular, such calculations could guide the identification of electrically active defect sites in 2D materials, thus helping to achieve the necessary reliability to make industrial production of FETs employing 2D semiconductors feasible. Additionally to the hole trapping investigated in 1L-WSe $_{2}$ in this thesis, defects in different 2D transition metal dichalcogenide compositions could be theoretically analyzed to compare their affinity to reliability degrading phenomena related to NMP charge transitions.