Modeling Si/SiGe quantum dot variability induced by interface disorder reconstructed from multiperspective microscopy

SiGe heteroepitaxial growth yields pristine host material for quantum dot qubits, but residual interface disorder can lead to qubit-to-qubit variability that might pose an obstacle to reliable SiGe-based quantum computing. We demonstrate a technique to reconstruct 3D interfacial atomic structure spanning multiqubit areas by combining data from two verifiably atomic-resolution microscopy techniques. Utilizing scanning tunneling microscopy (STM) to track molecular beam epitaxy (MBE) growth, we image surface atomic structure following deposition of each heterostructure layer revealing nanosized SiGe undulations, disordered strained-Si atomic steps, and nonconformal uncorrelated roughness between interfaces. Since phenomena such as atomic intermixing during subsequent overgrowth inevitably modify interfaces, we measure post-growth structure via cross-sectional high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM). Features such as nanosized roughness remain intact, but atomic step structure is indiscernible in $1.0\pm 0.4$~nm-wide intermixing at interfaces. Convolving STM and HAADF-STEM data yields 3D structures capturing interface roughness and intermixing. We utilize the structures in an atomistic multivalley effective mass theory to quantify qubit spectral variability. The results indicate (1) appreciable valley splitting (VS) variability of roughly $\pm$ $50\%$ owing to alloy disorder, and (2) roughness-induced double-dot detuning bias energy variability of order $1-10$ meV depending on well thickness. For measured intermixing, atomic steps have negligible influence on VS, and uncorrelated roughness causes spatially fluctuating energy biases in double-dot detunings potentially incorrectly attributed to charge disorder.