Materials for Quantum Computing
Quantum computing is emerging as the next digital revolution. In solid-state implementations computing errors ultimately originate from the materials comprising the quantum bits (qubits). Fault tolerant qubits may be encoded in exotic quantum states that exist in topological low dimensional systems. One of the most promising materials platforms for this purpose is hybrid superconductor/ semiconductor/ ferromagnet interfaces. The desired quantum computing functionality, or even the very existence of exotic topological quantum states, depend critically on the nature of the interface at the atomistic level. Epitaxial, atomically sharp, defect free interfaces are particularly desirable.
Our goal is to explore the materials space of hybrid superconductor/ semiconductor/ ferromagnet interfaces in the context of quantum computing. We aim to discover new material combinations for epitaxial interfaces, in which topological phases may exist, as well as metastable interface phases with desirable properties that may be stabilized by epitaxial templating. To this end, we employ a computational approach combining density functional theory (DFT) with lattice matching and genetic algorithm optimization. Lattice matching is used for screening a large number of candidate interfaces and genetic algorithms are used for detailed studies of the most promising candidates. This may guide the experiments of our collaborators in promising directions and advance the realization of new quantum computing schemes.
Machine learning the Hubbard U parameter in DFT+U
Within density functional theory (DFT) approximate exchange-correlation functionals are used to describe the many-body interactions between electrons. Semi-local functionals, such as the generalized gradient approximation (GGA) of Perdew, Burke, and Ernzerhof (PBE) only contain a dependence on the local density and its gradient. Therefore, they are computationally efficient. However, these functionals fail to describe correctly the electronic properties of some materials, owing to the self-interaction error (SIE), the spurious electrostatic interaction of an electron with itself. For example, for the semiconductor InAs, one of the key materials currently considered for quantum computing applications, the PBE functional produces no band gap, as shown below. The effect of SIE may be mitigated by using hybrid functionals, which contain a fraction of exact (Fock) exchange, such as the Heyd-Scuseria-Ernzerhof (HSE) functional, as shown below for InAs. However, owing to the non-locality of the exact exchange, the computational cost of hybrid functionals is too high for simulations of e.g., large interface models with hundreds of atoms. The DFT+U approach offers a balance between accuracy and efficiency by adding a Hubbard U correction to certain orbitals. The value of U is a parameter, which is often determined by fitting to experiments. We have developed a new method of machine learning the Hubbard U parameter by Bayesian optimization (BO). The objective function is formulated to find the value(s) of U that reproduce as closely as possible the band structure obtained from a hybrid functional. DFT+U(BO) produces band structures of comparable quality to hybrid functionals at the computational cost of a semi-local functional, as shown below for InAs. This enables performing unprecedented simulations of interfaces between materials that could not be described reliably with semi-local functionals.
Electronic structure of InAs and InSb surfaces
Superconductor/ semiconductor interfaces are among the leading materials platforms for the realization of Majorana-based quantum computing schemes. The superconductor is often grown epitaxially on top of the narrow-gap semiconductors InAs and InSb. It is important to understand the electronic structure of InAs and InSb surfaces because it plays a key role in the properties of quantum devices. We have studied the electronic structure of InAs(001), InAs(111), and InSb(110) surfaces using a combination of density functional theory (DFT) and angle-resolved photoemission spectroscopy (ARPES). DFT simulations of large surface models are enabled by the DFT+U(BO) method we have developed. To facilitate direct comparison with ARPES results, we have implemented a “bulk unfolding” scheme by projecting the calculated band structure of a supercell surface slab model onto the bulk primitive cell. The band structures obtained from DFT+U(BO) are overall in excellent agreement with ARPES experiments. The combination of theory and experiment has enabled us to elucidate the effects of surface reconstructions on the electronic structure of the InAs(001) surface and the effect of oxidation on the electronic structure of the InAs(111) and InSb(110) surfaces. For InAs(111), shown here, oxidation leads to significant band bending and produces an electron pocket. This is attributed to charge transfer from the surface In atom to the adsorbed O atom.
Topological properties of the SnSe/EuS(111) interface
SnSe in a topological crystalline insulator (TCI), a material that is insulating in the bulk but has a conducting state at the surface, which is topologically protected by time reversal symmetry and the crystal symmetry. We conducted DFT simulations of an epitaxial interface between SnSe and the ferromagnetic insulator (FMI), EuS. The magnetic proximity effect breaks the time reversal symmetry and opens a gap of 21 meV at Γ and 9 meV at M in the topological interface state of SnSe, shown in red. Because the magnetic proximity effect is short ranged and confined to the interface, the topological state at the top surface of SnSe, shown in green, in unperturbed. Charge transfer at the interface leads to band bending and shifts the gapped interface states below the Fermi level by 88 meV at Γ and above the Fermi level by 47 meV at M. The topological interface state of SnSe/EuS (111) has a larger gap and a smaller binding energy than e.g., Bi2Se3/EuS. Hence, it could be a potential candidate for implementing quantum/ topological devices.