(Invited) Isotopically Engineered Silicon Testbeds for Advanced CMOS and Quantum Information

Abstract

Silicon isotope engineering invokes a new paradigm into material science for nanoscale device applications [1]. The development of planer CMOS devices confronts an immediate issue of its downscaling limit, requiring a basal changeover to three-dimensional (3D) vertical transistors as a possible solution. General gate-all-around transistors involve formation of reliable gate-oxide film surrounding Si nanopillar structures. For the precise 3D process simulator, silicon isotopes are used as a self-diffusive marker for experimental determination of its model parameters. Recent advancement of atom probe tomography enables nearly atomic-resolution mass spectroscopy for 3D limited areas, together with great sensitivity to different Si isotopic mass [2]. By extending this technique to isotopically programmed Si nanopillars, individual self-diffused Si isotopes were visualized at MBE-grown isotopic layer interface [3, 4]. Recently, our quantitative evaluation revealed that Si self-diffusivity in gate-oxidized nanopillars was the same as the one measured for standard planar oxidation despite of its high interface-to-volume ratio [4]. In addition, silicon isotope engineering provides a crucial solution to eliminate unwanted decoherence for silicon quantum computers (QCs). Strong compatibility with industrial CMOS technology is expected to satisfy a scalability criterion significant for realizing practical QCs. In order to support the Si-QC research, we adopted silicon isotope engineering to tailor high-quality CVD-grown 28Si epi-wafers [1], where state-of-the-art Si-MOS nanoelectronic devices were implemented to demonstrate two-qubits operations encoded by single donor electron and nuclear spins [5]. Following an original Kane’s proposal, selective control of the electron spins was successfully achieved by adding a local gate bias thanks to the Stark tuning [6], while the nuclear-spin resonance has been recently shown to be driven by local electric fields in the presence of lattice strain [7]. Additionally, the isotopically enriched wafers were, in parallel, employed to demonstrate single and two spin-qubit operations in 28Si-MOS quantum dots (QDs) that were defined by standard CMOS lithography [8]. A series of proof-of-concepts experiments has given a boost to several proposals of Si-QC architectures based on advanced manufacturing technology [9]. Meanwhile, an alternative approach following conventional III-V QDs was made to obtain buried-type QDs utilizing strained-Si/SiGe quantum-well (QW) heterostructures, which are placed away from gate-oxide interface traps that can work as sources for charge noise to disturb spin-qubit operation. In fact, we accomplished isotope enrichment of Si-QW layer to efficiently remove magnetic nuclear-spin noise. As a consequence, the electron-spin dephasing time was prolonged to T 2 * ~ 20 ms for 28Si/SiGe QDs, where local single-spin controllability was introduced by a magnetic-field gradient from an on-chip micro-magnet [10]. These 28Si-QD devices amenable to several technical improvements have recorded the highest single-spin gate fidelities of 99.93% for 28Si/SiGe QDs [10] and 99.96% for 28Si-MOS QDs [11], respectively. Concurrently, it was revealed that inherent charge noise and/or residual nuclear-spin noise induced spin-resonance frequency fluctuations during the benchmark test executions. For higher-fidelity operations, including spin-qubit readout and initialization, low-energy excited states responsible for Si valley degeneracy should be also well-lifted by rendering the QD confinement atomically sharp. In these contexts, we will present recent development of high-quality 28Si/SiGe QWs appropriate for the silicon quantum information research. This work has been in part supported by MEXT Quantum Leap Flagship Program (MEXT Q-LEAP) Grant Number JPMXS0118069228 and the Center for Spintronics Research Network, Keio University. The authors acknowledge fruitful collaborations with Yasuyoshi Nagai’s group, Seigo Tarucha’s group, Mark Eriksson’s group, Andrew Dzurak’s group and Andrea Morello’s group.[1] K. M. Itoh and H. Watanabe, MRS Communications 4, 143 (2014).[2] Y. Shimizu et al., J. Appl. Phys. 106, 076102 (2009).[3] T. Südkamp et al., J. Appl. Phys. 123, 161515 (2018).[4] R. Kiga et al., J. Appl. Phys. under review.[5] J.T. Mohonen et al., Nature Nanotechnol. 9, 986 (2014).[6] A. Laucht et al., Sci. Adv. 1, e1500022 (2015).[7] S. Asaad et al., arXiv:1906.01086.[8] M. Veldhorst et al., Nature Nanotechnol. 9, 981 (2014); M. Veldhorst et al., Nature 526, 410 (2015).[9] L.M.K. Vandersypen et al., npj Quantum Information 3, 34 (2017); M. Veldhorst et al., Nature Comm. 8, 1766 (2017); R. Li et al., Sci. Adv. 4, eaar3960 (2018).[10] J. Yoneda et al., Nature Nanotechnol. 13, 102 (2018).[11] C.H. Yang et al., Nature Electronics 2, 151 (2019).