Enabling large-scale silicon spin qubit platform using memristor-based neuromorphic circuits for quantum dots auto-tuning

自1997年首次展示基于核磁共振波谱的量子计算以来，该领域取得了巨大的进展，现在有多种技术可以获得高质量的量子比特(Qubit)。现在，人们正致力于量子比特的大规模集成。在这方面，2007年首次演示了硅中的自旋操纵，确定使用硅技术进行基于自旋的量子计算是最具诱惑力的方法之一。硅确实是现代电子的基础，50多年来基于cmos的超大规模集成电路(VLSI)的高产量制造可以被用于逻辑量子比特和大规模量子计算。此外，在同位素富集型28Si器件中，单电子自旋表现出了特殊的量子相干性。然而，为了迈向大规模量子计算，一个可扩展的集成量子比特系统还有待开发。使用目前可用的室温仪器在低温环境中操作量子设备只适用于当前的几个量子比特系统。我们知道，如今通过几个控制门调整十几个量子比特是一项费力但可行的任务，很明显，在这些条件下，不可能管理数量急剧增加的量子比特和I/O。因此，随着量子比特数量的增加，互连和控制线的比例被认为是阻碍实际量子计算机创建的主要瓶颈之一。拟议的研究计划旨在通过研究忆阻器和基于忆阻器的神经形态电路的使用，与量子点共同集成，极大地简化量子点的形成和控制，同时降低必要的I/O数量，从而实现大规模硅自旋量子比特平台。通过i)提供可扩展的高密度和高质量的基于cmos的量子点集成，ii)在忆阻器中存储静电形成量子点所需的栅电压值，iii)嵌入基于忆阻器的神经形态自动调谐系统，以及iv)极大地减少低温恒温器内部和外部之间所需的物理连接数量，在量子系统附近的这种存储器和机器学习技术的这种集成将同时解决阻碍主流量子计算出现的物理尺寸、控制和连接问题。

Since the first demonstrations of quantum computing based on nuclear magnetic resonance spectroscopy in 1997, tremendous progress has been made in the field and multiple technologies are now available to obtain high quality quantum bits (qubits). Great efforts are now channeled toward large-scale integration of qubits. In that regards, the first demonstration of spin manipulation in silicon in 2007 has identified the use of silicon technologies for spin-based quantum computing as one of the most seducing approaches. Silicon is indeed the foundation of modern electronics, from which more than 50 years of high-yield manufacturing of CMOS-based very large-scale integrated circuits (VLSI) can be leveraged towards logical qubits and large-scale quantum computing. Moreover, exceptional quantum coherence has been demonstrated with single electron spin in isotopically-enriched 28Si device. However, to make the step to large-scale quantum computation, an extensible integrated qubit system has yet to be developed. Using currently available room-temperature instrumentation to operate quantum devices in the cryogenic environment is only practical for current few-qubit systems. Knowing that nowadays the tuning of a dozen of qubits through several control gates is a laborious but feasible task, it becomes clear that a drastically higher number of qubits and I/Os is impossible to manage in these conditions. Scaling of interconnections and control lines with the number of qubits is thus considered as one of the main bottleneck preventing the creation of an actual quantum computer. The proposed research program seeks to enable large-scale silicon spin qubits platform by investigating the use of memristors and memristor-based neuromorphic circuits, co-integrated with quantum dots to greatly ease their formation and control while lowering the number of necessary I/Os. Such integration of memory and machine learning technologies in close vicinity of the quantum system would address at the same time the physical size, control and connection issues hindering the advent of mainstream quantum computing by i) offering scalable high-density and high-quality CMOS-based quantum dot integration, ii) storing in memristors the gate voltage values required to electrostatically form the quantum dots, iii) embedding memristor-based neuromorphic auto-tuning system and iv) dramatically reducing the number of required physical connections between the inside and the outside of the cryostat.