Lecturer: Mikhail Belogolovskii, G.V. Kurdyumov Institute for Metal Physics, Kyiv, Ukraine
Abstract:
I present a brief state-of-the-art overview on the rapidly developing field of superconducting supercomputing which could offer an alternative to semiconducting devices. The editorial note in Nature [1] argues the need to enhance further performance of classical computers, which encode data into conventional binary digits, simultaneously with the development of quantum counterparts. In particular, it relates large, centralized computing facilities (supercomputers) used for a wide range of computational tasks. Now the effort to push high-performance CMOS-based computing to the next level faced with the problem of heat generation [2]. Superconducting computing could offer an attractive low-power alternative to CMOS switching devices and normal-metal interconnects with many potential advantages.
Four years ago the US intelligence agency IARPA launched related program. In 2014, “The Next Wave,” the USA National Security Agency’s review on emerging technologies, published a foresight report on the actual state and prospective development of superconductive electronics [3]. As was emphasized in the report, two promising research directions are urgent for a further progress of Josephson-junction circuitry. The first one relates self-shunted devices able to eliminate the need for external-shunt resistors and the second direction concerns hybrid structures with ferromagnetic interlayers which could realize junctions with a built-in -phase shift of the superconductive wave function and so to create a superconducting memory element.
Concerning the first problem, I present results of our joint activities on this problem which were summarized in the paper [4] where we provided a complete view on the self-shunting problem in Josephson junctions, relating it to specific features of a multichannel weak link between electrodes where averaging over the channels yields a bimodal distribution of transparencies with maxima near unity and zero. I shall discuss two examples of such internally shunted devices: (i) four-layered Nb/Al - Al oxide - Nb junctions with strongly disordered nanometer-thick insulating layers where stochastic distribution of transparencies takes place on a local rather than a global scale and (2) MoRe - W-doped Si - MoRe devices with strongly inhomogeneous silicon interlayers partly doped by metallic nanoclusters where the main charge transport occurs across resonance-percolating trajectories. The predicted universal distribution function of transmission coefficients has been verified experimentally without any fitting parameters. Our conclusions were confirmed by advanced experiments performed in the USA within the IARPA program aimed to establish superconducting supercomputing [5]. I shall discuss also the way to realize an energy-efficient superconducting memory referring to our results [4] and those recently obtained in the NIST [6].
At the end, I plan to tell very briefly about activities of Fluxonics, a non-profit European society whose goal is to develop Superconductive Electronics in Europe and to promote associated technological innovations through research, training and transfer of knowledge, see http://www.fluxonics.de/.