High-Performance Quantum Computing — The Convergence Forthcoming

Last week saw a much-anticipated announcement, the specs for the LUMI supercomputer: LUMI will be the leading computing facility in Europe when fired up in mid-2021

Yes, with over 550 peak petaflops, LUMI is faster than the current #1 in the world and will be the European champion by a large margin. However, even more fascinating is that all of these petaflops are powered by 100% renewable energy in an overall carbon negative setup! A true testament to both engineering ingenuity and geographical advantage. Now, more than ever, we need the aid of computers to solve some of the most pressing problems of society. In order to avoid aggravating the situation, the supercomputers have to operate with as small a carbon footprint as possible. As service providers, it is our responsibility to host green datacentres, not in a decade, but already today!

Another aspect where LUMI excels is in pan-European collaboration. LUMI is the joint effort of ten European countries, with the support of the entire European high-performance computing (HPC) community via the EuroHPC programme.

A tightly-knit coalition of top expertise in the field of HPC is formed

Quantum computing will be a cornerstone of HPC ecosystems of the future. Where traditional computers rely on classical bits, zeros or ones, quantum computers extract their power from manipulating quantum bits, qubits. Simplifying somewhat, qubits can be zero and one and everything in-between at the same time. By exploiting counterintuitive quantum physical phenomena like superposition and entanglement, a sufficiently mature quantum computer can solve certain computational problems that will forever remain intractable to classical supercomputers. The potential for scientific and industrial breakthroughs is therefore immense.

Still, much work lies ahead before quantum computing can compete with traditional computer technologies, or even successfully integrate and complement them. Right now, it is not yet even clear which manufacturing approach should be used to produce the qubits. The two most promising qubit implementations are, arguably, based on trapped ions and superconducting circuits. Presently, superconductors lead the qubit race with Google’s Bristlecone (72) and Sycamore (53), and IBM’s Hummingbird (65) at the front; Sycamore demonstrated quantum supremacy in October 2019. IBM predicts to reach 100+ qubits next year, and 400+ qubits in 2022. The race has only just begun.

The technologies have complementary features. Ions have the longest quantum coherence times, while superconducting circuits have superior operating speed. Trapped ions have a somewhat lower error rate, while superconducting circuits are based on established production technology. Other, less established qubit approaches like photon and neutral atom techniques also have potential advantages and corresponding drawbacks, e.g., higher error rates and slower operations. Further, note that the oft-cited qubit count is not everything, fidelity counts even more. The field is evolving rapidly and to date, no single technique is clearly superior; all of them need further refinement. Given this diversity, we need to focus on a variety of designs in order to drive the technology and support early user adoption. Overarching platforms that enable wide experimentation and also integrate with classical systems are paramount.

The LUMI consortium has brought together local high-tech skill-centres across Europe, not only for traditional HPC, but also for quantum technologies. What is more, the network is rapidly expanding, growing new academic, public, and industrial branches. Just last week, the work on on-chip measurement devices for trapped ions at ETH Zürich, Switzerland, was published in Nature [1]. Earlier this month, also in Nature, a boost for future measurements in superconducting quantum circuits using graphene bolometers was reported [2] by Aalto University and VTT, Finland, and IQM. IQM is a rapidly growing company developing, among other things, hybrid analog/digital quantum computers in Helsinki and Munich. Work performed at the University of Hasselt, Belgium, on spin state measurements in charged nitrogen vacancies, published in Science [3], outlines how technologically less-demanding approaches could be integrated into current nanoelectronics. All of these example breakthroughs pave the way for truly scalable quantum computers.

The convergence of quantum and classical

Here enter convergence and synergy. Quantum computers will not replace traditional supercomputers. Instead, they will become an integral part of supercomputing solutions. This hybrid HPC+QC or HPQC approach is where real-world applications will find their quantum advantage. In the cloud, naturally, where we at CSC have hosted our computing capacity for a full five decades come next year.

Building fast, fault-tolerant quantum computers is only one part of the final equation. To utilise a quantum computer, useful algorithms need to be found, and usable programming tools developed. This requires coordinated integration work and know-how of both novel quantum algorithm development and traditional parallel programming, all of which are available within the extended LUMI collaboration. In addition, quantum computers need to be matched with a truly powerful HPC solution. For this purpose, the pan-European LUMI solution is unrivalled in both flexibility and raw computing power. The HPC ecosystem is in place, let’s exploit it to the fullest extent possible!

Quantum advantage is still some time away. This is no reason to hold the horses, however. We need to spur all of our quantum equines to reach the first real-world applications that make a difference. We cannot postpone, wait-and-see: every year that we can shave off from the waiting time counts. For when the quantum advantage is reached, it will grow exponentially. Keep in mind that every additional scalable high-quality qubit roughly doubles the performance of a quantum computer! A 101-qubit device is twice as powerful as a 100-qubit device; a 1000 qubit device is already solving some useful problems even LUMI could not tackle.

LUMI compiled HPC expertise into something remarkable, it is high time to do the same for European quantum!

Depiction of the LUMI-Q concept, where several different quantum computing solutions are integrated with the LUMI supercomputing ecosystem.



Many thanks to Prof. Jonathan Home (ETH Zürich), Dr. Pekka Manninen (CSC), Prof. Milos Nesladek (U Hasselt), Dr. Pekka Pursula (VTT), Prof. Martin Schulz (TU München), Prof. Göran Wendin (Chalmers), and Dr. Joanna Johansson (Viessmann Refrigeration Systems) for inspiring comments.


 Read more


Author: Mikael Johansson
The author explores and enables quantum technologies at CSC.