In an attempt to keep better track of what I am reading and should read, I thought I’d throw down some links to papers of interest from the past few weeks.
Benchmarking
Application-Oriented Performance Benchmarks for Quantum Computing
Sporting an author list with affiliations sampling many of the QC startups out there today, this paper suggests a new kind of benchmark for quantum computers. Namely, one that is more focused on algorithms or applications similar to what one might expect useful QCs to execute.
The IBM crew adds a new benchmark to extend Quantum Volume (QV) called ciruit-layer operations per second or CLOPS. While QV is supposed to holistically represent the ‘quality’ of a quantum computer, CLOPS is presented as a way to holistically represent the speed of the same device, which is certainly an important consideration. Ali Javadi-Abhari, an author on this paper, was even kind enough to field my ignorant questions about this paper on twitter.
The team at Honeywell deliver a beefy deep-dive into the details and characteristics of the quantum volume test. Since I am not a benchmarking expert, the most helpful part of this paper for me was the nice FAQ the authors included! I’m a big fan of the FAQ, especially because many of the answers point the reader to the section of the paper that goes into much more detail on the sub-topic of interest.
Open Source Hardware
The QICK (Quantum Instrumentation Control Kit): Readout and control for qubits and detectors
I’ve only skimmed these, but I wanted to highlight them because I love the idea of open source hardware for qubit characterization. As far as I can tell, both systems can take a transmon from zero to randomized benchmarking without requiring very expensive arbitrary waveform generators.
While commercial QC endeavors may not choose to go this route, instead partnering with companies like Keysight to source their specialized hardware, I think this will end up being a boon to smaller operations that may not want to blow tens or hundreds of thousands of dollars on waveform generators.
Misc
A Chinese high-performance classical computing collaboration smashes Google’s estimate of 10,000 years to simulate their results in the ‘Quantum Supremacy’ paper. Turns out that for a few exaflops and a lot of researcher brainsweat, you can turn 10,000 years into 304 seconds. That’s a reduction on the order of about 1,000,000,000x.
I like this paper, and this sort of endeavor, for reasons that the authors spell out in their intro:
Second, for a leading-edge supercomputer, we think that taking a challenge that is inherently exponentially more complex could potentially bring algorithmic and architectural innovations within the traditional supercomputing community. There is an old saying from the Classic of Poetry (Shijing) of China, “a stone from a remote mountain could serve to polish a local jade.” The completely different way of computing in a quantum environment could potentially lead to a rethinking of algorithms for many cases on classical computers.
Effects of surface treatments on flux tunable transmon qubits
If I told you that this paper is published in a nature journal, you might reasonably expect that there is a very flashy result inside about some amazing new surface treatment for transmons. Instead, surprisingly, you will find a pretty thorough investigation of 8 different surface treatment methods for CJJ transmons and the effects these treatments have on lifetime, coherence time, low-frequency noise amplitude and power.
Spoiler alert: None of the treatments have a beneficial effect on T1 or T2, while one or two seem to lower the low-frequency noise amplitude. Also, a few of these surface treatments appear to reliably alter the qubit frequency without damaging its other desirable properties.
The take-away is that you probably don’t need to deploy any of these treatments, unless maybe you need to move the qubit frequency or beat down low-frequency noise.
If you want more of The Quantum Observer, smash that subscribe button.