On Wine and Scalable Quantum Computers
2 February 2018
by Matt Reagor
Rigetti’s technical team is constantly working to improve the quantum logic operations that underpin algorithms on Forest. In today’s edition of Science Advances, our team shared one of the exciting breakthroughs that make our quantum algorithms better.
Each silicon wafer that we manufacture at Fab 1 hosts many hundreds of qubits. But we actually dice that wafer into smaller chips, containing fewer qubits (e.g. 19Q), before integrating with our Forest stack. Essentially, the rest of our full-stack system is playing catch-up to that scale. The control problem, in particular, is a hard problem at scale. Each new circuit element on a device compounds potential sources of error.
We now have a technique that enables us to reduce interference between qubits as we add more and more qubits to a chip, thus retaining the ability to perform logical operations that are independent of the state of a (large) quantum register. My goal here is to explain how that all works.
Fortunately, the basic problem and our solution are purely classical effects, most of which are highly intuitive. The fundamental concept at play is resonance.
Clink a wine glass, and you will hear it ring at its resonant frequency (usually around 400 Hz). Likewise, soundwaves at that frequency will cause the same glass to vibrate. This is why opera singers don’t get invited to dinner parties — though I am willing to forgive Walter Lewin (his demo is amazing!).
You already know a few other important things about wine glasses. Different shapes or amounts of liquid in a glass will produce different clinks, i.e. different resonance frequencies (see, for example, Fur Elise with glassware). You may have noticed, too, that a clinked wine glass will cause identical, nearby glasses vibrate. But when the glasses that are different shapes, these off-resonant glasses will not vibrate much at all (like an out of tune opera singer).
Here’s the connection. Each physical qubit on a superconducting quantum processor stores energy in the form of an oscillating electric current. Essentially, each qubit is a wine glass. The logical state of a qubit (e.g. “0” or “1”) is encoded by the state of its corresponding electric currents (in this metaphor, this is equivalent to whether or not a wine glass is vibrating.).
A highly successful class of entangling gates for superconducting qubits operate by tuning two or more qubits into resonance with each other, or close by resonance. At this tuning point, the “wine glasses” pick up on one another’s “vibrations”. This effect can be strong enough to produce significant, conditional vibration changes that we can leverage as conditional logic. You can imagine pouring or siphoning off wine from one of the glasses to make this tuning happen. With qubits, we have tunable circuit elements that fulfill the same purpose.
Here’s the problem. As we scale up quantum processors, there are more and more wine glasses to manage when executing a specific conditional logic gate. Imagine lining up a handful of identical wine glasses with increasing amounts of wine. Now we want to tune one glass into resonance with another, without disturbing any of the other glasses. To do that, you could try to equalize the wine levels of the glasses (for instance, by pouring a precise amount of wine). But that transfer needs to be instantaneous to not shake the rest of the glasses along the way.
Our solution is a clever bar trick. Continuing in the language of this analogy, the Rigetti quantum processors are comprised of wine glasses having various shapes and sizes. Let’s say one glass has a resonance at one frequency (call it 400 Hz) while another, nearby glass has a different one (e.g. 380 Hz). Now, we make use of a somewhat subtle musical effect. We are actually going to fill and deplete one of the glasses repeatedly.
Furthermore, we repeat that filling operation at the difference frequency between the glasses (here, 20 times per second, or 20 Hz). By doing so, we create a beat-note for this glass that is exactly resonant with the other. (Physicists sometimes call this a parametric process.) Amazingly, our beat-note is “pure” — it does not have frequency content that interferes with the other glasses. That’s exactly what we demonstrated in our recent work, where we navigated a complex eight-qubit processor with parametric two-qubit gates.
While this analogy may sound somewhat fanciful, its mapping onto our specific technology, from a mathematical standpoint, is surprisingly accurate. Fortunately, wine country is right up the road from our lab in Berkeley, California. For more details, see the paper!
Some of the members of the team who wrote “Demonstration of Universal Parametric Entangling Gates on a Multi-Qubit Lattice” shared a bottle of wine to celebrate. From left to right, Matt Reagor, Nik Tezak, Alexa Staley, Chris Osborn.
Originally published at rigetticomputing.github.io on February 2, 2018.