Simulating the Next Generation of Quantum Computers
April 29, 2015 Nicole Hemsoth
Just as modeling and simulation tools are being used to design and test prototypes of everything from rocket engines to safer tools, so too are they being used in novel ways to model technologies that rest on the edge of commercial reality.
Quantum computers are in production at only a few centers currently, but the next stage of design at the processor level is still in the works. The goal of quantum computing company, D-Wave, is to add greater performance for its quantum computing machines, but prototyping at the circuit and component level, especially with the environmental extremes required for these systems, is an expensive and labor-intensive task.
The team at D-Wave is working toward simulating its quantum processors using complex multiphysics software that will allow them to enhance circuit design against the stringent temperature and magnetic environment of the quantum system. Quantum processors based on superconducting integrated circuits can take advantage of existing semiconductor fabrication capabilities, but because of the superconducting nature of these circuits and the fact that they access ultra-high frequencies, there are special considerations that need to be made in overall design. Many of these designs can be tested without physical prototypes using complex modeling and simulation software, which can help processor designers—even for these novel devices—to understand interactions and tweak their specs accordingly.
Of course, this larger processor design challenge does not happen in a void. Without going into the obligatory “how it works” conversation about quantum computers, at the processor design level, these superconducting circuits hinge on quantum mechanics for speed, which means they need to be operated in extreme environments. For instance, the processors are kept at near absolute zero temperatures and further, require an ultra-low magnetic vacuum. For reference, as Jeremy Hilton, Vice President of Processor Development at quantum computer maker, D-Wave, told The Next Platform “the temperature is about 30,000 times colder than room temperature and the magnetic environment is about 50,000 times lower than on earth.
The potential complications in mere processor design are well-known, but add in the fine-tuned balance of extremities and the fact that physical prototyping of these systems is enormously expensive, and one is left with all the right conditions for a complex software simulation to help virtually design the next wave of quantum computers.
D-Wave is using ANSYS software for its modeling of several different aspects of the quantum processor and piecing their findings into a coherent picture of where potential design tweaks might need to be made, especially as they relate to the performance of the processor. As Hilton noted, “These are multiphysics simulations. We can separate the chip from the system, in a sense, an there are also different modeling tools from ANSYS around these ultra-low temperatures and magnetic environments as well as for tools to simulate some of what we need around the high frequencies for the circuit design itself.”
It is early days for the D-Wave team in simulating quantum processor performance, but Hilton says the most success has been in magnetic shield engineering. The team had to simulate the ultra-low temperature environment and the impact that has on the materials being used. The geometry and materials progress made following simulations of these factors led to some breakthroughs that helped the team look behind some legacy approaches that were causing weaknesses in the shields and having an impact on performance.
What D-Wave is working on with just one component of the overall system, the magnetic shielding, is evolving rapidly with new needs that can drive performance improvements. “We’ve already been through several product generations and with each, the ‘payload’ inside the shield is changing substantially as well. The regions of magnetic environments that need to be controlled in terms of stability and our ability to tune that environment are important factors. But we’ve reached a point now where there is a contradiction in that the payload space requirements are increasing and at the same time, our desire to reduce the magnetic environment further is increasing. We’re anticipating the challenges of the next several years and all of this depends on high-powered modeling tools,” explains Hilton.
They are just getting started with circuit modeling and so while D-Wave does not have a sense on what that dataflow and might look like, the component-level experiments have been useful. Still, Hilton says the real benefits will shine through when they move the simulations from the device to fine-tuning and optimization efforts across the system. But at this stage, Hilton says, they are mostly running experiments that target materials, temperature, and magnetics, among other elements of a quantum computing system.
“We are still in the process of fully developing this simulation program now. The objective is to shorten our development cycle for system design. In terms of optimizing system design, minimizing costs, these are the goals, but the real emphasis is on the performance of the quantum processor, which is affected by the environment of the system, circuit design, and other factors.”
ANSYS has a range of tools that solve difficult physics problems for applications in manufacturing, science, aerospace and beyond, as well as multiphysics software that can simulate circuit designs and test for effectiveness and performance.
According to Larry Williams, who directs product management at ANSYS, there are some connections between various physics packages we provide but D-Wave is pushing the envelope in some ways. “We were looking at some of their high-frequency analysis and to get the proper behavior of the thermal materials and environment, we had to look into specialized material properties. And for the magnetics we looked at coupling mechanical simulation with low-frequency magnetics, some of which required some custom scripting. They’re using a standard physics solver, but certainly customizations are going on to make it solve these new and challenging problems.”
Interestingly, ANSYS has had to go out on a limb in the past for other new, challenging simulation problems, including the exact inverse of what D-Wave is trying to do. “Sometimes we look at high power microwave problems or circuit drives that drive electric motors as you might find in a hybrid vehicle. In that case, you’re looking at the electromagnetics and electronics getting hot and we need to calculate what the thermal behavior is and there’s a lot of interaction there. With D-Wave, it’s the inverse, we what the thermal profile and distribution, then we run a magnetic simulation, and because of the changing fields in those region we can have a range of temperature distributions, so we have to close that loop,” Williams said.
Although these are demanding simulations, for now, since they are at the component level, the team is running ANSYS on a single node at D-Wave, but Hilton says as they look to more challenging system-level problems, they will likely scale the hardware with the problem and put more cores on the task.
Hilton told The Next Platform that this is a fairly new relationship between D-Wave and ANSYS, but the teams have been working for several months, starting with the component pieces. “There is a lot of value in getting our scientists and engineers trained up on the performance level of those modeling tools. But once those pieces are working, there are ways for us to stitch those things together.”