Commercially viable quantum computers are at least a few years away, but some researchers are already wondering if existing EDA tools will suffice for the design of quantum chips and systems. That’s because quantum design requirements sometimes go beyond classical rules about materials, temperature, and structure—rules that are fundamental to most EDA products on the market.
This situation could accelerate EDA’s move toward customization as quantum computing becomes a more attractive business opportunity. McKinsey & Co. estimates that the quantum computing market could reach $700 billion as early as 2035. Scientists say quantum computers could drive truly disruptive innovations in a variety of fields, including pharmaceuticals and AI. But there are still a lot of technical hurdles that need to be overcome first, most notably the production of long-lived, error-correcting qubits in larger quantities than is possible today.
Quantum researchers are currently using a variety of different materials, techniques and approaches to develop quantum computers. Since there is no consensus on basic questions like the best approach to create a qubit, there are also different ways to design the chips. For example, some systems rely on electron spin while others use photon polarization.
Existing EDA tools are generally better suited to the electronic (superconducting) than the photonic (optical) domain of quantum chip design, but neither is perfect.
“One of the challenging parts of superconducting chip design is the complexity of electromagnetic simulation with the increased number of qubits,” said Mohamed Hassan, planning manager for quantum solutions at Keysight Technologies. “Furthermore, tuning qubits and resonators and their coupling implies an iterative cycle of electromagnetic simulation, which significantly lengthens the design cycle and increases computational costs.”
Hassan explained that the design of superconducting qubits composed of Josephson junctions follows a typical microwave circuit design process. “The quantum microwave circuit usually consists of coplanar waveguide resonators coupled to qubits. The resonators are used for two purposes – to read the state of the qubits and to let them talk to each other. In quantum language, that’s entanglement. From a microwave design perspective, the circuit will have many resonant frequencies corresponding to the resonators and the qubits. In fact, qubits can be abstracted by their lumped inductances, which are expected in the quantum domain at very low signal excitation.”
Fig. 1: The top image shows a four-qubit circuit created in IBM Qiskit Metal and imported into PathWave ADS2023. The square islands represent qubits and the tangled lines represent entanglement/bus resonators. The image below shows EM simulation results using the moment method. Source: Keysight
While this incorporates some elements of existing semiconductor design, the quantum world requires entirely different optimization techniques, some of which are still evolving. As a result, there are no agreed best practices and methods, and it is not clear when there will be any. However, both will be essential for wide commercial acceptance.
“Chip design involves constructing the cavities and qubits to bring the different frequencies of the circuit to target values, typically in the range of a few GHz, and adjusting the coupling between the cavities and qubits to meet specific quantum parameters,” Hassan said . “When the qubit changes state, we can see this in the microwave signal response of its associated readout cavity with a shift in resonant frequency.”
Planar microwave circuits lend themselves to a very low-cost electromagnetic simulation solution using the method of moments, which resolves only for the currents on the metal surface rather than the full-volume electric field. This approach significantly reduces the computational effort. Quantum planar microwave circuits can reap the same benefits and enable faster design cycles for large quantum circuits. But scaling and accuracy requirements are more demanding for quantum circuits, necessitating the need for innovative solution techniques and skilled engineers to meet the new challenge.
Using optical design approaches requires developing nonlinear shapes for photons—curved waveguides instead of an electrical connection to allow optical signals to move smoothly through a chip. “[These are] very large curvilinear structures that not only have to be very precisely shaped circles and waveguides, but also have to be very tightly coupled to other waveguides if you’re making photonic interconnection at the quantum scale or even at room temperature,” said Ted Letavic, CTO and VP for Computer and wireless infrastructure at GlobalFoundries. He pointed out that EDA tools are generally sufficient to design waveguides and coupling structures for photonic receivers and data center communications. “However, given the uniqueness of the quantum requirements, significant updates to EDA tools are required in the way we handle non-rectangular shapes.”
Other important changes will be required to reflect the nonlinear optical properties of the materials involved in the photonic domain, which today includes barium titanate, strontium titanate, molybdenum and silicon alloys.
In contrast, the design of quantum chips in the electronic field is mainly about “optimization”, according to Letavic. “It’s very similar to other electronic systems,” he said. “There are normal EDA layout rules for some restricted zones to look out for depending on the noise. We need to be aware of radio frequency (RF) shielding and such. But the techniques are known. I wouldn’t say there are big showstoppers there.”
That changes when you look beyond chips to consider packaging simulation design and the impact of temperature. “When you have a quantum computer that has three or four different temperature regimes, from 4 millikelvin (-273 °C) all the way down to room temperature, the stresses and strains on the packaging are pretty extreme,” he said. “There is still work to be done to quantify the reliability and robustness of the different packaging solutions required to implement the next generation of quantum computers.”
The energy needed to keep quantum systems cool could lead to a chiplet model, which also has implications for EDA. “What needs to happen in these quantum systems is a chiplet-based format,” he said. “Some of these chiplets will be at very low temperatures, and some of them will be at room temperature. The solution will be chiplet segmentation. The EDA tools need to be able to very accurately and with high fidelity simulate a single chip in the system, and they also need to very accurately model the connections to chips and other parts of the system at different temperatures.”
Of course, not everyone in the quantum world questions the adequacy of the existing EDA ecosystem. William McGann, CTO and COO of Quantum Computing Inc., said the company recently licensed an existing EDA product which he believes will offer more than is required to develop the company’s chip designs. According to McGann, the qubits are photonic, and the design uses lithium niobate and certain challenges associated with transferring the technology to a mass-produced chip.
“If you’re cutting intelligently, there’s a certain amount of damage based on the energy and cleavage plane you’re cutting the crystal across,” McGann said. “How do you polish these defects out of the surface? How do you measure the active volume of your device when building photonic band gap materials? Certainly, purity is very important from an optical point of view. Some of the metrology challenges, such as how we measure what we know to be good, can likely push some of the laser technology we need beyond what we have today. Most of the processes we expect to use have already been publicly demonstrated, so now it’s really about developing our recipes.”
Still, GlobalFoundries’ Letavic predicts that considerations like the quantum researchers he describes could force them to look for custom solutions. This could amplify existing EDA trends toward increased customization and bespoke silicon. “The EDA tool environment is often just focused on the hardware part of the ecosystem, but what will really determine whether these quantum computing architectures deliver on the promise of quantum is the rest of the stack,” Letavic said. “There will be so many differences in software, algorithms and quantum sources that I don’t see this narrowing down to a standard solution in the next ten years.”
He predicts that these EDA solutions will be extremely customized, starting with the properties of the qubit source itself — photonic qubits, with their nonlinear properties, and electronic qubits, with their extreme temperature requirements.
John Ferguson, director of product management at Siemens Digital Industries Software, said that much of the design flow in quantum computing is similar to classical computing. “It’s like the EDA in the mid to late 1980s. The automation is not there. What is forcing the industry to innovate is the problems with curvilinear shapes, which have long been difficult to design with existing tools. We knew it was coming for a long time, but everyone said, ‘Let’s wait until it’s really necessary.’ Now we can’t avoid it and everyone is working on it.”
However, traditional EDA tools for checking design rules don’t do well with curved shapes, since wires and transistors are drawn as rectangles in the IC world.
“But curved things are increasingly becoming a fact of life, not only in quantum computing but also in photonics and other fields. There are suggestions to change the Oasis format so we can render things that are curvilinear,” Ferguson said. “Quantum design is currently more of a mix of art and science, with a bit of intuition. It’s more like creating a custom analog block where you know exactly what you want and how you want them to interact with each other and you move them graphically and then run some kind of simulation. It’s a lot of what I call “construction by correction”. Developing EDA tools to meet the needs of quantum computing researchers is an important and difficult problem. And it’s far less difficult than some of the engineering challenges involved in scaling quantum computing, such as B. developing more qubits of the right type. It’s an interesting challenge, but it’s doable.”
