Shana Massar, engineer in the quantum computing programme at Imec, states: “The goal of quantum computers is not to replace our already known classical computers for performing our daily tasks. We need quantum computers for a very particular set of problems, problems that have a high degree of complexity.”
One example of a use case for quantum computing is solving optimisation problems; another is simulating molecular systems. This can be done to gain a better understanding of materials science and can also be done to help discover new drugs.
In a quantum computer, information is manipulated in a fundamentally different way than in a classical computer. In a classical computer, the logic element is a bit, which can take on one of two states: zero or one. In a quantum computer, the logic element is a qubit, or quantum bit, which is defined as any coherent two-level system that can be initialized, manipulated, and read.
“If I look at the state of a bit, the state is either zero or one and this leads to deterministic measurement, while the qubit has a superposition of state,” says Massar.
“It is a linear combination of zero and one simultaneously. But after readout, it’s either zero or one along with a certain probability – and this leads to probabilistic measurement.
“The quantum computer has another feature, entanglement. The classical bit states are independent of each other, which leads to the fact that N bits store N states. But qubits can be entangled. They can be coupled, which means N qubits can ‘process’ in some sense up to two to the power of N states. When we apply a logical operation to all those states at the same time, we get massive parallelisation and a very high computational power.”
But none of these promises of quantum computing will ever come to fruition until somebody finds a way of producing reliable qubits in a repeatable manner. Qubits are currently implemented in labs in a customised fashion, but researchers at Imec would like to change that. They have started looking for ways to produce qubits on an industrial scale.
Developing a fab process for qubits
“To build a one-million qubit system, or just a meaningful quantum computer, you have to reduce the qubit variability and increase the production yield, while maintaining the fidelity and coherence,” says Kristiaan De Greve, scientific director and programme director for Quantum Computing at Imec.
“The methods that some of the best research labs in the world have been using will likely not allow you to go all the way. We have a different approach and are trying to see if we can use existing tools from the semiconductor industry, where they have produced very complex circuits, with low variability and high yield.”
There are several different approaches to implementing qubits: quantum optics, trapped ions, magnetic resonance, superconductors, nitrogen vacancy in diamond, and quantum dots. Researchers at Imec focus on two technologies – the superconducting devices and the semiconductor quantum dots.
One reason for these choices is that Imec sees those technologies as promising ways to make high-quality qubits. But the second reason – the biggest reason for Imec – is that qubits in those two technologies can be fabricated in a way that is first order compatible with complementary metal-oxide-semiconductor (CMOS) facilities, facilities that Imec has in very high quality.
One challenge with both approaches is that they operate at very low temperatures. For this reason, Imec is also doing research in cryo electronics, electronics that can work at very low temperatures.
Imec aims to build suitable and stable qubits and qubit arrays along with the necessary electronic interfaces, which allow programmers to setup the qubits to run a program and then to read the results.
Discovering the right materials and techniques
To discover optimal production techniques, Imec has set up a research process, where they try different materials, architectures and production techniques to produce qubits and then test the results to measure which techniques work best.
The first phase of its research is the design phase, where a team of experts run simulations to find the best design, given different materials and the required dimensions. When the design phase is completed, they move to the second phase, the fabrication phase, which begins by running other simulations to find optimal ways of creating the qubits, determining the most accurate process flow and the best settings and recipes.
Imec then process its sample in the fab, closely monitoring the different processing steps using inline characterisation. When the fabrication of the samples is successful, they move to the last phase – cryo characterisation or characterisation at low temperature.
In the end, they wind up with a wafer full of dies, sub dies, and chips that they mount on a sample holder to put it in a refrigerator for measurements at very low temperature. The temperatures go down to just a few thousandths of a Kelvin, which is much cooler than outer space. Using the cryo measurements, Imec researchers extract qubit performance and characteristics and assess how well a given design and fabrication process works.
“We are currently focusing our research on the fabrication of devices, and we are investigating different gate stack materials and patterning technology,” says Massar. “We are also investigating different substrate materials and formation recipes. And we look at the overall thermal budget of our processes and the consequence it has on the qubit quality.
“At the same time, we’re working on the qubit control and design. We’re improving the design of our devices, the controlling devices of the qubit and the measurement setup quality. As an example, over the past few months, we have worked on decreasing the electromagnetic noise in our measurement setup. This leads to a better quality on the qubit read.
“At the other end, we’re also looking at the characterisation setup quality. We want to improve the qubit read and also improve our setup in terms of both the quantity of measurements and the quality of each measurement.”
Imec has made big progress. Last year, it demonstrated a fab-compatible process to manufacture high-coherence superconducting qubits and are now transferring the process from the lab to the fab. By doing this, they hope to open new possibilities for manufacturing fab qubits with high coherence and low variability.
Who knows? Maybe one day this will lead to a one-million qubit quantum computer.