X

Science fiction writer William Gibson once said: *the future is already here, it’s just not very evenly distributed.* These were the opening words of Rich Uhlig, Intel senior fellow, vice president, and director of Intel Labs, at the recently held Intel Labs Day. During the event, Intel spotlighted research initiatives across multiple domains. During the discussion on Quantum Computing, Intel introduced the 2nd-Gen Horse Ridge Cryogenic Quantum Control Chip.

Gibson was referring to the way futuristic technologies and other developments propagate. At Intel’s event, Uhlig offered this example: genotyping technologies and home DNA tests are fully accessible to some, but they’re not available at scale to the world’s populations; this “future” (genotyping technology) is unevenly distributed. The evolution to big data is similar. From an IT point of view, huge amounts of data have to be moved, stored, analyzed, secured, and calculated on a large scale. The planet is increasingly covered with sensors, and huge amounts of new data are being generated every day. New challenges are coming.

We need a huge improvements across different areas of technology spanning computing and memory interconnections. Intel believes that these kinds of gains require a new way of thinking, and they come when experts can collaborate by intersecting multi-disciplinary knowledge, science, and technology domains. One of those domains is quantum computing.

Quantum computing has been a really active field of research in recent years, with a number of different companies working on it. Quantum computers have some unique properties that gives them the potential to be far more powerful than the computers we have today.

In an interview with EE Times, Anne Matsuura, director of quantum applications and architecture at Intel Labs, who presented on the topic at the event along with James Clarke, director of quantum hardware at Intel, highlighted applications for the disruptive technology and offered a deeper look at Intel’s quantum efforts.

The discovery and design of new materials is one of the early applications where the use quantum technology could have meaningful impact. “We believe that quantum computers on a commercial scale will allow simulation of these materials so that, in the future, we can also design materials, chemicals with desired properties. Today’s 100 qubits or even thousands of qubits will not lead us to this result. We will need a commercial-scale quantum computer system of millions of qubits to achieve quantum practicality for this kind of ambitious problem solving,” said Matsuura.

**What’s a qubit? Quantum vs. classical computing**

Qubits are “bits” of information for quantum systems and some elements of quantum mechanics. But how are qubits physically created? How can electronics control qubits efficiently in the quantum system?”

Unlike the conventional bit, rather than existing in a zero or one state, a qubit exists in a “superposition” of both states at the same time. Moreover, in a quantum processor, there can be more qubits in a state of superposition connected among them, to the point that they express a group behavior, called entanglement. This state of entanglement is the basis for the incredible computing power of quantum computers, and the source of their potential to solve complex tasks beyond the capabilities of traditional supercomputers.

As part of Rich Uhlig’s Labs Day keynote, Matsuura provided an introductory speech for the Quantum section, introducing what Intel is doing and the latest market developments.

“A simple way to gain intuition about quantum computing is to think of a computer bit as a coin. It can be in either state — state-heads or state-tails. It’s in either one state or the other. Now imagine that the coin is spinning. While it’s spinning, it is in a sense in both state-heads and state-tails at the same time; it is in a superposition of the two states,” said Matsuura.

This is similar to a quantum bit or qubit. She added, “Now imagine that we bring two qubits together and entangle them. Now we have four states at the same time. So two entangled qubits represent a mixture of four states at the same time. And more generally, n qubits represent 2^{n} states.”

A quantum computer’s computing power grows exponentially with the number of qubits. In theory, if we have 50 of these entangled qubits, we could access more states than any possible supercomputer. If we have 300 entangled qubits, we could represent more states than atoms in the universe at the same time.

Like a classical computer, a quantum computer is made up of quantum circuits consisting of elementary quantum logic gates. Quantum computers have the potential to address problems that conventional computing solutions cannot handle. The underlying technology is quantum physics; because a quantum bit (or qubit) can exist simultaneously in multiple states, it can be used to calculate on all possible states at the same time, significantly speeding up the resolution of complex problems.

She added, “Qubits don’t have very long lifetimes, and noise or observation causes a loss of information. So in reality, we’ll need hundreds of thousands, or even more likely, probably millions of high-quality qubits for a commercial-sized quantum computer. In other words, we need quantum to scale to be useful for practical applications. There are key areas that Intel’s quantum research program is focused on: spin qubit technologies, cryogenic control technology, and full-stack innovation. Each of these areas addresses critical challenges related to scaling quantum, and Intel is tackling each one systematically to achieve scaling.”

**Qubit technologies**

Qubits are available in the small quantum computing systems we see today. Their quality and numbers simply not high enough for them to be on a commercial-scale system. We will need to have many stable or noise-resilient qubits and highly efficient connectivity between qubits to scale to a commercial-sized quantum computer with millions of qubits capable of running quantum algorithms. After Matsuura’s talk, Intel Labs offered an interesting speech by James Clarke.

Clarke pointed out that Intel is strongly interested in building a quantum computer of millions of qubits. “We are building our qubit chips using the same technology in transistors. We are controlling our qubits using CMOS control electronics, which are made with our transistor technology,” said Clarke.

Short-term applications will be in the fields of chemistry, materials, biology, medicine. In the long term, applications such as optimization algorithms, cryptography, and machine learning are being sought.

Clarke pointed out that there are many ways to make quantum bits for a quantum computer. The first is a trapped ion, in which you study the excited states of a metal ion with a laser. These excited states can represent zero and one of the qubits. This technology is actually very similar to that of an atomic clock, which won the Nobel Prize for physics in 2012.

The second technology is that of superconducting qubits, where small rings of a superconducting metal are used to essentially create an artificial atom whose state represents zero and one of the systems.

The third technology is that of the silicon quantum dots or spin qubits, which essentially controls the spin of an electron and that spin represents the zero and one state of the qubit. “We think this is a very interesting technology for Intel,” said Clarke.

The analogy with the transistor allows us to explain this latest technology. The transistor is essentially a switch; when you apply an electrical voltage or potential, you have an electron current flowing through the device.

“Transistors are the most omnipresent man-made object on Earth. At Intel, we believe we ship 800 quadrillion transistors a year; it’s an incredible number. And the fact is, there are essentially a few transistors for everyone on the planet for every minute of every day. There have been predictions that by the middle of this decade, there will be more transistors on Earth than there are human cells. Transistors are everywhere,” said Clarke.

Instead of having a current of many electrons flowing through the device, a single electron is trapped. In that device a single electron transistor is created. By putting together many of these single transistors, we can create a network of electrons. And by controlling the potential between the individual transistors, we can actually control the interaction between two adjacent electrons.

The involvement of a single electron transistor in a magnetic field will cause that single electron to have two energy states that will be used for a qubit. “With two states, we are essentially controlling a single electron. The way we create our rotational cubits is the same way we create our transistors. With two states, we’re essentially controlling a single electron. The way we make our spin cubits is the same way that we make our transistors. Our current technologies are based on the fin-FET geometry, we’re making our qubit structures in the same way based on fins,” said Clarke.

At this point, improving qubit technology involves solving the same challenges we have with transistors, i.e., variability in size, gate oxide defects, variability in voltages. During the process, it’s important to characterize qubits quickly.

**Qubit Control**

One of the challenges for quantum computing is qubit control. These quantum bits of information are very fragile, as Matsuura pointed out.

Today’s qubits are controlled by many electronics racks with complex wiring that connect to the qubits, which reside in a cryogenic refrigerator to shield the fragile qubits from thermal and electrical noise for commercial-scale quantum computing. “This is an area that Intel is addressing with our cryogenic qubit control chip technology with scalable interconnects,” said Matsuura.

Making more qubits leads to the creation of millions of wirings that make the hardware too complicated. One of the approaches Intel is taking to make that wiring more efficient is to bring control very close to the quantum chip, with a technology that is based on CMOS technology. In the long term, it will allow having fewer wires and a more elegant interconnect system. The technology is Intel’s Horse Ridge, produced with Intel’s 22FFL CMOS process. Its functionality has been verified at 4 kelvins.

During the Intel Labs event, the company introduced Horse Ridge II, its 2nd-Gen Horse Ridge Cryogenic Quantum Control Chip. Horse Ridge II is built on the ability of first-generation SoC and it generates radio frequency pulses to manipulate the state of the qubit, known as qubit drive. It introduces two additional control features: Qubit readout, a function that grants the ability to read the current qubit state; and Multigate pulsing, which enables simultaneous control of many qubit gates. The addition of a programmable microcontroller operating within the integrated circuit enables Horse Ridge II to deliver higher levels of flexibility and sophisticated controls in how these control functions are executed.

With Horse Ridge, Intel wants to improve the scalability of quantum computers to thousands, even millions, of qubits by reducing the complexity of quantum system interconnects, which is one of the critical barriers to achieving quantum practicality and solving real-world problems through quantum computers.

**Programming a quantum computer**

A quantum computer doesn’t work like a classical one: instead of operating on binary arithmetic, a quantum computer manipulates the probability amplitude of quantum wave functions, and then you sample resulting probability distributions. “So, programming a quantum computer is very different from programming a classical one. Qubits are really fragile, the ability to correct qubit errors as they occur will be very important. But since today’s quantum computers don’t have an error correction system implemented, we’re developing noise resilient quantum algorithms and error mitigation techniques to help us run those algorithms on today’s small qubit systems,” said Matsuura.

She added, “The qubit control processor sends microcode to the control electronics. And it translates all the logical operations from the algorithm that needs to occur to run this quantum algorithm into microcode. That tells the control electronics what pulses to send, and when to send them to the qubits. The runtime software, which runs on a classical processor, loads and executes the quantum program, your algorithm, and it feeds those sequences of quantum operation instructions to the qubit control processor to execute. The program code which is comprised of classical and quantum instructions, is generated by a quantum compiler. The compiler takes the algorithm, compiles it, and it figures out how to map and schedule your quantum up qubits given the connectivity between the qubits and the specific properties of the qubits.”

The program code, made of both classical and quantum instructions, is generated by the quantum compiler. It calculates how to map and program quantum operations on qubits, given the connectivity between qubits and their properties. Quantum compilers have a challenging task. They have to choreograph a sort of qubit dance that positions and moves the qubits to the right place at the right time, and the algorithm that works against deadlines is that qubits have a very short life, usually fractions of a second, and the operations require significant and often variable timescales.

Error correction is a critical element of quantum computing; it requires a lot of work to choose the correct logical qubit without error for commercial-scale systems with millions of qubits. Meanwhile, Intel is developing noise-resilient quantum algorithms and error mitigation techniques to help make these algorithms work on all qubit systems today.

The correction of a quantum error is fundamental to most quantum computer projects as it helps preserve the fragile quantum states on which quantum calculation depends. The necessary operations for error correction are not only very complex, but must also keep the quantum information unchanged.

One way to improve error tolerance is to delegate part of the calculation to a CPU. And, in reality, this classical-quantum hybrid approach is needed throughout the stack. “We expect many of the algorithms we use for commercial full-scale quantum computing systems to be hybrids composed of classical and quantum parts, leveraging the unique strengths of each of the two computing models,” said Matsuura.

**Validation of a Quantum Computer**

Research and development begin with a deep understanding of the workloads that the system is about to perform. The nature of the workloads guides the design of a complete computer system.

Anne Matsuura’s talk after Clarke’s speech highlighted the fact that even with a large number of qubits, without building all the elements of a compute stack, we could never achieve our goal of running useful applications on quantum computers. “There’s a long way to go before we actually have a practical, commercial size, useful, impactful quantum computer. And without the complete stack, hardware and software, we won’t have a quantum computing system,” said Matsuura.

She added, “Intel has introduced equipment that helps to test our qubits very quickly on CMOS wafers in our fabs. We are talking hours instead of days to get information; we are essentially imitating the cycle of information rotation that we have in standard transistor research and development. With our new cryoprober, a custom-designed equipment we developed with our partners Bluefors and Afore, we can get test data and learn from our research devices 1000 times faster, significantly accelerating qubit development.”

During his talk, Clarke explained: “At Intel, we are making our qubits on the same technology that we use with our advanced transistors technology. This is done in the plant located in Oregon. And we’re making these devices using 300mm wafers; every wafer we get, we produce thousands and thousands of quantum devices to test these qubits. We use the so-called dilution refrigerator to cool our quantum chips at very low temperatures to preserve the quantum effects. The temperatures we are considering are a fraction of a degree above absolute zero.”

In the process, there are probes to characterize these transistors. Through Intel’s cryo prober, it is possible to scan a 300mm wafer to characterize qubits quickly. One way to analyze the situation is by applying small microwave pulses near the qubit to observe the Upstate and Downstate state changes.

Future quantum computing applications have captured the imagination of the entire world, and, as a result, it has been the subject of considerable hype from the press. Potentially, quantum computers may one day have an impact on shipping and routing logistics, designing new drugs and areas such as protein folding perhaps even modeling climate risk analysis, and the price of financial options, and of course, one of the original applications that caught the interest on quantum computing: cryptography. In reality, though, these uses depend on discoveries in quantum computing, hardware and software, and will take many years to realize.

Since quantum computing is a completely new type of computing, which runs programs in a completely different way, we need hardware, software, and applications developed specifically for quantum computing.

“This means that quantum computing requires new components on all levels of the compute stack from the control electronics to the compiler to the qubit control processor and the qubit chip device. Intel is developing all the components of the complete quantum computing stack,” said Matsuura.

Making these quantum components work together is some sort of quantum choreography, as Matsuura pointed out. She added that collaboration with external entities could provide that extra momentum for progress in the field. “We have collaborated with various universities such as the University of Chicago. We have recently committed to participating in the announced National Quantum Information Science Research Center of the Department of Energy called Q-NEXT, where Intel will provide research partners with access to a full quantum stack,” said Matsuura.

There are still many challenges in quantum computing technology, as Matsuura pointed out during our interview and during the Intel Labs Day event. One is scalability. Today’s quantum computing systems are scalable brute force versions that avoid the problems that come with scaling to millions of qubits. “We’re trying to figure out how to scale to a really large number of qubits,” said Matsuura.