Welcome to Space Race, where we explore the impact of breakthroughs in science research on the intensifying global contest of superpowers in the 21st century. Subscribe for weekly breakdowns of the scientific breakthroughs that will shape the future.
This week’s newsletter focuses on the foundational unit of quantum computing — the qubit. We will motivate qubits by first examining some of the incredible applications of quantum computing, and explore the different ways to actually make the qubits that will power the quantum computers of tomorrow.
One of the most exciting trends underway today is the second quantum revolution. The first quantum wave gave us groundbreaking inventions like the transistor and semiconductors which have now become vital to our society. This second revolution holds equal promise, with the emergence of many exciting applications such as quantum sensing and communication. Despite these myriad exciting applications, one has singularly dominated the space — quantum computing.
Once we examine the impact of building an actual quantum computer, it becomes clear why this has become the holy grail of quantum research. By exploiting exotic effects that only manifest at the quantum scale, quantum computers could complete a whole class of computing tasks which would be very difficult for the fastest conventional supercomputers. For example, quantum computers would be invaluable in applications such as drug discovery and materials design because they could simulate the quantum dynamics of atoms and molecules far more efficiently than classical computers.
Another example of an application of quantum computers is Grover’s algorithm, which can search through data provably faster than any other algorithm. Yet another is Shor’s algorithm, which can break numbers into their unique prime factors, e.g given 15 as input, it could tell you that 15 = 3 x 5. Why is this important? Because all public key encryption (including your favourite cryptocurrency) relies on the fact that factorising large numbers is basically impossible for classical computers. If a quantum computer could implement Shor’s algorithm and factorise the large numbers behind encryption protocols, it could break almost all public key encryption in the world. This is a long way off yet, but just the prospect of breaking encryption is already causing problems for the US government.
What is it exactly about a quantum computer that makes it fundamentally different from a classical computers and capable of doing such amazing things? The answer lies in how both encode information at the lowest levels. Normal computers encode information in the familiar bit. Short for binary digit, bits can be either 1 or 0, or sometimes true/false or maybe on/off. The label doesn’t matter — it is the two options which represent a logical state and store the information we care about. In a quantum computer on the other hand, we have quantum bits or qubits. The special aspect of a qubit is that it can be 1, 0, or a combination of both 1 and 0 at the same time.
If that sounds weird, it’s because it is. It is extremely weird. There simply is no analogous system in our normal classical world which has the same behaviour. This feature — called superposition — is a uniquely quantum phenomenon which only manifests at very small length scales, where the laws of quantum mechanics dictate what happens. Superposition is so strange that it confused even the founders of quantum theory. Erwin Schrödinger, whose namesake equation predicts superposition, was so perturbed by this that he thought it was as bizarre as a cat that could be both dead and alive at the same time — the infamous Schrödinger’s Cat. While we probably won’t ever see superposition in cats (more on this below), experiments have confirmed that superposition exists. Time and time again.
By harnessing effects such as superposition1, we make the leap from classical to quantum and from bit to qubit. All these extra quantum effects are the origin of quantum computing.
Quantum Computing’s Bottleneck
However, it will be a long time till we can get quantum computers that do useful things, let alone break all public key encryption in the world. The issue is not in the theory, which has already been robustly tested over the past century. The problem is that these algorithms require quantum computers with a large number of qubits, just as classical computers have a high number of bits. Therefore, the key bottleneck right now for quantum computing is making high quality qubits at scale.
The problem with quantum weirdness is that it is incredibly fragile. When quantum systems interact with the environment, all the noise and heat tends to destroy any interesting quantum behaviour, leaving normal classical behaviour. This is called decoherence and it’s why your cat will never ever be in a superposition of dead and alive. It’s simply too large and interacts too much with the environment to show quantum effects.
Unlike bits, which are fairly easy to make, qubits need to first be small enough to display quantum behaviour. Even then, qubits can decohere and become useless. To manage this, qubits need to be completely isolated from the environment. One example of this is putting them in giant fridges to keep them cold. For example, Google and IBM put their qubits in a device called a dilution refrigerator, which is capable of cooling things to within mere tenths of a degree near absolute zero. For some perspective, this is about 10x colder than the cold and empty vacuum of outer space.
How to Make Qubits
Even after they have been well isolated, qubits can still decohere. At this point, the problem is not in the isolation, but in the design of the qubit itself. Much like there are various ways of making a classical bit, there are various ways of making qubits with their own tradeoffs to consider. The key tradeoff is how long the qubit implementation takes to decohere vs how easy it is to manufacture scalably.
Any quantum system with two states can technically function as a qubit. Since this is such a general prescription, there are many plausible qubit mechanisms. To get a sense of what qubits today look like, we can examine two of the main candidates — ion trap and superconducting qubits.
Ion Trap Qubits
One approach when making qubits is to base them on the systems that nature itself has provided for us — atoms.
Many of us are familiar with the classic picture of an atom with a nucleus of protons and neutrons at the centre and electrons moving around it in well defined orbits.
While this simple picture isn’t exactly what happens, it does capture one very important detail: the electrons can only occupy certain specific orbits around the nucleus. This is because quantum mechanics dictates that the electrons can have only certain specific energy values. We say that the electron energy levels are quantised.
The electrons aren’t stuck in their levels. They can jump between them by absorbing or emiting the exact right amount of energy. For instance, if the electron is at an energy level and the level above it is 2 units higher, giving the electron 2 units of energy will cause it to jump up. Giving it 1 unit will do nothing. It will still stay in the same energy level since there is no level for it to go to at 1 unit above.
The way these energy level transitions happen in nature is when the electrons absorb and emit the most fundamental form of energy — light. If electrons go from higher energy levels to lower ones, they need to give up the right amount of energy and emit light. Hydrogen, for example, is known to emit red, blue, and violet light when its electrons fall down. The specific colours correspond to the different energy spacings between hydrogen’s energy levels. To excite an electron to a higher energy level, we can simply do the reverse and instead shine light of the right colour on it. Like hydrogen, all elements of the periodic table have their own characteristic emissions of light, which depend on their own individual atomic energy level structure.
Now we have a quantum system where we have (at least) two states accessible, so we can make a qubit out of it. Simply choose any two of these states and use them as the 1 and 0 of the qubit. IonQ, one of the main companies using these types of qubits, uses the element Ytterbium since it is a very stable atom which can remain in the same state for long periods of time.
Technically, there is nothing wrong with using normal atoms as qubits. However, in a practical implementation, we need a way to trap the qubits and isolate them. The only way to do this for individual atoms is electronically since they are too small for mechanical methods. The problem is that atoms are electrically neutral. They have no electric charge since the number of negatively charged electrons is the same as the number of positively charged protons. To resolve this, we can use lasers to break electrons free of the atom. Now our atom will have less electrons than protons and have an electrical charge so it can be manipulated with electric forces. It is not an atom anymore but an ion. This is why IonQ actually uses Ytterbium ions instead of atoms.
Now that the qubits are isolated, they can be manipulated using lasers to perform computations. One advantage of IonQ’s system is its flexibility — their trap can house a varying number of ions. This means it can run different number of qubits without having to change the underlying hardware. IonQ claims to have run simple operations on a 79 qubit chain and complex algorithms on 11 qubits.
Another advantage of ion trap qubits is that every single qubit is exactly identical since all the ions are from the same element. Therefore, we avoid errors that might stem from inconsistencies in qubit manufacture since they come to us from nature directly. Due to this, ion trap qubits are some of the most stable and decoherence resistant in the field.
One downside however is that it takes a lot of infrastructure to create and trap the ions. For example, it requires a sophisticated array of lasers to create the ions and perform the quantum operations on the trapped qubits. While this leads to robust qubits, it causes problems in scalability.
79 stable qubits is impressive, but nowhere near enough to break encryption. Shor’s algorithm nominally requires closer to 1,000—10,000 qubits. This is already a lot, but the tendency of qubits to have errors inflates this number even further. In order to be able to correct errors, each of those 1,000—10,000 logical qubits needs to be made up of many more actual physical qubits which we can average over to remove errors. So to break encryption, we might need a chain of 100,000 or even more physical qubits. Given ion trap qubits’ weakness at scaling, this might prove to be quite difficult.
Superconducting Qubits
An alternate architecture preferred by some of the industrial heavyweights such as Google and IBM are superconducting qubits. Superconductivity is another uniquely quantum phenomenon which causes certain metals to lose all electrical resistance at very low temperatures. This means that currents can essentially flow forever. Superconducting qubits rely on quantum scale circuits on chips made with superconducting metal to reduce noise and protect the qubit from decoherence. The requirement of low temperatures is why IBM and Google need large dilution refrigerators.
There are many possible circuits we could make for our qubit since all we need is two energy levels, but the most common architecture consists of a capacitor and inductor. A capacitor is a component that can be thought of as a storage device for electric energy, and an inductor stores magnetic energy. When we connect them together, the stored energy oscillates between the two components and creates a system known as a harmonic oscillator.
The harmonic oscillator is an extremely well understood system in both the classical and quantum cases. The quantum harmonic oscillator, like the ions and atoms we discussed above, has quantised energy levels. Unlike atomic energy levels however, which can be quite messy in their structure, harmonic oscillator levels are all equally spaced like rungs on a ladder. The spacing of the rungs is determined by the values of the capacitor and inductor we use to make the qubit.
In this ladder of possible energies, we can play the same game and choose the lowest two rungs to be our 0 and 1 states. Unfortunately, this system isn’t a good qubit. To see why, suppose we are in the 0 level and want to excite the qubit up to 1. Just as with atoms, we can send in an external signal to supply energy and make it jump. However, since the spacing is uniform, this means that our signal will not only trigger jumps between 0 and 1, but also between 1 and 2, 2 and 3 etc. Very quickly, these other higher levels start to interfere and our qubit becomes a useless mess. This isn’t a problem with ions because they naturally have a complicated and non-uniform energy level spacing.
If we can break this uniformity in the oscillator’s energy level spacings, then we can excite only the 0 and 1 states and have a good qubit. To do this, we replace the inductor with a component called a Josephson Junction. If you read my earlier article on quantum sensing, you may recall that a Josephson Junction is a quantum device used in quantum sensors like SQUIDs. The overall effect of this substitution is to change the energy levels slightly so that the spacings stop being uniform. Since Josephson Junctions have been used for decades, they are well understood, easy to fabricate, and a good candidate to replace inductors. This almost harmonic system with non-uniform energy spacing is now called an anharmonic oscillator.
Now the superconducting qubit is ready, and as with ion trap qubits we can use external signals to manipulate them. While ion trap qubits use lasers, superconducting qubits use microwaves.
One advantage that superconducting qubits have over ion trap qubits is that they are more scalable. They rely on already existing high tech supply chains and so can be produced en masse. Their design is also elegantly simple and relies on components that have been in use for decades, such as Josephson Junctions. This is one reason why industrial players have thrown their support behind this architecture.
However, unlike ion trap qubits which are given to us by nature and are all perfectly identical, superconducting qubits have to be made in very specialised fabrication labs. Although the techniques that are used, such as lithography, are highly sophisticated and refined, there is always possibility for errors and imperfections. As a result, superconducting qubits are not as high quality and are much more susceptible to decoherence.
Despite this, advances in fabrication technology are quickly improving the quality of superconducting qubits. Google’s state of the art Sycamore quantum processor, which underlies its claims of quantum supremacy, used 53 physical qubits to solve a nontrivial problem involving random numbers. Once again, we are far from real world impacts like breaking encryption but these are important intermediate steps.
Ion trap and superconducting qubits are just two of the many possible architectures. Since the requirement for a system to be a qubit is so general, it is also entirely possible that new architectures might appear that have more favourable tradeoffs. There is also the possibility of combining current architectures into hybrid qubit systems that can benefit from the advantages of their constituents. In many ways, qubits are still in their early days and there is still a lot of interesting new research to be done.
There is no doubt that the pursuit of a quantum computer has become a holy grail of research today. With the power of quantum phenomena such as superposition, quantum computers can solve problems classical computers could never such as breaking public key encryption. To do this though, quantum computers need large amounts of high quality, decoherence resistant qubits. This is the bottleneck that is the main driving force behind the intense competition in the space today. Because of this, various qubit architectures such as ion trap and superconducting qubits are in a race for both quality and scalability. While current state of the art quantum processors aren’t capable of doing anything beyond toy computations, it appears to be only a matter of time till qubit architectures advance to the level where quantum computers will start having a real impact on the world.
Superposition is not the only quantum effect that is needed. We need others such as entanglement, which I will talk about in a future article. Obligatory footnote to make sure the quantum computing folks don’t get angry at me!