If you buy into the hype, quantum computers promise to revolutionize the world. Scientists hope future versions of this nascent technology could allow pharmaceutical companies to discover new drugs in weeks instead of years or enable governments to calculate the logistics of how to feed the world’s growing population. Tech giants, start-up companies and university laboratories are vying to create bigger and better quantum computers. But so far none have conclusively achieved “quantum advantage”—the point where a quantum computer can solve a problem no classical computer can. The key quantum computing element is the qubit, and different groups are betting on different qubits to make quantum computers roar.
A complete quantum computing system could be as large as a two-car garage when one factors in all the paraphernalia required for smooth operation. But the entire processing unit, made of qubits, would barely cover the tip of your finger.
Today’s smartphones, laptops and supercomputers contain billions of tiny electronic processing elements called transistors that are either switched on or off, signifying a 1 or 0, the binary language computers use to express and calculate all information. Qubits are essentially quantum transistors. They can exist in two well-defined states—say, up and down—which represent the 1 and 0. But they can also occupy both of those states at the same time, which adds to their computing prowess. And two—or more—qubits can be entangled, a strange quantum phenomenon where particles’ states correlate even if the particles lie across the universe from each other. This ability completely changes how computations can be carried out, and it is part of what makes quantum computers so powerful, says Nathalie de Leon, a quantum physicist at Princeton University. Furthermore, simply observing a qubit can change its behavior, a feature that de Leon says might create even more of a quantum benefit. “Qubits are pretty strange. But we can exploit that strangeness to develop new kinds of algorithms that do things classical computers can’t do,” she says.
Scientists are trying a variety of materials to make qubits. They range from nanosized crystals to defects in diamond to particles that are their own antiparticles. Each comes with pros and cons. “It's too early to call which one is the best,” says Marina Radulaski of the University of California, Davis. De Leon agrees. Let’s take a look.
The Superconducting Qubit
The qubit that most frequently grabs headlines is the “superconducting” qubit, preferred by Google and IBM, two of the largest companies that are developing quantum computers. Google’s largest functioning computer has 53 superconducting qubits, and IBM’s has 433, though right now more isn’t necessarily better. Google’s qubits are made of aluminum; IBM uses a mix of aluminum and niobium, the two most often used materials for this qubit type.
A superconducting qubit is typically a tiny loop or line of metal that behaves like an atom—an inherently quantum object. The two states of the qubit correspond to two energy states of this artificial atom: its lowest energy state, which is known as the ground state, and the next one up. The states are initiated and controlled using microwave pulses.
The superconducting qubit became an early quantum-computing front-runner in part because it can be manufactured and operated with existing technologies used to make and operate electronic transistors. The qubit processor is about as wide as the thickness of a fingernail. Small is key because estimates indicate that a world-changing quantum computer will need one million to 10 million qubits, says quantum computing engineer Jeff Thompson, also at Princeton.
Although superconducting qubits are small and inexpensive, the hardware needed to operate them is neither. To behave like an atom, a superconducting qubit must be cooled to a few hundredths of a degree above absolute zero, which is –273 degrees Celsius. Doing that requires a dilution refrigerator, a machine bigger than a residential fridge and significantly more costly to buy and operate. Also, at least two wires and other hardware are needed for every qubit. “The cost per qubit ends up pretty high,” Thompson says.
Right now scientists can only support tens to hundreds of superconducting qubits in a dilution refrigerator, but they have plans for getting to the thousands. For millions of connected qubits, engineers will, among other things, either need to build bigger refrigerators, a challenge they are working on, or quantumly connect—that is, entangle the signals—of superconducting qubit arrays housed in different refrigerators, a currently unachievable feat. Finding a way to get the quantum information out of one cold qubit and into another, is “the holy grail” of this technology, says Britton Plourde, a superconducting qubit expert at Syracuse University. “It’s a very hard problem.”
A popular alternative to the superconducting qubit is the trapped-ion qubit: a charged atom or molecule that behaves like a tiny bar magnet. The two states of the ion correspond to two orientations of this magnet, say up and down, and they can be set by hitting the ion with a laser beam thinner than a human hair. Companies that are exploring this technology have popped up in the past few years, including Alpine Quantum Technologies (which has a computer with 24 qubits), IonQ (which has 29 qubits) and Quantinuum (which has 32 qubits). AQT uses calcium ions; the other two use ytterbium ions.
Each company’s trapped-ion computer looks slightly different, but they all contain the same elements: a computing chip the size of a dime or bigger, a cylindrical vacuum chamber the size of a large beer can around that chip, a handful of lasers and a light detector. The chip houses the ions and traps them using electric fields in the voids between its tiny printed circuits. The lasers shoot through the windows of the vacuum chamber, cooling the ions and operating the qubits.
The vacuity of the space around an ion qubit means that its state (0, 1 or both) is relatively unimpacted by state-destroying air particles. As a result, it can retain its quantum information for minutes to hours—compared with a few hundred microseconds for a superconducting qubit. That long life is great for quantum data storage, but it can be problematic for performing complex calculations. That’s because the property that gives the qubit good data storage abilities—low interaction with the environment—makes it hard to control interactions among the qubits. So computations on this system take much longer.
Trapped-ion qubit computers also have a scalability problem. Each chip can contain at most a few tens of ions without the interactions among them becoming too complex to control. Reaching millions of qubits will require moving ions between modules, a feat scientists have yet to reliably achieve. Chips in multiple vacuum chambers will also need to be connected to reach the million-qubit mark.
Scalability is less of a problem for the so-called neutral-atom qubit. A neutral-atom quantum computer is like a charged-atom one, but light rather than electricity holds the atoms in place. To make the light traps, scientists shine a laser through a lens above a chamber containing neutral atoms. The lens splits the incoming beam into a multitude of light spots, each of which can hold an atom in place. The same splitting occurs for other laser beams, which are used to operate the qubits.
Scientists have created arrays of two to 1,000 neutral-atom qubits, and QuEra has a quantum computer with 256 neutral rubidium atoms, a common atom choice. Next-generation lenses and lasers will likely take that number to 10,000 or more, Thompson says. “Adding qubits just requires splitting the laser into more beams, which simplifies scaling,” he says. Neutral-atom qubits also have a reasonably long lifetime, retaining their information for tens of seconds.
One reason the neutral-atom qubit isn’t the front-runner in the quantum computer race is speed. Thus far they can perform a few calculations per second—comparable to trapped-ion quantum computers but more than 1,000 times slower than a superconducting-qubit system. And although it’s easy to make neutral-atom qubits, they are finicky to operate. Performing a complex calculation requires hitting the atom with a carefully timed sequence of laser pulses; researchers have yet to work out how to efficiently and rapidly operate more than a handful of neutral-atom qubits. Once they do—and Thompson is confident they will—neutral atoms could surpass superconducting qubits in computing power, grabbing the headlines. “I think that transition point will be pretty soon, barring any unforeseen events,” Thompson says.
Even with such a breakthrough, a system that reliably performs calculations that a classical computer can’t is still a ways off, says Luca Dellantonio, a quantum physicist at the University of Exeter in England, who develops algorithms to run on quantum hardware. It turns out that operating the qubits is the “easy” part of the challenge. The “difficult” part is correcting errors that pop up as quantum calculations progress, which happens at a significantly higher rate than in classical systems. An error occurs when the state of a qubit is altered by some influence outside of those used to run the computer. “It can be anything really,” Dellantonio says. The factor could be heat, radiation from space or a sneeze made by someone sneezing on the other side of the world. “The problem of building a quantum computer is much harder than scientists tend to admit,” he adds. “It will happen, just not as quickly as people thought.”