On Wednesday, researchers at the Joint Quantum Institute at the University of Maryland unveiled a first-of-its-kind fully programmable and reconfigurable quantum computer. The five-qubit machine, which is described in the journal Nature, represents a dramatic step toward general-purpose quantum computing—and, with it, an upending of what we can even consider to be computable.
It's often remarked rather abstractly that the rather abstract power of future quantum computers will nuke our most fundamental layer of digital security by virtue of their very existence. How will they do this? By being very powerful, goes the nigh-universal pop science answer. A very powerful computer, of the sort that has never been seen before, may use that great power to factor numbers much more quickly than could be accomplished using even non-quantum supercomputers. The RSA algorithm, which guards most of our digital data, is based on not being able to do this.
The mechanism behind this RSA crippling has had a name since 1994, one year before the very first quantum logic gate had even been realized at a NIST laboratory in Boulder, Colorado. It's called Shor's algorithm, and is a method of factoring very large whole numbers using quantum hardware. Like many quantum algorithms, it relies on a mathematical operation known as the quantum Fourier transform (QFT), which decomposes a given quantum state into its constituent parts.
The QFT was one of three algorithms that Shantanu Debnath and his team at the Joint Quantum Institute successfully implemented using their quantum computing module. While currently limited to just five qubits, the group's computer could potentially be scaled up to as many as 100 qubits, and, moreover, could be linked to other computers (hence the modularity), possibly by using photon channels. Linkages of these modules would constitute larger and larger quantum machines.
Debnath and co.'s quantum computer is based on trapped ions, or atoms that have either a positive or negative charge. This charge is manipulated such that the ions can be shoved around using magnetic fields—here, they're arranged into a line. The tight arrangement means that the ions wind up acting like particles in a crystal, which means that it's possible to get them all vibrating coherently. Getting all of the ions humming in just the right way results in quantum entanglement, a scenario in which the particles, from certain perspectives, become indistinguishable.
So, the entangled particles are all sharing the same state and, as such, wind up acting like mirrors of each other. Entanglement in such a line-based arrangement has the helpful property of not needing to be passed from neighboring particle to neighboring particle. Instead, you can skip around. This is unique to trapped-ion quantum computers—other quantum architectures are limited to neighbor-to-neighbor entanglement, which is intuitively less scalable.
Everything in Debnath and co.'s machine is done with lasers of different colors. At the beginning of a computation, one laser is used to increase the energy of the qubits (the ions) until they've reached a desired quantum state. Then, the particles are scooted through a series of quantum gates, which are themselves implemented with laser beams (which is part of what makes the system reconfigurable). Functioning analogously to switches and transistors, these gates are what represent the actual computation. The states of the particles on the other side of the gates represent the computation's result.
"By reducing an algorithm into a series of laser pulses that push on the appropriate ions, we can reconfigure the wiring between these qubits from the outside," Debnath offers in a statement. "It becomes a software problem, and no other quantum computing architecture has this flexibility."
The researchers were able to reconfigure the computer, which had a mean fidelity 98 percent, to run a variety of new algorithms without changing the hardware itself. They ran the QFT algorithm with about 70 percent accuracy, which is not ideal. The two other algorithms tested, neither of which are especially useful for real-world problems, were each about to return results that were 90 percent and 95 percent accurate on average. A general-purpose quantum computer useful in the real-world will need to both scale the number of qubits up and reduce this error to negligible levels. Neither of these things are impossible hurdles.