Quantum computers require many quantum bits to perform complex calculations, but devices with more than a few bits are difficult to program. A device based on five atomic quantum bits shows a way forward. See Letter p.63
Quantum-savvy entrepreneurs are already bringing the first quantum computer processors out of the physics laboratory and onto the market. But these devices are mostly designed to perform just one function and cannot be programmed to run different algorithms. It would therefore be advantageous to build a fully fledged quantum computer that could be programmed to run anything we might want. In particular, it might execute the complex quantum algorithms that researchers think will solve today's intractable problems in quantum chemistry, materials science and data security. On page 63, Debnath et al.1 present a small but fully programmable quantum computer consisting of five quantum bits (qubits), and they demonstrate its functionality by running several simple quantum algorithms.
Debnath and colleagues' computer is based on one of the oldest and most developed quantum architectures, which dates back to a design2 proposed by physicists Ignacio Cirac and Peter Zoller in 1995. In this design, the computer's qubits are individual atomic ions that are trapped in a line using magnetic fields and manipulated with lasers (Fig. 1). The trapped ions behave like a tiny crystal, and precisely controlled vibrations along this line can cause the ions to become 'entangled'. Entanglement is a key ingredient of quantum computing whereby two or more particles share a common state, such that each particle can no longer be described independently. Unlike most other quantum-computer architectures, the operations used to entangle the particles are not restricted only to neighbouring qubits.
Decades of research into precision metrology, such as the development of atomic clocks, now allow the quantum electronic states of trapped ions to be manipulated at an exquisite level of control and stability. Debnath and collaborators took advantage of this work and have also made several improvements to Cirac and Zoller's design, including the ability to target each ion (in this case, five ytterbium ions) individually with optical lasers. The net result is an elementary quantum processor in which every basic operation — initializing the states of the qubits, transforming them, entangling any pair of ions, and reading out the ions' quantum state — can be performed with errors occurring less than 2% of the time.
The accuracy of the authors' quantum processor allowed them to develop preset quantum logic gates that enact a desired sequence of laser pulses to generate an elementary component of a quantum circuit. In addition, they built a compiler that can take a quantum program — an algorithm designed to exploit some aspect of quantum mechanics to solve a mathematical problem — and determine how to operate the hardware to run the program.
The problems that can be solved by a small computer with only five qubits are limited — they could be solved faster with even the most sluggish conventional laptop. But nonetheless, running simple algorithms can yield valuable information about the performance of the quantum processor as a whole, even when the outcome of the algorithm is already known. Why? A key concern for quantum architects is that qubits may seem to operate well when viewed individually, but can fail in unknown ways when required to work in tandem with many other qubits as part of a complex system. Simple algorithms are therefore used as a benchmark to see how several qubits function when combined in a larger circuit.
Debnath et al. demonstrate several algorithms. These include the Deutsch–Jozsa3 and Bernstein–Vazirani4 algorithms, which both use quantum effects to perform a mathematical calculation in a single step, whereas a conventional computer would require several operations. They also demonstrate a quantum Fourier transform5,6, which is a key component of many of the heftier quantum algorithms, such as those used to break encryption. In all of these demonstrations, the resulting error rate is consistent with the authors' observations of how their qubits work in isolation, showing that the qubits can be used together in more-sophisticated algorithms in the future.
There is still a long way to go before quantum computers can reach their full potential. For the trapped-ion architecture explored here, researchers have already hit the limit of the number of ions that can be placed in a line in a single trap — around a dozen7. The future of this field is believed to involve either joining many such traps together using optical quantum couplers, or shuttling ions between interaction zones in microfabricated traps that have a 2D layout8. The latter approach also offers the tantalizing possibility of low error rates for basic logic operations, perhaps even just one error in every thousand operations — a figure commonly thought to be the highest error rate that a large-scale quantum computer could tolerate. Research in these directions has been encouraging, but it may be a while before these scalable approaches can reproduce even the five-qubit results demonstrated by Debnath and colleagues' quantum computer.
Trapped-ion quantum architectures are not the only game in town. A range of other solid-state, atomic and optical quantum systems each have different advantages for quantum computing. Notably, an approach using qubits built of superconducting circuitry — considered the dark horse of quantum computing research only a decade ago — has shown enormous recent success9,10. Not only can superconducting technologies now compete with the phenomenal precision that has been shown with trapped ions, but they can also operate at much higher speeds and may have a clearer pathway to being scaled up.
A programmable five-qubit quantum computer built using superconducting circuits has now also been demonstrated11, and has similar capabilities to Debnath and colleagues' device. Both the superconducting-circuit and ion-trap approaches seem to be capable of being scaled up to larger devices that have more quantum bits. The next challenge for all of these technologies is to demonstrate that quantum error correction can bring error rates down to negligible levels.
Debnath, S. et al. Nature 536, 63–66 (2016).
Cirac, J. I. & Zoller, P. Phys. Rev. Lett. 74, 4091–4094 (1995).
Deutsch, D. & Jozsa, R. Proc. R. Soc. Lond. A 439, 553–558 (1992).
Bernstein, E. & Vazirani, U. SIAM J. Comput. 26, 1411–1473 (1997).
Shor, P. SIAM J. Comput. 26, 1484–1509 (1997).
Nielsen, M. A. & Chuang, I. L. Quantum Computation and Quantum Information 2nd edn (Cambridge Univ. Press, 2002).
Monz, T. et al. Phys. Rev. Lett. 106, 130506 (2011).
Kielpinski, D., Monroe, C. & Wineland, D. J. Nature 417, 709–711 (2002).
Barends, R. et al. Nature 508, 500–503 (2014).
Chow, J. M. et al. Nature Commun. 5, 4015 (2014).
Takita, M. et al. Preprint at https://arxiv.org/abs/1605.01351 (2016).