Despite the widespread research on quantum computing, nobody has built a machine that uses quantum mechanics to solve a computational problem faster than a classical silicon-based computer. Now scientists from universities in England and Australia have developed a device called a boson-sampling computer that rivals a quantum computer. Although boson-sampling computers are not believed to have all the problem-solving ability of a full quantum computer, they can solve some problems faster than today’s machines, and can be much easier to build experimentally with existing photonic technology. The device could pave the way to larger ones offering the first definitive quantum-enhanced computation. Boson sampling requires three main ingredients: single bosons, the large-scale linear manipulation of bosons, and single-boson-sensitive detectors. Photons are identical at a fundamental level, exhibiting a strong quantum level of entanglement. If two sufficiently identical photons come together, they behave in a connected way – almost as if they cling together. When scaled up to multiple input photons, these entanglements cause the outputs of a boson-sampling circuit to clump together in a characteristic way, predictable by quantum mechanics but difficult to calculate using conventional computers. The 8-cm-long silica-on-silicon photonic chip in the center of the picture served as the four-photon quantum boson-sampling machine. Arrays of single-mode fibers are glued to the left and right sides of the chip. For viewing purposes, a red laser is coupled into two of the single-mode fibers (right side of picture), which illuminate a portion of the on-chip interferometric network. For the boson sampling experiment, the red laser was replaced with single-photon sources. There are five thermal phase shifting elements on top of the chip, although they were not used in this experiment. This image relates to the paper by Dr. Justin Spring and colleagues. In their experiment, Oxford University’s Justin Spring and colleagues used single photons and quantum interference to perform a calculation that is believed to be very difficult on a classical computer. “Boson sampling provides a model of quantum-enhanced computation that is experimentally feasible with existing photonic technology,” Spring said. “Future generations of boson-sampling machines will benefit from ongoing advances in integrated photonics.” The experiment was performed on a photonic chip developed by professor Peter Smith and Dr. James Gates from the Optoelectronics Research Center at the University of Southampton. “The chip offers a scalable route … to build large linear systems required for larger boson-sampling machines,” Gates said. “If one is going to eventually need to move ‘on chip’ with more complex boson-sampling machines, there is obvious benefit in building the proof-of-principle devices ‘on chip’ as well. The move to optical processing on a chip format can be likened to the shift to integrated silicon chips in electronics.” In a separate experiment, Dr. Matthew Broome and colleagues at the University of Queensland built a device they called BosonSampling to determine whether quantum computers are the only way to perform efficient computations, or whether conventional computers can solve the problem almost as quickly. The device implemented a form of quantum computation where a handful of single photons were sent through a photonic network and then researchers sampled how often they exited the network outputs. The logo of the Quantum Technology Lab spelled out with the laser beams used in the BosonSampling device. This image relates to the paper by Dr. Matthew Broome and colleagues. “Although this sounds simple, for large devices and many photons, it becomes extremely difficult to predict the outcomes using a conventional computer, whereas our measurements remain straightforward to do,” Broome said. The device – proposed in late 2010 by associate professor Scott Aaronson and Dr. Alex Arkhipov of MIT – will provide strong evidence that quantum computers do indeed have an exponential advantage over conventional computers. “Scott and Alex’s proposal was a 94-page mathematical tour de force,” said experimental team leader Andrew White of the University of Queensland. “We genuinely didn’t know if it would implement nicely in the lab, where we have to worry about real-world effects like lossy circuits, and imperfect single photon sources and detectors.” The BosonSampling device behaves as expected, paving the way for larger and larger instances of this experiment. The prediction is that, with just tens of photons, it can outperform any of today’s supercomputers. “The first proof-of-principle demonstrations of BosonSampling have been shown – even if only with three photons, rather than the 30 or so required to outperform a classical computer,” Aaronson said. “I did not expect this to happen so quickly.” The studies appeared in Science (doi: 10.1126/science.1231692 and doi: 10.1126/science.1231440).