Quantum information

An integrated light circuit

There's a long wish list for a workable quantum computer: a viable system must be fast, compact and stable. The first integrated optical quantum logic circuits are a step in the right direction.

Systems that use photons are one of several candidates for making a working quantum computer. One challenge (there are many) is to move beyond the current bulky experiments that involve just a few photons to the kind of stable, miniaturized circuits with very many logic gates that lie at the heart of conventional, classical computers. Writing in Science, Politi et al.1 report how they have used integrated optics for the first time to demonstrate several primitive functions of quantum logic.

Quite generally, the promise of quantum computing lies in making use of the 'parallelism' and 'connectedness' of quantum-mechanical phenomena such as superposition (that objects can be in many states at once) and entanglement (that the properties of two objects can be tied to one another, despite being separated in space). These features can be used to solve information-processing tasks that would be impossible, or at least very difficult, for processors denied these quantum advantages2. Instances include problems related to factoring large numbers into their prime constituents (think 21 = 3 × 7, writ large); efficiently searching an unsorted database; and even hard problems in quantum mechanics. This last application might well end up being the most important of all, helping to form a robust framework for understanding future problems in science and technology.

How can quantum computers achieve such feats? At a basic level, every quantum processor is essentially a complex, interwoven web of nested interferometers in actual physical space (for optical approaches) or in some abstract space. An optical interferometer is made of mirrors and partially reflective, partially transmissive elements known as beam splitters. These are arranged into an optical obstacle course that an incident photon can traverse in two or more ways.

This multiplicity of paths (very) loosely corresponds to the parallel operation of a computer algorithm. By arranging elements in a certain way, one can construct logic gates for single quantum bits (qubits), preparing them in arbitrary superpositions of the logical 0 and 1. Other arrangements can entangle two qubits, so that the state of one is altered by the state of the other3 (Fig. 1a).

Figure 1: Light circuits.
figure1

a, This circuit represents a basic two-qubit quantum logic gate. Blue beam splitters reflect 50% of the incident light, green beam splitters 33%. In this configuration, a combination of quantum and classical interference flips the target qubit states 0 and 1 if the control qubit is a 1. b, Politi and colleagues' first implementation1 of this scheme as an integrated optical circuit uses silica-on-silicon waveguides. c, A bulk-optics implementation of a circuit of similar complexity involves various mirrors, beam splitters and fibre inputs; the result is considerably bigger and less stable.

To understand where Politi and colleagues' achievement fits in, we need first to consider classical interference effects. We see these, for example, in light reflecting off a puddle with oil in it. The colours we see depend on the direction in which we are looking, and in the details of the oil–water non-mixture (the effect depends on the varying thickness of the thin oil film lying on top of the water). This kind of swirling, shifting pattern is exactly what we don't want for a reliable computer. Experiments on quantum computing thus far have done a tolerably good job of controlling the analogous instability, and thus realizing basic quantum operations4,5,6.

State-of-the-art processors for optical quantum computing currently use just six photonic qubits7. But the large-scale quantum algorithms of the sort we dream of running some day (or, at least, I dream of running) will require hundreds or thousands of stable, interconnected interferometers and low-loss, high-performance components. There are ways of correcting quantum-computational errors as they arise8,9, but if they are to work properly, they require a high base-level of performance that would be nearly impossible to achieve on such a large scale. And quite apart from questions of reliability, there are also questions of speed and bulk. Just as we couldn't imagine running a modern computer application on an old-school processor that uses vacuum tubes — it would be far too bulky and slow — the complex, scaled-up quantum computer of tomorrow will require a more integrated approach10.

Enter Politi and colleagues1, with their first stab at quantum-logic operations in an (optical) integrated circuit (Fig. 1b). They succeed in replacing the bulky mirrors and protracted path lengths of traditional experiments by micrometre-scale optical waveguides — pipes for light — fabricated into a silicon wafer. Much as fibre-optic cables guide light in modern telecommunications applications, these waveguides direct the photons along their desired trajectories — rather like an irrigation system to deliver photons where and when you need them.

Such integrated optical circuits are smaller and cheaper than bulk-optic systems (or at least they will be once we have settled on an optimized design and fabrication). They are also inherently more stable; temperature variations that could alter the finely balanced path lengths in particular areas of a chip tend to cancel out. In addition, once in the waveguides, photons tend to go only where they should, so there are no optical paths to align, as there are in conventional optics. Politi et al. demonstrate high-quality classical and quantum interference with their system, as well as basic one- and two-qubit logic, and entanglement with better than 92% fidelity. Their results are both a significant and a necessary step towards optical quantum computing.

Lest the reader gain an unhealthily rosy perspective, significant challenges must still be faced. There is much less opportunity to adjust these integrated circuits once they are fabricated — there is nothing to be twiddled with, unlike in conventional optical systems (Fig. 1c) — and so their design must be optimized very carefully. In addition, low-loss interconnects are a must: at present, the efficiency with which light is injected into and extracted from the circuit is 60%. That is good, but not great, and low enough to make the 99% transmission rate inside the device significantly less meaningful.

Outside the integrated circuit, other components also require work before we can achieve large-scale quantum computing using light. For a start, we need high-quality, on-demand sources of single and entangled photons, along with a low-loss device in which to store them. Then we need photon detectors of very high efficiency (so much the better if they can actually count the number of incident photons). All of these problems are currently being attacked by various research efforts around the world. Stay tuned for more big news from tiny circuits.

References

  1. 1

    Politi, A., Cryan, M. J., Rarity, J. G., Yu, S. & O'Brien, J. L. Science 320, 646–649 (2008).

  2. 2

    Nielsen, M. A. & Chuang, I. L. Quantum Computation and Quantum Information (Cambridge Univ. Press, 2000).

  3. 3

    O'Brien, J. L. Science 318, 1567–1570 (2007).

  4. 4

    Langford, N. K. et al. Phys. Rev. Lett. 95, 210504 (2005).

  5. 5

    Kiesel, N. et al. Phys. Rev. Lett. 95, 210505 (2005).

  6. 6

    Okamoto, R. et al. Phys. Rev. Lett. 95, 210506 (2005).

  7. 7

    Lu, C.-Y. et al. Nature Phys. 3, 91–95 (2007).

  8. 8

    Varnava, M., Browne, D. E. & Rudolph, T. Phys. Rev. Lett. 100, 060502 (2008).

  9. 9

    Dawson, C. M., Haselgrove, H. L. & Nielsen, M. A. Phys. Rev. A 73, 052306 (2006).

  10. 10

    Kielpinski, D., Monroe, C. & Wineland, D. J. Nature 417, 709–711 (2002).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Kwiat, P. An integrated light circuit. Nature 453, 294–295 (2008). https://doi.org/10.1038/453294a

Download citation

Further reading

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.