Certain features of quantum mechanics are undeniably strange, and it has taken a long time for physicists to understand them — a necessary step before being able to accept them. The predictions of quantum physics often seem to contradict the intuitive, and seemingly reasonable, assumptions about how the world should behave that can be deduced from classical physics. Not surprisingly, many experiments have been done to explore these differences, and the results have tended to support quantum mechanics. But there have been persistent loopholes in these experiments, allowing alternative interpretations of the data. On page 791 of this issue, however, Rowe et al.1 describe how they measured the quantum correlations between two beryllium ions with nearly perfect detection efficiency — thereby closing the last experimental loophole, and providing compelling evidence that some basic ideas inherited from classical physics must be abandoned.
Non-separability, or entanglement, has emerged as the most emblematic feature of quantum mechanics. Briefly stated, it says that one can define 'entangled' quantum states of two particles in such a way that their global state is perfectly defined, whereas the states of the separate particles remain totally undefined. In other words, the information contained in such an entangled state is all about the correlation between the two particles, and nothing is said — even more, nothing can be known — about the states of the individual particles.
Some thinking is necessary to realize how weird this is. In 1935 it led Einstein, Podolsky and Rosen to suggest that quantum mechanics is incomplete, on the basis that any theory of nature must be both 'local' and 'realistic'. Local realism is the idea that, because the properties of one particle cannot be affected by a particle that is sufficiently far away, all properties of each particle must exist before they are measured. But non-separability contradicts this sort of local realism.
For a long time, this debate about the nature of quantum mechanics was largely theoretical, but in 1964 Bell introduced a set of mathematical equations — Bell's inequalities — that could be tested experimentally. The inequalities must be obeyed by any local realistic (classical) theory, whereas they are violated by quantum mechanics. It became clear at the beginning of the 1980s that the overwhelming majority of experiments supported the predictions of quantum mechanics. But two important experimental loopholes meant the evidence was inconclusive — at least to supporters of local realism.
The first of these loopholes, the so-called locality loophole, arises whenever measurements are performed on two spatially separated particles, because any possibility of communication between the two parts of the apparatus must be excluded. This can be achieved if the two measurements occur in different 'light cones', which means they cannot be connected by a signal travelling at a speed equal to or less than the velocity of light. There were considerable difficulties with doing this sort of experiment, but finally all were solved, and quantum mechanics, once again, turned out to be right2.
There still remained a second loophole, based on the statistical nature of previous experiments. The 'detection-efficiency' loophole argues that, in most experiments, only a very small fraction of the particles generated are actually detected. So it is possible that, for each measurement, the statistical sample provided by the detected pairs is biased. For example, in experiments using pairs of photons emitted by atomic cascades2, only one pair in every million was used in the measurement. So, to extract a meaningful conclusion from the observed data, it was necessary to assume that a small fraction of data provides a fair statistical sample. This would be similar to allowing one hundred votes to decide a ballot of one hundred million. Not surprisingly, several local realistic models were built to mimic the experimental results, and the detection-efficiency loophole became the Achilles' heel of experimental tests of Bell's inequalities.
It was first thought that improving the detection efficiency in experiments with pairs of entangled photons2 would solve this problem. But this proved more difficult than expected and, despite several proposals3, 4, no conclusive experiment has been done. Another avenue is to use correlated measurements on pairs of massive particles, with the hope that their quantum states would be easier to detect than those of photons. Theoretical proposals considered pairs of atoms produced through a photodissociation process5, 6, or pairs of Rydberg atoms7. But such experiments are not easy, and no conclusive test of Bell's inequalities has been carried out in these systems.
In their experiment, Rowe et al.1 manipulate the quantum states of two massive entangled particles — in this case trapped ions. For an efficient measurement, it is necessary to detect and compare the individual quantum states of the two ions, which are separated by only a few micrometres. The authors use a two-step measurement process, which results in four possible outcomes (Fig. 1). Contrary to the photon- correlation experiments, in which many photon pairs are missed, here every ion pair is included in the measurement. Remaining errors are attributed to an incorrect preparation of the initial state (less than 12%), a phase error in the rotations (less than 6%), or a mistake in the identification of the final states (less than 2%). The agreement with quantum mechanics is excellent, making this the first violation of Bell's inequalities with high enough efficiency (80% overall) to close the detection loophole.
Figure 1: Testing our quantum picture of the world.
Rowe et al.1 prepared two ions in an entangled state by using Raman laser beams. Next, they applied an arbitrary rotation to each ion by simultaneously controlling the phase of the Raman beams and the strength of the ion trap. Finally, they determined the quantum state of each ion by shining light (the probe laser) on the ion pair to identify whether each ion fluoresces. There are four possible outcomes: two in which the states are the same (both ions are bright or both are dark), and two in which they are different (only one ion or the other fluoresces). The measurements give the ratio of (Nsame–Ndifferent)/(Nsame+Ndifferent), where N is the number of detected ion pairs. The experimental result violates Bell's inequalities — thereby confirming the predictions of quantum mechanics — in a way that escapes the 'detection-efficiency' loophole.
High resolution image and legend (38K)What conclusions can be drawn from this experiment? It does not close the locality loophole, whereas the experiments that do, cited in ref. 2, do not close the efficiency loophole. Closing both loopholes in the same experiment remains a challenge for the future, and would lead to a full, logically consistent rejection of any local realistic hypothesis. Even so, the overall agreement with quantum mechanics seen in all the experimental tests of Bell's inequalities is already outstanding. Moreover, each time a parameter is changed that was considered to be crucial (for example, using time-varying measurements, or increasing the detection efficiency), the experiments show that these changes have no consequence: the results continue to agree with quantum-mechanical predictions. This appears rather compelling evidence to me that quantum mechanics is right, and cannot be reconciled with classical physics.
Finally, Rowe et al.'s experiment is a vivid illustration of the high degree of sophistication that has been reached in the control of 'engineered' quantum systems. The entangled-ion pair was prepared on purpose, and can be observed at will. This opens fascinating possibilities for manipulating the quantum state of entangled many-particle quantum systems. Such systems are the basic elements for achieving long-term goals in quantum information processing, such as building a quantum computer.
