Non classicality criteria A key point of the quantum theory is the existence of nonclassical states displaying properties incompatible with any classical theory. These states are clearly interesting since they support the existence of the quantum theory. Thus, the discovery, detection and characterization of nonclassical states is always an area of interest in quantum physics, including quantum optics in particular. At this stage it is convenient to recall that in quantum physics the system states play two different but complementary roles. A) They provide complete information about all properties of any system. B) They determine the statistics of any measurements by projection on the state of the measured system. To distinguish both roles we may refer to them as measured states (case A) and measuring states (case B). In quantum physics the statistics p(B) of the result B of a measurement is usually given by the formula p ( B) B A 2 AB 2 . This is the scalar product between an state in role A and another in role B. There is a complete symmetry between measuring and measured states despite their very different roles. It is worth noting that that the nonclassicality of measuring states (this is the nonclassicallity of observables) has been rarely studied since all the effort in this subject has been devoted to the nonclassical properties of measured states. Since the outcomes of quantum measurements are almost by definition random the detection of the nonclassicality depends on the evaluation of an statistical estimator. Most often this is variance, as it is the case of squeezing of quadratures or subPoissonian photon-number statistics. The variance X of the observable X is X 2 x j x 2 P( j ) , j x x j P( j ) , j where x j are the possible values for X with probabilities P ( j ) that we have assumed discrete for simplicity. Although variance is an estimator with many good properties it has also some noticeable drawbacks. The main difficulty is that it requires some more or less complex calculus that in general will amplify unavoidable errors of practical origin. For example, in the variance the outcomes with increasingly large | x j x | are exceedingly overvalued and multiply by very large values the probabilities P ( j ) of very unlikely outcomes that then experimentally determined with large relative errors. We may refer to this as tail effect. This leads to paradoxical results since distributions P ( j ) highly concentrated may have diverging variances. In this context the purpose of our work is twofold: i) Discover the most simple non classicality criteria. ii) Apply the above criteria to both measured and measuring states. The clssical versión of the above formula for p(B) is p( B) dzPA ( z ) PB ( z ) where z represents coordinates in the phase space of the problem (set of variables that competely specify the classical system state), and PA, B are the probability distribution on the phase space associated to the system states A, B , whatever they are. Since they are true probability distributions we have classically PA, B ( z ) 0 and dzPA, B ( z ) 1 . Since PA ( z) 0 we have PA ( z ) PB ( z ) PA ( z ) PB max where PB max is the maximum of PB (z ) so that using that dzPA ( z ) 1 we get p( B) dzPA ( z ) PB ( z ) PB max . Similarly, since PB ( z) 0 we have PA ( z ) PB ( z ) PB ( z ) PA max where PA max is the maximum of PA (z ) so that using that dzPB ( z ) 1 we get p( B) dzPA ( z ) PB ( z ) PA max . These equations provide upper bounds for the probability of the outcome B provided that PA ( z) 0 and PB ( z) 0 , respectively. Therefore, the violation of these bounds would mean that the key classical fact PA, B ( z ) 0 is infringed so that the state mut be nonclassical. The first bound above will be violated by nonclassical measured states while the second bound will be infringed by nonclassical measuring states. It is worth noting in passing that the definition of PA, B ( z ) in the quantum domain is far from being simple problem. Nevertheless, when this is solved the conclusions obtained above derived from simple reasoning are still valid. The main properties the above nonclassicality criteria that distinguish this from other approaches are: ▪ These nonclassical tests requires no data analysis whatsoever. This is just to perform the measurement and compare the number of outcomes B with the corresponding bound. ▪ This simplicity implies that the criteria are exceedingly robust under practical imperfections such as losses, inneficient detection, or low number of repetitions of the experiment. As a matter of fact, we have shown that, unlike other tests, in this case the practical imperfections may help to observe nonclassical behaviours. ▪ These criteria are very general in the sense that they apply equally well to every system and observable. They are equally valid for measuring and measured states. ▪ These criteria are independent of previous nonclassical critera. Nonclassicality of states and measurements by breaking classical bounds on statistics A. Rivas y A. Luis, Phys. Rev. A 79, 042105 (2009) As we have discussed above, the definition of the phase space distributions PA, B ( z ) are far from being trivial in the quantum case. A consequence of these difficulties is that there is more than one definition depending on satisfaction of different properties. The above analysis was based on the so called P and Q functions defined in terms of the quadrature coherent states. In a further work we have analyzed nonclassical effects PA, B ( z ) 0 when the definition used for PA, B ( z ) is the one introduced by Terletsky-Margenau-HillKikwood. This definition naturally arises when performing weak measurements so that the measurement of a given observable is performed causing the minimum perturbation on the observed system. An important result is that in this approach the quadrature coherent states display nonclassical behaviour. This is relevant since these states are always presented as a basic universal example of classical states. Nonclassicality in weak measurements Lars M. Johansen, A. Luis, Phys. Rev. A 70, 052115 (2004)