Classical correlations are described consistently within classical information theory. This thesis presents a consistent quantum information theory of purely quantum correlations, i.e. entanglement. The main problem arises when we consider mixed states, for which it is difficult to separate quantum from purely classical correlations. This problem is the main subject of the thesis and is undertaken from two different perspectives. The first approach follows Shannon’s own approach, where we define a number of intuitively clear and physically sound conditions that a “good” measure of entanglement has to satisfy, and then search for measures satisfying these conditions. Our second approach is to extend the classical idea of distinguishing two probability distributions to quantum physics. The amount of entanglement will then determine the experimental ability to distinguish a given entangled state from a classical, disentangled state. We show that these two approaches have a number of features in common, leading to the same measures of entanglement.
Classical information can be spoilt due to interactions with the environment. Classical information theory has a branch dealing with methods for protecting information called classical error correction. Quantum information is even more fragile and here we develop the quantum analogue of error correction. We develop a code that protects quantum states in the presence of spontaneous emission. We then show how to protect entanglement using this method.
We also present a cavity QED implementation of various schemes aiming at increasing and protecting entanglement between two cavities using the standard Jaynes-Cummings interaction model between an atom and a cavity.