We demonstrate efficient and reversible mapping of a light field onto a thulium-doped crystal using an atomic frequency comb (AFC). Thanks to an accurate spectral preparation of the sample, we reach an efficiency of 9%. Our interpretation of the data is based on an original spectral analysis of the AFC. By independently measuring the absorption spectrum, we show that the efficiency is both limited by the available optical thickness and the preparation procedure at large absorption depth for a given bandwidth. The experiment is repeated with less than one photon per pulse and single photon counting detectors. We clearly observe that the AFC protocol is compatible with the noise level required for weak quantum field storage.

We study the efficiency of the Atomic Frequency Comb storage protocol. We show that for a given optical depth, the preparation procedure can be optimize to significantly improve the retrieval. Our prediction is well supported by the experimental implementation of the protocol in a Tm 3+ :YAG crystal. We observe a net gain in efficiency from 10% to 17% by applying the optimized preparation procedure. In the perspective of high bandwidth storage, we investigate the protocol under different magnetic fields. We analyze the effect of the Zeeman and superhyperfine interaction.

We consider in this paper a two-pulse photon echo sequence as a potential quantum light storage protocol. It is widely believed that a two-pulse scheme should lead to very low efficiency and is then not relevant for this specific application. We show experimentally by using a Tm${}^{3+}$:YAG crystal that such a protocol is on contrary very efficient and even too efficient to be considered as a good quantum storage protocol. Our experimental work allows us to point out on one side the real limitations of this scheme and on the other side its benefits which can be a source of inspiration to conceive more promising procedures with rare-earth ion doped crystals