Photonic quantum information processing is a key element for scalable quantum technologies, and has applications in secure long-distance quantum communication and connecting nodes of a quantum computation network. However, logical photon-photon gates and state-of-the-art single photon sources rely on probabilistic processes. Quantum memories are devices that enable storage and on-demand recall of quantum states of light, and have been highlighted as a vital component in photonic networks to overcome the scaling problem by synchronising probabilistic processes.

The Raman memory has a large storage bandwidth and high synchronising capacity, and is an ideal candidate for local synchronisation. However, previous demonstrations of the Raman memory suffer from four-wave mixing noise, which prohibits quantum level operation. In this thesis I investigate methods to increase the signal to noise ratio in the Raman memory. I investigate increasing the light-matter coupling strength to boost the memory efficiency, and then explore two different methods to suppress four-wave mixing noise. I demonstrate that operating the Raman memory in a cavity is successful in reducing four-wave mixing, but it is technically challenging to maintain a high memory efficiency. I investigate a new method of noise suppression by introducing an absorption feature at the frequency of the unwanted noise field. This technically simple method is successful in reducing the noise by an order of magnitude, and will be applicable to many quantum memory protocols.

In the final section of this thesis I explore the temporal mode properties of the Raman memory. I demonstrate that the Raman memory is single mode and can be used to separate and manipulate temporal modes of light. This positions the Raman memory as a key device for enabling high-dimensional photonic quantum information processing, and enhancing light-matter interactions.

These results pave the way towards an efficient, low-noise, mode-selective quantum memory.