The limits of multiplexing quantum and classical channels: Case study of a 2.5 GHz discrete variable quantum key distribution system

Fadri Grünenfelder, Rebecka Sax, Alberto Boaron, Hugo Zbinden

Applied Physics Letters 119, 124001 (2021)

You can read the full article here.

In order to make QKD a successful technology, it must be possible to easily integrate it into existing, classical networks. The problem, however, is that typical signals in existing network environments are of much higher intensity than QKD signals. The network signals usually have a power of tens of milliwatt, while the QKD signals are below a nanowatt. Because of this huge difference in power, the network signals easily add noise to the QKD signal via Raman scattering and other processes, therefore reducing the secret key rate. The only way to avoid this is by reducing the amount of noise from the network signal which ends up in the QKD signal.


Schematic of the Experimental Setup

Schematic of the Experimental Setup for integrating QKD into existing networks


In our article, we investigated how our state-of-the-art QKD system at the University of Geneva performs in such a network environment. Specifically, we are interested in how much power in the network can be tolerated by the QKD system before the creation of a secret key becomes impossible.

By using optimized spectral filters, our high qubit repetition rate of 2.5 GHz, and using a light wavelength of 1310 nm for the QKD signals, we were able to increase the power of network signals which the QKD signal can tolerate. In a 91 km link, we could create a secret key until the co-propagating signals reached a power of 7.8 mW. For a lossy 51 km link, our system could tolerate up to 46.8 mW. We also showed that an ideal system with perfect filters could potentially tolerate up to 500mW.

In our article we also discussed previous experiments by other research groups, including Toshiba Europe, the University of Science and Technology of China, and Télécom ParisTech. By comparing these results to the ones of our work, we concluded that for short distances (below 100 km) and high throughput networks, a wavelength of 1310nm is favourable. For long distance (about 150 km) but low throughput networks, a wavelength of 1550 nm is better suited.

In conclusion, we demonstrated how a state-of-the-art QKD system would perform in a network environment similar to those of existing classical networks, and what potential improvements can be gained in the future. Successful integration of QKD in existing networks is a key feature to future-proof our way of private communication.