Secure optical communication
Adaptive coherent receivers for quantum communication with machine learning
Motivation
QKD principle
Adaptive homodyne receiver
Due to the Heisenberg uncertainty relation a lower phase uncertainty and hence a lower error probability is achievable by measuring the phase quadrature only ( m-PSK signals). A balanced homodyne detection with an adaptively controlled local oscillator phase is needed. During each signal pulse the local oscillator phase is adjusted by a controller to extinguish the BHD photocurrent. The output signal of the controller contains the phase information.
Displacement Receiver
Coherent states carrying information encoded in the phase are detected by a displacement (Fig. 2) and a single photon counter ( SPC, Fig. 1). In the simplest case, Kennedy Receiver [2], a BPSK signal in the coherent states | ±α⟩ is displaced by the amplitude +α (translates phase difference into amplitude degree of freedom).
Errors only occur from not detecting the state |2α⟩ having a non-vanishing probability to be projected onto the vacuum state. The error probability (Fig.3) is even lower using an optimized displacement amplitude β, derived from minimizing the term of the overall error probability [3].
are measurement dependent POVM elements
Applying m-PSK signals the displacement angle (LO phase, Fig. 1) has to be adjusted m-times during one symbol period to probe all signal states. Advanced probing strategies accessible through adaptive feedback.
Machine learning techniques for heterodyne quantum communications*
* This work will be carried out in cooperation with Prof. Darko Zibar, DTU Fotonik 
In quantum communications, accurate phase estimation of signals containing less than 1 photon is required. With a static heterodyne receiver, this can be done by the DSP. Laser phase noise complicates this problem. A severe performance improvement is expected, when the laser characteristics are known in detail and taken into account [4].
- First approach: Bayes filter
-
- Model with state x, measurement y, quantum noise q and measurement noise r :
- Kalman filter
Optimum filter for classical signals with white Gaussian noise - Extended Kalman filter
Kalman filter with linearized measurement function h(.) - Unscented Kalman filter
- Approximation of joint distributions using the unscented transform
- Particle Filter
Monte Carlo approximation of joint distributions
- Model with state x, measurement y, quantum noise q and measurement noise r :
- Artificial neural networks (ANNs)
- In contrast to the above model, no Markov sequence assumption is made
Experimental investigation of the influence of laser phase noise
- Detailed laser statistics have to be taken into account. The experimental variation of all rate equation parameters is therefore necessary.
- Characterization of the adaptive receivers and machine learning techniques with the above setup
QKD Demonstrator
- At the end of the project, the adaptive receivers and machine learning techniques are evaluated and compared. Based on this, a QKD demonstrator is built, which employs the most promising approach
- Implementation of error correction based on LDPC codes
- D. Huang, H. Peng, L. Dakai and Z. Guihua,
Long-distance continuous-variable quantum key distribution by controlling excess noise, Nature Scientific Reports, 2016 - R.S. Kennedy,
MIT Research Laboratory of Electronics, Quarterly Progress Report No. 108, 1973 - C. Wittman, M. Takeoka, K. N. Cassemiro, M. Sasaki, G. Leuchs and U. L. Andersen,
Demonstration of Near-Optimal Discrimination of Optical Coherent States, Phys. Rev. Lett., 2008 - M. Piels, M. I. Olmedo, W. Xue, X. Pang, C. Schäffer, R. Schatz, G. Jacobsen, I. T. Monroy, J. Mørk, S. Popov and D. Zibar,
Laser Rate Equation-Based Filtering for Carrier Recovery in Characterization and Communication, Journal of Lightwave Technology, 2015
Persons involved in this project:
Dipl.-Wirtsch.- Ing. S. Kleis , M.Sc. M. Rückmann , Univ.-Prof. Dr.-Ing. C. G. Schäffer
Co-operation partner:
Letzte Änderung: 16. November 2017