Narrow Linewidth Tunable Laser Source for Quantum Key Distribution (QKD)

Quantum key distribution is unusual among photonic applications in that the source laser carries a security argument, not only a performance one. The protocols that underpin QKD assume the transmitted quantum states are what the security proof says they are — indistinguishable in every degree of freedom except the one deliberately used to encode information. A source whose wavelength drifts, or whose emission carries spectral impurity or residual side modes, can silently break that assumption, because a spectral signature correlated with the encoded state creates a side channel an eavesdropper could in principle exploit, while the same imperfections also depress the achievable secret key rate through elevated error rates. Sub-1 kHz linewidth addresses both halves of that problem at once. Spectrally, the emission occupies a very narrow band with strong side-mode suppression, so states prepared at different settings remain spectrally indistinguishable. Temporally, the corresponding long coherence time is what makes phase-encoded and interferometric schemes viable, since the relative phase between pulses must stay well defined across the delay used to encode it. Continuous-variable QKD depends on this even more directly: homodyne or heterodyne detection of quadratures against a local oscillator only reaches shot-noise-limited performance if signal and local oscillator remain mutually coherent, and excess laser phase noise appears in the measurement as added quadrature noise, which is indistinguishable from eavesdropper-induced excess noise and therefore eats straight into the key rate. This variant is accordingly usable as the QKD source laser itself, and equally as the trusted reference against which prototype transmitter hardware is developed and characterised — a role that matters because a QKD transmitter cannot be validated against a source no cleaner than itself. For discrete-variable protocols the same stability underpins the interferometric visibility that sets the quantum bit error rate floor, and every tenth of a percent shaved off that floor translates into usable key at longer distances.

Wavelength placement is the second reason this source suits QKD work, and it becomes critical the moment a quantum channel has to share fibre with anything else. Practical deployments rarely get dark fibre to themselves; a quantum channel is usually multiplexed alongside classical traffic and the accompanying classical post-processing link, and the dominant threat to key rate in that configuration is not loss but noise photons — Raman scattering from classical channels and crosstalk leaking into the single-photon detector’s window. Defending against that is a wavelength-engineering exercise: the quantum channel must be placed accurately on the ITU-T grid, held there, and separated from neighbouring channels with enough margin that filtering can reject the rest. Coarse tuning allows a researcher to move the source to the intended grid channel quickly, while fine tuning places it precisely at the centre of the filter passband and keeps it there, so filter rejection is not being wasted on avoidable misalignment. Low residual drift matters because key accumulation runs for hours or days, and a slow wavelength walk toward a filter edge shows up as a gradually declining key rate that is easily misattributed to the channel or to the detectors. Multiplexed and multi-channel experiments also need several sources behaving in the same understood way, and CoBrite DX, DX2 and MX chassis options let a laboratory host multiple laser modules in one benchtop, compact or 19-inch rack unit — keeping them on a shared thermal and mechanical platform, under one control interface, and available as signal, local oscillator and reference within a single ecosystem rather than assembled from unrelated instruments. Programmatic control of wavelength and power lets these settings be scripted into automated runs and logged alongside key-rate data, so a later dip in performance can be correlated against the source’s own recorded state instead of reconstructed from memory. Field trials over installed fibre add a further requirement, since a source that returns to the same operating point after transport and power cycling saves scarce trial time.