Fig. 2
From: Customizing supercontinuum generation via on-chip adaptive temporal pulse-splitting

Operational principle of the on-chip optical pulse-splitter. a Schematic diagram: The sample comprises cascaded Mach-Zehnder interferometers (MZI) and different delay lines made of high-refractive-index silica glass (white)—see Methods for details. Adjustable splitting and routing of the optical pulses into the various waveguide paths can be controlled by tuning each MZI arm phase difference ΔφN (where N is the interferometer number) via a refractive index change, thermally-induced using electrodes (yellow). We show in the figure a diagram comprising four interferometers used to adjust the relative splitting of pulses into the two arms of the subsequent unbalanced waveguide section. The last interferometer (Δφout) is used to recombine the train of delayed pulses and regulate the overall output power. b–e Examples of generated optical pulse patterns characterized using intensity autocorrelation (AC—top) and frequency-resolved optical gating (FROG—bottom) measurements53. Multiple pulses (up to 16), featured by equal power but different pulse-to-pulse separation (as low as 1 ps), were obtained by setting each of the N = 4 interferometers to predefined splitting conditions, in order to sequentially split and interleave pulse trains with variable relative delays. These simple pulse pattern examples were achieved by setting each MZI relative phase difference to one of three specific values only: ΔφN = 0, so that the incoming signal (i.e. pulse or train of pulses) entirely propagates within the short arm of the unbalanced waveguide section; ΔφN = π, so that the incoming signal entirely propagates within the long arm of the unbalanced waveguide section; ΔφN = ±π/2, so that the incoming signal is equally split between the two arms of the unbalanced waveguide section (leading to a relative delay corresponding to the unbalance between these two arms). Parameters: Δφ1–4 = 0 for (b); Δφ1–3 = 0 and Δφ4 = π/2 for (c); Δφ1–3 = π/2 and Δφ4 = 0 for (d); Δφ1–4 = π/2 for (e)