Fiber dispersion plays a significant role in spectral broadening of incoherent continuous-wave light. We develop a self-consistent stochastic model for spectral broadening of incoherent continuous-wave light through nonlinear wave mixing and apply this model to numerical simulations of spectral broadening in a continuous-wave fiber Raman laser. The results of these numerical simulations agree very well with carefully conducted laboratory measurements. Under a wide range of operating conditions, these numerical simulations also exhibit striking features, such as damped oscillatory spectral broadening (during the initial stages of propagation) and eventual convergence to a stationary, steady-state spectral distribution at sufficiently long propagation distances. We analyze the important role of fiber dispersion in such phenomena. We also derive an analytical rate equation expression for spectral broadening, whose numerical evaluation is far less computationally intensive than the fully stochastic simulation, and a mathematical criterion for the applicability of this analytical expression. C 2011 Society of Photo-Optical Instrumentation Engineers (SPIE).
IntroductionConsider what is entailed in a full blown numerical simulation of incoherent cw wave mixing. For the simpler case of coherent light containing a comblike distribution of N optical frequencies of known initial phase, one could start at the beginning of fiber (z = 0) and calculate the extent to which power is exchanged between the various spectral components during propagation over a minute distance δz. In the limit of small δz, the increase/decrease in power for each spectral component over the distance δz can be calculated independently. For N spectral components, there are N 3 four-wave mixing interactions that must be considered for each propagation step δz.To further generalize to the case of incoherent light, one must make N relatively large to simulate the effect of a continuous spectral distribution. However, the situation is further complicated by the fact that the initial amplitude and phase of each spectral component is neither known nor deterministic. For each spectral component, to calculate the increase/decrease in power over a distance δz, we must calculate the ensemble average of this observable for a large number of statistically independent trials whose initial amplitude and phase are randomly assigned at z = 0 on the basis of a probability distribution that correctly accounts for the statistical properties of incoherent light. Such an ensemble average is valid because incoherent light is ergodic in the sense that the expectation value for a continuous sequence of a random quantity (such as optical power or spectrum) over a period of time is well represented by the ensemble average of a large number of independent random trials.Thus, in principle, starting with the nonlinear wave mixing equation, derivation of the differential equation for the