The origin of the power-law decay measured in the power spectra of low Prandtl number RayleighBénard convection near the onset of chaos is addressed using long time numerical simulations of the three-dimensional Boussinesq equations in cylindrical domains. The power law is found to arise from quasidiscontinuous changes in the slope of the time series of the heat transport associated with the nucleation of dislocation pairs and roll pinch-off events. For larger frequencies, the power spectra decay exponentially as expected for time continuous deterministic dynamics. DOI: 10.1103/PhysRevLett.87.154501 PACS numbers: 47.54. +r, 47.20.Bp, 47.27.Te, 47.52. +j Significant insight into the onset of chaotic dynamics in fluid systems, and continuum systems in general, has been gained from cryogenic Rayleigh-Bénard convection experiments [1][2][3][4]; for a review, see [5,6]. Two of the most dramatic discoveries were the observation of time dependence almost immediately above the onset of convective flow, and the power-law falloff in frequency for the power spectral density derived from time series of a global measurement of the temperature difference across the fluid at fixed heat flow [1]. However, these and other important observations remain poorly understood although further insight has been gained from room temperature argon experiments allowing flow visualization [7 -9]. The power-law behavior is unexpected, since bounded deterministic models typically show an exponential falloff at high frequency [10]. Phenomenological stochastic models were proposed to explain the spectra [11,12], but no understanding of the origin of the ad hoc stochastic driving has followed.In this Letter, we use numerical simulations of the threedimensional Boussinesq equations for the fluid flow and heat transport in the cylindrical geometries of the experiments with realistic boundary conditions to investigate the power spectrum in more detail. The numerical simulations allow us to determine the spatial structure of the flow field in the aperiodic dynamics, and the absence of experimental or measurement noise provides us with more complete results for the power spectra. Our completely deterministic simulations yield results consistent with the experimental observations, including a power-law falloff of the power spectrum over the range accessible to the experiment. Using knowledge of the flow field, we are able to associate this power-law behavior with specific events in the dynamics, namely, the creation and annihilation of defects in the convection roll structure, which occur on a time scale rapid compared with the slow pattern evolution. At higher frequencies, the power spectra decay exponentially, consistent with the behavior expected for smooth deterministic time evolution. The low amplitude region of the spectra was inaccessible experimentally due to the noise floor.Our simulations in a cylindrical geometry are performed using an efficient spectral element algorithm (described in detail elsewhere [13]). The velocity u, temperature ...