| In this study, nickel carbonyl decomposition in a hot wall tube reactor was investigated using mathematical modeling and experimental observations.; A simple first principles approach was used to develop a model for particle growth inside the reactor. The mass balance was composed from published mechanisms and constants for the simultaneous homogeneous and heterogeneous decomposition of nickel carbonyl. The particle population balance was constituted using a general expression for Brownian coagulation, which is valid across the range of particle size. The model assumed that the particle size was monodisperse, coalescence was immediate and that the reactor was plugflow. The model predicted a rapid nucleation burst, followed by surface reaction and an asymptotic increase in particle size.; The reactor in this study was equipped with samplers that extracted samples at S axial positions and 3 radial positions during the reaction. Special care in the design and operation of these samplers insured that thermophoresis was the dominant mode of particle collection. Each of the samples was measured by an appropriate microscope technique and the particles were sized and counted by image analysis. These measurements were converted into particle size distributions, and particle number concentrations to obtain detailed snapshots of the particle evolution at all points within the reactor.; In this study, nucleation occurred within 0.14 seconds or less. Particles grew extremely quickly by the surface reaction, to a size of 20--30 nm and a number concentration of 1015--1016 particle/gram of gas. These particles continued to grow by coagulation. As this happened, nucleation briefly continued, which created a temporary bimodal distribution, that persisted as residual nickel carbonyl gas was available. When the carbonyl gas became depleted, the bimodal distribution was transformed to a log-normal distribution by coagulation and approached the self-preserving limit. The simple aerosol model developed earlier described the experimental results well when the size distribution was log-normal, but not so well when the distribution was strongly bimodal. |
