To model-fit each of the 8.4 GHz datasets the Caltech VLBI task MODELFIT was utilised. MODELFIT uses a non-linear least-squares algorithm to fit the parameters of a set of simple components in the image plane (a model) to the visibility amplitudes and closure phases of the VLBI data and reports a measure of the goodness-of-fit (defined as the square root of the reduced chi-square value for the fit of the model to the data). MODELFIT can undertake a gradient search for the best match of model to data by finding the gradient of the goodness-of-fit function and adjusting the parameters of the model so that the gradient is followed towards the local minimum. A brute force method is also available which finds the best match of model to data by adjusting each of the model parameters individually and in turn, to minimise the goodness-of-fit.
The models initially chosen to represent the source were a minimum representation of the main features which could be seen in the images; the core and components C1 and C2, with flux densities, separations, position angles, and component extents estimated from the images.
These starting models, along with the corresponding self-calibrated data from the imaging process, formed the input to the MODELFIT program. Initially the gradient search mode was used, allowing only one parameter to vary, the core flux density, and then one by one the flux densities of the remaining components. Then the other parameters of the model components were allowed to vary, one by one, during individual runs of MODELFIT. Once this process was complete all of the model parameters were allowed to vary together with full freedom in the gradient search mode, and then finally, using the brute force method to ensure convergence.
In all cases, the resulting model did not fit the data well and the inclusion of further model components was required at each epoch. Guided once more by the images, single components were added to the starting models and the model-fitting process initiated once more from first principles. Again when the model-fitting was completed, the data at some epochs was still not well fit by the models. Again another component was added to the starting models at these epochs and model-fitting initiated a third time.
In this way, good fits to the visibility amplitudes and closure phases of the data at all nine epochs were finally obtained. At four epochs, models with 4 components were required, and at five epochs, models with 5 components were required. Thus the data could consistently be quantified with a reasonably small number of parameters, considering the degree of complexity of the source. Accompanying each of the images in Figures 5.4 to 5.12 are the best fit models for the structure of the source. Each model consists of a number of components which are identified in the right hand column of each table. Each component has 7 parameters associated with it:
Figure 5.4: VLBI image of Centaurus A at 8.4 GHz from 1991 March 6. Map peak, 3.0 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64% of peak. Beam FWHM, 7.6 
3.0 mas @ -81.7 
.
Table 5.2: Best fit model for Centaurus A data from 1991 March 6
Figure 5.5: VLBI image of Centaurus A at 8.4 GHz from 1991 Nov. 24. Map peak, 2.7 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64% of peak. Beam FWHM, 7.2 3.0 mas @ -89.3 .
Table 5.3: Best fit model for Centaurus A data from 1991 Nov. 24
Figure 5.6: VLBI image of Centaurus A at 8.4 GHz from 1992 March 26. Map peak 2.5 Jy/beam. Contours, -0.5, 0.5, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 8.7 3.0 mas @ 82.6 .
Table 5.4: Best fit model for Centaurus A data from 1992 March 26
Figure 5.7: VLBI image of Centaurus A at 8.4 GHz from 1992 Nov. 22. Map peak, 3.1 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 6.9 3.1 mas @ -79.6 .
Table 5.5: Best fit model for Centaurus A data from 1992 Nov. 22
Figure 5.8: VLBI image of Centaurus A at 8.4 GHz from 1993 July 3. Map peak, 1.7 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 8.0 3.0 mas @ -74.6 .
Table 5.6: Best fit model for Centaurus A data from 1993 July 3
Figure 5.9: VLBI image of Centaurus A at 8.4 GHz from 1993 Oct. 20. Map peak, 2.4 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 6.8 3.2 mas @ -83.1 .
Table 5.7: Best fit model for Centaurus A data from 1993 Oct. 20
Figure 5.10: VLBI image of Centaurus A at 8.4 GHz from 1994 Feb. 27. Map peak, 2.6 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 6.3 3.2 mas @ -83.3 .
Table 5.8: Best fit model for Centaurus A data from 1994 Feb. 27
Figure 5.11: VLBI image of Centaurus A at 8.4 GHz from 1994 June 20. Map peak, 2.1 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 6.4 3.0 mas @ -84.2 .
Table 5.9: Best fit model for Centaurus A data from 1994 June 20
Figure 5.12: VLBI image of Centaurus A at 8.4 GHz from 1995 July 3. Map peak, 4.1 Jy/beam. Contours, -1, 1, 2, 4, 8, 16, 32, 64%. Beam FWHM, 12.2 2.3 mas @ -4.2 .
Table 5.10: Best fit model for Centaurus A data from 1995 July 3