5. Discussion

5.6 Impact of numerical model assumptions

Given the demands of three dimensional simulations, it is useful to explore ways in which the global problem can be restricted without too much loss of significance. One such simplification is the cylindrical disk limit, and another is the restriction to a smaller domain in $\phi $. This paper has examined the influence of the later approximation for a model with an initial vertical field, and one with an initial toroidal field.

The explicit comparison between simulations using the full $2\pi$ and ones that span only $\pi/2$ in angle suggests that the restricted azimuthal domain is indeed a useful approximation. The vertical field simulations are qualitatively very similar throughout. Although the initial growth stage in the toroidal field cases are noticeably different in the two different computational domains, the final turbulent states are again very similar. With vertical fields the magnetic azimuthal power spectrum has a significant break at relatively high m wavenumbers, consistent with the fastest growing modes of the vertical MRI. With toroidal fields the power spectra level out toward low m and do not show the same break. Again, this is consistent with the MRI: weak toroidal fields are unstable for all m less than $\approx v_\phi/v_A$. In either case, however, the power spectra of the $2\pi$ and $\pi/2$ simulations are very similar where their wavenumbers overlap. Total stress and accretion rate values are reduced by about 10% in the restricted domains, indicating that 10% of the energy is to be found on the largest azimuthal scales. This is not entirely negligible, but the qualitative difference does not appear to be profound.

A significant factor in global simulations remains the grid resolution which will always fall short of what one can achieve with a local simulation. Here the average radial grid zone size is $\Delta R =
0.23$. The standard radial resolution in the local simulations of HGB95 was almost a factor of 10 better. Because of the lack of stratification, the vertical resolution in the cylindrical simulations is comparable to the local model. A resolution comparison done by HK found larger magnetic energies with better resolution, a result consistent with more detailed resolution studies in the local shearing box. Even with the grid resolution that is presently possible, however, global simulations generate and maintain MHD turbulence with physically significant stress levels.


Title Title Page   |   Method 5.5 Coherent Structures   |   Discussion 6. Conclusion