Imfit Frequently Asked Questions ================================ This is preliminary version of a page for assorted (potentially) Frequently Asked Questions about using Imfit. How do I include a true point source (star, AGN, etc.) in a model? ------------------------------------------------------------------ Although point sources can be approximated by something like a Gaussian function with a very small size (e.g, 0.1 pixels), the best way to include point sources is with the PointSource function: :: FUNCTION PointSource I_0 This uses the user-supplied PSF image, shifting it (via bicubic interpolation) and scaling its brightness using the ``I_0`` parameter (which corresponds to the total flux of the source). If you are worried about possible rotation of the PSF for different positions within the image plane, you can use the PointSourceRot function, which adds a position-angle parameter. How can I determine B/T (bulge/total) and other flux ratios? ------------------------------------------------------------ Assuming you have *either* an ``imfit``/``makeimage`` config file *or* the best-fit-parameters output file produced by running ``imfit`` (which has the same general format), you can run ``makeimage`` in “–print-fluxes” mode, which will compute the fluxes of the individual components and their fractions of the total flux of the model: :: $ makeimage config_or_bestfit_file.dat --print-fluxes (In versions earlier than 1.9, you should also supply the “–nosave” option, which tells ``makeimage`` to skip saving the model image; this is the default in version 1.9 and later.) The output will look something like this: :: Component Flux Magnitude Fraction Label Sersic 1.0800e+06 --- 0.09038 Exponential 4.8448e+06 --- 0.40545 nuclear disk Sersic_GenEllipse 5.6323e+05 --- 0.04713 bar? Exponential 5.4612e+06 --- 0.45703 disk Total 1.1949e+07 Note that the flux units are in counts (i.e., the units of individual pixel values, summed over all pixels). To get outputs in magnitudes as well, see the next question. (The “Label” entries take their names from the optional “# LABEL ” comments in the config/best-fit-parameters file.) By default, ``makeimage`` computes component fluxes by generating a 5000 x 5000 pixel internal image and summing up the pixel values (a few functions – PointSource, Gaussian, Exponential, and Sersic – are able to compute their total fluxes analytically, which is faster). Should you be working with images larger than this size – or with models whose flux will extend significantly outside the default size – you can specify a larger (or smaller) summation image size with the “–estimation-size=”, where the specified value is the size of the (square) image. Finally, fluxes for “background” image functions (e.g., FlatSky), which are unbounded and would just scale with the summation image size, are treated as zero. How can I determine magnitudes of model components? --------------------------------------------------- Imfit works in units of counts/pixel, so it does not by default worry about magnitude systems, zero points, etc. However, you can ask the ``makeimage`` program, as part of the “–print-fluxes” mode, to compute magnitudes for individual components (and the whole model): :: $ makeimage config_file.dat --print-fluxes --zero-point= The “ZP” value will be used to compute the magnitude of each component, as :: m = ZP - 2.5 log10(flux) The output will look something like this (using “–zero-point=25”): :: Component Flux Magnitude Fraction Label Sersic 1.0800e+06 9.9164 0.09038 Exponential 4.8448e+06 8.2868 0.40545 nuclear disk Sersic_GenEllipse 5.6323e+05 10.6233 0.04713 bar? Exponential 5.4612e+06 8.1568 0.45703 disk Total 1.1949e+07 7.3066