IDL vs. Python Spectrum Fitting

Background Information

The surface brightness data in the following spectra are from a southern region of NGC 2023. To learn more about the region of NGC 2023 examined, refer to The PAH Emission Characteristics of the Reflection Nebula NGC 2023 by Els Peeters et al. (found in The Astrophysical Journal, Volume 836, Issue 2, article id. 198, 40 pp. (2017) or found in the ADS at https://ui.adsabs.harvard.edu/abs/2017ApJ…836..198P/abstract)

Note

Several changes were made to the default package scipack_ExGal_SpitzerIRSSLLL.ipac when fitting the infrared spectrum of NGC 2023. Of note, the amp values of H2 S(4), H2 S(6) and H2 S(7) were changed to 0. Additionally, the amp_fixed values of the same features were changed to True.

Python vs. IDL PAHFIT Results

The graphs below compare the fit results of given pixels of NGC 2023. The top graph shows the Python and IDL spectrum fit along with the data points used. The second graph shows the residuals resulting from the fits. Lastly, the third graph compares the calculated integrated intensities of PAH features obtained from the fits. In both graphs, the resulting fit came from default maxiter values (the IDL version of PAHFIT uses a default maxiter of 200 while the Python version uses a default maxiter of 1000).

../_images/NGC2023_pixel_14_18.png

The above graphs of pixel [14, 18] represent a ‘good’ comparison between the IDL and Python fit. That is, the fits resulted in similar residuals and integrated intensities. Of note, the output of the Python version of PAHFIT fit did not contain a dust feature at 5.27 µm (DF0 amp = 0) while the IDL fit did. Additionally, both the Python and IDL PAHFIT outputs did not contain a dust feature at 14.04 µm.

../_images/NGC2023_pixel_21_37.png

Contrastingly, the graphs of pixel [21, 37] represent more dissimilar IDL and Python fits. Interestingly, in this case, the Python PAHFIT output did not contain dust features for 5.27 and 5.70 µm. Conversely, while it did contain the 5.70 µm feature, the IDL PAHFIT output did not contain the 10.68 µm dust feature.

In general, for NGC 2023, Python spectrum fitting seemed to be the least accurate below 6 µm and above 14 µm. Further, in both Python and IDL, the spectrum fit around 11 µm does not always appear accurate. In fact, the accuracy of the fit tends to decrease as the 11.0 µm PAH feature increases.