Lineshape issues at low pressure with DIT min-RMS

[erwanp]

🐛 Describe the bug

For very narrow broadening (typically mbar), the lineshape obtained with min-RMS weights of the DIT algorithm ( #116 ) give wrong lineshape.

This happens both with FFT and Analytical Voigt.

🖥 Steps To Reproduce

for broadening_method in ["fft", "voigt"]:
    
    for optimization in ["simple", None, "min-RMS"]:
        
        if broadening_method == "fft" and optimization is None:
            continue # not Implemented
        
        import matplotlib.pyplot as plt
        from radis import calc_spectrum
        s = calc_spectrum(4355, 4365,         # cm-1
                          molecule='CO',
                          pressure=1e-3,   # bar
                          Tgas=1000,           # K
                          databank='hitran',  # or use 'hitemp',
                          optimization=optimization,
                          wstep='auto',
                          verbose=False,
                          broadening_method=broadening_method,
                          name="DIT: {0}, Lineshape : {1}".format(str(optimization), broadening_method)
                          )
        s.plot('xsection')
        plt.title(s.name)
        plt.yscale("log")

Analytical Voigt :

FFT :

💡 Possible solutions

Fix coefficients @dcmvdbekerom ?

in the meantime :

• Switching default weights to "simple"
• add a warning if pressure/wstep too small ?

[anandxkumar]

The temporary fix can be a good first issue for newcomers!

[dcmvdbekerom]

This also happened in #186 (comment)

Even after all corrections, the optimized weights do this at too narrow widths.

My preference would be to make the simple weights default, they are more reliable if a little less accurate, but you can compensate by increasing width grid resolution

[erwanp]

@anandxkumar already done in #422 :)

[ishumi007]

Hi @erwanp, is this issue still open if yes then I'd like to contribute and can you please tell me how can I do so I mean I know basic python but I am ready to learn anything that will be required here. Thank you

[anandxkumar]

Hello @ishumi007, thanks for showing interest in the issue. Erwan actually created a Pull request regarding the same, so I would suggest you look into other issues and let us know if you need any help regarding it :)