Variation in the measured [OIII]5007/[OIII]4959 flux ratio.

[dirkscholte]

I found that for some low redshift galaxies the measured [OIII]5007/[OIII]4959 ratio deviates from the theoretical ratio by a small fraction.

The deviation can be seen for a small sample of high S/N [OIII] [LINE]_FLUX measurements:

This also seems to be present when using [LINE]_BOXFLUX:

[moustakas]

Thanks for the report @dirkscholte.

@jdbuhler would you be willing to take a look at this issue? I've verified that the MODELAMP [OIII] 5007 / 4959 ratio is the fixed, theoretical 2.993 ratio built into the code:
https://github.com/desihub/fastspecfit/blob/3.1.0/py/fastspecfit/emlines.py#L284

Note that because of the resolution matrix convolution we expect the AMP (observed) ratio to be close to, but not exactly 2.993.

However, since the line-width SIGMA is also an "intrinsic" quantity (i.e., the physical line-width, before the resolution matrix is applied) then we should definitely expect the FLUX ratio to be exactly 2.993, too.

I think the issue is here, where we render the (model) line-profile for each line:
https://github.com/desihub/fastspecfit/blob/3.1.0/py/fastspecfit/emlines.py#L856-L864

This is the line-profile we integrate to compute the (Gaussian) flux, but I think we need the line-profile without the resolution matrix:
https://github.com/desihub/fastspecfit/blob/3.1.0/py/fastspecfit/emlines.py#L811-L840

[dirkscholte]

@moustakas I also realised that I am using a criterium in my selection that only selects sources where [OIII]5007 BOXFLUX and FLUX are similar. So the result in BOXFLUX could be a result of that selection.

[dirkscholte]

@moustakas @jdbuhler I am just adding another figure that John suggested to make. Here still showing the [OIII]5007/[OIII]4959 ratio but now with NARROW_SIGMA on the x-axis.

Which shows a clear trend!