Opened 4 years ago

Last modified 4 years ago

#260 new defect - wrong answer

feature longward of Balmer, Paschen, etc jumps due to finite levels in H atom

Reported by: Gary J. Ferland Owned by: nobody
Priority: major Milestone:
Component: radiative transfer Version: trunk
Keywords: Cc:

Description

User inquired into origin of absorption feature visible longward of Balmer jump. Attached smoke.jpg shows emitted continuum for smoke test with feature marked. Produced with

test
save emitted continuum units Angstroms

Figure high_resolution.jpg shows origin of figure. This has continuum resolution increased by factor of ten and is hydrogen-only for simplicity. The emission with the default number of H I levels is black. The gap between the highest emission and the Balmer Jump is due to our finite number of levels. The strong emission in the highest level is due to our topoff of the model. This was done with

test
init "honly.ini"
set continuum resolution 0.1
save diffuse continuum units Angstroms "hemis.con"

The red line shows the effects of increasing the number of collapsed levels so that we go above n = 100. The gap is largely filled in, but the high-n lines do not smoothly join onto the peak of the BJ. Probably many more levels would increase the merged line emission to bring it up to continuously join onto the BJ. This plots was done with

test
init "honly.ini"
set continuum resolution 0.1
atom h-like collapsed levels 100
save diffuse continuum units Angstroms "large.con"

In nature there is no feature at the join between the highest line emission and the continuum emission in the BJ - the absorption oscillator strength (used to get emission) is continuous between high-n Balmer lines and the BJ. With enough levels we should be able to reproduce this.

We do have extra lyman lines in the iso sequences to get smoother behavior in the absorption spectra. These are not done for higher n and are not done for emission since the lines would need to be transferred and have good population data. LTE would be a good assumption for highly excited n levels.

Calculations done with trunk at r7798.

Attachments (3)

smoke.jpg (34.1 KB) - added by Gary J. Ferland 4 years ago.
smoke test emission
high_resolution.jpg (43.9 KB) - added by Gary J. Ferland 4 years ago.
high resolution around Balmer jump, with default and large H I levels
Liu+00.jpg (43.7 KB) - added by Gary J. Ferland 4 years ago.
BJ from Liu+-00

Download all attachments as: .zip

Change History (6)

Changed 4 years ago by Gary J. Ferland

Attachment: smoke.jpg added

smoke test emission

Changed 4 years ago by Gary J. Ferland

Attachment: high_resolution.jpg added

high resolution around Balmer jump, with default and large H I levels

comment:1 Changed 4 years ago by Gary J. Ferland

I have posted all of my helper files, including output and veusz files, on nublado.org data_area under etc / hemis_ticket260.zip

comment:2 Changed 4 years ago by Gary J. Ferland

further not from OP

I don't think it needs a time-consuming fix. It's also not

worth more CPU cycles calculating more lines either. You've already got more high-n lines in CLOUDY than any sane person ought to want! I don't think anyone has paid attention to the high-order lines question in over 40 years. By the time it was easy to observe the lines with modern linear detectors everyone had lost interest in the problem. (I remember 'Gene Capriotti saying to Bob Williams and I during a talk at some meeting that in his younger days they would have killed to get spectra as good as some speaker was showing of high-n Paschen lines, but the speaker couldn't care less about them; he was interested in something else in his spectra.)

I think the fix is easy. First, have a warning that CLOUDY is

only explicitly calculating lines up to such and such a point. Then beyond that treat the blended lines as a pseudo-continuum. You can calculate the number of photons in the lines per frequency/wavelength/energy unit (whatever CLOUDY likes) and then extrapolate to the series limit. This is what I did in my late 1970s program. There is actually astrophysical information in this pseudo-continuum, of course. It has the same slope as the bound-free continuum on the blue side, but as you go to lower n there comes an inflection point where it starts to dip down. The point where it curves down mostly depends on the density. In the bottom of some box somewhere I still have my old program for doing this but it would be faster for me to re-write it than to figure out which box the printout is in.

All the best,

Martín

Changed 4 years ago by Gary J. Ferland

Attachment: Liu+00.jpg added

BJ from Liu+-00

comment:3 Changed 4 years ago by Gary J. Ferland

figure Liu+00 from http://adsabs.harvard.edu/abs/2000MNRAS.312..585L does look a lot like our predicted BJ when we go up to n=100

Note: See TracTickets for help on using tickets.