Документ взят из кэша поисковой машины. Адрес оригинального документа : http://star.arm.ac.uk/~csj/software_store/resources/history.1997 Дата изменения: Wed Jun 12 12:58:29 2002 Дата индексирования: Tue Oct 2 11:22:17 2012 Кодировка: Поисковые слова: http astrokuban.info astrokuban

SPECTRUM - bug reports and modification history 1993 - 1997.

C.S.Jeffery

---------------------------------------------------------------------------
Executive Summary

Major changes:
- U: Partition functions tidied up
- HEI: improved wavelength sampling in HeI profiles (4026,4388,4471,4921)
- VCS: new hydrogen data from VCS, incl. Lyman,Balmer,Paschen lines
- COMMON_ALL: several arrays increased in size
- S2S: a separate utility for formatting line lists from Hubeny's data
- binary input data converted to ascii to enable dual O/S support
- extended tabular data for diffuse HeI profiles

Major bugs fixed:
- ATSCAN: unit conflict resolved
- OUTDIS: OPEN statement conflict resolved
- LHEII: correcte partition function treatment
- EWABUND: convergence failures trapped
- ATDATA: changed to free format read
- Kiel: argument type mismatch in SEATON

Persistent problems:
- CONFL: overflow in C/L ratio
- LSYNT: partition functions not correct
- LSYNT: continuous opacity needs improving
- LSYNT: problems with HeII lines
- ATSCAN: requires check for existing ATSCAN
- ATCOMP: need better way of defining composition for synthesis
- HEII: need ability to treat as blends with Balmer lines.
- HEI: need better wavelength sampling for 4026 and 4388.

------------------------------------------------------------------------
Journal of actions


2-FEB-1993

Some perceived changes to MAIN to create SPECTRUM (a different main tool)
Can these be done without disturbing MAIN ?

1. Make filename conventions standard (internally and externally)
done 4-FEB-1993 and applied to all files

2. Get model filename and linelist filename internally ?

3. Disallow cycling through models ?

4. Specify options and constants more conventionally ...


2-MAR-1993

Function U (-lkiel) has been modified to use TAP_PARFN (-ltap), because it
is only sensible to maintain a single version. Data for Sr, Y and Zr had
been added.

Function QAS (-lkiel) is redundant


4-MAR-1993

Increased paramter KI (number of atoms) to 40.

Changed format statement 201 in atdata.f, 100 in outdis.f and outter.f,
to allow for 10 character element names.

Changed data type of ELEM from REAL*8 to CHARACTER*10.

Moved ELEM from common block /IONS/ to new block /IONSC/

Shuffled order of variables in common blocks /LINE/ and /HE2/ (? forgotten its proper name) to remove misalignment warnings. Still some left !

LTE_LINES .... increased length of element identifier field from 8 to 10 characters in file atoms.d.


27-MAR-1995

ATSCAN, COMMON_ALL:
Unit NSCAN added for output fropm ATSCAN (conflict with NDISC in OUTDIS)

OUTDIS:
OPEN statement changed to STATUS='UNKNOWN' to resolve conflict with
existing logical link (SPECTRUM.TMP) defined in script Spectrum.

kiel / dp:
Both libraries contain u.f.
dp - is p.f. routines in line
kiel - link to TAP p.f. routines
The first needs to be removed, then ALL tested to ensure TAP is version correct.

CONFL:
When calculating abundances from equivalent widtsh, the output
for C/L RATIO=******* always overflows. I suspect iterative calculation
of CONFL in EQUIV or similar. Seems ok in normal mode.

HEI: (June 1995)
The forbidden component of HeI4921 (4910) was not adequately sampled
in wavelength. I increased the sampling. (Yes some stars show this
component).


VCS: (August 1995)
Michael Lemke supplied VCS tables for MANY Lyman, Balmer and
Paschen lines. I am implementing code to use these.
VCS_READ, VCS2, plus a modification to HSTARK.
LHI and LSYNT will also need to be modified.

3-AUG-1995

LHEII:
still used constant partition functions. Changed to use variable
partition functions.


VCS_READ: written / tested
VCS2: written / tested
HSTARK: modified (eventually: VCS->BALMER, VCS2->VCS)
LHI: 'lh1_new.f' written. / tested
LHI used constant partition functions. Changed.
LSYNT: Could leave just Balmer lines for now. LSYNT needs looking
at anyway because it doesn't deal with partition functions
properly and does not treat the continuous opacity properly.

7-AUG-1995

VCS_READ, VCS2, HSTARK, LHI, tested and give good results.

note - Lyman-alph fails in Hdef model
- Why are 'Kurucz' profiles for Halpha-delta 'better' than VCS
- What do we do about natural damping and resonance broadening in
hydrogen lines ?


14-AUG-1995

ATHEII

added He II 4859 and 4338 to lheii_red.d


16-AUG-1995

EWABUND

convergence also fails under other conditions, put a limit
of 25 iterations in.

17-AUG-1995

ATDATA

altered read format 101 (line data) to free format (*)
in order to accommodate variable length precision in the wavelength
(and other) data. Watch out for problems ...

1-DEC-1995

HEI:

Changed both 4921 and 4471 wavelength grids to asymmetric (ie read in all
wavelengths) so that 4910 and 4517 were both fully sampled.

note - problem with extrapolation of He I profiles to high density
currently switched off HeI extrapolation

12-DEC-1995

ATSCAN, LHI, HSEARCH:

In order to improve spectrum synthesis I included code for dealing
with VCS profiles in synthesis routines. I added some code to improve
frequency sampling in sparse line regions.

note - synthesis not valid for large spectral regions - can be fixed
- problems including 4686 and 4861 in synthesis
- problems including other HeII lines in synthesis


13-DEC-1995

SYNTH:

Proposal for a new structure.

subroutine SYNTH should parallel WHOLE
o wavelength scale (done in ATSCAN) done
o continuous opacity grid (CONT_OP)
o line opacity (LIN_OP) done
o radiative transfer (SYN_INTEG)
o remainder as in WHOLE.

ATSCAN:

crashes if ATSCAN already exists - need to check

ATCOMP:

Proposal for new structure

It should be possible to read in a file which defines the
composition for a synthetic spectrum separately from
having to edit the wavelength grid.


ATSCAN:

Proposal for modification to grid definition.

Addition of wavelength points for metal lines should only
proceed up to half the interval between the current line and
the next nearest lines on either side.


COMMON_OPAC:

B star synthetic spectra were giving wrong results. Found
that common_all.f had evolved beyond common_opac.f, so the
qub library was entirely out of date.

Easily corrected by recompiling and linking.



15-MAR-1996:

S2S:

I wrote a natty little utility for converting parts of Hubeny's versions
of the Kurucz linelists into a format suitable for inserting into
the linelist data file. It's not an ideal way to do it, but it works
and allows me to do spectrum synthesis for limited sections of UV spectrum.
Works very nicely, although the spectral resolution in the code
is currently lower than the GHRS!

KIEL opacity routines:

Tried to synthesize UV spectra for Kiel model atmosphere (H-rich).
I got no lines at 1335 or at 1398/1402, but did get some weak CIV
lines at 1550. I got good results at 4267. Where I got no lines,
TAUC had a very low value (6.3E-7 = lower limit of atmosphere)
right across the line.
I got good results at 1340-45 A for a QUB standard model (t25g40).
I 'editted' the Kiel model to make Spectrum think it was a QUB model,
and got good results at 1340-45 A (similar to above).

I conclude there is a bug in the Kiel opacity routines which is
setting the continuous opacity to a very high value below or
around some critical frequency.

By making tests with a series of dummy lines, I tracked the critical
wavelength to 1420 A. The error turns out to be a call to
SEATON by sb OPKCA2 where the type of constants in the argument list
do not match the double precision type of the dummy arguments in
SEATON.

- The problem is fixed.

Lyman-alpha:

Still having problems with some of the new VCS profiles,
including Lyman alpha.

20-MAR-1996

Lyman-alpha:

LHI_ID

The code for identifying lines was incorrect, and is now fixed.
Lyman-alpha was still appearing much weaker than expected.

LHI

The code for the Boltzmann equation in LHI has the excitation
potential hardwired for the Balmer series!! Changed the equation
to include EH(NLOW) instead of 0.75*CHYDEV, and entered data for
EH into COMMON_VCS.

I have a nagging doubt about the expression for FEX, because
for resonance lines the quantity exp(-CEVTOT*EXCIT(NLINE)/T) = 1,
so the sum over all levels will be greater than 1...
IS THIS RIGHT ?

Otherwise the Ly-alpha models are now looking good.

18-APR-1997

VCS and ATHEI

Three data files (balmer, prof_4471 and prof_4922) were being read as
binary files. This makes portability a minor problem. The convenient
solution (not the computationally most efficient) was to convert the
binary profiles to ascii and to convert the code to read the ascii files.

29-APR-1997

TRANSFER, I_INTEG and F_INTEG

There are problems -
1) redesign the makefile(s) so that the code builds correctly under
both Solaris and Dec/Unix.

2) model :
t20g40
lines : csj pld ml-vcs extrap-off
1 4101 12 5 6.262 6.262 6.260
1 4343 12 5 6.024 6.025 6.02
1 4861 12 5 5.856 5.856 5.847
1 6560 12 5 4.322 4.322
2 4471 11 5 1.641 2.738 2.738
2 4922 11 5 0.920 1.596 1.596

runs ok, although equivalent widths do not look great, they
are identical on Solaris and Dec systems. My results - csj
Checking equivalent widths with Philip Dufton - pld

3) model :
/simon/stars/models/0099001/1725L.q
lines :
1 4101 8 10
1 4343 8 10

causes run-time problems but only for Feutrier solution of transfer
equation. Problem traced to alignment of variables in
COMMON /source_block/ which is only referenced by subroutines
TRANSFER, I_INTEG and F_INTEG. Problem was cleared.

30-APR-1997

Cleared out redundant files from all folders, especially from
/data and data/utils/. The latter were put into /archive.

NOTE: Spectrum only reads in a reduced set of HeII line profiles
at present, including 10125, 5412, 4859, 4542, 4339, 4200 and 4686.
What is it doing for blends with Hydrogen Balmer lines ? The HeII
data is available in a more complete dataset.

1-MAY-1997

Numerical profiles for Balmer lines and HeI lines are wrong -
see above (4343, 4722 and 4922)

2-MAY-1997

HSTARK

Following a bug report from pld, corrected HSTARK to use old
versions of tables for Halpha - Hepsilon, rather than Halpha - Hdelta.

ATHEI

Converted the disk HeI profiles to logarithmic ascii form. Saves some
arithmetic, but doesn't fix the numerical problem with the equivalent widths.

LHI_ID - H-gamma

I changed HSTARK over so that Hbeta - Hepsilon are computed
from Lemke's VCS tables - and I get exactly the same error occurring
for Hgamma. Hbeta and Hdelta are OK to within errors.

I had been using 4343 as the wavelength for H-gamma! When I changed it
to 4340.5 (hey presto!) the answers started to come out right. On
experiment I found I could put in things like
1 4097 12 5
and it would recognise it as H-delta, but give an equivalent width that
was slightly incorrect. The failure comes at exactly 2A error in wavelength,
but there are minor changes in equivalent width for smaller errors.
I have inserted a trap in LHI_ID to detect wavelengths out by more than
1 Angstrom. A warning is given, but the program will attempt to continue
without taking corrective action.

This does not seem to be the cause of the problem for 4471 and 4922

5-MAY-1997

HEIDIF

There is a section in HEIDIF where I switch extrapolation in electron
density off in order to avoid problems (see above). The
version I had running had the switching code commented out (for what
reason I cannot imagine), so that I was extrapolating the HeI
profiles at high electron density. When I replaced the code, and thus
switched off the extrapolation I reproduced pld's results exactly.

I obtained the following equivalent widths for a series of models
with and without high-density extrapolation for HeI 4471. Note that
WITH extrapolation, 4471 is weaker than 4026. It is important
to realise that line shifts cause high-ne extrapolation to REDUCE the
opacity in the line cores, which is opposite to that expected. It is not
really better to avoid extrapolation; it is simply less wrong. The
lines should probably be stronger than we have calculated, but we
do not have the necessary theoretical profiles. I think this is IMPORTANT for
B stars as well as peculiar objects like helium stars.

Alain Beauchamp (1997: ApJ Supp 108, 559) has calculated profiles for 21
HeI lines and their forbidden components to high electron densities
(6.E17/cm^3). The BCS tables only go to 1.E16/cm^3.
However, I have not yet had success in obtaining the data from Montreal.
Beachamp has left astronomy.

Note that at present extrapolation is still done for 4026 and 4388

.....
Source function: Planck function.
He abundance: 11.0, microturbulent velocity 5 km/s:

4471 extrap-on extrap-off 4026 4388
t20g45 2.018 3.457 2.310 1.167
t20g40 1.641 2.738 1.854 0.965
t20g35 1.267 2.078 1.388 0.751
t20g30 0.885 1.496 0.941 0.539
......

*** 20-MAY-1997
***
I had left another bug in HEIDIF which prevented interpolation
at lower densities than the maximum density tabulated.
So the extrap-off version was also giving suspect results.

***

EWABUND

Under rare circumstances; the iteration for abundance would
oscillate about a solution with a radius slightly greater than
the convergence criterion. On exit from the DO loop, WHOLE
would crash - for reasons not yet determined.

The convergence criterion has been relaxed for the time being
to give equivalent width better than 0.1% or abundance better than 1%.

HEI DIFFUSE LINES

- spectrum -
CHECK
4026 Shamey 1969 - log ne = 14 - 16; 40,000 K
4388 Shamey 1969 - log ne = 14 - 16; 40,000 K
4471 Barnard Cooper & Smith 1974; log ne = 13 - 16; 4 temperatures OK
4921 Barnard Cooper & Smith 1975; log ne = 13 - 16; 4 temperatures

- synspec -

4026 Shamey 1969 - log ne = 14 - 17.5; 4 temperatures
4388 Shamey 1969 - log ne = 14 - 17.5; 4 temperatures
4471 Barnard Cooper & Smith 1974; log ne = 13 - 16; 4 temperatures
4921 Barnard Cooper & Shamey 1969; log ne = 14 - 17.5; 4 temperatures

Note: BCS69 do not print intermediate densities (15.5,16.5,... etc),
but offer them for private communication. I do not have them


Changes:
Incorporate high density profiles for 4471 and 4921
Incorporate high densities and all temperatures for 4388 and 4026

The file names are changed as follows

prof_4921.d -> he1_4921.bcs75
prof_4471.d -> he1_4471.bcs74

he1_4471.bcs74 + he1_4471.bcs69 -> prof_4471.d
he1_4921.bcs75 + he1_4921.bcs69 -> prof_4921.d
he1_4026.sha69 -> prof_4026.d
he1_4388.sha69 -> prof_4388.d

HeI line profile file formats are defined by extensions .1 and .2, thus
xxx.1 - Hubeny et al. format - kappa
xxx.2 - Spectrum format - log(kappa)
Convert using program hei_conv.f

All but the prof_files are moved to the archive folder

20-MAY-1997 COMMON_ALL, ATHEI, HEIDIF

common block HEIPROF -> HEI and integrated,
same procedures used for reading and interpolating in all four profiles

Startlingly it works - EXCEPT 4471 and 4388 look more like
'extrapolated' results (see above). This is because I fixed a
bug in the 'extrapolation' limit which was also preventing
interpolation at high density.

4471 4921 4026 4388
t20g45 2.017 1.146 2.202 1.195
t20g40 1.640 0.919 1.767 0.948
t20g35 1.265 0.699 1.335 0.731
t20g30 0.883 0.499 0.915 0.520

I "think" this is now working.

21-MAY-1997

WAVEG

Modified to generate better wavelength grids for 4026 and
(in particular) 4388. Both
have forbidden components, well-separated from the main
line centre. 4383.3 shows up in computed profiles (Shamey data)
for helium stars.
The component at 4045.2 is not present in the Shamey tables.

3-JUNE-1997

LSYNT

Modified to allow for inclusion of HeII lines.
Line opacity constant changed for each line in synthesis
(affected H and HeII)

INTEG_FEAUT

Uses local continuous opacity for spectral synthesis.
It is recalculated every DELTA * lambda. The continuum flux
is interpolated for every wavelength point, so that normalization
is done accurately. Currently DELTA = 0.001.

EQUIV

New normalization ofr synthetic spectra (NMODE=10, NSOURCE=3)


Other routines to change:

ATSCAN - include all Balmer lines ?
ATSCAN - search for HeII lines - INTEG_FEAUT gives rubbish answers if a
HeII line is included in the synthesis.
HESEARCH - not compatible with HEIDIF: i.e. gf-values and wavelength grids